Manufacturing Cost Analysis 100 and 250 KW Fuel Cell Systems Primary Power
Manufacturing Cost Analysis 100 and 250 KW Fuel Cell Systems Primary Power
Manufacturing Cost Analysis 100 and 250 KW Fuel Cell Systems Primary Power
Prepared for:
U.S. Department of Energy
Golden Field Office
Golden, CO
DOE Contract No. DE-EE0005250
January 2016
This report is a work prepared for the United States Government by Battelle. In no event shall either the
United States Government or Battelle have any responsibility or liability for any consequences of any use,
misuse, inability to use, or reliance upon the information contained herein, nor does either warrant or
otherwise represent in any way the accuracy, adequacy, efficacy, or applicability of the contents hereof.
Manufacturing Cost Analysis of 100 and 250 kW Fuel Cell Systems for Primary Power and Combined Heat and Power Applications /
DOE Contract No. DE-EE0005250
Acknowledgements
Funding and support of this work by the U.S. Department of Energy, Fuel Cell Technologies Office is
gratefully acknowledged.
Collaborators
The following have provided assistance in the form of design inputs, cost inputs, design review and
manufacturing review. Their valuable assistance is greatly appreciated.
Hydrogenics
Ballard
Nexceris
Johnson Matthey/Catacel
US Hybrid
Panasonic
Advanced Power Associates
Ideal Power
Innovatek
Zahn Electronics
Watt Fuel Cell
Vicor Power
SMA-America
Outback Power Technologies
Tranter
API Heat Transfer
Dry Coolers
Cain Industries
Battelle would also like to thank Kevin McMurphy, Adria Wilson and Tom Benjamin for strengthening the
report through their thorough reviews.
Executive Summary
Fuel cell power systems may be beneficially used to offset all or a portion of grid-purchased electrical
power and supplement on-site heating or cooling requirements. For this application the fuel of choice is
usually pipeline natural gas or on-site propane storage. These fuel sources generally have much higher
reliability than utility electric power, being less subject to damage-related outages, and can therefore
provide for some continued operation in the event of grid outage – performing both primary power and
back-up power functions. Battelle evaluated low temperature polymer electrolyte membrane (LTPEM)
and solid oxide fuel cell (SOFC) systems for use as continuous power (primary power) and as a source
for auxiliary heating in combined heat and power (CHP) configurations. The power levels considered for
this portion of the project were 100 kW and 250 kW. A significant primary-power/CHP commercial market
has not yet developed in this size range; however, our analysis suggests an attractive business
opportunity under the right conditions. Indeed, FuelCell Energy, Bloom, and Doosan are approaching the
primary power market with three different fuel cell technologies (molten carbonate, solid oxide, and
phosphoric acid respectively) and are achieving some market penetration. Based on the installed
applications from these companies, we believe that fuel cell systems of this size will primarily serve
medium to large commercial buildings, data centers, product distribution centers and light industrial
fabrication/manufacturing sites. We also suggest that utility operators and municipality operated power
companies may find these systems attractive for distributed generation and electric power islanding
support.
In the absence of a developed market and publicly available system configuration information, Battelle’s
evaluation included the definition of representative systems that could serve this market. The
representative system concepts were subjected to detailed cost evaluation based on industry feedback
and the application of standard design for manufacturing and assembly analysis methods, including the
application of the Boothroyd-Dewhurst DFMA® software for specific hardware and assembly evaluation.
A sensitivity analysis was performed to evaluate the influence of high cost/custom subassemblies and
components with a high degree of cost uncertainty. Commercially available components and commodity
materials generally have a low level of variability and do not significantly impact the overall economics.
Because some of the commercially available balance of plant (BOP) hardware for systems in this range is
commonly configured for different operating conditions (typically higher pressure), some assumptions are
required regarding customization of hardware such as heat exchangers.
The evaluation shows that the highest cost category for these systems is the balance of plant hardware
directly related to connecting to the grid – the inverter. For this evaluation we selected recently
developed hybrid inverters which incorporate a separate direct current I/O port for connection to batteries.
These inverters, while initially more expensive than a more conventional inverter, eliminate the need for a
DC/DC converter to match battery voltage as we have used in prior system analyses. Eliminating the
DC/DC converter saves significant cost though the power electronics still represent the highest cost
system component. Hybrid inverters are relatively new technology developed primarily for solar
applications, and as such, their costs, particularly for higher volumes, are not well defined. Therefore,
while the selection of the hybrid inverter does represent some uncertainty, the benefit is so clear for both
fuel cell and solar installations that we believe some version of hybrid inverter will be available for fuel cell
systems.
Heat exchangers, particularly high temperature heat exchangers, also represent a major portion of the
balance of plant. For the PEM system the most expensive heat exchanger was the steam
superheater/reformate cooler. Our cost for this heat exchanger was based on a commercially available
heat exchanger that is overdesigned for the application and is hence probably more expensive than
necessary. A designed-to-purpose heat exchanger may offer notable savings.
PEM stack costs were less than 30% of overall system cost for all sizes and production volumes
considered – typically less than 20% at the higher production volumes. SOFC stacks, however,
represented a notably larger fraction of overall costs as a result of a much simpler balance of plant.
A sensitivity analysis of some of the major cost contributors shows the potential for further cost reductions
through design-to-purpose engineering for some high cost items, especially for heat exchangers. Unlike
prior studies, we found the price of platinum to have a minor overall impact on the PEM system cost. This
primarily results from the relatively small overall contribution of the stack to system costs at this size. As
noted above, a significant cost benefit at the system level was achieved by using a hybrid converter in
place of a DC/DC converter and separate DC/AC inverter. Although not specifically considered in the
sensitivity analysis, we outline a possible method of directly connecting batteries in parallel with the fuel
cell that may be satisfactory for some applications, providing another path to cost reduction.
A lifecycle cost analysis (LCCA) was performed for a nominal commercial installation requiring electrical
power and additional heating or cooling energy input. With or without a supplemental heat requirement
the LCCA showed a significant positive return on investment for locations where electricity prices are high
and natural gas prices are low. For applications with a need for supplemental space or process heating
the number of attractive geographic installation sites increases. The analysis considered the utility costs
for specific states, finding several to be attractive for fuel cell installations. However, when the overall gas
and electricity cost averages for the US were considered, the return on investment was not good until
high production volumes are reached suggesting that additional capital cost reduction is needed to
achieve widespread acceptance.
Our analysis shows that the stack costs, while a significant portion of the overall system costs, may not be
the most critical area in need of development for achieving attractive initial costs. Instead, engineering
activities focused on addressing cost reductions for power electronics (all systems) and heat exchangers
(primarily PEM systems) are appropriate. SOFC systems were found to have better return on investment
(ROI) than PEM systems based on initial costs and overall efficiency, however, the lifetime for SOFC
stacks has not yet been demonstrated to the necessary level under deployed operating conditions.
Table of Contents
Page
Acknowledgements ...................................................................................................................... i
Executive Summary .................................................................................................................... ii
1. Introduction ............................................................................................................................. 1
2. Approach ................................................................................................................................ 1
3. Market Analysis ....................................................................................................................... 3
3.1 Market Requirements and Desired Features ................................................................. 4
3.2 Technology Selection .................................................................................................... 5
3.3 Market Analysis Conclusion .......................................................................................... 5
4. System Specifications ............................................................................................................. 6
4.1 General Description ...................................................................................................... 6
4.2 Nominal Metrics ............................................................................................................ 7
4.3 System Sizing and Operation ........................................................................................ 9
4.4 System Configuration ...................................................................................................11
4.4.1 LTPEM System .....................................................................................................12
4.4.2 SOFC System.......................................................................................................15
4.5 Electrical System..........................................................................................................18
4.5.1 Overview ..............................................................................................................18
4.5.2 On-Grid Operation ................................................................................................18
4.5.3 Off-Grid Operation ................................................................................................19
4.5.4 System Configurations ..........................................................................................20
4.5.5 Thermal Management...........................................................................................26
4.5.6 Wiring and Ancillary Components .........................................................................26
5. Manufacturing Cost Analysis ................................................................................................. 26
5.1 System Cost Scope ......................................................................................................26
5.2 System Cost Approach.................................................................................................26
5.2.1 System Manufacturing Cost Assumptions .............................................................27
5.2.2 Machine Costs ......................................................................................................27
5.2.3 Material Costs.......................................................................................................28
5.3 PEM System Manufacturing Costs ...............................................................................28
5.3.1 PEM Stack Manufacturing Process.......................................................................30
5.3.2 PEM Systems BOP Manufacturing Cost Assessment ...........................................41
5.3.3 PEM BOP Cost Assumptions................................................................................47
5.3.4 PEM System Assembly and Learning Curve Assumptions ...................................51
5.3.5 PEM CHP System Testing ....................................................................................51
5.3.6 PEM Capital Cost Assumptions ............................................................................52
5.4 SOFC System Manufacturing Costs .............................................................................53
5.4.1 SOFC Stack Manufacturing Process ....................................................................55
5.4.2 SOFC Systems BOP Manufacturing Cost Assessment .........................................70
5.4.3 SOFC BOP Cost Assumptions .............................................................................73
5.4.4 SOFC System Assembly and Learning Curve Assumptions .................................77
5.4.5 SOFC System Testing ..........................................................................................77
List of Tables
Page
Table 4-1. Nominal Design Basis. .............................................................................................. 8
Table 5-1. General Process Cost Assumptions .........................................................................27
Table 5-2. PEM Fuel Cell Design Parameters ...........................................................................29
Table 5-3. PEM MEA Material Cost Summary - 100 kW and 250 kW ........................................33
Table 5-4. PEM MEA Cost Summary - 100 kW and 250 kW .....................................................33
Table 5-5. PEM End Plate Cost Summary - 100 kW and 250 kW..............................................35
Table 5-6. PEM Anode Bipolar Plate Cost Summary - 100 kW and 250 kW ..............................35
Table 5-7. PEM Cathode Bipolar Plate Cost Summary – 100 kW and 250 kW ..........................36
Table 5-8. PEM Anode and Cooling Seal Cost Summary - 100 kW and 250 kW .......................36
Table 5-9. PEM Cathode Seal Cost Summary – 100 and 250 kW Systems ..............................37
Table 5-10. PEM Stack Assembly Costs - 100 kW and 250 kW Systems..................................37
Table 5-11. PEM Stack Testing and Conditioning Cost Summary - 100 kW and 250 kW ..........38
Table 5-12. PEM Stack Component Cost Summary - 100 kW and 250 kW ...............................39
Table 5-13. PEM Stack Manufacturing Cost Summary - 100 kW and 250 kW ...........................39
Table 5-14. PEM Reformer Sizing Parameters..........................................................................44
Table 5-15. PEM Reformer Dimensional Summary ...................................................................44
Table 5-16. PEM System Reformer Cost Summary - 100 kW and 250 kW ...............................44
Table 5-17. PEM Shift/PrOx Reactor Sizing Recommendations ................................................45
Table 5-18. PEM Shift/PrOx Reactor Dimensional Summary ....................................................45
Table 5-19. Shift and PrOx Reactor Manufacturing Costs .........................................................47
Table 5-20. PEM BOP Cost Summary for 100 kW and 250 kW systems...................................48
Table 5-21. PEM System Assembly Costs - 100 kW and 250 kW .............................................51
Table 5-22. PEM System Testing Cost Summary - 100 kW and 250 kW...................................51
Table 5-23. Summary of PEM Capital Cost Assumptions ..........................................................52
Table 5-24. PEM Capital Cost Summary - 100 kW and 250 kW ................................................53
Table 5-25. SOFC Fuel Cell Design Parameters .......................................................................54
Table 5-27. SOFC End Plate Cost Summary - 100 kW and 250 kW .........................................61
Table 5-28. SOFC Interconnect Cost Summary - 100 kW and 250 kW .....................................61
Table 5-29. SOFC Anode Frame Cost Summary - 100 kW and 250 kW ...................................62
Table 5-30. SOFC Picture Frame Cost Summary - 100 kW and 250 kW...................................63
Table 5-31. SOFC Cathode Frame Cost Summary - 100 kW and 250 kW ................................63
Table 5-32. SOFC Laser Welding Cost Summary - 100 kW and 250 kW ..................................64
Table 5-33. SOFC Ceramic-Glass Sealing Cost Summary - 100 kW and 250 kW .....................64
Table 5-34. SOFC Anode Mesh Cost Summary - 100 kW and 250 kW .....................................65
Table 5-35. SOFC Cathode Mesh Cost Summary - 100 kW and 250 kW ..................................65
Table 5-36. SOFC Stack Assembly Cost Summary - 100 kW and 250 kW................................66
Table 5-37. SOFC Stack Brazing Cost Summary - 100 kW and 250 kW ...................................66
Table 5-38. SOFC Stack Testing and Conditioning Cost Summary - 100 kW and 250 kW ........67
Table 5-39. SOFC Stack Component Cost Summary - 100 kW and 250 kW.............................68
Table 5-40. SOFC Stack Manufacturing Cost Summary - 100 kW and 250 kW.........................68
Table 5-41. SOFC Reformer Sizing Parameters .......................................................................72
Table 5-42. SOFC Reformer Dimensional Summary .................................................................72
Table 5-43. SOFC System Reformer Cost Summary - 100 kW and 250 kW .............................73
Table 5-44. SOFC BOP Cost Summary for 100 kW and 250 kW Systems ................................74
Table 5-45. SOFC System Assembly Costs - 100 kW and 250 kW ...........................................77
Table 5-46. SOFC System Testing Cost Summary - 100 kW and 250 kW ................................77
Table 5-47. Summary of SOFC Capital Cost Assumptions .......................................................78
Table 5-48. SOFC Capital Cost Summary - 100 kW and 250 kW ..............................................78
Table 5-49. DC/DC Converter Costs per Watt ...........................................................................80
Table 5-50. DC/AC Inverter Costs per Watt ..............................................................................80
Table 5-51. Hybrid DC/AC Inverter Cost per Watt .....................................................................80
Table 5-52. Controller and Sensors Cost ..................................................................................81
Table 5-53. Minimum Battery Requirements .............................................................................81
Table 5-54. Battery Cost per System ........................................................................................82
Table 5-55. Load Bank Cost per System ...................................................................................82
Table 5-56. Electrical Cost Summary ........................................................................................83
Table 6-1. PEM Manufacturing Processes Evaluated................................................................84
Table 6-2. SOFC Manufacturing Processes Evaluated. ............................................................86
Table 7-1. Cost summary for 100 kW PEM CHP Fuel Cell System. ..........................................89
Table 7-2. Cost summary for 250 kW PEM CHP Fuel Cell System. ..........................................90
Table 7-3. Cost Summary for 100 kW CHP SOFC Fuel Cell System ........................................94
Table 7-4. Cost Summary for 250 kW CHP SOFC Fuel Cell System ........................................95
Table 9-1. Cost Basis for California .........................................................................................108
Table 9-2. Operating Costs for California ................................................................................109
Table 9-3. Annual Cash Flows, IRR and Payback Period (100 kW SOFC system with waste
heat recovery) ..................................................................................................................110
Table 9-4. Annual Cash Flows, IRR and Payback Period (100 kW PEM system without waste
heat recovery) ..................................................................................................................111
Table 9-5. IRR and Payback for Installations in California .......................................................112
Table 9-6. IRR and Payback for Installations in New York State .............................................113
Table 9-7. IRR and Payback for Installations Based on Average U.S. Energy Costs ..............114
Table 9-8. Direct Utility Price Comparison ...............................................................................115
Table 10-1. PEM System Cost ................................................................................................119
Table 10-2. SOFC System Cost ..............................................................................................119
Table 10-3. IRR in California based on average commercial rates ..........................................122
List of Figures
Page
Figure 2-1. Battelle’s cost analysis approach. ............................................................................ 2
Figure 4-1. Notional load curve for illustration. ..........................................................................10
Figure 4-2. High level fuel cell system outline............................................................................11
Figure 4-3. Representative Multi-stack LTPEM CHP system.....................................................14
Figure 4-4. Representative SOFC CHP system. .......................................................................16
Figure 4-5. DC/DC Converter Electrical System Schematic. .....................................................22
Figure 4-6. Hybrid Inverter Electrical System Schematic. ..........................................................23
Figure 4-7. Directly-Connected Battery System Schematic. ......................................................24
Figure 4-8. PEM Fuel cell polarization curve with battery state of charge. .................................25
Figure 5-1. PEM fuel cell stack manufacturing process. ............................................................31
Figure 5-2. PEM MEA configuration for 780 cm2 active area. ....................................................32
Figure 5-3. PEM end plate size. ................................................................................................34
Figure 5-4. Breakdown of 100 kW system – 60 kW fuel cell costs and production volume trends.
............................................................................................................................................40
Figure 5-5. Breakdown of 250 kW system – 50 kW fuel cell costs and production volume trends.
............................................................................................................................................40
Figure 5-6. Catacel SSR® schematic diagram after Patent 7,501,102. ......................................42
Figure 5-7. Three-tube 100 kW reformer configuration. .............................................................43
Figure 5-8. Seven-tube 250 kW reformer configuration .............................................................43
Figure 5-9a. Shift/PrOx reactor configuration. ...........................................................................46
Figure 5-9b. Shift/PrOx reactor typical dimensions. ...................................................................46
Figure 5-10. 100 kW PEM BOP cost distribution. ......................................................................49
Figure 5-11. 250 kW PEM BOP cost distribution. ......................................................................49
Figure 5-12. 100 kW PEM BOP cost volume trends. .................................................................50
Figure 5-13. 250 kW PEM BOP cost volume trends. .................................................................50
Figure 5-14. Detail assembly of SOFC Cell (not to scale, parts depicted as shown for clarity) ..55
Figure 5-15. Cell repeat unit showing function of major components.........................................56
Figure 5-16. SOFC stack manufacturing process ......................................................................57
Figure 5-17. SOFC cell size. .....................................................................................................58
Figure 5-18. SOFC end plate size .............................................................................................60
Figure 5-19. 100 kW SOFC stack cost volume trends. ..............................................................69
Figure 5-20. 250 kW SOFC stack cost volume trends. ..............................................................69
Figure 5-21. Catacel SSR® schematic diagram after Patent 7,501,102. ....................................70
Figure 5-22. Three-tube 100 kW reformer configuration. ...........................................................71
Figure 5-23. Seven-tube 250 kW reformer configuration. ..........................................................71
Figure 5-24. 100 kW SOFC BOP cost distribution. ....................................................................75
Figure 5-25. 250 kW SOFC BOP cost distribution. ....................................................................75
Figure 5-26. 100 kW SOFC BOP cost volume trends. ...............................................................76
Figure 5-27. 250 kW SOFC BOP cost volume trends. ...............................................................76
Figure 7-1. 100 kW PEM system costs at 1,000 units per year. ................................................88
Figure 7-2. 250 kW PEM system costs at 1,000 units per year. ................................................88
1. Introduction
Battelle is conducting manufacturing cost assessments of fuel cells for stationary and non-automotive
applications to identify the primary cost drivers impacting successful product commercialization. Battelle,
under a five-year cooperative agreement with the Department of Energy’s (DOE’s) Fuel Cell Program, will
provide an independent assessment of fuel cell manufacturing costs at various volumes and alternative
system designs. This report provides cost estimates for the manufacture of 100-kW and 250-kW fuel cell
systems for combined heat and power (CHP) and primary power applications. Both polymer electrolyte
membrane (PEM) fuel cell stacks and solid oxide fuel cell (SOFC) stacks are considered. This report
identifies the manufacturing costs of fuel cell systems using scale-appropriate manufacturing processes
at annual production volumes of 100, 1000, 10,000 and 50,000 units. The manufacturing volumes were
defined by DOE and are used for all systems being evaluated within the overall project. For 100-kW and
larger systems, the higher manufacturing volumes would represent a significant fraction of the new power
generating capacity brought on line in a typical year. Thus the higher volumes represent a mature market
with fuel cells drawing market share from the gas turbine systems that now dominate new generation
capacity. In such a market, we believe utilities and municipal power companies will find fuel cells
advantageous for distributed generation to enable islanding and localized damage response.
The system designs were defined based on Battelle’s fuel cell system integration expertise and were
refined through discussion with industry partners. The report presents our representative designs for both
SOFC and PEM systems, including the basic sizing and configuration design assumptions. For SOFC
systems we elected to analyze planer configurations as these seem to be preferred over tubular
configurations by the active companies working at this size range (100 to 250 kW). Key components of
the example designs were evaluated using manufacturing processes modeled with the Boothroyd-
Dewhurst Design for Manufacture and Assembly (DFMA®) software. Costs of the system, sub-system,
and specific components were determined by obtaining quotes from candidate manufacturers and the
main cost drivers were identified through a sensitivity analysis. The sensitivity analysis includes the costs
of some of the more expensive components and those for which the included cost is less certain. A
summary of possible opportunities for cost reduction is included. Because balance of plant hardware for
the scale of equipment required for 100 and 250 kW systems is typically manufactured at a few to
perhaps as many as a few hundred units/year, the assumptions for the higher production volumes must
be understood to reflect engineering judgement as to the level of cost reduction possible through specific
design for mass-manufacturing that would necessarily occur to support the higher volume production
rates.
2. Approach
Battelle’s cost analysis methodology is a four-step approach (Figure 2-1):
Step 1—Market Assessment
Step 2—System Design
Step 3—Cost Modeling
Step 4—Sensitivity Analysis/Lifecycle Cost Analysis
This approach has been successfully applied to previous cost analyses developed by Battelle. 1,2
The first step in our methodology, Market Assessment, assures that we have selected the right fuel cell
type and appropriate production volumes to meet market requirements. In this step, Battelle identified the
operational and performance requirements (e.g., hours of operation, frequency, expected lifetime) of the
target application and market. Using this information, we defined an assessment of the user requirements
for a fuel cell product. For this phase of the project Battelle completed a quick survey of the market
through dialogue with industry to estimate the number of potential units in the market and the expected
market growth for fuel cells in CHP and primary power applications in the 100kW to 250kW range. This
information formed the basis for selecting the system design and fuel cell types best suited to meet user
requirements and the appropriate production volumes to consider in the modeling exercise.
Step 2, System Design, began with a literature review of fuel cell designs for CHP and primary power
applications including component design and manufacturing processes. Possible improvements in system
design and manufacturing were identified and incorporated into the example systems. From these results,
the basic construction and operational parameters for a fuel cell stack and system were defined, along
with potential improvements. The fuel cell system design did not focus on an individual manufacturer’s
designs, but was instead representative of typical design based on literature and Battelle’s engineering
expertise. The stack and system design were vetted with industry stakeholders to ensure feasibility of the
design, identify possible improvements, and determine current and alternative manufacturing approaches.
The finalized design and projected improvements were consolidated to form the basis for developing the
bill of materials (BOM). Decisions were then made about which components would be manufactured
internally and which would be outsourced. For internally manufactured components (including applicable
balance of plant (BOP) components), manufacturing processes and production equipment were defined
in detail.
In Step 3, Cost Modeling, Battelle gathered vendor quotes for material costs, production equipment and
outsourced components. Where necessary, custom manufacturing process models were defined and
parametrically modeled based on knowledge of the machine, energy and labor requirements for the
individual steps that comprise the custom process. The sequence of actions required to assemble the
1
Battelle. 2011. The High Volume Manufacture Cost Analysis of 5 kW Direct Hydrogen Polymer Electrolyte Membrane (PEM) Fuel Cell for
Backup Power Applications. Contract No. DE-FC36GO13110.
2
H. Stone, K. Mahadevan, K. Judd, H. Stein, V. Contini, J. Myers, J. Sanford, J. Amaya, and D. Paul. 2006. Economics of
Stationary Proton Exchange Membrane Fuel Cells, Interim Report. Contract No. DE-FC36GO13110.
components and test the final fuel cell system was developed and analyzed for cost reduction
opportunities through component consolidation and process optimization. Manufacturing quality control
requirements were based on input from equipment vendors and Battelle’s experience with product
manufacturing. Outsourced component costs were estimated through vendor quotes. Mathematical
functions for scaling factors were developed to estimate the changes to outsourced components and
material costs as a function of production volumes when vendor quotes for higher volumes were not
available. These were derived using engineering rules of thumb and estimates from other manufacturing
processes and considered impacts on system design.
Using the DFMA® software, component costs calculated from both custom and library manufacturing
processes and the outsourced components were incorporated into the assembly and test sequence
models to determine the final cost of producing the fuel cell systems. The output of the DFMA® models
was also used to calculate production line utilization. The calculated value determined the number of
individual process lines required to support various product demand levels. The manufacturing capital
cost model is based on the number of process lines required. We assumed that capital equipment
expenditures for production would be amortized over a 20-year period and that the annual amortized cost
would be distributed over the production volume for that year. The financial assumptions that were used
are consistent with the DOE Hydrogen Analysis (H2A) model. Total fuel cell system costs including capital
expenditures were then estimated for the baseline system and projected improvements. Details of these
costing calculations appear in Appendices A-1 through A-26.
The Sensitivity Analysis (Step 4) was then performed to determine which design parameters or
assumptions have the most effect upon the stack and system cost. Single factor sensitivity analysis were
performed and helped determine the impact of individual parameters on system costs. Based on these
results, Battelle outlined possible design optimization approaches to reduce the total fuel cell system cost
and total cost of ownership.
A preliminary Life Cycle Cost Analysis was developed for the systems evaluated in this portion of the
overall project. Life cycle analyses are necessarily tied to specific applications. For this study we
considered installations of alternative fuel cell power generation in the range of 100 kW to 1 MW of total
power requirements. Most installations depend on the fuel cell system to reduce demand charges and
overall electrical power costs and to provide back-up power in the event of grid outage. We believe that
using the waste heat may be a decision factor in selecting a fuel cell power system even for those
applications where the primary interest is reduced grid power demand or back-up power during grid
outage. The waste heat may be used directly for space or water heating (e.g. hotels, conference centers)
or indirectly to provide cooling through adsorption chillers (e.g. data and communication
centers). Therefore, we consider some hypothetical installations where the heat may be used to reduce
overall energy consumption.
3. Market Analysis
The fuel cell primary power and CHP markets in the United States are young, although their use has
been growing over the past twenty years. The market is served by three primary companies, each taking
different approaches in their market offerings. None of the companies directly align with the systems
examined in this report (explicitly, 100 and 250 kW PEM and SOFC CHP systems), although similar
market needs are served. FuelCell Energy offers large primary power and CHP systems in the 250 kW
operating range, utilizing a Molten Carbonate Fuel Cell (MCFC) system. Doosan (ClearEdge/UTC
Power) serves the market with its PureCell® Phosphoric Acid CHP system, providing 440 kW of power as
well as substantial usable heat. The final primary market player, Bloom Energy, utilizes SOFC technology
to produce power in 200+ kW systems which can be grouped to produce larger quantities of power.
Unlike the other market competitors, Bloom Energy does not offer CHP systems. Each of these fuel cell
systems serves several markets, and while few systems on the market align within the constraints of this
study (100 and 250 kW PEM/SOFC CHP systems), they cater to similar markets; therefore they were
used as guidance when market requirements and criteria were evaluated. Each of these systems may be
operated in parallel to provide larger quantities of power, if required, based on the end user’s demands.
Primary power and CHP applications will typically use readily available gaseous fuel—primarily pipeline
natural gas with some potential use of propane or anaerobic digester biogas, though the cost benefits are
typically less for propane than for natural gas. To be attractive, particularly for an emerging technology
like fuel cells, the savings compared to grid electricity must be significant: anecdotally, the payback period
must be less than or equal to 2 years, per personal communications with the president of a national
restaurant chain. To provide for more attractive return on investments, many manufacturers are offering
Purchase Power Agreements (PPA) through which the customer does not need to invest in the capital
equipment of the fuel cell system, but rather agrees to purchase power at an established rate for a given
period of time3.
Many of the early CHP installations will likely be at locations with significant heat loads that can be served
by the heat rejected by the fuel cell system. For example, large restaurants, hotels and resorts have
significant water heating loads with desired water temperatures near 135°F (57°C) for washing
applications (hands, dishes, linens, showers). Restaurants are also subject to significant electric demand
charges so offsetting the peaks which naturally occur at meal times can have notable economic benefit.
Both PEM and SOFC systems can support the hot water demand at either restaurants or hotels. Typical
water heating applications include gas-fired storage-type water heaters. For sterilization, a booster heater
is typically used to heat water from 135°F to 180°F (82°C). An SOFC system could support the higher
temperature water while a PEM system would provide only the 135°F water. Alternate applications could
include locations with significant cooling load, e.g., data centers. By utilizing heat rejected from the fuel
processor to increase the temperature of the coolant for a PEM system, the cooling water can be used to
drive an absorption chiller.
A significant consideration for CHP and primary power systems is the nature of the utility grid connection.
Some CHP installations are connected to the grid to primarily provide heat and partially offset base power
but depend on the grid for transient management and starting. If the grid goes down, so do these
systems. In discussions with some potential end users we found that back-up operation during grid
3http://www.bloomenergy.com/fuel-cell-purchase-options/,
http://www.doosanfuelcell.com/en/services/financial.do.
outage is considered to be a significant benefit. This seems to be a basis for at least some of the Bloom
installations. We believe it is relevant in particular for restaurants, grocery stores and warehouses
(refrigeration), hotels and hospitals (emergency egress/elevators, restaurant refrigeration), medium
commercial/industrial (critical process continuation), large schools or universities (heating), and data
centers (backup power, cooling) even if all business/normal operations cannot be continued during the
outage. Grid connection and operation during grid outages asserts significant influence on system design.
With the grid available, the fuel cell system can be operated at steady state power appropriate for the
application. When the grid is not available a back-up power system must effectively follow applied load
and respond to transients, for example refrigeration compressor operation, and thus requires appropriate
electrical energy storage: typically batteries. Because site use during a power outage will vary
significantly from normal use, a CHP system serving as back-up power must be able to operate with lower
heat load – perhaps with no heat load in some cases. Thus the fuel cell cooling system much be
configured to manage rejected heat with or without a CHP heating load.
SOFC and PEM have different operating characteristics and may, therefore, serve different segments of
the CHP/primary power market. SOFC systems will be less expensive than PEM because of the
reformate clean-up hardware required for PEM systems and the lower use of precious metals. However,
where off-grid operation is expected to occur with some frequency, the relatively slow load following
capability of SOFC may result in a higher cost for energy storage required to allow transient
management. LTPEM stack technology is relatively mature and will benefit from future cost reductions
achieved by the automotive industry—though the lifetime requirements are much higher for CHP so
automotive cost targets are not relevant to CHP. LTPEM is well suited to load management in off-grid
conditions—but transients are not as fast for a reformer based system as for a stored hydrogen system so
additional energy storage in the form of lead acid batteries is required. The demonstrated durability of
PEM systems is approaching 40,000 hours. SOFC durability lags but may eventually reach the same
level.
LTPEM stacks will typically be operated at 60°C to 70°C for long life and are only appropriate for low
temperature CHP. However, in a reformer based system some heat is recovered from intermediate
cooling within the fuel processor and from the reformer exhaust, enabling delivered temperatures to
exceed 70°C. In SOFC CHP systems, delivered temperature may be higher since the stack operates at
over 650°C. Recuperative heat exchangers are typically employed to improve electrical efficiency by
reducing uncontrolled heat lost but temperatures in excess of 200°C should be achievable with SOFC
systems.
light industrial applications. Demonstrated durability for PEM stacks is considerably higher than for SOFC
and load following for applications that experience significant load swings is superior. The greater
flexibility of PEM will be beneficial for restaurant and some other applications at the smaller scales where
a greater exposure to uncontrolled operating conditions exists and where individual loads may represent
a greater fraction of the total load. SOFC will ultimately be less expensive, higher efficiency, and capable
of serving higher temperature thermal loads. At this point in the development of the CHP/Primary power
market, there is no reason to down select to a single technology.
4. System Specifications
This section provides a general description of the systems selected for analysis. Both SOFC and LTPEM
were considered. The systems analyzed are representative of potential system configurations but do not
reflect any specific commercial system. They reflect Battelle’s judgment on an appropriate balance
between efficiency and cost and between proven and developing technology. In conversation with fuel
cell stack manufacturers we concluded that 100 kW and larger single-stacks would not be the preferred
approach. This is particularly true for SOFC systems where the entire stack must be scrapped if it
includes a defect found either in factory testing or during a warranty period in the field. For PEM systems
it is generally possible to repair a defective stack, but the cost and manufacturing uncertainty associated
with the larger components discouraged single-stack approaches. Hence the systems considered for this
analysis incorporate multiple PEM stacks at 50 or 60 kW per stack. The SOFC systems evaluated used
30 kW rated stacks essentially equivalent to the stacks used in our 2014 report 4 on smaller sized CHP
systems. The choice of multiple stack systems has significant impact on the overall analysis because
production volumes of specific stack components quickly reach threshold quantities where significant
automation is appropriate. The use of multiple stacks also impacts overall system design and operation
potentially affording operational flexibility. However, incorporating the hardware necessary to allow
operation of a subset of the stacks in order to achieve partial load or damage recovery would represent a
significant cost adder and was not considered in this analysis – the stacks are assumed to be balanced at
the factory and remain so during their lifetime.
A key assumption for the systems evaluated in this study is that the utility grid is available so the fuel cell
system is offsetting grid power and in some cases excess power generated may be exported onto the
grid. This assumption allows the fuel cell system to be operated to meet a specific heat demand or to be
operated at an optimum condition for offsetting grid power at the owner’s discretion. However, a key
benefit of a fuel cell CHP system is continued operation of at least critical loads during periods of grid
outage. This may be particularly important, and in fact may be the primary reason for the purchase of a
CHP system, for some applications with critical power requirements, including commercial buildings,
health service facilities, server/telecom sites, and critical industrial operations. Therefore, the designs
4Battelle 2015. “Manufacturing Cost Analysis of 1, 5, 10, and 25 kW Fuel Cell Systems for Primary Power and
Combined Heat and Power Applications.” DOE Contract No. DE-EE0005250
considered include features to enable off-grid operation—most importantly, battery support for transients
and load management. However, we did not take credit for off-grid operation when analyzing the life
cycle costs, which could vary widely between applications.
For many applications there will be some form of “hard-start” electrical hardware such as refrigeration or
air conditioning compressors. These devices have very high power demand for a short period as the
system is coming up to speed. Inrush current can be six to ten times the normal run current. The duration
may only be a few cycles (AC power assumed) to a few seconds, but the power must be available. If the
voltage decreases due to the heavy current draw, the high load period may be extended due to slower
motor spin-up, which may also damage the motors if the voltage sag is extreme. For a grid-connected
system these transients are not applied to the fuel cell but are instead handled by the grid. The opposite
situation arises if a heavy load is turned off: the sudden drop in current may cause a sharp increase in
voltage. With the grid available, the voltage spike is largely absorbed by the grid. The specifics of these
transients will vary widely with application with some installations having few, if any, spikes while others
may experience frequent and significant spikes. Systems at the 100 kW and larger scale are likely to be
at least partially custom engineered systems. The specifics of spike management would likely be
customized to the application. In developing the system design we assumed that load picks and drops
will be limited to 20% of nominal capacity; however, we have included capability to handle starting loads
of up to five times nominal for a period not exceeding 5 seconds.
During a grid outage the primary objective is to meet the critical load (heat and/or power), efficiency is
somewhat less important. Since grid outages in the US occur infrequently and typically have durations
less than 1 week, our system designs favor lower first cost over high efficiency during outages. We did
not include the capability for black start—that is, off-grid starting from a cold condition—as the CHP
systems are intended for near continuous operation. CHP systems would therefore be expected to be in
operation when a grid outage occurs.
Metric/Feature Objective
Input, Fuel Utility Natural Gas or Propane
(>30 psig preferred)
Input, Air Ambient air (-20° to 50°C)
Input, Other N/A
Output 480 3-phase VAC
Net Power Output 100, 250 kW
System Efficiency (electrical)
LTPEM 30%
SOFC 40%
System Efficiency Overall (including heat delivered)
LTPEM 80%
SOFC 90%
System Life 50,000 hours
System Maintenance Interval
1 year
(filter change: sulfur trap, air filter, fuel filter)
Grid Connection Yes, local and/or utility
Operate off-grid Yes, critical load back-up
Start off-grid No
Figure 4-1 shows a notional load curve (red) along with solid horizontal lines representing the peak,
minimum and critical loads and dashed horizontal lines representing possible fuel cell designs to be
discussed below. Critical loads are safety or process critical loads (e.g. elevator power and stairwell/exit
lighting for commercial, operational minimums for industrial processes) that should be available at all
times—though they may not be continuous. As shown in Figure 4-1, the critical load is shown above the
red load curve for a significant portion of the time representing loads such as refrigeration which operate
intermittently but which must be available when needed. The illustrated load curve does not include
transient (surge) power required for compressor starting or voltage spikes caused by switch opening to
turn off high power loads. Those issues are addressed separately.
With the grid available and power export allowed, the fuel cell system may be sized to meet either the
heat load or some selected fraction of the total electrical load. So long as the grid is available, the fuel cell
is assumed to be operated continuously and at steady near-full load with the grid providing any load
power beyond the fuel cell rating and absorbing any excess power generated. Two options for fuel cell
sizing are illustrated in Figure 4-1. Both options are shown at rated power and at minimum power
(approximately 50% turndown).
Option 1 is sized to cover a significant portion of the total load during the high load portion of the day.
Option 2 is sized so that the power at fuel cell minimum is less than the minimum load expected for the
system.
In Figure 4-1, purchased power is represented by the area under the load curve and above the fuel cell
rated power. Any demand charge would be based on the maximum difference between load curve and
fuel cell power. Option 1 would clearly make a significant change in both purchased power and demand
charges. When the fuel cell power is above the load curve, power is assumed to be exported to the grid.
The fuel cell power might be reduced to minimum during low-load conditions unless an attractive power
purchase arrangement is in place with the utility for those times. If the grid is not available (during a grid
outage or if power exporting is not allowed), the difference between the minimum fuel cell output power
and load must be rejected to ambient in some way (e.g., an air cooled resistor bank). Storage of the
excess power from the fuel cell could be considered, but storage would add significant cost, space, and
weight requirements making the economics site-specific. As illustrated in Figure 4-1, with Option 1
significant power must be stored by some means (not considered here), rejected or returned to the grid
during a significant portion of the day even at minimum power. For occasional grid outages, the excess
power wasted might represent an acceptable cost of maintaining operation and reducing demand/power
charges over the remainder of the year. For instances in which grid export is disallowed by regulation or
the absence of a utility agreement, the “wasted” energy represents a significant and likely unacceptable
cost.
Option 2 offsets significantly less of the overall load for the profile in Figure 4-1 but still maintains
sufficient power to handle the critical loads continuously. In fact, as shown it may be larger than needed
to cover the critical loads since some of the intermittent load might be addressed with energy storage
(e.g., batteries, which are required to manage start/stop transients for off grid operation). As illustrated,
the minimum fuel cell power level is below the overall minimum load power—an intentional design choice.
Thus, it should not be necessary to export or dump power at any time with this configuration, and the cost
of power export hardware or power dump hardware can be avoided. For applications in which the
difference between peak load and minimum load is much less significant than illustrated, Option 2 may be
very attractive because it will offset a greater portion of the average load without requiring power export or
dump while still maintaining the ability to operate critical loads during grid outage. Applications with near
constant heat or power load are well suited to the Option 2 approach – that is sizing the system to meet
minimum demand at the lowest steady state operating condition.
Since the sizing is typically specific to the load profile of the application, we assume a nominal resistor
bank will be provided (and included in the cost of installation) for most applications to allow sizing based
on normal, grid-available operation (Option 1 above).
Fuel Input
(Natural Gas Sulfur Removal Reformer
or Propane)
Power
Electrical Reformate Clean-
Management and Fuel Cell
Power up (LTPEM only)
batteries
Heat Recovery
Tied to Grid and Export Exhaust
via Inverter
Both propane and natural gas typically contain mercaptan odorant, a sulfur-bearing compound. While
generally present in low levels, even low levels of sulfur are damaging to the catalyst in the reformer and
fuel cell so sulfur removal is included in all systems considered.
The key differentiating factor between SOFC and LTPEM is the composition of the anode inlet stream.
LTPEM systems prefer pure hydrogen. However, some LTPEM stacks can be designed to operate on
reformate so long as the CO content of the gas is low (<50 ppm). We have based our systems on
reformate-capable LTPEM stacks in order to avoid the high cost of hydrogen purification. SOFC stacks
can accommodate percentage levels of CO and CH4 in the reformate, so they require minimal reformate
clean-up and tolerate less effective reforming so long as the higher hydrocarbons that may be present in
natural gas are reformed to methane. Excess methane and/or percent levels of higher hydrocarbons in
the reformate reaching an SOFC stack can lead to carbon deposition on the stack anode, capacity loss
and eventual failure.
As indicated by the dashed line in Figure 4-2, the anode and/or cathode exhaust from the fuel cell (SOFC
or PEM) may be routed to the reformer for energy and water recovery. While both systems use steam
reforming in order to achieve high efficiency, the details of how the fuel cell exhaust energy is recovered
for each differ significantly. Water produced appears primarily on the cathode side for LTPEM systems
and on the anode side for SOFC systems. For SOFC systems, some of the anode exhaust is directed to
the reformer where the water content (as steam) directly interacts with the input fuel to achieve steam
reforming. For efficient operation of PEM systems, particularly open anode (once through anode)
systems, the residual heating value in the anode effluent is recovered by combusting this gas in a
separate burner to heat the reformer. Water is condensed from the combustion and cathode exhausts
and re-vaporized for use as steam in the reformer. The energy required to vaporize the water for LTPEM
systems is a factor in the overall system efficiency being lower for LTPEM than for SOFC.
Other features of the overall system configuration that differentiate SOFC from LTPEM will be discussed
below.
Fuel supply including: fuel filter, fuel flow management (control valve) and sulfur sorption
reactor.
Fuel processor including steam generator, reformer and reformate CO removal reactors (two
shift and two preferential oxidation [PrOx]) along with several heat exchangers. All of these
components are typically hot, albeit at different temperatures; integrating them reduces heat
loss and improves system efficiency.
Air supply including filter, three blowers (combustion, preferential oxidation (PrOx), and
cathode), and recuperative cathode humidifier.
Cooling and heat export system including several heat exchangers operating on low-
electrical-conductivity water/glycol coolant and a CHP load heat exchanger that couples the
fuel cell coolant system to the external thermal (heat) load. The heating load is assumed to
be independent of the fuel cell system—that is, the fuel cell system cannot adjust the external
heat load, it can only respond to it. Low conductivity glycol coolant is required for the LTPEM
stack to avoid shorting the stack. A different configuration using conventional automotive
antifreeze for the other heat exchangers is possible but would add another cooling loop and
pump along with additional control hardware.
Electrical system including batteries, resistor bank, overall controls, and voltage/current
management. The electrical system also includes the grid connection electronics (inverter
and any required physical disconnects.) Additional detail, including a schematic of the
electrical subsystem, appears in Section 4.5. The electrical system is similar for both LTPEM
and SOFC; though the energy storage required is larger for the SOFC and the voltages vary
somewhat depending on system size.
Multi-stack assembly including manifolds and electrical connections. Figure 4-3 shows that
the stacks are parallel for anode, cathode, and coolant flows. Figure 4-3 suggests that the
stacks are electrically parallel. However, to achieve the voltage preferred by the power
electronics typically used in this size range, the stacks will be paired in series. For the 100
kW system there are two 60 kW stacks in series. For the 250 kW system there are six 50 kW
stacks arranged in with pairs in series yielding three parallel sets. Cell active area is the
same for all LTPEM stacks so the 50 kW stacks have a lower output voltage.
As shown in Figure 4-3, fuel enters the system and passes immediately through a sulfur sorption trap to
remove mercaptan odorant and any other sulfur compounds present. The sulfur sorbent is considered a
consumable and, if space permits, would typically be designed for a 1 or 2 year replacement cycle based
on local conditions of sulfur in the gas. For start-up, desulfurized fuel is routed directly to the reformer
burner and combusted to preheat the reformer and steam generator (vaporizer) directly. Once
combustion gas temperature at the vaporizer is adequate (~150°C), water is started and the system is
preheated by flowing steam through the downstream reactors. Once the system is adequately preheated,
fuel is diverted from the start-up burner and mixed with hot steam before being routed to the reformer.
Upon leaving the reformer, the reformate is cooled by steam from the vaporizer to approximately 350°C
before entering the high temperature shift reactor. A water cooled heat exchanger reduces the reformate
temperature between the high temperature and low temperature shift reactors and before the PrOx
reactor. Additional heat exchangers reduce the temperature between the high temperature and low
temperature PrOx reactors and after the PrOx reactors. The shift and PrOx reactors are packaged with
the reformer to minimize heat loss. Reformate leaving the final PrOx reactor will be cooled to stack
operating temperature—typically ~60°C.
Electrical System
Controls and Sensors Local or
Batteries, Inverter Utility Grid
Main Air Filter
BOP Power Mgmt.
LT
PrOx
Combustion
Air Blower
Burner
HT
PrOx Additional
LT PEM LT PEM LT PEM Stacks?
Blower
Stack Stack Stack
PrOx
Recoupertive
Humidifier
Fuel Gas
Air
Start-up fuel
Reformer
bypass
Reformate/H2 HT
Shift
Desulfurizer
Reformate
DI Water/steam
Cooler
Combustion
Electrical Exhaust
Heat Recovery
Natural Gas Subsystem
Generator
Control DI polish
Steam
Export Heat
Water Tank: Overflow
Load
Combustion and
Stack Condensate
Water
Natural Gas or
Fuel Processor Pump
Filter LT PEM mCHP System
Propane from
Supply
The cathode air entering the stack is humidified by adsorbing water across a membrane (or via enthalpy
wheel or some other form of recuperative humidifier) from the stack cathode exhaust. Anode exhaust is
routed to the reformer burner. Stack fuel utilization is adjusted to balance the system so that the anode
exhaust has sufficient energy content to support reforming. Because the anode effluent is burned to
provide the energy to support the reforming process, the system must be balanced to avoid overheating
the reformer while providing adequate reformate for the fuel cell load. The large number of heat
exchangers and reactors in the LTPEM system represent significant thermal inertia. Therefore, the
reformer is expected to respond slowly to load changes – due to both thermal lag and system control
stability requirements. In considering system dynamic design we allowed for up to 15 minutes for the 100
kW system to respond to a 25% of nominal load change; 25 minutes was allowed for the 250 kW system.
A low-electrical-conductivity glycol/water mixture is used to cool the PEM stack. Coolant should enter the
stack at about 50°C. As shown in Figure 4-3, coolant leaving the circulating pump is preheated slightly by
the final combustion exhaust. This preheating serves to condense some residual water out of the
combustion exhaust for re-use in the system. After the stack, the glycol coolant is directed to the initial
combustion gas condenser where the coolant temperature is increased above the stack operating
temperature by combustion gases and where the initial condensation of water from the combustion
exhaust occurs. After passing through the initial condenser, the coolant is routed to the pre-stack
reformate cooler and then to the shift and PrOx reactor heat exchangers. This somewhat arbitrary heat
exchanger sequence was developed based on a simplified ChemCad® model of the PEM system that
allowed several heating and cooling scenarios to be considered. Of the systems evaluated, this
configuration provided the highest temperature coolant to the CHP system. An important feature of a
reformer-based PEM fuel cell system is that the waste heat can be provided at a temperature significantly
above the stack temperature. A second coolant loop could be used to deliver even higher temperature
waste heat by separating the stack cooling from fuel processor cooling which would allow lower coolant
flow and higher delivered temperature from the fuel processor. However, for many CHP applications the
additional cost of this approach is not needed and so it was not considered in the costing analysis. The
PEM system is equipped with a radiator on the stack coolant loop to reject any heat not absorbed by the
CHP system. Since the coolant loop is required to maintain system temperatures, especially the stack
temperature, the radiator must be sized to reject all heat from the system if a CHP thermal load is not
present. A smaller radiator might be specified for systems with known continuous CHP load. Because the
temperature of the coolant to the heat export heat exchanger is projected to be less than 90°C, it is not
necessary to bypass the export heat exchanger if little or no CHP load is present. For this analysis, all
heat exchangers are assumed to be counter-flow heat exchangers (typically either tube-in-tube or plate
frame type) with the exception of the radiator, which is automotive style cross-flow.
Figure 4.3 includes high temperature and low temperature shift reactors and high temperature and low
temperature PrOx reactors. This follows advice from a Shift/PrOx catalyst supplier to cool between
reactors. Based on the advice of the catalyst supplier, the shift and PrOx reactors are ceramic monoliths
coated with an appropriate catalyst and housed within stainless steel ducts. The monoliths are available
in various diameters and in two lengths (full: ~6 inch and half: ~3 inch). The lead CO and PrOx reactors
include one monolith each for the 100 kW system. The trailing CO and PrOx reactors include
1.5 monoliths. For the 250 kW system the lead reactors include 2.5 monoliths and the trailing reactors
include 4 monoliths each. Additional information on these reactors is included in Section 5.
Fuel supply including the fuel filter, fuel flow management (control valve) and sulfur sorption
reactor. Because the SOFC system uses an eductor for anode gas recirculation, natural gas
pressure must be 30 psig or greater for all systems.
Fuel processor including natural gas/anode recirculation eductor/mixer, preheater/reformer
and tail gas combustor.
Air supply including filter and two blowers (start-up and primary cathode) and recuperative
cathode preheater. The start-up blower is used as a secondary cathode blower to achieve
full stack load.
Heat export system including an exhaust gas to CHP heat exchanger and a bypass valve
combination to manage the heat delivered to the CHP system. The heating load is assumed
to be independent of the fuel cell system—that is, the fuel cell system cannot adjust the
external heat load, it can only respond to it. Therefore, the valve system provides for diverting
high temperature exhaust gas away from the CHP system if required to avoid overheating the
CHP fluid.
Electrical system including batteries, overall controls, and voltage/current management. The
electrical system also includes the grid connection electronics (inverter and any required
physical disconnects.) Additional detail, including a schematic of the electrical subsystem, is
included In Section 4.5. The electrical system is essentially identical for either LTPEM or
SOFC with minor variations associated with the additional energy storage needed for SOFC
transient management.
Multi-stack assembly including manifolds and electrical connections. Figure 4-4 shows that
the stacks are parallel for anode and cathode flows. Figure 4-4 shows two pairs of series-
connected stacks to yield a high voltage. . The stack pairs are electrically parallel so that the
current is shared between stack pairs. For the 100 kW system there are four 30 kW stacks
as illustrated in Figure 4-4. For the 250 kW system there are ten 30 kW stacks arranged as
series pairs yielding five parallel sets.
No shift or PrOx reactors are required, no PrOx blower or PrOx heat exchanger is needed
No recuperative humidifier is required
No steam generator is required
No special glycol cooling loop is required
Several heat exchangers are not needed
No water supply subsystem is required
The relative simplicity of an SOFC system is attractive from a cost perspective. The higher quality heat
also facilitates SOFC use in a broader range of applications.
As shown in Figure 4-4, the fuel is first passed through an absorber identical to the one used for the
LTPEM system to remove sulfur. Fuel would be routed directly to the tail gas combustor for start-up,
where it is burned with cathode air to preheat the cathode indirectly via the cathode recuperative heat
exchanger. The combustion products also preheat the reformer in preparation for shifting to CPOx mode
during the next phase of start-up. As the stack assembly warms above ~100°C, the fuel is shifted to the
preheater/reformer where it is burned in catalytic partial oxidation (CPOx) mode to continue heating from
both the anode and cathode sides of the stack. Combustion is completed in the tail gas burner for indirect
heating of the reformer and preheating of the anode. Careful control of flow rates and burner
temperatures is necessary to heat the stack uniformly. When fuel is initially routed to the
preheater/reformer, anode exit gas begins recirculating to mix with the fuel. Initial recirculation will be
limited because the fuel flow for start-up will be significantly lower than for final operation, causing the
aspirator used to operate significantly off-design.
Electrical
Preheater/
Reformer
SOFC SOFC SOFC SOFC
Desulfurizer
Additional
Stack Stack Stack Stack Stacks?
Natural Gas A1 A2 B1 B2
Control
Start-up System
Air Blower Exhaust
Natural Gas or
Propane from Supply
Cathode
Air Blower
As the stack heats and fuel can be increased, more anode recirculation will occur. Current draw can
begin at a stack temperature determined by the manufacturer. The start-up blower will be reduced and
eventually turned off to shift the reformer into steam reforming mode using water in the recirculated anode
for reforming and heat from the tail gas burner to support the endothermic reaction. Once the stack
reaches operating temperature, full power may be drawn from the stack. As shown in Figure 4-4, the
start-up air blower output may be diverted to combine with the primary cathode blower. This is more
economical than having a full size cathode blower and a start-up blower that is seldom used. Stack and
other system temperatures are managed by fuel and cathode air flow. Anode recirculation is passive and
depends on the design of the aspirator. There is a potential for limiting the available stack turndown if
insufficient steam is recirculated to support reforming. In developing our system performance estimates
we assume the aspirator will be sized for approximately 50% load with the higher fraction of recirculation
at full load being generally beneficial. Alternatively, a hot gas blower might be used to provide greater
turndown in exchange for additional cost and operating complexity. We did not include this option in our
cost estimates.
Residual heat from the cathode is available to the CHP load as soon as the system has warmed
sufficiently—typically before actually drawing power from the stack. Unlike the PEM system, the CHP
system is not expected to cool the stack. Stack cooling depends primarily on cathode air flow so
additional heat can be directed to the heat load by increasing cathode air flow. To prevent the hot cathode
outlet from overheating the CHP system at times of low demand, provision is made in the system to
bypass cathode air directly to exhaust. Therefore, system exhaust may be hot and appropriate safety
precautions and exit locations requirements must be defined and observed. Routing of the hot exhaust is
outside the scope of this study.
The ability of an SOFC system to follow electrical load is somewhat restricted because long-term
reliability requires that thermal ramp-rates be limited. Thus, an SOFC system will typically operate at near
full load with the grid either making up any un-met requirement or accepting any excess power. If the grid
is not available, the SOFC system will need to keep running at or above its minimum power output, even
if the power must be dumped to an air cooled resistor bank. A hot standby condition is plausible but
somewhat difficult for SOFC because of the sensitivity of the stack to internal temperature gradients.
Generally the SOFC stack will remain on and above minimum power to provide for critical loads and any
other loads that may be within the overall capability of the stack. Generated power in excess of the
available load will be rejected through an air cooled resistor bank.
Because transients are slower for the SOFC system than for the LTPEM system, the energy storage
requirements are greater. The load pick/drop assumption for off-grid operations is the same for both
systems – 20% maximum single pick. However, the SOFC system is assumed to require significantly
longer to respond to the load pick (or drop) while maintaining stack health and control system stability.
We assume a 20% load change would require approximately 30 minutes to re-establish steady state for
the 100 kW system and 45 minutes for the 250 kW system. Thus the battery and resistor bank
subsystems are larger for the SOFC system than for the LTPEM system.
4.5.1 Overview
The electrical system provides the interface between the fuel cell stack and the local electric utility grid as
well as providing auxiliary power for system components and, when hybridized with batteries, the
capability for operation during grid outages. All systems considered in this cost study assume that a utility
grid is available and that for normal operation, the fuel cell system will use the grid to manage start-up
and load changes. To provide for critical load continuance during grid outage, the systems considered
include battery capacity to allow rapid electrical load changes while limiting the speed of the fuel cell
system response to achieve stable and reliable long-term operation. The batteries are maintained on
trickle charge when the grid is available.
The main challenge in designing a fuel cell system for off-grid operation is matching the stack variable
voltage over the desired load range with the battery system while keeping it in an acceptable range for
the DC/AC inverter. The most straightforward design is to have a DC/DC converter between the fuel cell
and battery bus to keep the stack output voltage in a proper range for the battery and the DC/AC inverter.
However, a DC/DC converter adds a significant cost to the system. Another option is to use a three-port
hybrid converter. Hybrid converters are an emerging technology that is more economical than discrete
components. These converters are designed to integrate energy sources and storage. A three-port
inverter includes two bidirectional DC ports – one for the primary source and one for the storage source -
with an additional bi-directional AC port that ties the system to the grid. The currently available three-port
inverters are designed and marketed for solar PV applications. However, a fuel cell system can be
configured to provide comparable input voltages that would allow three-port inverters to be used. Three-
port inverters are more expensive than simple inverters; however, they are less expensive than a DC/DC
converter followed by an independent DC/AC inverter.
An off-grid fuel cell system could be designed with a simple wide-input-range DC/AC inverter (these are
typically used for solar systems) by directly connecting the battery in parallel with the fuel cell if the battery
voltage can be matched to the fuel cell output. This is somewhat problematic as most battery chemistries
have a preferred voltage range that is narrower than a typical fuel cell polarization curve. For this report
we have assumed the use of a three-port inverter as the base cost assumption. In section 4.5.2.3 below
we suggest a means to directly connect batteries to the fuel cell output and DC/AC inverter input as a
possible future cost reduction approach appropriate for some applications requiring wide turndown for off
grid operation.
If the grid is considered to be always present when the fuel cell is operating, no energy storage or rapid
transient response is needed by the fuel cell system, greatly simplifying the electrical system. The only
major components necessary are a DC/AC inverter configured to accept the fuel cell input voltage range
and an appropriate disconnect switch on the AC power.
Commercially available DC/AC grid-tied inverters in the 100 to 500 kW power range are primarily
designed for solar photovoltaic systems with input voltages in the 300 to 1000 VDC range. They are
typically unidirectional (DC to AC only) because there is no need for routing power back to the solar array.
The same conditions apply for fuel cell operation. However, solar inverters include features intended to
optimize solar power output based on transient solar conditions which may interfere with their direct use
by fuel cell systems. Based on phone conversations with hybrid inverter manufacturers, the production
costs associated with solar inverters are representative even if some features must be revised so long as
the input voltages remain within the inverter allowed operating range. While some features of the solar
inverters (e.g. power optimization for variable solar incident radiation) are not required for fuel cells, these
are apparently minor variations on the overall inverter cost so the fuel cell versions will be comparable.
Fuel cell systems may be designed to have a voltage that matches commercially available solar inverters.
So long as there is no need for off-grid operation, the inverter is the major cost driver within the power
electronics area. If off-grid operation is desired (as it typically will be), additional storage and interface
requirements will add significant cost.
The more challenging load changes are electrical as these occur very quickly and frequently include short
duration spikes of many times the nominal equipment power rating; starting compressor motors for
example. While a PEM fuel cell operated on compressed hydrogen can respond quickly – within a few
seconds, a reformer based fuel cell is limited by the thermal inertia of the reformer and support hardware.
An SOFC fuel cell stack cannot respond as quickly as a PEM stack without potentially damaging the
stack due to thermal expansion issues. As shown in Section 4.4.2, an SOFC system is not burdened with
as many heat exchangers or secondary reactors as a PEM system; however, the higher temperatures
make response times slow, and since the stack is not repairable if damaged, transients are typically slow
to minimize the potential for thermal stress damage to the stack.
For this analysis we have assumed a PEM system can respond to a change that is 25% of nominal load
change within 15 minutes for the 100 kW system (a change from a 50 kW to a 75 kW load) and within 25
minutes for the 250 kW system. The SOFC system is assumed to require roughly twice as long to
respond to equivalent load changes – 30 minutes and 45 minutes for 100 kW and 250 kW systems,
respectively. With these assumed constraints on system response, an electrical energy storage system
(battery bank) is required to manage the load. The batteries must provide at least sufficient amount of
stored energy to accommodate a single 25% load step up, and preferably two 25% steps up in
succession.
The energy storage requirements may be relaxed in some cases where system controls and facility
operation permit the demand side management of loads and where no single load represents more than
(for example) 5% of the nominal power. As will be shown later in the cost section, energy storage costs
make up a noticeable percentage of the system installation costs. Furthermore, the batteries represent
an ongoing cost as their lifetimes, particularly if deeply cycled frequently, may be shorter than the
presumed life of the fuel cell system. Some systems which have frequent high power spikes associated
with equipment starting may benefit from ultracapacitors in place of or in coordination with battery
storage. These would be specialized applications and are not considered in this report.
There are three possible methods for connecting a battery assembly to a fuel cell system. All relate to
methods of managing the voltage associated with the energy storage so that the batteries are neither
overcharged nor undercharged. The most conventional is to place a DC/DC converter on the output of
the fuel cell to regulate the voltage on the battery bus to match that of the battery system. Emerging
hybrid DC/AC inverters that incorporate one or more auxiliary bi-directional DC ports represent a lower
cost approach to managing the voltage to the batteries. In some cases it is possible that the batteries
may be connected directly to the output of the fuel cell and operate in parallel with it. This option
represents the lowest cost approach but brings significant uncertainty associated with battery life and
overall system management. Directly connected batteries would not be appropriate for all applications
and would require specific engineering to purpose in most cases. All three approaches are described
below. Because 3-port hybrid inverters are commercially available, would serve a wide range of potential
system configurations, and are lower cost than a high-power DC/DC converter + DC/AC inverter
configuration, they were chosen as the basis for our cost analysis.
Lead acid batteries are used for energy storage. Lead acid batteries are widely available, relatively
inexpensive and well understood. They tolerate rapid charge and discharge cycles well and achieve
acceptable lifetimes when properly managed. For lead acid batteries, the battery management system
(BMS) can be relatively simple. State of charge is reasonably well represented by battery open-circuit
voltage or by a polarization curve of voltage versus current. The BMS regulates charging rate based on
the implied state-of-charge, dropping to a trickle charge as battery voltage increases. The BMS also limits
the minimum terminal voltage difference, tripping the system should the battery terminal voltage become
too low indicating excess current draw for the state of charge of the battery.
Other energy storage options exist. Lithium ion (LI) batteries have a high energy/power density relative to
other battery technologies but they cost more than lead acid for an equivalent amount of energy storage
and require a more sophisticated BMS. For mobile/transportation applications, LI batteries are attractive
due to their low weight and small footprint, but for stationary applications the minimal premium for smaller
size and weight is not enough to overcome the cost advantage enjoyed by lead-acid batteries.
Ultracapacitors are also an option, particularly for high surge power applications. The main drawback of
ultracapacitor technology is limited energy density. This limitation can be overcome by hybridization with
either lead acid or LI batteries. For this study we assumed that lead-acid batteries alone would be
sufficient to manage the transients; alternative technologies were not considered.
A secondary DC/DC converter is required to power the control electronics and miscellaneous support
equipment in the system. The DC bus is assumed to be ~400 VDC for both the 100 kW and 250 kW
systems. Most of the equipment is 24 VDC so we have included a 400/24 DC/DC converter rated for
10% of nominal load for 100 kW and 250 kW. Buck converters, such as the one used for auxiliary power,
are well defined, consist of minimal components and can be very efficient at high power levels. The
range of equipment available for 24 VDC is generally good so the controls will typically include internal
DC/DC converters to achieve the tightly regulated voltage required for the sensing electronics and
hardware. These “brick” converters, which have a relatively wide input voltage range, are typically
included in the cost of the control hardware and would usually accept a supply voltage of 24 VDC nominal
or 120 VAC depending on the system design – both are available for a typical system.
For grid connected service, the DC/AC inverter may allow charging the batteries if the fuel cell is not able
to provide adequate power or is off-line for maintenance or some other reason. As indicated in Figure
4.5, the battery bus voltage is nominally 408 VDC, corresponding to 34 twelve-volt batteries in series. A
similar battery bank assembly is used for all three system configurations with higher power batteries being
used for the 250 kW system. As will be explained below, the specific number of batteries used for the
directly connected option is varied to match the stack output voltage range.
An additional feature that may be required for the hybrid converters is a means to adjust the balance
between the batteries and the fuel cell so that excess power is not drawn from the fuel cell during
transients as the fuel processor ramps up. This feature would seem to be comparable to the load
optimization feature for solar; therefore, we assume the cost would be comparable.
high, significant increases in load are unlikely but significant decreases in load are expected as
equipment turns off. In that case, the batteries should be in a low state of charge to absorb the extra
power produced by the fuel cell while the reformer ramps down.
Stack
Voltage
375-550 VDC DC
Fuel Cell
AC Main
Disconnect/
lockout
Natural Gas/
Propane Reformer
DC On-site AC Load
DC
PWM
Switch
Batteries
370-470 VDC AC BOP loads
Resistor Bank DC BOP loads
Fxn(SOC) (if used)
Figure 4-8 shows a typical PEM fuel cell polarization curve (extrapolated to the stack full power conditions
for the 60 kW fuel cell stacks used for 100 kW PEM5). Also shown on the graph are the gross power from
a single fuel cell stack and the state of charge of a set of 17 (in this case) lead acid batteries represented
by the stack voltage at each power level. For this illustration the state of charge of a nominal 12 volt
battery is presumed to be linear from 10.5 VDC at 0% state of charge to 13.8 VDC at maximum charge.
Voltage above 13.8 VDC may be acceptable if the voltage doesn’t exceed that usually used for trickle
charging the batteries. However, Figure 4-8 shows that the fuel cell will output excessive voltage at low
power. Therefore, as power drops below ~35% of nominal, action should be taken to protect the batteries
from excess voltage. For extreme conditions (system trip), the batteries should be disconnected from the
fuel cell outlet bus until the fuel cell can be fully taken off-line. Alternatively, the resistor bank may be
used to dump power as fuel cell voltage approaches the fully charged limit.
Figure 4-8. PEM Fuel cell polarization curve with battery state of charge.
For the system in Figure 4-8 a bi-directional DC/AC inverter is preferred to assure that the batteries are
maintained in a relatively high state of charge during a fuel cell outage. The illustration in Figure 4-8 is for
a single fuel cell stack matched with a set of 17 batteries. However, to achieve the appropriate voltages
for typical solar-derivative inverters we assume that two stacks will be connected electrically in series and
be matched with a series connected set of 34 nominally 12 volt batteries. Although not explicitly shown in
the schematic, appropriate disconnects are required for personnel safety given the high voltages
associated with this system.
Although illustrated here for PEM, a similar approach is possible for SOFC. However, in both cases
significant engineering and testing is needed to confirm successful operation over the life of the fuel cell
and battery assembly as both may experience changes with age that could result in a mismatch in the
state of charge versus nominal power. Because of the uncertainty we have chosen not to include this
case in the base costing but rather consider it as possible improvement in the sensitivity analysis.
subsystem/component costs are shown in Appendices A-6 through A-26. Purchased components were
incorporated into the assembly sequence models to determine the assembly cost for each size and
production volume. The final cost of producing the fuel cell systems includes a testing and burn-in
sequence for the overall system. The output of the manufacturing models was also used to calculate the
number of individual assembly stations or production lines required to support various product demand
levels. From this, manufacturing equipment utilization was calculated and used to determine machine
rates for the various manufacturing processes. The manufacturing capital cost model is also based on the
number of process lines required, which provides the basis for calculating factory floor space and
personnel requirements. We assumed that capital equipment expenditures for production would be
amortized over a 20-year period (see Appendix A-1) and that the annual amortized cost would be
distributed over the production volume for that year. Total system costs including capital expenditures
were then estimated for the baseline system and projected improvements.
Because each system considered in this analysis incorporates multiple stacks (anywhere from 2 to 10
discrete stacks), stack production volumes increase rapidly, reaching 500,000 stacks/year for the 250 kW
SOFC systems at 50,000 systems/year. Since each stack includes over 200 cells, stack component
manufacturing volumes quickly reach millions of units/year even at relatively low system shipment levels
(in excess of 2 million cells for one-thousand 250 kW SOFC units shipped.) These high volumes affect
the cost estimation process and very few stack related operations are considered to occur as outsourced
supply.
At the 100 kW to 250 kW scale, most purchased components (e.g. heat exchangers) are not mass
produced. In fact, for many of the purchased components the manufacturer (or perhaps more accurately
engineer/fabricator) typically sells less than 10 of any specific item in a year. Thus achieving reliable high
volume cost quotes or rough order of magnitude ROM estimates was very difficult, and in many cases
impossible. In order to establish the relative importance of the various purchased components we used
the best available quotes and conventional extrapolation methods (see Appendix A-2) to estimate higher
production volume costs. It is likely that substantial redesign of many of the low production volume
components would be required for significantly larger volume purchases. Designs specific to larger
volume production would enable additional automation and more rapid cost reduction than we could
confidently include in this analysis.
followed the machine cost protocols described in James et al. (2014)6. Appendix A-1 provides details of
our machine rate calculations for the various production processes used to manufacture the CHP
systems.
For each production line or machine station, utilization is calculated as the fraction of the total available
time required to produce the required annual volume of systems. We assume that total available
manufacturing time consists of three 8-hour shifts per day for 250 days per year, or 6000 hours per year.
The total required machine time is the product of the number of systems to be produced and the time
required to produce the required components for each system. The number of machines required is
calculated as:
For each machine, utilization is calculated as the fraction of the total available time required to produce
the required annual volume of stacks:
The base (100% utilization) machine rate is divided by the utilization to determine the machine rate used
to produce the components for that level of system production.
At low utilizations, job shops may make parts at a lower cost because their machines are used by multiple
customers. This is particularly true for flexible Computer Numerically Controlled (CNC) tooling that can be
applied to diverse industries. Additional job shop costs include the profit charged by the job shop and any
overhead incurred by the manufacturer as a result of contract administration, shipping, and incoming
parts inspection. For consistency across all types of tooling, we assume a job shop will base their cost on
65% machine utilization overall and 40% markup for profit plus overhead when calculating their rate.
Refer to Appendix A-1 for details of the job shop machine rate calculations and the details of the make vs.
buy decision.
6
James, B.D., Spisak, A.B., Colella, W.G., 2014, “Design for Manufacturing and Assembly (DFMA®) Cost Estimates of
Transportation Fuel Cell Systems,” ASME Journal of Manufacturing Science and Engineering, New York, NY: ASME, Volume 136,
Issue 2, p. 024503.
appropriate voltages and overall system power the fuel cells are arranged in pairs connected in series:
two 60 kW stacks yielding 120 kW gross for the 100 kW systems and six 50 kW stacks in three pairs
yielding 300 kW gross for the 250 kW system. The only difference in the stacks is the number of cells –
283 to yield 60 kW and 236 to yield 50 kW. The number of cells affects the stack voltage and impacts
some aspects of the balance of plant – primarily the electrical power management and energy storage
components.
Only the primary manufacturing and assembly processes are shown. As indicated in Figure 5-1 and Table
5-2, a stack consists of two end plates and the appropriate number of repeat units. Repeat units include:
One MEA
One each Cathode and Anode bipolar plate
o When stacked, the Anode and Cathode plates are back to back to provide coolant flow
channels)
Seals between each item (three seals per each repeat unit).
A356
Aluminum
Liquid
Silicone
Rubber
Injection Mold
BMC940
Graphite
Composite
Compression
Pre-form Mold Post Bake
(12 sec) (800 tons for 180 (15 min @ 175°C)
sec @ 160°C) End Plate
Gasket
Bipolar Plate C
Gasket
MEA
Gasket Repeat
Platinum Assembly
Membrane Bipolar Plate A /
Cooling Manifold
Gasket
XC-72
Carbon
Black Gas
Coat Anode Diffusion
Layer
Bipolar Plate C
Nafion Ball Mill
DE-521 (10 hrs) Gasket
MEA
Hot Press
Decal Transfer
Coat Cathode (1000 PSI for 120 Die Cut Gasket
(Roll Calendar)
sec @ 100°C
Bipolar Plate A /
DI Water
Cooling Manifold
Bipolar Plate C
Gas Gasket
Transfer Transfer
Diffusion
Methanol Substrate Substrate End Plate
Layer
BMC940
Graphite
Composite
Compression
Pre-form Mold Post Bake
(12 sec) (800 tons for 180 (15 min @ 175°C)
sec @ 160°C)
20mm.
100mm. 30mm.
Ø 10.0mm.
Anode Cooling typ
Gas Fluid
5mm.
318mm.
20mm.
Cathode Gas
Cathode Gas
245mm.
305mm.
Active Area
5mm.
Cooling Anode
Fluid Gas
378mm.
Figure 5-2. PEM MEA configuration for 780 cm2 active area.
membrane is then hot pressed between two gas diffusion layers and die cut to final cell dimensions. The
catalysts and gas diffusion layers are only applied to the active area. The die cutting process includes
cutouts for the manifolds as shown in Figure 5-2. Details of the analysis are shown in Appendices A-7
(catalyst application) and A-8 (gas diffusion layer application). For all production volumes, the component
reject rate was assumed to be 2.5% for catalyst production, 2.5% for catalyst application, 3.0% for decal
transfer, 0.5% for hot pressing and 3.0% for die cutting. The MEA material cost summary is provided in
Table 5-3 and the part production cost summary is given in Table 5-4.
Table 5-3. PEM MEA Material Cost Summary - 100 kW and 250 kW
Although the cells are identical for the 60 kW and 50 kW stacks, the higher volume production associated
with the 50 kW stacks used for the 250 kW system (six stacks/system versus two) results in material and
labor cost savings. Note that the membrane and gas diffusion layers dominate the cost at low volume,
becoming much less important at high production rates, where catalyst cost becomes the dominant
factor. Because of the precious metal content of the catalyst cost, it does not reduce as quickly with
production volume as do those of the other materials. In fact, a potential concern is that the high
production volumes, particularly for the larger systems, could cause increases in catalyst cost. While
some recycling programs to reclaim the catalyst from retired fuel cells do exist, they still lack the
availability and robustness of other precious metal programs, such as those in place for automotive
catalytic converters. Additionally, due the to the nature of PEM fuel cell construction, recovering high
value materials would likely be easier than is the case for catalytic converters.
5mm.
Ø 10.0mm.
30mm. 100mm. typ 15mm. 12mm.
20mm.
80mm.
159mm.
M14 x 1.25 typ
80mm.
20mm.
6mm.
7mm.
M30 x 1.5
245mm.
305mm.
30mm.
23mm. 140mm.
5mm.
15mm.
5mm.
377mm.
3mm.
12mm.
Table 5-5. PEM End Plate Cost Summary - 100 kW and 250 kW
Table 5-6. PEM Anode Bipolar Plate Cost Summary - 100 kW and 250 kW
Table 5-7. PEM Cathode Bipolar Plate Cost Summary – 100 kW and 250 kW
Table 5-8. PEM Anode and Cooling Seal Cost Summary - 100 kW and 250 kW
Table 5-9. PEM Cathode Seal Cost Summary – 100 and 250 kW Systems
Table 5-10. PEM Stack Assembly Costs - 100 kW and 250 kW Systems
being produced currently. We have assumed a failure rate of 5% for this analysis (lower than the industry
reported values, but still high for a mature production process) regardless of production volume. Stacks
failing the test are reworked by disassembling the stack, replacing the defective part and reassembling
the stack. The cost of the rework is included in the scrap cost. Details of the analysis are shown in
Appendix A-11. The stack testing and conditioning costs were calculated as shown in Table 5-11. The
high stack failure rate would usually be expected to come down as higher volumes are reached and
additional automation and quality control measures are instituted. In the absence of information on why
the stack failure rates are high, we have to assume that the rate does not change with production volume.
There is a sharp drop in machine cost per stack that occurs as production volumes change from 100 to
1000 units/year; this reflects the low utilization rate in the case of the 100 unit volumes, and the
assumption that all stack testing is performed in-house.
Table 5-11. PEM Stack Testing and Conditioning Cost Summary - 100 kW and 250 kW
Table 5-12. PEM Stack Component Cost Summary - 100 kW and 250 kW
Table 5-13. PEM Stack Manufacturing Cost Summary - 100 kW and 250 kW
Breakdowns of stack cost volume trends appear in Figures 5-4 and 5-5.
Figure 5-4. Breakdown of 100 kW system – 60 kW fuel cell costs and production volume trends.
Figure 5-5. Breakdown of 250 kW system – 50 kW fuel cell costs and production volume trends.
The reformer concept pricing was based on a reformer design patented by Catacel Corporation (now part
of Johnson Matthey). The patent describes a single tube, single burner configuration suitable for systems
up to approximately 25 kW. For the 100 and 250 kW versions described here, Battelle envisioned a
multiple tube, single burner configuration. The steam generator is integrated with the reformer as a coil of
finned tubing surrounding the multi-tube assembly and directly using the combustion gas from the
reformer. The shift and PrOx reactors are pipe reactors using catalyst coated ceramic monoliths.
Additional detail on these components is included below.
7Battelle 2015. “Manufacturing Cost Analysis of 1, 5, 10, and 25 kW Fuel Cell Systems for Primary Power and Combined Heat and Power
Applications.” Report to DOE. DOE contract no. DE-EE0005250.
Figure 5-7 shows the conceptual reformer configuration for a 100 kW system and Figure 5-8 shows the
reformer configuration for a 250 kW system. The reformer tube lengths and overall length are the same
for both systems. The number of tubes and overall diameter is increased for the 250 kW system. In both
systems the steam/fuel-reformate flow through the CC fan stack is opposite the combustion gas flow.
Reformate flows to the exit manifold through a central tube within the CC fan stack.
Also illustrated in Figures 5-7 and 5-8 is a steam generator coil wrapped around the reformer and heated
by the effluent combustion gas from the reformer. Integrating the steam generator in this way reduces
both cost and heat loss compared to a separate steam generator. The steam generator coil is assumed to
be a 25.4 mm OD tube with 9.5 mm tall stainless steel fins continuously wound and brazed to the tube.
Although shown as a loosely wound coil in Figures 5-7 and 5-8, for costing purposes the coils were
assumed to be closely wound – that is, with the fin tips contacting adjacent fins on each coil and the coil
occupying the available length.
Steam Out
3608mm
2794mm 610mm
Fuel In
102mm
318mm
406mm
Reformate Out
609mm
Burner
Heat In
51mm
Water In Heat Out
51mm
Ø102mm
Ø318mm
2326mm
1461mm 610mm
51mm
Fuel In
102mm
432mm
Reformate Out
724
Burner
Heat In
51mm
102mm
Ø102mm
The required catalyst coated surface area for each of the reformers is calculated based on a catalyst
content of approximately 3850 cm 2 per kW as used by Catacel in their SSR® design. For the systems
under consideration, we started with the reformer sizing parameters shown in Table 5-14.
System Size (kW) Approximate Reformer Size Target Catalyst Area (cm2)
100 3 tubes @ 2663 cm active fan length [email protected] = 4.62x105 total
250 7 tubes @ 2663 cm active fan length [email protected] x105 = 1.16x106 total
Each reformer tube includes 168 CC fans. Using fixed fan dimensions and the target coated areas listed
in Table 5-14, we established the final reformer dimensions for each system as shown in Table 5-15. The
tubes and housings are rolled and welded from high alloy or stainless steel sheets. Appropriate locating
features are assumed to be incorporated into the tube to position the CC fans. The fans and associated
separator washers are loaded into the tubes prior to the endplates being welded on and before the tubes
are welded into the tube sheet. The center tubes are inserted through the inlet tube sheet and through the
CC fan stack before the assembly is welded into the overall shell. The reactor cost summary is shown in
Table 5-16.
100 kW 250 kW
Overall Diameter (mm) 609 724
Includes steam generator and insulation
Overall Length. (mm) 3710 3710
Flow Stages 3 tubes @ 168 = 504 7 tubes @ 168 = 1176
Final Catalyst Area (cm2) 485,300 1,132,365
Table 5-16. PEM System Reformer Cost Summary - 100 kW and 250 kW
The supplier also recommended that each reactor be split into two sections with a heat exchanger
between. The 100 kW reactors (both shift and PrOx) were divided into two reactors with 2.5 monolith
assemblies in each followed by a heat exchanger. For the 250 kW assembly the lead reactor includes 6
monoliths and the trailing reactor (after the heat exchanger) includes 6.5 monoliths. The 250 kW system
uses the same overall can size for all shift and PrOx reactors. The reactor dimensions for each system
are shown in Table 5-18.
100 kW 250 kW
Shift Diameter (OD-cm) 152.5 152.5
Shift Length (each reactor) (mm) (2 required) 1170 1930
PrOx Diameter (OD-mm) 152. 5 152.5
PrOx Length (each reactor) (mm) (2 required) 1170 1930
25mm
Catalyst
Catalyst Monolith Catalyst Monolith Monolith
Half-Block
64mm
25mm
25mm
533mm 533mm/1295mm
25mm
51mm
100 kW 250 kW
100 1,000 10,000 50,000 100 1,000 10,000 50,000
Material $96.69 $94.47 $94.10 $94.10 $183.86 $181.65 $181.27 $181.27
Process $12.50 $94.47 $94.10 $94.10 $39.43 $181.65 $181.27 $181.27
Scrap $5.46 $9.45 $9.41 $9.41 $11.16 $18.16 $18.13 $18.13
Tooling $82.46 $8.25 $0.82 $0.16 $157.86 $15.79 $1.58 $0.32
Manufactured
$197.11 $206.64 $198.44 $197.78 $392.31 $397.24 $382.25 $380.99
Parts
Catalyst $4,566.73 $3,784.36 $3,136.02 $2,750.00 $10,982.76 $9,101.18 $7,541.96 $6,756.75
Purchased
$13.39 $13.39 $13.39 $13.39 $26.71 $26.71 $26.71 $26.71
Parts
Assembly $7.43 $5.94 $5.79 $5.77 $12.35 $9.87 $9.62 $9.60
Reactor Cost $4,784.67 $4,010.33 $3,353.63 $2,966.94 $11,414.13 $9,535.00 $7,960.53 $7,174.04
# per System 4 4 4 4 4 4 4 4
Total Cost per
$19,138.69 $16,041.31 $13,414.53 $11,867.77 $45,656.52 $38,139.99 $31,842.13 $28,696.16
System
A category titled “Additional Work Estimate” is included to capture small contingencies not specifically
itemized in this report. These include components such as heat sinks and fans for additional electrical
cooling, supplementary temperature or pressure sensors and any extra assembly hardware. This
estimate is based on a 20% buffer to the electrical subsystem cost and a 10% buffer to all remaining
hardware.
8
Battelle 2015. “Manufacturing Cost Analysis of 1, 5, 10, and 25 kW Fuel Cell Systems for Primary Power and Combined Heat and
Power Applications.” Report to DOE. DOE contract no. DE-EE0005250.
Table 5-20. PEM BOP Cost Summary for 100 kW and 250 kW systems.
Annual Production of 100kW PEM Systems Annual Production of 250kW PEM Systems
System Component Description
(100) (1,000) (10,000) (50,000) (100) (1,000) (10,000) (50,000)
Nat. Gas Filter $ 97 $ 81 $ 68 $ 57 $ 305 $ 254 $ 241 $ 178
Fuel supply
Natural Gas Control Valve $ 4,561 $ 3,882 $ 3,397 $ 3,154 $ 4,501 $ 3,830 $ 3,352 $ 3,112
Natual Gas Flowmeter $ 1,697 $ 1,619 $ 1,416 $ 1,214 $ 1,697 $ 1,619 $ 1,416 $ 1,214
Desulfurizer $ 2,427 $ 1,457 $ 1,269 $ 1,142 $ 4,855 $ 3,424 $ 3,027 $ 2,724
3-way valve $ 1,320 $ 1,247 $ 1,174 $ 1,174 $ 3,521 $ 3,325 $ 3,130 $ 3,130
Reformer Water Pump $ 2,011 $ 1,910 $ 1,810 $ 1,709 $ 2,011 $ 1,910 $ 1,810 $ 1,709
Water Supply
WGS & PrOx Reactors $ 19,139 $ 16,041 $ 13,415 $ 11,868 $ 45,657 $ 38,140 $ 31,842 $ 28,696
Superheater $ 24,306 $ 23,091 $ 21,936 $ 21,497 $ 28,229 $ 26,818 $ 25,477 $ 24,967
Air Filter $ 510 $ 479 $ 447 $ 415 $ 1,177 $ 1,108 $ 1,039 $ 970
Combustion Air Blower $ 3,049 $ 2,818 $ 2,586 $ 2,471 $ 3,696 $ 3,416 $ 3,136 $ 2,996
Combustion Air Flowmeter $ 1,493 $ 1,405 $ 1,317 $ 1,229 $ 1,493 $ 1,405 $ 1,317 $ 1,229
Air Supply
PrOx blower $ 805 $ 770 $ 735 $ 717 $ 2,961 $ 2,730 $ 2,499 $ 2,383
PrOx Flowmeter $ 1,493 $ 1,405 $ 1,317 $ 1,229 $ 1,493 $ 1,405 $ 1,317 $ 1,229
Cathode Air Blower $ 3,049 $ 2,818 $ 2,586 $ 2,471 $ 3,696 $ 3,416 $ 3,136 $ 2,996
Cathode Air Flowmeter $ 1,493 $ 1,405 $ 1,317 $ 1,229 $ 1,493 $ 1,405 $ 1,317 $ 1,185
Recouperative Humidifier $ 498 $ 376 $ 285 $ 248 $ 1,245 $ 966 $ 712 $ 619
WGS Outlet Cooler 1 $ 2,948 $ 2,653 $ 2,388 $ 2,268 $ 4,412 $ 3,971 $ 3,574 $ 3,395
WGS Outlet Cooler 2 $ 2,948 $ 2,653 $ 2,388 $ 2,268 $ 4,412 $ 3,971 $ 3,574 $ 3,395
First Combustion Condenser $ 3,870 $ 3,483 $ 3,135 $ 2,978 $ 4,964 $ 4,467 $ 4,020 $ 3,819
Heat Recovery
Second Combustion Condenser $ 5,490 $ 4,941 $ 4,447 $ 4,225 $ 7,683 $ 6,915 $ 6,223 $ 5,912
PrOx Outlet Cooler 1 $ 5,335 $ 4,801 $ 4,321 $ 4,105 $ 7,984 $ 7,185 $ 6,467 $ 6,143
PrOx Outlet Cooler 2 $ 5,335 $ 4,801 $ 4,321 $ 4,105 $ 7,984 $ 7,185 $ 6,467 $ 6,143
CHP Export Exchanger $ 4,860 $ 4,374 $ 3,937 $ 3,740 $ 5,512 $ 4,960 $ 4,464 $ 4,241
Radiator $ 5,957 $ 5,659 $ 5,376 $ 5,268 $ 12,513 $ 11,888 $ 11,293 $ 11,067
Glycol Pump $ 698 $ 629 $ 556 $ 509 $ 752 $ 676 $ 598 $ 548
Hybrid Inverter $ 42,000 $ 33,600 $ 26,400 $ 21,600 $ 96,000 $ 75,000 $ 57,000 $ 45,000
Electronics
Power
Control Module $ 719 $ 647 $ 582 $ 565 $ 719 $ 647 $ 582 $ 565
Control and
Assembly Hardware $ 1,030 $ 935 $ 840 $ 755 $ 1,610 $ 1,465 $ 1,320 $ 1,190
Frame and Housing $ 3,085 $ 2,805 $ 2,525 $ 2,275 $ 4,835 $ 4,395 $ 3,955 $ 3,560
Additional Work Estimate $ 15,100 $ 13,700 $ 12,300 $ 11,100 $ 24,300 $ 22,100 $ 19,900 $ 17,900
+
TOTAL BOP COST $ 200,924 $ 174,868 $ 154,728 $ 142,047 $ 350,048 $ 298,714 $ 258,865 $ 234,045
Table 5-22. PEM System Testing Cost Summary - 100 kW and 250 kW
Production line use was estimated to determine the number of individual process lines required to support
various product demand levels. This information, along with equipment cost quotes, was used to
determine production line equipment costs. The production facility estimation is based on the floor area
required for production equipment, equipment operators, and support personnel. Guidelines used for this
analysis were developed by Prof. Jose Ventura at Penn State University, and are detailed in Appendix
A-4. Capital cost summaries appear in Table 5-24. A production line in Table 5-24 refers to the equipment
needed to produce a subcomponent or to assemble an entire system from components produced on
support production lines. For example, bipolar plate production requires a hot press and a post-press heat
treating oven. Because each pair (press and oven) can only produce ~27 parts/hour, a significant number
of bipolar plate production lines are required (e.g. 869 bipolar plate production lines to produce 50,000 of
the 250 kW systems/year).
One interconnect and anode frame (these may be integrated into a single piece)
One Cell (anode, electrolyte, cathode) and picture frame (The picture frame supports the entire
periphery of the electrolyte
Cathode Frame
One Cathode Mesh and one Anode Mesh
o The Anode and Cathode meshes sandwich the cell to provide flow cavities and a
compliant electrical path from the cell to the interconnects
Seals between each frame, the cell and interconnect
Figures 5-14 and 5-15 illustrate the layer configuration and orientation.
Figure 5-14. Detail assembly of SOFC Cell (not to scale, parts depicted as shown for clarity)
This study only focuses on the primary manufacturing and assembly processes shown in Figures 5-14,
5-15, and 5-16, and in Table 5-25.
Hastelloy X
Ferritic
Stainless Perovskite
Steel Powder
SS-441 Suspension
Aerosol Spray
Heat Treat
Blank/Punch Deposition
(4 hours @ 800°C)
2-3 µm
Ferritic
Glass-
Stainless
Ceramic
Steel
Sealant
SS-441
End Plate
Interconnect
Stainless
Steel Anode Frame
Mesh
Anode Mesh
Cell Repeat
Unit
Picture Frame
Laser Cut
Cathode Frame
Cathode Mesh
Ferritic
Stainless
Steel
SS-441 Interconnect
End Plate
Blank/Punch
Ferritic
Glass-
Stainless
Ceramic
Steel
Sealant
SS-441
Stainless
Steel
Mesh
Laser Cut
5.4.1.1 SOFC Stack Component Size and Ceramic Cell Manufacturing Setup
Ceramic cell assemblies for the 30 kW stacks (used in both the 100 kW and 250 kW systems) are
assumed to have 414 cm 2 active area. Using a length to width ratio of 1.5, the active cell size was
determined to be 167 mm by 248 mm. Using a 10 mm margin on all sides to allow for sealing to the
picture frame, the overall cell size was determined to be 187 mm by 268 mm.
268mm.
10mm.
10mm.
248mm.
167mm.
187mm.
12mm.
6mm.
66mm.
Ø 10.0mm.
typ 12mm.
40mm.
15mm.
15mm.
249mm. 58mm.
6mm.
10mm.
130mm.
10mm.
149.5mm.
80mm.
15mm.
167mm.
299mm.
M24 x 1.5
M16 x 1.5
130mm.
80mm.
364mm.
The process selected to produce the end plates was sand casting Hasteloy X. Following cooling, the cast
plate is moved to a CNC drilling center to face one side, drill and ream the eight tie rod holes, and drill,
ream and tap the gas connector holes as needed. The end plate cost analysis is detailed in Appendix A-
24, and summarized in Table 5-27.
Table 5-27. SOFC End Plate Cost Summary - 100 kW and 250 kW
The anode frame supports the interconnect on the anode side and provides space for the anode
and anode mesh
The picture frame supports the entire periphery of the electrolyte and is sealed (weld or glass
seal) to the anode frame
The cathode frame supports the interconnect on the cathode side and provides space for the
cathode mesh
The anode, picture and cathode frames are manufactured from 1.033mm, 0.25mm, and 0.25mm thick
ferritic stainless steel (SS-441) sheet. The material is stamped into a rectangular blank, then punched to
provide the required gas path openings and active area relief. For all volumes, the process scrap rate for
the stamping operation was assumed to be 0.5%. Details of the analysis are shown in Appendix A-19.
The frame cost summaries are provided in Tables 5-29, 5-30, and 5-31. The tooling cost differences are
primarily the result of different configurations resulting in different total shearing lengths of the tooling.
Table 5-29. SOFC Anode Frame Cost Summary - 100 kW and 250 kW
Table 5-30. SOFC Picture Frame Cost Summary - 100 kW and 250 kW
Table 5-31. SOFC Cathode Frame Cost Summary - 100 kW and 250 kW
Table 5-32. SOFC Laser Welding Cost Summary - 100 kW and 250 kW
Table 5-33. SOFC Ceramic-Glass Sealing Cost Summary - 100 kW and 250 kW
from 0.08mm Stainless Steel that has been expanded and corrugated to the proper height, 0.5mm for the
anode and 0.75mm for the cathode. The material is then laser cut to final dimension. Details of the
analysis are shown in Appendix A-25. The mesh cost summaries are provided in Tables 5-34 and 5-35.
Table 5-34. SOFC Anode Mesh Cost Summary - 100 kW and 250 kW
Table 5-35. SOFC Cathode Mesh Cost Summary - 100 kW and 250 kW
applying learning curve analysis to the assembly times and multiplying by the standard labor rate of
$45.00/hour, the average stack assembly costs were calculated as detailed in Appendix A-3 and
summarized in Table 5-36.
Table 5-36. SOFC Stack Assembly Cost Summary - 100 kW and 250 kW
Table 5-37. SOFC Stack Brazing Cost Summary - 100 kW and 250 kW
Table 5-38. SOFC Stack Testing and Conditioning Cost Summary - 100 kW and 250 kW
Table 5-39. SOFC Stack Component Cost Summary - 100 kW and 250 kW
Table 5-40. SOFC Stack Manufacturing Cost Summary - 100 kW and 250 kW
SOFC stack cost volume trends appear in Figures 5-19 and 5-20.
Figure 5-22 shows the conceptual reformer configuration for a 100 kW system and Figure 5-23 shows the
reformer configuration for a 250 kW system. The reformer tube lengths and overall length are the same
for both systems. The number of tubes and overall assembly diameter are increased for the 250 kW
9Battelle 2015. “Manufacturing Cost Analysis of 1, 5, 10, and 25 kW Fuel Cell Systems for Primary Power and Combined Heat and Power
Applications.” Report to DOE. DOE contract no. DE-EE0005250.
system. In both systems the steam/fuel-reformate flow through the CC fan stack is opposite the
combustion gas flow. Reformate flows to the exit manifold through a central tube within the CC fan stack.
2326mm
1461mm 610mm
51mm
Fuel In
102mm
406mm
609mm
Reformate Out
12.5"
Burner
Heat In
51mm
51mm Heat Out
102mm
51mm
Ø102mm
Ø318mm
2326mm
1461mm 610mm
51mm
Fuel In
102mm
432mm
635mm
Reformate Out
Burner
Heat In
51mm
102mm
Ø102mm
The required catalyst coated surface area for each of the reformers is calculated based on a catalyst
content of approximately 1925 cm2 per kW. This value is half that used by Catacel in their SSR® design
for PEM systems because some methane breakthrough is acceptable, and in fact desirable, for SOFC
stack cooling from internal reforming. An additional factor in the downsizing is the higher electrical
efficiency of the SOFC system. For the systems under consideration, we started with the reformer sizing
parameters shown in Table 5-41.
System Size (kW) Approximate Reformer Size Target Catalyst Area (cm2)
100 3 tubes @ 1331 cm active fan length [email protected] = 2.43x105 total
250 7 tubes @ 1331 cm active fan length [email protected] x105 = 5.67x105 total
Each reformer tube includes 168 CC fans. Using fixed fan dimensions and the target coated areas listed
in Table 5-41, we established the final reformer dimensions for each system as shown in Table 5-42. The
tubes and housings are rolled and welded from high alloy or stainless steel sheet. Appropriate locating
features are assumed to be incorporated into the tube to position the CC fans. The fans and separator
washers are loaded into the tubes prior to the endplates being welded on and before the tubes are
welded into the tube sheet. The center tubes are inserted through the inlet tube sheet and through the CC
fan stack before the assembly is welded into the overall shell. The reformer cost summary is shown in
Table 5-43.
100 kW 250 kW
Overall Diameter (mm) 609 635
Overall Length (mm) 2326 2326
Flow Stages 3 tubes @ 168 = 504 7 tubes @ 168 = 1176
Final Catalyst Area (cm2) 232,550 565,950
Table 5-43. SOFC System Reformer Cost Summary - 100 kW and 250 kW
Aside from the reformer, BOP items are assumed to be commercially available and therefore quotes or
budgetary pricing was used.
Table 5-44. SOFC BOP Cost Summary for 100 kW and 250 kW Systems
Annual Production of 100kW SOFC Systems Annual Production of 250kW SOFC Systems
System Component Description
(100) (1,000) (10,000) (50,000) (100) (1,000) (10,000) (50,000)
Nat. Gas Filter $ 97 $ 81 $ 68 $ 60 $ 342 $ 266 $ 251 $ 188
Fuel supply
Natural Gas Control Valve $ 1,788 $ 1,453 $ 1,229 $ 1,182 $ 2,619 $ 2,287 $ 1,997 $ 1,816
Desulfurizer $ 2,427 $ 1,457 $ 1,269 $ 1,142 $ 4,855 $ 3,424 $ 3,027 $ 2,724
3-way valve $ 137 $ 115 $ 97 $ 86 $ 137 $ 115 $ 97 $ 86
Processor
Start-up Air Flowmeter $ 1,493 $ 1,405 $ 1,317 $ 1,229 $ 1,493 $ 1,405 $ 1,317 $ 1,229
Start-up Diverter Valve $ 241 $ 204 $ 172 $ 152 $ 241 $ 204 $ 172 $ 152
Cathode Air Blower $ 961 $ 854 $ 758 $ 698 $ 3,049 $ 2,818 $ 2,586 $ 2,586
Cathode Air Flowmeter $ 1,493 $ 1,405 $ 1,317 $ 1,229 $ 1,493 $ 1,405 $ 1,317 $ 1,229
Cathode Recuperator $ 18,800 $ 17,580 $ 16,439 $ 15,685 $ 28,200 $ 26,369 $ 24,658 $ 23,528
Recovery
Heat
System Exhaust Valves $ 241 $ 204 $ 172 $ 152 $ 241 $ 204 $ 172 $ 152
CHP Export Exchanger $ 2,016 $ 1,915 $ 1,819 $ 1,783 $ 5,416 $ 5,145 $ 4,888 $ 4,790
Hybrid Inverter $ 42,000 $ 33,600 $ 26,400 $ 21,600 $ 96,000 $ 75,000 $ 57,000 $ 45,000
Electronics
Power
Assembly Hardware $ 335 $ 305 $ 275 $ 250 $ 550 $ 500 $ 450 $ 405
Frame and Housing $ 1,005 $ 915 $ 825 $ 745 $ 1,650 $ 1,500 $ 1,350 $ 1,215
Additional Work Estimate $ 6,400 $ 5,800 $ 5,200 $ 4,700 $ 11,400 $ 10,400 $ 9,400 $ 8,500
+
TOTAL BOP COST $ 103,503 $ 87,405 $ 75,818 $ 67,927 $ 203,575 $ 169,675 $ 143,793 $ 126,677
Figures 5-26 and 5-27 show SOFC BOP cost trends across the annual production volumes studied.
Table 5-46. SOFC System Testing Cost Summary - 100 kW and 250 kW
Machine utilization was used to determine the number of machines required to support various product
demand levels. This information, along with equipment cost quotes, was used to determine production
line equipment costs. The production facility estimation is based on the floor area required for production
equipment, equipment operators, and support personnel. Guidelines used for this analysis were
developed by Prof. Jose Ventura at Penn State University, and are detailed in Appendix A-5. Capital cost
breakdown appears in Table 5-48. As was the case for the PEM estimates, a production line refers to the
set of equipment necessary to produce a component, subassembly, or system.
Most of the commercially available grid-tie DC/AC inverters are designed for PV systems and may not be
directly applicable to fuel cells due to their incorporation of specific features for maximizing power output
from a variable solar input. However, several solar inverter manufacturers noted that they could design an
appropriate inverter for fuel cells. They expected the cost of fuel cell inverters to be similar to if not less
than PV inverters since designing an inverter for a fuel cell is less complicated. Design simplification
results from relatively steady input at known voltage which removes the need for maximum power point
tracking (MPPT) system control as is needed to manage solar arrays. The simplification is somewhat
offset by the larger conductor size and heat management required for the higher current associated with
the lower input voltage likely with a fuel cell system. For cost estimating, we assumed that PV inverter
costs would be representative. Inverters are sized based on 120% of the nominal output of the system
size to allow for parasitic loads and short-term overload.
Recently, hybrid DC/AC inverters have been developed to integrate a secondary high power DC/DC port
into the primary DC/AC inverter specifically for connection to storage systems operating in parallel with
the primary power source (typically solar for the existing commercial products). The cost is comparable to
that of a single DC/AC inverter.
The DC/DC, DC/AC and hybrid DC/AC hardware costs are estimated based on the costs of similar
hardware already in limited production that were obtained from the manufacturers. These costs were
blended with costs for systems appropriate for PV systems currently being produced in high volume. The
PV systems operate at somewhat higher voltage, but are a reasonable approximation for the necessary
systems. Table 5-49, 5-50, and 5-51 show the cost breakdown for increasing power and increasing
production volumes on a $/W basis for the DC/DC converter, the DC/AC inverter and the DC/AC hybrid
inverter, respectively. Note that the hybrid inverter costs are considered to be significantly uncertain since
current production levels and field experience are limited.
We received estimates from one company for hybrid inverters that were significantly lower than for
conventional inverters at higher production volumes. In the absence of corroborating estimates, we
believe it unlikely that the companies fabricating hybrid inverters will sell them at substantially reduced
prices compared to conventional inverters already on the market; therefore, we have used the current
price quote for the hybrid inverter and applied an 80% discount rate for higher volumes. This brings the
hybrid inverters near to, but not below, conventional inverters on a cost-per-watt basis.
Control Module $719 $647 $582 $565 $719 $647 $582 $565
H2S sensor $243 $219 $210 $204 $243 $219 $210 $204
Anode/cathode Pressure
$452 $406 $364 $354 $1,356 $1,218 $1,092 $1,062
Sensor
5.5.3.1 Batteries
Energy storage is required to provide critical load support during grid outages. As outlined in Section 4,
the batteries were sized for direct connection to provide 25% of nominal power. An alternative design
approach might include sizing the batteries to provide near full operational power for the high power
portion of the day followed by recharging at night. Battery costs were obtained from battery manufacturers
on the basis of the design required to provide 25% nominal power for 20 and 40 minutes for PEM and
SOFC respectively. Table 5-53 shows the minimum requirements of the battery. For costing purposes
both PEM and SOFC use the same battery assembly. However, the assumption is that the PEM batteries
will experience a 50% discharge (50% remaining charge) during a typical outage event, whereas SOFC
batteries will experience an ~80% discharge associated with the same event. The battery capacity is
shown in Table 5-54. The batteries specified were 12V, 58 Amp-hr at C/100 Hr Rate for 100kW system
and 12V, 137 Amp-hr at C/100 Hr Rate for 250kW.
10
Battelle. 2011. The High Volume Manufacture Cost Analysis of 5 KW Direct Hydrogen Polymer Electrolyte (PEM) Membrane Fuel
Cell for Backup Power Applications. Report to the DOE. DOE Contract No. DE-FC36-03GO13110.
Nominally 34 twelve-volt batteries connected in series are used for both the SOFC and PEM systems.
The costs are shown in Table 5-54.
Battery Power (kW) 100 units 1,000 units 10,000 units 50,000 units
100 $5,690 $5,464 $5,269 $5,060
250 $11,765 $11,298 $10,897 $10,465
The resistor bank is required in case of load decrease to dump some excess power as the reformer
decreases reformate output. During a load decrease, excess reformate from the fuel cell anode can result
in overheating the reformer and its support hardware if power ramp rate is not controlled. The resistor
bank combined with a pulse width modulated (PWM) switch has been designed to be able to handle a
load reduction of 25% of nominal power in a case that battery was fully charged. Table 5-55 shows the
cost of the load bank system.
11 Load bank costs sized based off 25% of the system net power output.
while these costs are high, the power management costs would be substantially greater if a DC/DC
converter were used between the stack and inverter. Hybrid inverters represent an emerging technology
which offers significant overall cost reduction to fuel cell systems that require a managed interface with
energy storage including batteries and ultracapacitors.
Resistor Bank $2,000 $1,610 $1,289 $1,030 $4,026 $3,220 $2,577 $2,061
Grid Interconnect
$1,204 $1,084 $975 $878 $2,645 $2,381 $2,142 $1,928
Switch
Control Module $719 $647 $582 $565 $719 $647 $582 $565
H2S sensor $243 $219 $210 $204 $243 $219 $210 $204
Anode/cathode
$452 $406 $364 $354 $1,356 $1,218 $1,092 $1,062
Pressure Sensor
Total Cost per System $52,536 $43,221 $35,258 $29,859 $117,058 $94,238 $74,725 $61,509
12 Hybrid inverter sized based on system gross power, 120kW and 300kW for the 100kW and 250kW systems,
respectively.
selection. Based on past experience and industry input, reasonable choices regarding the overall system
design were selected: a limitation of this analysis is dependence on representative system designs, not
field tested hardware.
Alternative and innovative manufacturing techniques were not evaluated. Industry feedback indicates that
the techniques used for the cost analysis are consistent with existing processes used by stack component
manufacturers. One possible exception is with the bipolar plates, for which some manufacturers use
compression molded graphite composite material and others use stamped and coated metal material. For
this analysis, graphite composite bipolar plates were chosen due to longevity concerns associated with
the CHP application. Table 6-1 summarizes the manufacturing processes that were evaluated.
Single head slot die with decal transfer (not Dual head slot die
chosen) Multi-pass slot die
Screen printing (not chosen)
Spray coating (not chosen)
Bipolar plate Compression molding Die stamping and coating (metal plates)
MEA forming Ruler blade die cutting Laser cutting
Gasket/seal forming Injection molding Laser cutting
The cost analysis assumed that membrane and gas diffusion layer (GDL) materials were purchased in roll
form. This could result in slightly higher stack cost compared to in-house production of these materials.
However, the membrane and GDL materials are manufactured using complex, highly specialized, multi-
step processes.13 Consequently, in-house production may not be justified until yearly volumes reach the
larger production volumes considered here. However, for consistency with prior reports we assumed both
membrane and GDL materials would be purchased materials for all production volumes.
Heat exchangers make up another significant cost category, especially for the PEM system. The PEM
heat exchangers were scaled on the basis of a ChemCad® model of the system and assumed to be sized
appropriately for their designed load heat duty. These heat exchangers represent an area of uncertainty
in the development of the system costs. Typical off-the-shelf (OTS) heat exchangers carry pressure
ratings that are much greater than required for the PEM CHP systems considered here. Although efforts
were made to work with suppliers to identify the lowest cost heat exchangers capable of meeting system
requirements, a full DFMA analysis was not conducted for these items, and significant potential exists for
their associated costs to decrease with a design-to-purpose engineering effort.
Blower selection depends significantly on system pressure drop at design flow rate. Absent an actual
design, we presumed modest pressure drop. As pressure requirements increase, either through the
cathode or through the combustion system for the reformer, blower requirements may increase notably
with an attendant increase in cost. In most cases, the system designer has some leeway in the design to
minimize pressure drop.
13
James, B.D., J.A. Kalinoski, and K.N. Baum. 2010. Mass Production Cost Estimation for Direct H2 PEM Fuel Cell Systems for
Automotive Applications: 2010 Update. NREL Report No. SR-5600-49933. Directed Technologies, Inc. Available at
http://www1.eere.energy.gov/hydrogenandfuelcells/pdfs/dti_80kwW_fc_system_cost_analysis_report_2010.pdf
Alternative and innovative manufacturing techniques were not evaluated. Based on industry feedback, the
techniques used for the cost analysis are consistent with existing processes used by SOFC stack
component manufacturers
As for the PEM system, SOFC BOP hardware cost estimates are strongly influenced by electrical system
components. The same limitations mentioned for the PEM analysis apply for SOFC. The SOFC system
may include additional energy storage (batteries) to compensate for the relatively slow response of the
stack to changes in electrical load. Since the specifics of a grid outage are not defined, for this analysis
we assumed that the batteries will be the same for either technology. However, the SOFC batteries may
be subjected to deeper discharge events during rapid electrical transients with the system off grid.
The BOP estimates for SOFC presume that lower effectiveness reforming is adequate since some
methane introduced to the stack will be reformed in situ, benefiting stack thermal management. We
assumed that a reformer approximately half the size of a similar scale PEM reformer would be
satisfactory. Appropriate scaling of the reformer to satisfy the specific requirements set by a stack
manufacturer could result in higher or lower costs. Similarly, we assume anode recirculation using an
eductor. Performance of the eductor could impact overall system performance and efficiency and will
likely limit turndown to 2:1 for the SOFC. Alternative methods, such as high temperature blower
recirculated anode gas, could provide better turndown if desired for some applications, but high
temperature blowers represent a significantly higher initial expense and have an unknown life. Proven
high temperature blowers for anode recirculation do not seem to exist at this scale so they were not
considered for this analysis.
BOP hardware cost estimates included here exhibit limited cost savings at high volume. This results from
the specialized nature of the BOP components, particularly the expense of the higher temperature
materials necessary to achieve adequate lifetime.
Bulk commodity materials used in much of the hardware have relatively fixed costs unless purchased at
very low quantities, where prices are very high. Conversely, certain specialty components (e.g., fuel
reformers and compact heat exchangers) required to meet the rigorous specifications of the CHP system
are currently custom designed and are not available at high volumes. The volume discounts applied were
based on general scaling parameters that may not reflect the benefits of industry standardization and/or
more specific design efforts yielding lower overall costs.
Since the fuel cell stacks dictate much of the equipment and space capital costs, all capital costs are
captured in the “Total Stack Manufacturing” category. Furthermore, the manufacturing capital cost (the
investment required to produce the systems) is relatively small on a per-system basis even for a limited
number of units, accounting for only 0.17% of total system cost at the most. Capital costs are assumed to
be amortized over the projected lifetime of the machine or 20 years, whichever is shorter. The number of
stacks required for all but the lowest volume production rates considered for this report results in most
fabrication work being done in house with the attendant capital expenditures necessary to obtain and
commission the production machinery. Stack and System testing is incorporated into the stack and
Assembly categories respectively. All systems and production volumes assume that stack and final
system testing and evaluation will be done in house as a quality control measure. The cost of dedicated
test equipment is rolled into the capital investment.
Figure 7-1. 100 kW PEM system costs at 1,000 units per year.
Figure 7-2. 250 kW PEM system costs at 1,000 units per year.
Tables 7-1 and 7-2 provide the estimated costs for each size and production volume. Figures 7-3 and 7-4
graphically show the pre-markup cost trend with increasing manufacturing volume that is represented in
Tables 7-1 and 7-2. 50% markup was used based on discussions with industry to cover overhead and
profit and to remain consistent with past reports.
Table 7-1. Cost summary for 100 kW PEM CHP Fuel Cell System.
Table 7-2. Cost summary for 250 kW PEM CHP Fuel Cell System.
Figure 7-5 shows the cost per kilowatt (excluding mark-up) for each of the sizes and production volumes.
As expected, there is benefit to increased total production and system size on a cost per kWnet basis.
The trends in Figure 7-5 influence the life cycle cost analysis of Section 9. When considering a 5 year life
for the system, the higher production rate, larger capacity systems offer attractive payback periods
because they are able to generate electrical power at rates competitive with or lower than utility rates in
many locations.
Figure 7-5. Cost per kW for 100 kW and 250 kW PEM system
Since the fuel cell stacks dictate much of the equipment and space capital costs, all capital costs are
captured in the “Total Stack Manufacturing” category. Furthermore, the manufacturing capital cost (the
investment required to produce the systems) is relatively small on a per-system basis even for limited
numbers of units, accounting for 0.64% of total system cost at the most. Capital costs are assumed to be
amortized over the projected lifetime of the machine or 20 years, whichever is shorter. As is the case with
PEM systems, most production volumes involved in SOFC system construction benefit from in-house
production facilities with the attendant capital expenditures necessary to obtain and commission the
machinery. This is somewhat more important for SOFC systems as the cells are fabricated from raw
materials (e.g. YSZ) rather than purchased semi-finished materials (e.g. gas diffusion later carbon paper)
as they are for PEM systems. All systems and sizes assume that final testing, evaluation, and burn-in (if
required) will be done in house as a quality control measure. Hence, the cost of dedicated test equipment
is also rolled into the capital investment for all sizes and volumes.
Figure 7-6. 100 kW SOFC system costs at 1000 units per year.
Figure 7-7. 250 kW SOFC system costs at 1000 units per year.
Tables 7-3 and 7-4 provide the estimated cost values. Figures 7-8 and 7-9 graphically show the pre-
markup cost trend with increasing manufacturing volume that is represented in Tables 7-3 and 7-4.
Table 7-3. Cost Summary for 100 kW CHP SOFC Fuel Cell System
Table 7-4. Cost Summary for 250 kW CHP SOFC Fuel Cell System
100 1,000 10,000 50,000
Description
Units/Year Units/Year Units/Year Units/Year
Total Stack Manufacturing $94,814 $73,566 $70,452 $70,113
Fuel and Air Supply Components $18,298 $15,700 $14,309 $13,556
Fuel Processor Components $14,347 $9,797 $8,604 $8,253
Heat Recovery Components $33,857 $31,718 $29,718 $28,470
Power Electronic, Control, and
$117,962 $95,050 $75,453 $62,217
Instrumentation Components
Assembly Components and Additional
$19,110 $17,410 $15,710 $14,180
Work Estimate
Total system cost, pre-markup $298,389 $243,241 $214,244 $196,789
System cost per KW net, pre-markup $1,193.56 $972.96 $856.98 $787.16
Sales markup 50.00% 50.00% 50.00% 50.00%
Total system cost, with markup $447,583 $364,861 $321,367 $295,184
System cost per KWnet, with markup $1,790.33 $1,459.45 $1,285.47 $1,180.74
Figure 7-10 shows the cost per kilowatt (excluding mark-up) for each of the sizes and production
volumes. As expected, there is benefit to increased total production and system size on cost per kW
capacity. The trends in Figure 7-10 figure into the life cycle cost analysis. When considering a 5 year life
for the system, the higher production rate, larger systems offer attractive payback periods because they
are able to generate electrical power at rates competitive with or lower than utility rates. The cost analysis
does not place any value on grid outage response, though that may be a significantly beneficial factor in
many locations.
Figure 7-10. Cost per kW for 100 kW and 250 kW SOFC system
Before considering specific cost reduction areas it is appropriate to note that neither the PEM nor SOFC
systems analyzed here have been optimized for cost. Neither design specifically reflects installed
equipment and is therefore amenable to improvement. Further, specific applications or installations will
apply different constraints and afford different opportunities in system design. A significant opportunity for
cost reduction likely exists in modifications to the system schematics to eliminate components by
integrating hardware components with one another or by advances in technology that eliminate the need
for some hardware. The first place to look for cost improvement is in the details of the system
configuration, giving attention to potential simplification and function integration. For example, a low cost
hydrogen purification membrane technology could potentially eliminate the PrOx reactor, PrOx air blower
and associated heat exchanger and control hardware, reducing costs in three of the primary categories
considered (control hardware savings would be minimal).
A review of the cost tables (Section 5) and the sensitivity analysis (Section 8) shows that power
electronics are major contributors to the overall cost. The costs presented here assume a hybrid (three- or
four-port) inverter has replaced discrete DC/DC and DC/AC hardware. An additional cost reduction may
be available in some cases by using batteries connected directly in parallel with the fuel cell stack.
However, the savings associated with incorporating a battery in parallel are not as notable as for
incorporating the hybrid inverter and there is greater risk and/or engineering required to match the system
to the application. If the system is to be operated only with the grid present and if an appropriate power
purchase agreement is in place to allow the sale of excess power to the utility the costs of the batteries
can be avoided and some savings is potentially available by having a single purpose inverter in place of
the hybrid inverter.
As the sensitivity analysis of Section 8 shows, additional BOP improvements (e.g. lower cost,
designed-to-purpose heat exchangers, blowers) have the potential to impact cost. Blowers are sensitive
to overall system pressure drop so system design changes that enable a lower overall system pressure
drop are potentially beneficial. The caveat is that we have assumed relatively aggressive pressure drop
characteristics so there is a potential for blower requirements to increase rather than decrease.
A final comment on the manufacturing process is appropriate. The scrap and reject rates used here are
those recommended by our industry contacts as representative of the state of the art. Five percent rate of
failure at final test, however, is an unacceptable rate for mass produced hardware. It is essential to
develop effective quality control measures and robust fabrication processes to reduce those rates to less
than 0.5%.
8. Sensitivity Analysis
8.1 PEM System
The sensitivity analysis of the costs for the 100 kW and 250 kW PEMFC systems at production volumes
of 1,000 and 10,000 units per year explores the impact of specific variations to the assumptions for the
major contributing cost factors and highlights their significance. The cost factors subjected to the
sensitivity analysis were chosen because of their significant contribution to the cost and/or the difficulty
associated with precisely assessing their magnitude, such as the cost of platinum. The analysis
demonstrates the effect of varying each with all else held equal on the overall cost of the system based
on reasonable variations to each factor. The results of the sensitivity analyses are shown in the following
charts (Figures 8-1 thru 8-4), which show the relative importance of the major cost drivers investigated.
For the 100 kW PEMFC system sensitivity analysis, the cost factors that were varied, along with the basis
for the change and the overall effect, include:
o The three-port DC/AC hybrid inverter is relatively new technology and has not been
proven in mass production for this application
o The result of including (if needed) the DC/DC Converter significantly increases the overall
system cost
Heat Exchangers
o Assumed total cost of $50,797 at 1,000 units/year
o Assumed total cost of $46,873 at 10,000 units/year
o Varied by -25%
o OTS heat exchangers are over designed for this application, yielding a conservative
estimate for their cost, thus, a 25% reduction yields a significant effect
o The accuracy of heat exchanger cost would benefit from a complete DFMA® on future
cost studies
DC/AC Hybrid Inverter
o Assumed to be $33,600 at 1,000 units/year
o Assumed to be $26,400 at 10,000 units/year
o Varied by -25%
o Some of the cost estimates received for the DC/AC hybrid inverter suggested even lower
overall costs are possible at higher volumes; however, since the technology has not been
proven for mass production a more conservative estimate was employed
o Despite the more conservative cost estimate, employing the DC/AC hybrid inverter yields
significant savings that will be even greater if the lower costs at high volumes are
achieved
Current Density
o Assumed to be 0.4 A/cm 2
o Adjusted to 0.6 A/cm2 to see effect
o The current density of 0.4A/cm2 was chosen to achieve the desired 50,000 hours of
durability
o If the current density could be raised to 0.6 A/cm2 while still maintaining durability, there
is a slight positive impact on system cost
Blowers
o Selection dependent on system pressure drop
o Assumed blower output pressure of ~2 psig
o Varied to blower output pressure of ~5 psig (resulting in specification of different blowers)
o There is a slightly negative impact on overall system cost if the assumed system
pressure drop of 2 psig is not achieved
Platinum Loading
o Assumed to be 0.4 mg/cm 2
o Varied by + 0.1 mg/cm2, - 0.25 mg/cm2
o For past system studies on lower power systems, platinum loading has shown a
significant cost impact
o There is only a minor impact on overall system cost for a system of this size
Platinum Cost
o Assumed to be $1,380/tr.oz.
o Varied by ±50%
o The cost of platinum is highly variable
o For past system studies on lower power systems, platinum cost has shown a significant
cost impact
o There is only a minor impact on overall system cost for a system of this size
Figure 8-1. PEM sensitivity analysis: 100 kW system cost – 1,000 production volume
Figure 8-2. PEM sensitivity analysis: 100 kW system cost – 10,000 production volume
For the 250 kW PEMFC system sensitivity analysis, the cost factors that were varied along with their
basis and effect include:
oThere is a slight negative impact on overall system cost if the assumed system pressure
drop of 2 psig is not achieved
Platinum Cost
o Assumed to be $1,380/tr.oz.
o Varied by ±50%
o The cost of platinum is highly variable
o For past system studies on lower power systems, platinum cost has shown a significant
cost impact
o There is only a minor impact on overall system cost for a system of this size
Figure 8-3. PEM sensitivity analysis: 250 kW system cost – 1,000 production volume
Figure 8-4. PEM sensitivity analysis: 250 kW system cost – 10,000 production volume
For the 100 kW SOFC system sensitivity analysis, the cost factors that were varied along with their basis
and effect include:
o Varied by -25%
o Some of the cost estimates received for the DC/AC hybrid inverter suggested even lower
overall costs are possible at higher volumes; however, since the technology has not been
proven for mass production a more conservative estimate was employed
o Despite the more conservative cost estimate, employing the DC/AC hybrid inverter yields
significant savings which will be even greater if the lower costs at high volumes are
achieved
SOFC Stack
o Assessed to be $31,070 at 1,000 units/year
o Assessed to be $27,779 at 10,000 units/year
o Varied by ±25%
o Stack is a significant cost contributor
o Varying the stack costs by 25% results in a moderate effect on system cost
Cathode Recuperator Heat Exchanger
o Assumed to be $17,580 at 1,000 units/year
o Assumed to be $16,439 at 10,000 units/year
o Varied by ±25%
o A COTS heat exchanger for this application was slightly over designed yielding a
conservative estimate for the cost thus a 25% reduction yields a slight positive effect
o The accuracy of heat exchanger cost would benefit from a complete DFMA® on future
cost studies
Fuel Processor
o Assessed to be $5,675 at 1,000 units/year
o Assessed to be $5,115 at 10,000 units/year
o Varied by ±25%
o Only a minor effect was seen by varying the fuel processor cost by 25%
Blowers
o Selection dependent on system pressure drop
o Assumed blower output pressure of ~2 psig
o Varied to blower output pressure of ~5 psig (resulting in specification of different blowers)
o There is a slight negative impact on overall system cost if the assumed system pressure
drop of 2 psig is not achieved
Figure 8-5. SOFC sensitivity analysis: 100 kW system cost – 1,000 production volume
Figure 8-6. SOFC sensitivity analysis: 100 kW system cost – 10,000 production volume
For the 250 kW SOFC system sensitivity analysis, the cost factors that were varied along with their basis
and effect include:
Figure 8-7. SOFC sensitivity analysis: 250 kW system cost – 1,000 production volume
Figure 8-8. SOFC sensitivity analysis: 250 kW system cost – 10,000 production volume
It should be noted that all power generation alternatives would be expected to operate continuously, or
near continuously. Continuous duty operation would likely bring them under environmental regulations
related to total site emissions, however, fuel cell systems would have significantly lower emissions than
other standard alternatives (with the obvious exception of solar and geothermal energy). However, for this
analysis we have not included any environmental externality costs, as they vary widely with location,
application, and site specifics. Additionally, although some recycling systems are in place to capture high
value materials such as those used in fuel cells, these programs are not mature enough to assume they
are easily or widely available and thus have been excluded from this analysis.
Average electrical load of 400 kW for primary operations (e.g. lighting, information
processing/servers, production equipment operation)
o 400 kW was selected as a suitable demand load based on an assumed electrical
baseline load at or near 100/250 kW
Average heating load of 200 kW for space, water, or process heating
o Usually provided by natural gas or electrically driven vapor-compression heat pumps
The lower cost and lower maintenance requirement of natural gas fired heating
equipment usually results in the selection of natural gas when available, unless
the natural gas cost is exceptionally high
For consistency with standard commercially installed systems , natural gas
heating was assumed to be 85% efficient as a basis for comparison
Average cooling load of 200 kW for space, water, or process cooling
o Usually provided by either electrically driven vapor-compression equipment or natural gas
fired absorption equipment.
Comparable equipment capital cost and maintenance requirement between
electrically or gas fired equipment results in a site specific choice
For consistency with standard commercially installed systems, an electrically
driven heat pump with Coefficient of Performance (COP) of 3 was used as the
basis for comparison
Natural gas provided at the site with rates based on Energy Information Agency (www.eia.gov) on-
line reported cost for commercial sites (varies with location, e.g. the cost in California is
$0.811/therm)
Electric power provided at the site with rates based on Energy Information Agency on-line reported
costs for commercial sites (varies with location, e.g. the cost in California is $0.173/kWh)
Using the conditions above, data from three states were analyzed: California, New York, and Illinois. We
also considered the average rates for the United States.
The basic fuel cell system parameters used for the analysis are:
PEM
Total system efficiency: 80%
System electrical efficiency: 30%
Net electrical output: 100 kW and 250 kW
Waste heat peak temperature: ~80°C
Recoverable waste heat:
o ~167 kW @ 100 kW electric output
o ~417 kW @ 250 kW electric output
SOFC
Total system efficiency: 90%
System electrical efficiency: 40%
Net electrical output: 100 kW and 250 kW
Waste heat peak temperature: ~600°C
Recoverable waste heat:
o ~125 kW @ 100 kW electric output
o ~313 kW @ 250 kW electric output
The cost basis and operating cost assumptions for California are shown in Tables 9-1 and 9-2.
Finally, the lifecycle cost analysis was performed for eight system configurations:
PEM CHP system without waste heat recovery
o 100 kW
o 250 kW
PEM CHP system with waste heat recovery
o 100 kW
o 250 kW
SOFC CHP system without waste heat recovery
o 100 kW
o 250 kW
SOFC CHP system with waste heat recovery
o 100 kW
o 250 kW
9.3 Results
An example comparison of the 100kW SOFC system with waste heat recovery in California is shown in
Table 9-3. The internal rate of return (IRR) presumes an initial cost to install the system followed by yearly
savings which are applied at the beginning of each year following the initial outlay. For systems analyzed
with waste heat recovery, additional cost is included to cover integration with facility heating/cooling
infrastructure; this cost is outside the scope of the FC CHP system since every installation would be
slightly different, but important when considering installation costs. The life of the system is presumed to
be 5 years of operation at the nominal power level. The payback period is the total installed cost divided
by the projected yearly savings.
Operating costs include:
Annual Electricity Cost – Cost of FC generated electricity and electricity purchased from the grid to
make up remainder of demand not covered by FC CHP system
Annual Natural Gas Cost – Cost of NG to provide heating/cooling
Annual Consumables – Cost of consumed items (filters, elements, desulfurizer media, etc)
Annual O&M – Cost for estimated time and materials for routine maintenance
The Annual Savings is the difference between the Annual Total Operating Costs and the typical operating
costs listed in Table 9-2.
For the case shown in Table 9-3 the internal rate of return increased from 28% to 56% as the number of
systems manufactured and installed increased from 100 to 50,000. While this case has an attractive
return on investment, not all cases were found to be beneficial. For example, Table 9-4 shows the data
for 100 kW PEM systems without waste heat recovery. Due to additional fuel costs (both for heating and
a reduced electrical efficiency compared to SOFC) the annual savings over grid electricity do not provide
a return on investment within the first 5 years at production volumes below 10,000 units/year.
Table 9-3. Annual Cash Flows, IRR and Payback Period (100 kW SOFC system with waste heat
recovery)
Table 9-4. Annual Cash Flows, IRR and Payback Period (100 kW PEM system without waste heat
recovery)
Table 9-5 summarizes the IRR and payback period for all eight California cases considered using the
base costs for electricity and natural gas shown in Table 9-1.
As illustrated in Table 9-5, 100 kW PEM systems without heat recovery are not profitable after the 5 year
payback period until production volumes increase substantially and still do not offer significant financial
benefit even at the highest production levels. The addition of heat recovery for all applications (in Table 9-
5 scenario) yields improved rates of return, particularly as production volumes increase for the 250 kW
systems.
SOFC systems offer much better financial benefit, particularly in California, for most production volumes
with or without heat recovery. The potential return in excess of 35% for the 250 kW system with waste
heat recovery even at low production volumes provides a strong business case for the initial investment;
this is significant as this application can provide an initial market entry for fuel cell systems that can
support additional cost reduction and market expansion in the future. A caution is appropriate, however,
because this rate of return is based on a 5 year lifetime for the fuel cell system. Current SOFC
technology has not yet been demonstrated to consistently achieve the required 40,000 hour lifetime.
The electric and gas prices shown in Table 9-1 are unique to California. Tables 9-6 and 9-7 provide cost
comparisons using New York State utility rates and the average U.S. utility rates, respectively. The New
York costs appear roughly comparable to the California installations because of the relatively higher
electricity costs compared to natural gas costs. However, when the national averages are considered,
none of the configurations offer a reasonable return on investment because the ratio between electricity
and gas cost is too small.
Table 9-6. IRR and Payback for Installations in New York State
New York Average Utility Rates for Commercial Installations
System Production rate 100/year 1,000/year 10,000/year 50,000/year
100 kW
PEM IRR after 5 years -12% -4% 1% 5%
(without Payback Period (years) 7.41 5.66 4.83 4.38
waste
heat 250 kW
recovery) IRR after 5 years -2% 7% 14% 18%
Payback Period (years) 5.24 4.07 3.46 3.11
100 kW
PEM
IRR after 5 years 1% 10% 17% 21%
(with
waste Payback Period (years) 4.84 3.71 3.17 2.88
heat 250 kW
recovery
IRR after 5 years 7% 17% 25% 30%
heating)
Payback Period (years) 4.04 3.14 2.67 2.40
100 kW
SOFC IRR after 5 years 13% 23% 29% 34%
(without Payback Period (years) 3.55 2.80 2.46 2.25
waste
heat 250 kW
recovery) IRR after 5 years 23% 34% 41% 47%
Payback Period (years) 2.81 2.27 2.01 1.83
100 kW
SOFC
IRR after 5 years 24% 36% 44% 49%
(with
waste Payback Period (years) 2.71 2.14 1.88 1.71
heat 250 kW
recovery
IRR after 5 years 33% 45% 53% 60%
heating)
Payback Period (years) 2.27 1.84 1.63 1.48
Table 9-7. IRR and Payback for Installations Based on Average U.S. Energy Costs
US Average Utility Rates for Commercial Installations
System Production rate 100/year 1,000/year 10,000/year 50,000/year
100 kW
PEM IRR after 5 years -49% -45% -42% -41%
(without Payback Period (years) 55.80 41.41 34.54 30.80
waste
heat 250 kW
recovery) IRR after 5 years -44% -40% -37% -35%
Payback Period (years) 39.34 29.70 24.64 21.81
100 kW
PEM
IRR after 5 years -19% -12% -8% -5%
(with
waste Payback Period (years) 9.63 7.35 6.27 5.68
heat 250 kW
recovery
IRR after 5 years -21% -15% -11% -8%
heating)
Payback Period (years) 10.65 8.23 6.97 6.25
100 kW
SOFC IRR after 5 years -19% -13% -9% -7%
(without Payback Period (years) 9.84 7.73 6.77 6.16
waste
heat 250 kW
recovery) IRR after 5 years -13% -7% -3% 0%
Payback Period (years) 7.76 6.26 5.52 5.01
100 kW
SOFC
IRR after 5 years -19% -12% -8% -5%
(with
waste Payback Period (years) 9.63 7.35 6.27 5.68
heat 250 kW
recovery
IRR after 5 years 4% 12% 17% 21%
heating)
Payback Period (years) 4.36 3.53 3.11 2.83
Although not included here, it is worth noting that some states have natural gas to electric price ratios
which are significantly disadvantageous to fuel cell installations, particularly for commercial customers.
These ratios are really another way of looking at utility spark spread14, the margin of electricity pricing vs.
price of fuel to produce the energy. Generally, the price of electricity is high enough to yield even a small
amount of savings from a relatively efficient fuel cell system, such as the 30% and 40% efficient systems
evaluated in this study. However, in some states, like Texas and Illinois for example, the commercial price
for natural gas is greater than 30% of the price for commercial electricity. This structure is such that the
cost of electricity generated by a PEM fuel cell using commercial gas is actually greater than the price of
14 “An introduction to spark spreads,” Energy Information Administration, last modified February 8, 2013,
https://www.eia.gov/todayinenergy/detail.cfm?id=9911.
electricity from the utility. On the other hand, “industrial” gas prices in these states do yield a potentially
profitable price margin. Therefore, the business and regulatory conditions in these states and others lead
to a natural gas rate15 for industrial customers that is more beneficial for onsite generation when
compared to the commercial rate. In contrast, the natural gas rates in California and New York are more
nearly comparable for commercial and industrial customers (although New York’s electricity price drops
significantly for industrial customers). Table 9-8 provides a comparison of the natural gas vs. electricity
prices for the states evaluated and highlights the general advantage the “industrial” price can have over
“commercial” prices in some states. The states with red ratios indicate that PEM systems without heat
recovery cannot be profitable.
Electricity ($/kWh) 0.1077 0.1729 0.0934 0.1555 0.0788 0.1388 0.0666 0.0620 0.0580
$NG/$Electricity 26% 16% 33% 15% 32% 16% 30% 36% 19%
Using the 100 kW SOFC system with waste heat recovery (at the 100 units per year production level) as
an example, Figure 9-1 provides a heat map for IRR indicating favorable utility prices for FC CHP
applications. Due to the significant impact of electricity costs, the IRR is more vertically aligned. For
example, any location with electricity cost below $0.09 per kWh will have a negative IRR, whereas, at any
gas price there is the potential to yield favorable returns, provided electricity costs for that area are high
enough.
$0.014 -16% -10% -5% 0% 5% 10% 14% 18% 22% 26% 30% 34% 37%
$0.016 -18% -12% -6% -1% 4% 9% 13% 17% 21% 25% 29% 33% 36%
$0.018 -20% -13% -7% -2% 3% 7% 12% 16% 20% 24% 28% 32% 36%
$0.020 -22% -15% -9% -3% 2% 6% 11% 15% 19% 23% 27% 31% 35%
$0.022 -23% -16% -10% -5% 0% 5% 10% 14% 18% 22% 26% 30% 34%
$0.024 -25% -18% -12% -6% -1% 4% 9% 13% 17% 21% 25% 29% 33%
$0.026 -27% -20% -13% -7% -2% 3% 7% 12% 16% 20% 24% 28% 32%
$0.028 -30% -22% -15% -9% -3% 2% 6% 11% 15% 19% 23% 27% 31%
$0.030 -32% -23% -16% -10% -5% 0% 5% 10% 14% 18% 22% 26% 30%
$0.032 -35% -25% -18% -12% -6% -1% 4% 9% 13% 17% 21% 25% 29%
Figure 9-1. IRR heat map for a 100-kW SOFC system with waste heat recovery at various costs of
electricity and natural gas
15 “Natural Gas Prices,” Energy Information Agency, last modified December, 31, 2015,
http://www.eia.gov/dnav/ng/ng_pri_sum_dcu_SCA_m.htm
Figure 9-2 uses the sample example to illustrate the payback period in years based on a range of utility
costs.
$0.014 8.81 6.98 5.78 4.94 4.30 3.82 3.43 3.11 2.85 2.63 2.44 2.27 2.13
$0.016 9.42 7.36 6.04 5.12 4.45 3.93 3.52 3.19 2.91 2.68 2.48 2.31 2.16
$0.018 10.13 7.79 6.33 5.33 4.60 4.05 3.61 3.26 2.97 2.73 2.53 2.35 2.20
$0.020 10.96 8.27 6.64 5.55 4.76 4.17 3.71 3.34 3.04 2.79 2.58 2.39 2.23
$0.022 11.93 8.81 6.98 5.78 4.94 4.30 3.82 3.43 3.11 2.85 2.63 2.44 2.27
$0.024 13.09 9.42 7.36 6.04 5.12 4.45 3.93 3.52 3.19 2.91 2.68 2.48 2.31
$0.026 14.49 10.13 7.79 6.33 5.33 4.60 4.05 3.61 3.26 2.97 2.73 2.53 2.35
$0.028 16.24 10.96 8.27 6.64 5.55 4.76 4.17 3.71 3.34 3.04 2.79 2.58 2.39
$0.030 18.47 11.93 8.81 6.98 5.78 4.94 4.30 3.82 3.43 3.11 2.85 2.63 2.44
$0.032 21.40 13.09 9.42 7.36 6.04 5.12 4.45 3.93 3.52 3.19 2.91 2.68 2.48
Figure 9-2. Payback heat map for a 100-kW SOFC system with waste heat recovery at various costs of
electricity and natural gas
Commercial rates were chosen for the analysis since the scenarios evaluated would probably be
classified as commercial, rather than industrial. If the site has, or can negotiate, industrial gas rates, the
financial benefits can be improved in those states with a notable difference between commercial and
industrial rates. Additionally, our analysis only considered state average rates. Local gas and electric
rates may vary significantly from state averages such that most installations will need to perform a similar
economic analysis before investing in fuel cell systems.
9.4 Conclusions
The initial capital cost for PEM systems inhibits, but does not eliminate, the development of attractive
investment opportunities for CHP systems. The addition of waste heat recovery, for applications where
the heat is useful, improves the return on investment significantly. An SOFC system that can achieve a 5
year lifetime (or 40,000 hours of operation) would provide more attractive returns, particularly for sites
with a use for recovered heat. Lifecycle cost analyses depend significantly on local electric and gas rates
and the rate structure associated with certain applications. As noted, if the industrial gas price applies,
then the installation of fuel cells becomes attractive in more locations. It is also important to factor in the
amount of baseline load covered by the fuel cell system.
As a final note regarding value proposition, this lifecycle cost analysis does not place any value on grid
outage response, though that may be a significantly beneficial factor in many locations.
10.0. Conclusions
10.1 System Cost Summary
Figures 10-1 and 10-2 show the summary pie charts for 1000 units per year, repeated from Section 7,
which emphasize that BOP costs dominate the final cost for both PEM and SOFC systems. The stack
represents a maximum of 31% of the total system cost with the stack contributing less cost as a percent
of total cost for most system production volumes. Within the BOP costs, the major contributor for all
systems is the power electronics. Heat recovery and fuel processing contribute significantly to the PEM
system, but less so to the SOFC system owing to the simpler BOP system required for it. For this size of
equipment, the power electronics would represent even a greater portion of the cost if a more
conventional DC/DC converter and DC/AC inverter were used to accommodate batteries for off-grid
operation, rather than the hybrid inverter used in this analysis. The reduced cost of solar photovoltaic
power is driving progress for grid-connected DC/AC inverters that can interface with storage. That
progress will benefit fuel cell systems with active management of fuel cell power (available 24 hours/day)
eventually bringing the cost of this hardware below that for solar systems. Heat exchanger costs
represent a significant area for evaluation and future cost reduction; particularly for the high temperature
heat exchangers that are required for both SOFC and PEM systems.
An important feature of primary power and CHP systems is that the space and weight requirements are
less stringent than they are for mobile applications (e.g., material handling equipment), therefore requiring
less customization and enabling the use of more commercially available hardware and production
equipment. However, at this early stage of system design, the industrial grade components used may be
over-designed for the application, forcing costs up. Cost reduction efforts at the 100 and 250 kW scales
must be mindful of codes and standards for industrial equipment and of local siting and permitting
requirements which may apply at this scale – a topic not explored in this study.
Because of the number of stacks to be produced for each system (2 to 10), most stack fabrication
processes are assumed to be brought in-house along with final assembly and testing. The calculation for
determining whether to outsource versus produce in house is shown in Appendix A-1. This ensures that
the capital investment, once made, will be distributed across an appropriate number of systems.
The total costs for each size at two representative production volumes are shown in Tables 10-1 and 10-
2. A sales markup of 50% was integrated into the overall cost and is called out separately. These tables
emphasize the cost difference of the PEM and SOFC systems on an installed cost per kW basis. PEM
systems cost 37% to 65% more than SOFC. For continuous power applications where the cyclic life
concerns for SOFC are not realized, SOFC clearly appears to be the preferred choice, having both higher
efficiency and lower initial cost. However, for applications where either operational characteristics or
diurnal spark spread suggest turning down or turning off the fuel cell system, the PEM system may be the
preferred choice. Applications with significant use for the waste heat may also favor PEM. As shown in
Section 9, PEM systems may offer an attractive value proposition, so the choice between these
technologies will be site specific with SOFC generally being preferred when applicable. This is consistent
with the expanding installed base of fuel cell systems designed for large capacity (400 kW and greater)
power offset and back-up power, like those being installed by Bloom ,for example. The life cycle cost
analysis of Section 9 uses the highlighted values inTables 10-1 and 10-2 for the 100 kW and 250 kW
systems.
Figures 10-3 and 10-4 show the after-markup cost per kW electrical and for kW total for PEM and SOFC
systems assuming the waste heat is used for space, water, or process heating. Figure 10-3 is based on
actual kW of heat delivered to the CHP system; it is not adjusted for the cost of electricity or natural gas
that would have been otherwise used to heat water. The heat delivered may have more or less value
depending on how the water would otherwise have been heated. A comparison of Figure 10-3 with Figure
10-2 shows a marked advantage of using the waste heat. However, some applications will not benefit
directly from waste heat, having no use for hot water, space heating, or process heat. Data Centers are
an example of an application that requires cooling, essentially to remove the heat generated by the power
supplied to the servers. To assess the potential for this application we assumed an absorption chiller
would be operated with the waste heat from the fuel cell system. Because of the heat collected from the
fuel processor, the PEM system waste heat is hot enough to use in chiller applications. The SOFC system
clearly has sufficient temperature but has a lower net heat available because of the higher electrical
efficiency. Figure 10-4 includes the cost per installed kW total assuming that some of the waste heat is used
for cooling. The estimate of cooling capacity and associated equipment cost was based on an online
calculator16 for a specific size and type of cooling. The available options were not optimized but provide
some guidance on the possibilities. The absorption chiller cost used in the calculations is discounted from
the on-line estimate by 50%, the assumed cost of an electric chiller that would otherwise be required to
provide similar cooling. Figure 10-5 shows a further improvement in the installed cost of fuel cell power
and some shift in the relative cost of total power provided between the two technologies.
When considering the value of the heat delivered in terms of natural gas and electric costs avoided, the
economic benefit of a continuously operating system versus a back-up power system becomes apparent
if the difference between electric power and natural gas cost on a comparable energy basis (the “spark
spread”) is characterized by electric prices higher than gas prices by at least $0.10/kWh. A particular
advantage of a fuel cell system over other alternatives (e.g., a natural gas reciprocating engine or gas
turbine) is longer expected life in continuous operation coupled with simple maintenance requirements.
Very low noise and emissions enable siting in urban and environmentally sensitive areas that would not
allow most lower-cost power generation options.
Figure 10-4. Final system cost of energy delivered (thermal and electric).
Figure 10-5. Total system cost of energy delivered including an absorption chiller.
These returns/payback values do not incorporate any adjustment for environmental externalities or the
value of continued operation during grid outages. For some businesses, the value of maintained
operations during an extended grid outage may exceed the total system cost (e.g., avoided frozen food
loss from a grocery store, or business loss for an on-line retailer). To the extent that electricity can be
generated on-site for less than grid-purchased electricity, the payback and IRR become very attractive.
Some sites in restrictive areas – like southern California – may only be amenable to fuel cell primary
power and CHP installations because of the relatively low environmental footprint of fuel cells when
compared to the available continuous-operation competitive options.
Although the investment appears attractive for many locations, there is risk in making the investment until
fuel cell systems have been demonstrated to consistently achieve 40,000 hours of trouble-free operation
in real world applications. SOFC systems (typically somewhat higher capacity) are currently being
installed in east and west coast locations. However, the financial arrangements, while proprietary, seem
to provide some level of protection to the business. The designs apparently utilize a smaller stack
configuration than used for this analysis, increasing system cost due to redundancy but also allowing a
certain percentage of stack failures before system routine maintenance and repair are necessary. The
analysis we have undertaken assumes that larger stacks can achieve the necessary lifetimes and
therefore expand the opportunities for fuel cell system installations through lower system first cost.
The choice to define a dual-use system capable of both grid-connected and off-grid operation increases
power electronics cost because batteries must be incorporated and coordinated with the fuel cell to permit
safe and effective transient response. Requiring that the system operate with or without an applied
thermal heat load also increases system cost slightly. For PEM systems, a full load radiator is required.
For SOFC systems, high temperature bypass valves are required and provisions must be made for safe
discharge of high temperature combustion gas when a CHP heat load is not present. The components
may not be appropriate for all applications and some cost could be eliminated by assuming that the
system would only operate when the grid is available. However, with a beneficial spark spread the back-
up power capable system pays for itself in a few years. For areas without a beneficial spark spread, the
site-specific life cycle cost analysis should include the value of avoided losses associated with typical grid
outage conditions. Grid outage operation will be more valuable in areas with lower grid reliability.
As fuel cell-powered material handling equipment increases market penetration and as automotive use of
fuel cells increases, we expect that the CHP/prime power market will continue to develop, initially for
areas with conducive utility pricing and critical grid outage conditions and then to applications where the
waste heat can add sufficient value to provide acceptable payback times. We note, however, that the
MHE and automotive developments only indirectly impact SOFC stationary applications because those
markets require load-following, high turn-down systems. Power electronics developments for MHE and
automotive systems assist in reducing fuel cell system cost. Development of power electronics for the
solar power industry may also provide future cost savings for fuel cell systems.
∗
∗ ∗
where:
1
1 1
1
To calculate a baseline production line cost, we assume that the line is capable of operation three, 8-hour
shifts per day for 250 days per year. Therefore, total available production time for both operation and
setup (TR + TS) at 100% utilization is 6000 hours per year. The actual production time based on
utilization (U) can be calculated as:
Input assumptions based on our previous work and on the assumptions shown in Table 3 of the ASME1
paper results in the following:
Total capital cost over the assumed 20 year production life is calculated as:
20
Where:
n = number of unique pieces of equipment making up production line
1
James, B.D., Spisak, A.B., Colella, W.G., 2014, “Design for Manufacturing and Assembly (DFMA®) Cost
Estimates of Transportation Fuel Cell Systems,” ASME Journal of Manufacturing Science and
Engineering, New York, NY: ASME, Volume 136, Issue 2, p. 024503.
1
Machine Rate Calculations
Manufacturing Cost Analysis of 100 and 250 kW Fuel Cell Systems for Primary Power and Combined Heat and Power
Applications / DOE Contract No. DE-EE0005250
As an example, the bipolar plate compression molding line consists of the following items:
Units
Expected
Cost Per
Life
line
Bipolar Plate Compression Molding
1000 ton fast‐acting press Wabash 1000H‐48 $650,000 1 20
Heated platens, 15"x12", 4.5 kW, Custom Engineering $12,500 1 10
controller
Arbor or hand‐operated hydraulic pre‐ Central Hydraulics 6 ton $400 1 20
mold press bench‐top press w/ pump
Electronic scale, industrial, gram Mettler‐Toledo WM3002 $6,000 1 10
resolution
Small industrial oven Grieve NBS‐400 $1,000 1 20
Ni Ci Li CCap
Bipolar Plate Compression Molding
1000 ton fast‐acting press 1 $650,000 20 $600,000
Heated platens, 15"x12", 4.5 kW, 1 $12,500 10 $25,000
controller
Arbor or hand‐operated hydraulic pre‐ 1 $400 20 $400
mold press
Electronic scale, industrial, gram 1 $6,000 10 $12,000
resolution
Small industrial oven 1 $1,000 20 $1,000
Total $638,400
Energy costs to operate the line are a function of the power required to operate each piece of equipment.
For cost estimating purposes, the total power draw of the production line can be calculated in similar
fashion to the total capital cost as follows:
Where:
n = unique pieces of equipment making up production line
Ni = number of item i required for production line
Vi = voltage supplied to item i
Ai = current draw of item i
Di = duty cycle of item i
2
Machine Rate Calculations
Manufacturing Cost Analysis of 100 and 250 kW Fuel Cell Systems for Primary Power and Combined Heat and Power
Applications / DOE Contract No. DE-EE0005250
Ni Vi Ai Di P
Bipolar Plate Compression Molding
1000 ton fast‐acting press 1 460 150 96% 66.24
Heated platens, 15"x12", 4.5 kW, 1 230 25 25% 1.44
controller
Arbor or hand‐operated hydraulic pre‐ 1 0 0 10% 0.00
mold press
Electronic scale, industrial, gram 1 120 1 100% 0.12
resolution
Small industrial oven 1 230 20 20% 0.92
Total 68.72
Where: U > 0
Graphically:
500
400
300
200
100
0
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Utilization
3
Machine Rate Calculations
Manufacturing Cost Analysis of 100 and 250 kW Fuel Cell Systems for Primary Power and Combined Heat and Power
Applications / DOE Contract No. DE-EE0005250
Applying the above to the remaining stack production lines yields the following:
100%
Power Labor
Baseline Utilization
LTPEM Production Line Cost Cost
Cost Machine
(/hr)) (/hr)
Rate
Bipolar Plate Compression Molding $638,400 $1.50 $45.00 $64.70
Platinum Catalyst Preparation $37,000 $0.01 $9.00 $11.18
Slot Die Coating $94,892 $0.90 $22.50 $62.20
Decal Transfer $58,400 $0.74 $22.50 $26.66
MEA Hot Pressing $289,000 $0.80 $22.50 $40.24
Die Cutting $125,000 $0.32 $22.50 $30.15
Gasket Injection Molding $48,000 $0.72 $45.00 $26.04
End Plates $416,000 $6.50 $90.00 $120.89
Stack Assembly $1,310 $0.00 $45.00 $45.08
Testing and Conditioning $300,000 $2.42 $14.85 $34.85
100%
Power Labor
Baseline Utilization
SOFC Production Line Cost Cost
Cost Machine
(/hr)) (/hr)
Rate
High Volume Slurry Production $80,000 $2.49 $4.50 $16.18
Low Volume Slurry Production $60,000 $1.03 $4.50 $13.54
Tape Casting $288,360 $0.66 $45.00 $62.56
Anode Blanking $125,000 $0.00 $45.00 $29.83
Screen Printing $66,612 $0.81 $45.00 $49.71
Kiln Firing $393,000 $16.10 $14.85 $53.99
Sintering $500,000 $16.10 $14.85 $60.26
Laser Cutting $35,000 $9.26 $14.85 $33.81
Sheet Metal Stamping $150,000 $1.61 $45.00 $55.40
Interconnect $365,860 $2.98 $90.00 $114.43
End Plates $416,000 $6.50 $90.00 $120.89
Sealing $27,500 $1.59 $14.85 $48.21
Stack Assembly $550 $0.00 $45.00 $45.03
Stack Brazing $500,000 $16.10 $14.85 $60.26
Testing and Conditioning $150,000 $2.49 $4.50 $16.18
4
Machine Rate Calculations
Manufacturing Cost Analysis of 100 and 250 kW Fuel Cell Systems for Primary Power and Combined Heat and Power
Applications / DOE Contract No. DE-EE0005250
As indicated in James (2014)2, at low utilizations, job shops may make parts at a lower cost by pooling
orders. Additional job shop costs include the profit charged by the job shop, and any overhead incurred
by the manufacturer as a result of contract administration, shipping, and incoming parts inspection.
Assuming a 65% minimum machine utilization and 40% markup for profit plus overhead, the job shop
maximum machine rate becomes:
∗
1.4 ∗ ∗ ∗
6000 ∗ 0.65
Assuming labor, energy and capital costs are the same, the maximum job shop machine rate for the
bipolar plate line above would be:
$37.42
1.4 ∗ $27.31 $118.83
0.65
$37.42
$27.31 $118.83
$37.42
0.409
$118.83 $27.31
In other words, for utilization rates of less than 40.9%, bipolar plate manufacturing should be sub-
contracted to a job shop instead of incurring the costs of manufacturing the plates in house. A similar
calculation was conducted for each stage of the manufacturing process for both PEM and SOFC systems,
and the appropriate costs (e.g. produced at a job shop or in house) are included in the cost analysis for
each annual production volume evaluated.
It should be noted that the make-buy strategy outlined above results in a discontinuity in the machine rate
curve (and by extension the total cost curve) since the job shop machine rate is unchanged up to the
critical utilization rate of 40.9%, as shown below.
2
Ibid.
5
Machine Rate Calculations
Manufacturing Cost Analysis of 100 and 250 kW Fuel Cell Systems for Primary Power and Combined Heat and Power
Applications / DOE Contract No. DE-EE0005250
This can be further illustrated by estimating the production per unit for bipolar plates. Each plate contains
0.691 kg of BMC940 composite. Material cost for a purchase quantity Q is computed using the formula
presented in Appendix A2. The throughput of the process is 50 parts/hr, yielding a maximum annual
capacity of 300,000 plates per year, and requires 0.5 operator time per machine. Using the above
equations, the bipolar plate unit cost as a function of line utilization is shown below:
5.00
Cost per Plate
4.00
3.00
2.00
1.00
0.00
0 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Utilization
Where multiple processes are closely coupled due to timing or handling constraints, the make-buy
decision needs to consider the overall cost of the entire process train and not just the cost of individual
processes within the train. In cases like these, the entire cost of the process train needs to be computed for
both in-house and outsourced manufacturing costs using the following formula:
6
Machine Rate Calculations
Manufacturing Cost Analysis of 100 and 250 kW Fuel Cell Systems for Primary Power and Combined Heat and Power
Applications / DOE Contract No. DE-EE0005250
A similar situation arises when a single machine can be used for multiple processes, such as a slot die
coater that can be used for both anode and cathode catalyst deposition. In this case, the utilization used in
the machine rate calculation is total time required to complete all of the processes divided by the total
machine time available:
7
Machine Rate Calculations
Manufacturing Cost Analysis of 100 and 250 kW Fuel Cell Systems for Primary Power and Combined Heat and Power Applications /
DOE Contract No. DE-EE0005250
Background
In general, material cost on a per unit basis (e.g. per kg, per m2) decreases with increasing
purchase volumes, due primarily to the manufacturer’s ability to produce larger volumes of
material from a single production run setup. It has been noted in previous work that material
cost estimates at various discreet purchase volumes could be estimated for the intermediate
volumes using a learning curve analysis.
From Cost Estimator’s Reference Manual, Stewart, R.M., et al, 2nd Ed., Wiley-Interscience,
1995, the general learning curve equation is:
Y = AXb
where:
Y = time or cost per cycle or unit
A = time or cost for first cycle or unit
X = number of cycles or units
b = log(m)/log(2)
m = slope of learning curve
If the material production is “learned” after 10,000 units (i.e., no substantial discounts are
available for higher volume purchases), then the cost of X is the cost of the 10,000th unit.
Preliminary Analysis
Where possible, quotes were obtained from both domestic and international suppliers for the
materials. Other material costs were obtained from previous third party fuel cell manufacturing
analysis reports.
Some materials, such as the silicone gasket material, are considered commodity items for which
manufacturing processes are well established and supplies are high enough to support most
available demand. One supplier provided the following quote for liquid silicone material of
$7.00 to $7.50 per pound ($15.40 to $16.50 per kg) for quantities ranging from 250 to 25,000
pounds.
For these materials, the cost curve is very flat, which means the value of m in the learning curve
equation is high. Iteration using the costs above led to a value of m=0.99, which results in:
b = log(0.99)/log(2) = -0.0154
Using a learned cost of $15.40/kg for a volume of 55,000 kg, then the cost of the first unit is:
For a purchase of 250 kg of material, the calculated cost per unit is:
12.00
10.00
8.00
6.00
4.00
2.00
0.00
For specialty materials, like the GDL in the LTPEM stack, the cost curve is steeper. One
supplier provided low volume quotes of $535.63/m2 for 3 m2, and $313.13/m2 for 45 m2.
Estimates obtained from previous fuel cell manufacturing cost analyses estimated high volume
costs to be in the range of $56.00/m2 for volumes up to 100,000 m2. Iteration using the costs
above led to a value of m=0.86, which results in:
b = log(0.86)/log(2) = -0.21759
Using a learned cost of $56.00/m2 for a volume of 100,000 m2, then the cost of the first unit is:
For a purchase of 40 and 100 m2 of material, the calculated cost per unit is:
GDL Cost
800.00
700.00
600.00
Cost($/m2)
500.00
400.00
300.00
200.00
100.00
0.00
1 50 100 150 200 250 300 350 400 450 500
Purchase Volume (m2)
Using the above approach, the following learning curve parameters were used for the cost
analysis:
For the annual system volumes used, the material purchase volume can be extremely large. For
example, to manufacture bipolar plates for 50,000 100 kW systems requires over 60,000 metric
tons of BMC940 material. According to the learning curve equation, a bulk purchase of this size
would cost $0.188/kg – a cost that is most likely unachievable and therefore unrealistic.
To address this problem, we have elected to assume that any additional volume discounts beyond
the bulk pricing represented by the cost Y and quantity X in the above tables would be no more
than that achieved by doubling the quantity X. Since the learning curve slope (m) represents the
amount of reduction in Y when quantity X is doubled, the minimum material price is simply
Y×m. Therefore, the material price for a given purchase quantity (q) is calculated as:
Yq = Max(A Xb, Y × m)
Special Cases
Platinum prices are dictated by precious metals spot markets and are generally not subject to
purchase volume reductions. This correlates to a learning curve slope value of m = 1.
The polyester film would seem to be a material that would have a commodity price profile like
that of the silicone sheet. However, price quotes received showed relatively high cost at low
purchase volumes of less than 100 m2, which falls over by over 97% at bulk purchase volumes
greater than 14,000 m2.
High volume purchase costs of DE-521 were very difficult to obtain. Since the cost of DE-521 is
driven primarily by the cost of DuPont’s Nafion polymer, the cost was calculated based on
learning curve analysis of the primary cost driver, and on the assumed values associated with
manufacturing and supplier markup.
Background
The DFMA® software produces assembly times based on hand assembly at its most efficient.
Using the 50 kW PEM as an example, the assembly time for the stack was estimated to be 3.070
hours.
The learning curve analysis essentially backs that number up to a time when bugs are still being
worked out of the assembly process.
From Cost Estimator’s Reference Manual, Stewart, R.M., et al, 2nd Ed., Wiley-Interscience,
1995, the general learning curve equation is:
Y = AXb
where:
Y = time or cost per cycle or unit
A = time or cost for first cycle or unit
X = number of cycles or units
b = log(m)/log(2)
m = slope of learning curve
Analysis
For stack assembly time, if we assume that m = 0.85 (typical for aerospace processes, from Cost
Estimator’s Reference Manual, Stewart, R.M., et al, 2nd Ed., Wiley-Interscience, 1995), then:
b = log(0.85)/log(2) = -0.23447
If the stack assembly process is “learned” after 100 units, and the assembly time for the X =
100th stack is the DFMA® time, then the time to assemble the first unit is:
The average time to assemble the first 100 units ( ) is calculated as:
. ∗ .
3.954
The time to assemble all subsequent units is assumed to be A, making the average time to
assemble n units (n > 100) is calculated as:
.
A∗i ∗ 100
Using the above equations, the average stack assembly times are:
The production facility estimation is based on the floor area required for production equipment,
equipment operators, and support personnel. Primary space allowance guidelines used for this
analysis were developed by Prof. Jose Ventura at Penn State University, and were downloaded
from http://www.personal.psu.edu/jav1on 10/18/2013.
Equipment Footprint
The line utilization calculations provide the equipment count for a particular production line.
Using the bipolar plate production as an example, each station consists of two pieces of
equipment: the 1000 ton fast acting press, and the post-bake oven, which have the following
footprint dimensions:
Press: 60 in x 70 in
Oven: 40 in x 40 in
Allowing a 3 foot (36 in) margin on all sides for maintenance access increases the total machine
footprints:
Three additional space allowances are made for each line for material, personnel, and aisles. The
production stations will require space for material receiving and part pickup, typically done using
pallets. We will assume one standard 40” by 48” pallet for receiving and pickup, adding to the
required area by:
Ventura recommends personnel space of 20 ft2 per person to allow for movement within the
work station during equipment operation. The bipolar plate pressing requires a single operator,
adding:
Aisle allowance is based on the largest transported load. Because we intend to transport material
and finished parts on standard pallets, our anticipated load size is 27 ft2, for which Ventura
recommends a 30% - 40% allowance for the net area required, which include personnel and
material. Using a value of 35% makes the aisle allowance for the bipolar plate station:
The total floor space allocation for the bipolar plate station is:
The PEM fuel cell stack production was broken up into 9 primary work stations with total floor
space allocations calculated using the above formulas as:
Floor Space
Production Station
Allocation (ft2)
Catalyst 262
Anode Slot Die Coating 296
Cathode Slot Die Coating 296
Decal Transfer 258
Hot Press & Die Cut 426
Die Cutting 178
Bipolar Plate 357
End Plate 1236
Anode / Cooling Gasket 233
Cathode Gasket 233
Assembly 258
Test and Conditioning 245
In addition to equipment, industrial facility space must be allocated for offices, food service,
restrooms and parking, all of which depend on the number of people present during operation.
For most automated or semi-automated production equipment, one operator can cover multiple
machines. In addition, some operations have long periods of unsupervised operation (e.g. the 10
hour milling time in catalyst production). Ventura estimates the number of required machine
operators using the formula:
n′ = (a + t) / (a + b), where
a = machine-operator concurrent activity time (load, unload)
b = independent operator activity time (inspect, package)
t = independent machine activity time
n′ = maximum number of machines per operator
The reciprocal of n′ would represent the minimum number of operators per machine. Using time
data (in seconds) extracted from the DFM process analyses for a and t, and estimating time for b,
resulted in the following:
To obtain a rough estimate of the number of operators required during any one shift, multiply the
required number of operators per station (combinations of either 1.0, 0.5, 0.33) by the number of
stations required to produce a particular annual volume and the line utilization (assuming a single
operator is trained to perform multiple tasks). Using the line utilization numbers for 10,000
systems per year we have:
Operators Operators
PEM Production Station
Stations Utilization per line Per Shift
Catalyst 1 0.066 0.33 0.02
Anode Slot Die Coating 1 0.670 0.50 0.34
Cathode Slot Die Coating 1 0.311 0.33 0.10
Decal Transfer 1 0.744 0.50 0.37
Hot Press 4 0.772 0.33 1.03
Die Cutting 1 0.490 1.00 0.49
Bipolar Plate 20 0.945 0.50 9.45
End Plate 1 0.288 0.33 0.10
Anode / Cooling Gasket 4 0.794 0.50 1.59
Cathode Gasket 2 0.635 0.50 0.64
Stack Assembly 6 0.855 1.00 5.13
Stack Test and Conditioning 3 0.944 0.33 0.94
System Assembly 6 0.855 1.00 5.13
System Test 3 0.944 0.33 0.94
Total 26.27
Rounding up to 27 machine operators per shift, and assuming approximately 1 support staff per 4
line operators for purchasing, QC, and maintenance, the facility needs to support a total of 34
employees. Ventura provides estimates for the following additional facilities:
In addition, office space for support personnel is estimated at 72 ft2/employee based on the State
of Wisconsin Facility Design Standard. Therefore, additional space requirements are:
Space Required
Facility
(ft2)
Food Service 405
Restrooms 200
Parking 7452
Office 504
Assuming a construction cost of $250/ ft2, the estimated cost of factory construction is
approximately $4,610,750.
Total real estate required can be estimated as building floor space plus parking and building set-
back (distance from building to streets and other structures). Assuming a 30 foot set-back on all
sides of a reasonably square facility gives a total real estate requirement of:
Assuming a real estate cost of $125,000/acre, the estimated total real estate cost is approximately
$140,051.
The SOFC production facility estimation is based on the floor area required for production
equipment, equipment operators, and support personnel. Primary space allowance guidelines
used for this analysis were developed by Prof. Jose Ventura at Penn State University, and were
downloaded from http://www.personal.psu.edu/jav1 on 10/18/2013.
Equipment Footprint
The line utilization calculations provide the equipment count for a particular production line.
Using the application of the cathode layer as an example, each station consists of two pieces of
equipment: the screen printer, and a heated conveyor for slurry drying, which have the following
footprint dimensions:
Screen printer: 63 in x 55 in
Heated conveyor: 24 in x 36 in
Allowing a 3 foot (36 in) margin on all sides for maintenance access increases the total machine
footprints:
Screen printer: (63 + (2 36)) (55 + (2 36)) / 144 in2/ft2 = 119 ft2
Heated conveyor: (24 + (2 36)) (36 + (2 36)) / 144 in2/ft2 = 72 ft2
Three additional space allowances are made for each line for material, personnel, and aisles. The
production stations will require space for material receiving and part pickup, typically done using
pallets. We will assume one standard 40” by 48” pallet for receiving and pickup, adding to the
required area by:
Ventura recommends personnel space of 20 ft2 per person to allow for movement within the
work station during equipment operation. The cathode layer screen printing requires a single
operator, adding:
Aisle allowance is based on the largest transported load. Because we intend to transport material
and finished parts on standard pallets, our anticipated load size is 27 ft2, for which Ventura
recommends a 30% - 40% allowance for the net area required, which include personnel and
material. Using a value of 35% makes the aisle allowance for the cathode layer screen printing
station:
The total floor space allocation for the screen printing station is:
SOFC fuel cell stack production was broken up into 15 primary work stations with total floor
space allocations calculated using the above formulas as:
Floor Space
Production Station
Allocation (ft2)
High Volume Slurry Preparation 457
Low Volume Slurry Preparation 336
Tape Casting 655
Anode Pressing 551
Anode Blanking 318
Screen Printing 321
Kiln Firing 622
Sintering 598
Laser Cutting 221
Sheet Metal Stamping 185
Interconnects 589
End Plates 1261
Sealing Line 395
Stack Assembly 422
Stack Brazing 598
Stack Test and Conditioning 252
In addition to equipment, industrial facility space must be allocated for offices, food service,
restrooms and parking, all of which depend on the number of people present during operation.
For most automated or semi-automated production equipment, one operator can cover multiple
machines. In addition, some operations have long periods of unsupervised operation (e.g. the 10
hour milling time in ceramic ink production).
Ventura estimates the number of required machine operators using the formula:
n′ = (a + t) / (a + b), where
a = machine-operator concurrent activity time (load, unload)
b = independent operator activity time (inspect, package)
t = independent machine activity time
n′ = maximum number of machines per operator
The reciprocal of n′ represents the minimum number of operators per machine. Using time data
(in seconds) extracted from the DFM process analyses for a and t, and estimating time for b,
resulted in the following:
To obtain a rough estimate of the number of operators required during any one shift, we multiply
the required number of operators per station (combinations of either 1.0, 0.5, 0.33) by the
number of stations required to produce a particular annual volume and the line utilization
(assuming a single operator is trained to perform multiple tasks). Using the line utilization
numbers for 10,000 stacks per year we have:
Operators Operators
Production Station
Stations Utilization per line Per Shift
High Volume Slurry Preparation 1 0.311 0.33 0.10
Low Volume Slurry Preparation 1 0.085 0.33 0.03
Tape Casting 8 0.95 0.33 2.53
Anode Pressing 1 0.309 0.33 0.10
Anode Blanking 1 0.141 0.50 0.07
Screen Printing 16 0.985 0.33 5.25
Kiln Firing 11 0.935 0.50 5.14
Sintering 9 0.897 0.33 2.69
Laser Cutting 3 0.871 1.00 2.61
Sheet Metal Stamping 1 0.421 1.33 0.56
Interconnects 27 0.986 0.33 8.87
End Plates 1 0.273 0.33 0.09
Sealing Line 5 0.86 0.33 1.43
Stack Assembly 6 0.983 1.00 5.90
Stack Brazing 1 0.596 0.33 0.20
Stack Test and Conditioning 4 0.903 0.33 1.20
Total 36.80
Rounding up to 37 machine operators per shift, and assuming approximately 1 support staff per 4
line operators for purchasing, QC, and maintenance, the facility needs to support 47 employees.
Ventura provides estimates the following additional facilities:
In addition, office space for support personnel is estimated at 72 ft2/employee based on the State
of Wisconsin Facility Design Standard. Therefore, additional space requirements are:
Space Required
Facility
(ft2)
Food Service 555
Restrooms 300
Parking 10212
Office 720
Assuming a construction cost of $250/ ft2, the estimated cost of factory construction is
approximately $1,249,050.
Total real estate required can be estimated as building floor space plus parking and building set-
back (distance from building to streets and other structures). Assuming a 30 foot set-back on all
sides of a reasonably square facility gives a total real estate requirement of:
Assuming a real estate cost of $125,000/acre, the estimated total real estate cost is approximately
$269,476.
Model Approach
Process Flow
Background
The BDI DFMA® software provides pre-programmed cost models for the casting and cell machining
operations used to manufacture the fuel cell stack end plates. The end plates need to be rigid in order to
apply even pressure across the face of the stack. The process selection for the LTPEM end plate was
automatic sand casting of A356 cast aluminum to near net shape, followed by finish machining of the
stack contact face on a Haas VF-3B vertical machining center, and reaming and tapping the holes for fuel,
exhaust, and cooling flows.
Preliminary Analysis
Ø 10.0mm.
5mm.
typ 15mm. 12mm.
30mm. 100mm.
20mm.
80mm.
159mm.
M14 x 1.25 typ
80mm. 20mm.
6mm.
7mm.
M30 x 1.5
245mm.
305mm.
30mm.
23mm. 140mm.
5mm.
15mm.
5mm.
377mm.
3mm.
12mm.
End Plate
The BDI DFMA® estimates a 18.35 hour machine setup time, and calculates the total manufacturing time
for the end plates as 375 sec, making the total machine time for annual production of 1,000 100 kW
systems:
Machine time = (375 sec/part / 3600) 4,000 parts + 18.35 = 435.0 hours
Machine utilization is:
Utilization = 435.0 / 6000 = 7.3%
Assuming 2 full-time operators (1 for casting, 1 for machining) per station, the total machine labor time is
equal to twice the machine time = 881.2 hours.
Material cost was determined in accordance with Appendix A2 as:
Material cost = $2.42/kg
Tooling cost is $19,786 and is assumed to be capable of producing 180,000 parts. Amortizing over a 5
year production life, the total annual tooling cost is:
ଵ
Annual tooling cost = (Tooling cost Number of tools purchased)
ହ
Model Approach
Catalyst ink preparation operation
o Machine setup labor time based on user input
o Compute required batch size based on part batch size and catalyst loading
o Compute catalyst ink material unit cost based on usage
o Compute catalyst ink processing time and machine utilization
Anode catalyst ink slot die deposition to membrane operation
o Compute processing time based on production size and substrate speed
o Compute number of setups based on purchased roll length
o Compute setup labor time based on user input and number of setups required
o Compute material unit cost based on usage
o Compute required heater area based on drying time and substrate speed
o Compute total anode ink deposition processing time and machine utilization
Cathode catalyst ink slot die deposition to transfer substrate operation
o Compute processing time based on batch size and substrate speed
o Compute number of setups based on purchased roll length
o Compute setup labor time based on user input and number of setups required
o Compute material unit cost based usage
o Compute required heater area based on drying time and substrate speed
o Compute total cathode ink deposition processing time and machine utilization
Cathode catalyst ink decal transfer calendaring operation
o Compute processing time based on batch size and substrate speed
o Compute number of setups based on purchased roll length
o Compute setup labor time based on user input and number of setups required
o Compute required heater area based on heating time and substrate speed
o Compute decal transfer processing time and machine utilization
Process Flow
Background
In their March, 2009 report, Directed Technologies, Inc. (DTI) reported that the wet platinum
catalyst composition as specified in US Patent 7,141,270 consists of:
6 wt% Pt
9 wt% Vulcan XC-72 (carbon black)
72 wt% Nafion DE-521 solution (5 wt% Nafion)
6.5 wt% DI water
6.5 wt% methanol
DTI also reported that, assuming that all solvents are driven off during the drying process, the
dry catalyst consists of:
48.4 wt% Vulcan XC-72 (carbon black)
32.3 wt% Pt
19.4 wt% Nafion
While DTI assumed that the ink slurry was mixed using ultrasonic processing, technical
literature and conversations with stack manufactures indicates that ball milling is used as the
primary means of grinding and homogenizing the catalyst ink, with milling times reported in the
range of 4 hours to “overnight.” US Patent 6,187,468 details a two-step preparation process of
mixing (milling) for 60 to 300 minutes, followed by 30 to 300 minutes in a “three-dimensional
vibrating stirrer.” Constant processing in a regular or planetary ball mill for 8-10 hours may
suffice for both the mixing and stirring parts of the process.
Manufacturers noted that there are significant losses during the ink production process, which
tends to occur when handling ink/slurry from one part of the process to the next (e.g. transfer of
final composition from mixing vessel to catalyst application method apparatus), but that much of
the platinum was subsequently recovered, reducing the platinum scrap rate to 1% or less.
Manufactures noted that past, low-volume catalyst application was performed using screen
printing, but that the current process is generally done roll-to-roll. At least one approach
involves a two-step process. One catalyst layer is applied directly to the membrane, and the other
catalyst layer is applied to a low-cost substrate material. The membrane is then turned over, and
the second catalyst layer is applied by hot press decal transfer.
Gore has proposed a 3-step MEA manufacturing process that involves sequential roll-to-roll
coating. The catalyst ink is applied to a backing material, dried and re-rolled. The membrane is
then applied to the first catalyst layer using a co-extrusion deposition, which is dried and re-
rolled. Catalyst ink is then applied to the membrane layer, dried and re-rolled. The 3-layer MEA
would then move to the hot-pressing operation to apply the GDL.
All of the above methods pay a material cost penalty by applying catalyst to the entire MEA
surface, including the non-active areas. A more economical approach as used in this analysis
involves using a slot-die patch coating process (see www.frontierindustrial.com), where anode
catalyst is applied to the membrane and the cathode catalyst applied to a transfer substrate in
rectangular patches sized to the active area. The cathode catalyst patches are then bonded to the
membrane using hot press decal transfer, followed by the hot-pressing operation to apply the
GDL.
Preliminary Analysis
Batch volume
Catalyst batch volume depends on the coated area, catalyst loading, and maximum catalyst batch size.
The cells for this analysis will have an active area size of:
Based on the wet platinum catalyst composition as specified above, 100 grams of wet catalyst
contains 6 grams of Pt and has a volume of:
To obtain a loading of 1 mg/cm2, the depth of the wet catalyst layer is:
Based on the dry platinum catalyst composition as specified above, 100 grams of dry catalyst
contains 32.3 grams of Pt and has a volume of:
To obtain a loading of 1 mg/cm2, the depth of the dry catalyst layer is:
The total Pt loading for this design is 0.4 mg/cm2 with cathode loading in a 2:1 ratio relative to
anode loading, making the loadings 0.267 mg/cm2 and 0.133 mg/cm2 for the cathode and anode,
respectively, which will require wet deposition to depths of 27 and 53 microns, respectively,
resulting in dry layer depths of 8 and 16 microns. Therefore, to coat both sides of the membrane
with a total loading 0.4 mg/cm2 will require a total coated depth of 80 microns (0.008 cm):
Wet catalyst weight = 0.85 g/cm3 (780.0 0.008) cm3 = 5.304 g/part
The 60 kW stack requires 283 cells. Based on producing 2,000 stacks per year (2 stacks per 100 kWnet
system, 1,000 systems per year), required annual production before scrap is:
Annual production = 283 parts/stack 2,000 stacks = 566,000 parts
Catalyst batch size = 566,000 parts 5.304 g/part 0.001 kg/g = 3,002 kg
Material cost of the ink is calculated using the weight percents of the slurry constituents
multiplied by the raw material cost to determine a cost per kilogram. Material pricing was
obtained from suppliers and supplier web sites in February, 2014. Platinum cost is assumed to
be equal to the 1 year Nymex futures market price as $1,380/tr.oz. ($44,368/kg) for April, 2015
delivery. Bulk cost for DE-521were estimated at $160, which translates to a price of about
$1,344/kg. Bulk costs for XC-72 catalyst grade carbon black was quoted by WeiKu Information
and Technology and others at around $900/MT ($0.90/kg). Bulk cost for Methanol was quoted
by Methenex and others at around $630/MT ($0.63/kg). The cost of DI water is based on
amortized distillation costs obtained from www.apswater.com.
Using the above quotes, learning curve analysis in accordance with Appendix A2 was applied to
determine the following material costs:
Platinum = $44,368/kg
Nafion DE-521 = $219.84/kg
Vulcan XC-72 = $0.990/kg
Methanol = $0.735/kg
DI Water = $0.094/kg
Material cost = (0.06 44,368) + (0.72 219.84) + (0.09 0.990) + (0.065 0.735) +
(0.065 0.094)
Material cost = $2,820.51/kg = $2.821/g
The primary cost for operating the ball mill is the energy input to the motor running the mill. Some
studies have looked into the cost of operating large ball mills used for cement and powder metallurgy
material processing, where the target parameter is the amount of energy required to process a given
amount of material, usually expressed in kW-hr/ton. The calculations are complex owing to the large
number of inputs to the calculations.
In “Technical Notes 8, Grinding,” R. P. King develops a relation based on fundamental physical models
of ball mill processing. He assumes a 35% volumetric loading ratio, of which milling balls represent 10%
of the total charge volume. Given a mill with diameter d and length l, the total catalyst charge volume is:
Patterson Industries offers simple torque drive batch ball mills in 42”d 48”l (1.067m d 1.219m l), and
48”d 60”l (1.219m d 1.524m l),. These provide maximum catalyst charge volumes of:
We note that production levels of 2,000 stacks per year will require 3,002 kg of catalyst production per
year, or about 11 batches per year in the smaller mill. At less than one batch per month, the smaller mill
would seem to be the prudent choice.
King presents a log-log plot showing that a mill with a diameter of 1 meter will consume about 10 kW of
power, where a mill with a diameter of 2 meters consumes about 100 kW. These two values yield the
equation:
Power = 10d3.32 kW
To estimate the power required to process a batch of catalyst with a density of 850 kg/m3, we plug the
mill diameter into the power equation to obtain:
Once processing is complete, the catalyst ink will need to be separated from the milling balls and
transferred to the coating machine. While we presently have no information about this part of the
process, one approach would be the use of a vacuum sieve (e.g., Farleygreene, Ltd. SM950 Sievmaster
Vacu-siev) to remove and separate the catalyst ink from the mill, and transfer the ink to a transport
container or directly to the coater reservoir.
ShopVac reports a sealed suction of 54 in-H2O (13.4 kPa) for their 2 HP (1.5 kW) unit. Using an
equivalent vacuum sieve with a 1.5” (0.038 m) diameter hose and 80% transfer efficiency, the flow rate
is:
Since the catalyst forms 90% of the charge volume, the total charge volume of
Charge volume (m3) = 1.11 (Catalyst weight (kg) / Catalyst density (kg/ m3))
Charge volume (m3) = 0.0013 Catalyst weight
Therefore, the optimal time required to remove the charge volume is:
Material removal time (sec) = Charge volume / Flow rate = 8.1 Catalyst weight
The optimal time to remove a full charge of catalyst from the mill would be:
We will estimate the total transfer time to remove the ink from the mill and transfer it to the coater as
twice the ink removal time.
The estimated total processing time is calculated as the sum of the setup time, material prep time, milling
time, and transfer time, multiplied by the total number of batches processed for annual production of
3,002 kg of catalyst, is:
Process time = 11 batches (10 + (1.8 / 60) + (2 (39.8 / 60))) hrs = 124.92 hrs
Given an availability of 6000 hours per year per machine, the number of mills required is:
As indicated previously, one approach to catalyst deposition involves a two-step process. The
anode catalyst is applied to the membrane and the cathode catalyst applied to a transfer substrate
in rectangular patches sized to the active area. The cathode catalyst patches are then bonded to
the membrane using hot press decal transfer. Both the membrane application and decal creation
are direct deposition processes onto a substrate material; one being the membrane itself, and the
other to a carrier substrate, commonly a polyester or polyimide material. The patches will be
centered in the full cell size envelope of 305.0 mm 378.4 mm.
We will assume a roll-to-roll slot die application process. Depending on the roll length and width,
multiple machine setups may be required to process the material for an entire production run. The length
of material being processed is a function of the batch size and the number of parts that can be produced
across the material width. Assuming no cutting margin for rectangular MEAs, the optimal part
orientation can be determined based on the fraction of material width left over as waste as:
Material Cost
Membrane material is sold in widths of 12” (0.305 m) and 24” (0.610 m) with lengths of 50 or
100 meters. Common thin films (polyimide, polyethylene) used as transfer media tend to be
either 0.4 or 0.8 meters, while lengths can be found up to a maximum of about 1000 meters.
GDL material is typically sold in either 0.4 or 0.8 meter widths, and is available up to a
maximum of 800 meter lengths.
The membrane roll has the smallest standard widths and is the most expensive, so it will be used
to determine the maximum coating width with minimum scrap. Sizing the cells across the 305.0
mm dimension, 2 cells will take up 610 mm of membrane width, leaving a 0.0 mm edge margin
on a 610 mm roll width for the membrane. Sizing across the 378.4 mm dimension, 1 cell will
take up 378.4 mm, leaving a 231.6 mm edge margin. Since 2 cell widths across results in lower
scrap, the material length required will be:
Material length = (566,000 parts / 2 part widths/part length) 378.4 mm part length /
1000 = 107,087 meters
Using learning curve analysis in accordance with Appendix A2, the material cost before scrap
can be estimated as:
Slot die coating machine setup consists of loading and threading the substrate and loading the
catalyst ink into the reservoir. For costing purposes, we will take the setup time as a user input
and assume a value of 0.5 hours. Bulk roll stock is available in 1000 meter lengths, so that the
number of setups required to run 400,000 parts is:
Slot die coating is capable of producing very thin coating thicknesses. The coated material
passes the slot die at a speed determined by the rheology of the coating material and the
thickness of the application. While the precise rheology of the catalyst ink is not known, we can
estimate the substrate speed using the tape casting estimating formula as follows:
The wet coating thickness was calculated above as 200 microns per 1 mg/cm2 of platinum
loading. The cathode/anode coating ratio is assumed to be 2:1. For a total loading of 0.4
mg/cm2 of platinum, the anode will be coated to a depth of 27 microns, while the cathode will be
coated to a depth of 53 microns, making the maximum coating speeds:
Anode Maximum coating speed = 157.18 0.98727 = 110.4 mm/sec = 6.62 m/min
Cathode Maximum coating speed = 157.18 0.98753 = 78.56 mm/sec = 4.71 m/min
Throughput (parts/hour) = Coating speed (m/min) Parts per part length (parts) /
Part length (m) 60 min/hour
Anode: Throughput = 6.62 2 / (318.4 / 1000) 60 = 2,495.0 parts/hour
Cathode: Throughput = 4.71 2 / (318.4 / 1000) 60 = 1,775.1 parts/hour
Anode machine time = (1,071 setups 0.5 hours/setup) + (566,000 parts / 2,495.0
parts/hour) = 762.4 hours
Cathode machine time = (108 setups 0.5 hours/setup) + (566,000 parts /
1,775.1parts/hour) = 372.9 hours
Given an availability of 6000 hours per year per machine, the number of coating systems required is:
Tooling Cost
Slot dies are precision machined and assembled to provide uniform coating thickness. The cost can vary
widely depending on the coating fluid properties and die size. Frontier Industries estimates a stainless
steel fixed die cost of $14,000 dollars, and is capable of delivering approximately 100,000 parts before
refurbishment at a cost of around $3500. Assuming 4 refurbishments before scrapping, and amortizing
over a 5 year production life, the total annual tooling cost is:
ଵ
Annual tooling cost = (Tooling cost Number of tools purchased)
ହ
where: Number of tools purchased = Roundup(Total production / Tool life)
Total production = Annual production 5
ଵ
Annual tooling cost = (($14,000 + (4 $3500)) Roundup((566,000 parts/year 5 years) /
ହ
500,000 parts/tool)) = $33,600
Following deposition, the catalyst ink is dried, usually by means of a tunnel dryer positioned
directly after the deposition step. The drying can be done by either radiant or convective heating.
For the cost analysis, we will assume radiant (infrared) heating and compute the cost of drying
by determining the required heater area based on the substrate speed and the drying time.
Infrared heating panels are generally sold with various energy watt densities and in standard
sized units and assembled to provide the necessary heating area. Using the Casso Solar Type FB
as an example, standard watt densities are 15 and 25 W/in2 (23 and 39 kW/m2) with a standard
width of 12” (0.305 m) and lengths in 12” increments up to 60” (1.524 m). They note that 25
W/in2 corresponds to an emitter temperature of 880°C, and that the conversion efficiency of
electrical power to usable radiant energy is up to 80%.
Drying time is a function of the evaporation rate of the solvent and is inversely and exponentially
proportional to the coating thickness. Experiments conducted by Mistler (Tape Casting of
Ceramics, Ceramic Processing Before Firing, 1978) indicate drying rates of 1.3510-5 g/cm2-sec
at room temperature for an air flow rate of 2 L/min, and 2.2210-5 g/cm2-sec at room temperature
for an air flow rate of 75 l/min.
The change in density from wet to dry catalyst is 0.335 g/cm3, making the liquid removed per
unit area a function of coating thickness as follows:
Anode liquid removed per area = 0.335 g/cm3 0.0027 cm = 0.0009 g/cm2
Cathode liquid removed per area = 0.335 g/cm3 0.0053 cm = 0.0018 g/cm2
For costing purposes, we will take drying time as an input and use the substrate speed and part
width to compute the theoretical required heater area.
At a rate of 2.010-5 g/cm2-sec drying rate, the estimated drying time is:
Anode drying time = 0.0009 g/cm2 / 2.010-5 g/cm2-sec = 45 sec = 0.75 min
Cathode drying time = 0.0018 g/cm2 / 2.010-5 g/cm2-sec = 90 sec = 1.50 min
Sizing for the maximum dryer length, and assuming 12” 36” panels fitted two across the
drying conveyor, we require 14 total IR panels.
The roll-to-roll decal transfer operation can be either a semi-continuous process where the
material is indexed into a standard heated platen press (see DTI, Mass Production Cost Estimation
for Direct H2 PEM Fuel Cell Systems for Automotive Applications: 2010 Update, Section 4.4.6.1) or
performed by pre-heating and passing through heated rollers in a calendaring process. For the
preliminary analysis we will assume a calendaring process
Setup
Decal transfer setup consists of loading, threading and aligning the anode and cathode into the
calendaring rollers. For costing purposes, we will take the setup time as a user input and assume
a value of 0.5 hours. The number of setups is a function of the shortest roll stock length, so that
the required setups to run 566,000 parts is the same as the number of setup for the anode slot die
coating:
The calendaring process consists of 2 main steps: preheating and rolling. We will assume that
the coated membrane and decal catalyst layers are brought together and passed through an
infrared tunnel oven for preheating. Assuming that the two layers need to reach 100°C, we can
estimate the oven dwell time as (noting that 1 W = 1 J/sec):
Oven dwell time = Part weight (g) Part specific heat (J/g-°C) Temperature rise (°C) /
Energy input (W)
If we assume that the same infrared heaters used for drying are used for preheating, the energy
rate impinging on the part is:
Common polymers (PTFE, polyester, polyimide) have specific heats in the range of 1.1-1.3 J/g-
°C and densities around 2.2 g/cm3. Specific heat capacities of the dry catalyst constituents are:
Catalyst specific heat = (0.194 4.2) + (0.323 0.13) + (0.484 4.18) = 2.88 J/g-°C
The volume of dry catalyst for the anode and cathode per part are:
The volume of substrate material (25 micron thickness) per part is:
The heating dwell time for each is then (dry catalyst density = 0.515 g/cm3):
Anode oven dwell time = ((2.2 g/cm3 1.02 cm3 1.2 J/g-°C) + (0.515 g/cm3 0.655
cm3 2.88 J/g-°C)) 75°C / 1,435 W = 0.318 sec
Cathode oven dwell time = ((2.2 g/cm3 1.02 cm3 1.2 J/g-°C) + (0.515 g/cm3 1.310
cm3 2.88 J/g-°C)) 75°C / 1,435 W = 0.366 sec
For the calendaring process, the layers will be moving together, so the worst case heating time of
0.366 seconds is used to determine the required oven length. At a substrate speed of 5 m/min
(8.33 cm/sec), the required heating length is about 0.31 meters, which can be accomplished using
four 12” by 24” IR panels (two for each layer).
Parts per hour = 300 m/hour / 0.318 m 2 parts per width = 1886.8 parts/hour
Once the material layers are preheated, they are compressed between steel rollers that bond the
catalyst decal layer to the membrane. The decal substrate is then peeled away from the decal
layer and collected on a roll or in a bin. Total machine time to setup and produce 400,000 parts
is:
Anode machine time = (1,071 setups 0.5 hours/setup) + (566,000 parts / 1886.8
parts/hour) = 835.5 hours
Given an availability of 6000 hours per year per machine, the number of coating systems
required is:
Process Flow
Gas
Diffusion
Layer
Gas
Diffusion
Layer
Background
In “Mass Production Cost Estimation for Direct H2 PEM Fuel Cell Systems for Automotive
Applications: 2010 Update,” Directed Technologies, Inc. (DTI) reported hot pressing conditions for MEA
fabrication as 160°C for 90 seconds using heated platens of 0.5 m wide by 1.5 m long for processing 0.5
m wide roll materials. They estimated a reset period of 3 seconds to open the press, index the materials,
and re-close the press.
In “Investigation of membrane electrode assembly (MEA) hot-pressing parameters for proton exchange
membrane fuel cell,” (Energy, 32:12, December 2007, Pages 2401–2411), Therdthianwong, et.al. found
that the most suitable hot pressing conditions for MEA fabrication to be 100°C and 1000 psi (70 kg/cm2)
for 2 minutes, stating that these conditions “…provided the highest maximum power density from the
MEA and the best contact at the interfaces between the gas diffusion layer, the active layer, and the
electrolyte membrane.”
Preliminary Analysis
The cells for this analysis will have a total size of:
Setup
GDL material is typically sold in either 0.4 or 0.8 meters widths, and is available up to a
maximum of 800 meter lengths. However, the total number of setup operations will be dictated
by the length of the shortest roll being processed, which is a membrane length of 100 meters. In
Appendix A7 on the platinum coating process, the number of roll setups was shown to be
Tooling
Tooling consists of the heated platens, which generally consist of 2”-2.5” thick aluminum plates loaded
with electric cartridge heaters spaced 3” (7.6 cm) apart. In 2010, DTI obtained a quote from Custom
Engineering Co. (platens.com) for heated platens used for compression molding of bipolar plates. The
quote estimated the cost at approximately $13,500/m2 of platen area and included platen, base plate, and
heater control electronics, estimated to be approximately $15,650 in 2015 dollars. Standard platen widths
are in 0.5 meter increments based on standard cartridge heater sizes. For the size and orientation of the
parts, the platen width will be 1 meter. Due to the indexing and alignment required for the patch coated
MEAs, the die length should be at least 1 meter long, and as close to a multiple of 378.4 mm as possible
while allowing for proper cartridge heater spacing. Six cells arranged 2 widthwise by 3 lengthwise take
up 114 cm of length, which fits 15 heaters.
An engineering estimate for tool life based on heater life would be around 100,000 cycles. Using
$15,600/m2 as a basis, and amortizing over a 5 year production life, the total annual tooling cost is:
ଵ
Annual tooling cost = (Tooling cost Number of tools purchased)
ହ
where: Number of tools purchased = Roundup (Total production / Tool life)
Total production = Annual production 5
ଵ
Annual tooling cost = (($15,600/m2 1.14 m2) Roundup(((566,000 parts/year / 6 parts/cycle)
ହ
5 years) / 100,000 cycles/tool)) = $17,191
Material
GDL material is typically sold in either 0.4 or 0.8 meters widths. The cost of the membrane is accounted
for in a previous process step, and is not included as part of the hot pressing operation. Assuming 2 GDL
layers per MEA, the GDL material usage is calculated as:
Material usage = 0.8 m 107,087 m 2 = 171.339 m2
Material cost is computed in accordance with Appendix A2 as:
Material cost = $39.14/m3
Hot Press
The hot press occurs in two steps: the material is moved into the press (handling time) and the press
operation (clamp time). The material handling time is computed using an empirical formula developed by
Boothroyd Dewhurst for automated handling with a 2.8 second minimum as follows:
Handling time = Max((0.012 (Platen length (cm) + Platen width (cm)) + 1.6), 2.8)
Handling time = Max((0.012 (114 + 100) + 1.6), 2.8) = 4.2 sec
Omega (http://www.omega.com/prodinfo/cartridgeheaters.html) estimates 0.5” cartridge heaters to have a
watt density of 50W per inch of heater length (about 20W per cm length). Calculating the total input
heater power for the platen:
Platen power input = Number of heaters (Platen width (cm) 20 (W/cm))
Platen power input = 15 heaters (100 cm 20 W/cm) = 30 kW
The heated platens need to maintain a temperature during pressing of about 100°C. A study conducted by
the food service industry indicates that 3 foot electric griddles with rated energy inputs of 8-16 kW
demonstrate a 25% duty cycle in actual use.
Platen sizing allows for processing 6 parts per press cycle (2 parts wide 3 parts long). Throughput can
be computed as:
Parts/hour = 6 parts/cycle / ((124 + 4.2) / 3600) hours/cycle = 168.5 parts/hour
The total machine time for processing and setup is:
Machine processing time = (566,00 parts / 168.5 parts/hour) + (1,071 setups 0.5 hrs/setup) =
3894.6 hours
Total machine labor time for processing and setup:
Machine labor time = 1 operator/machine 3894.6 hours = 3894.6 hours
Die Cutting
Following hot pressing, the MEA is die cut into the final shape as shown:
MEA:
20mm.
100mm. 30mm.
Ø 10.0mm.
typ
Anode Cooling
Gas Fluid
5mm.
318mm.
20mm.
Cathode Gas
Cathode Gas
245mm.
305mm.
Active Area
5mm.
Cooling Anode
Fluid Gas
377mm.
Tooling
The primary factor contributing to steel rule die cost is the total cutting length of the die. Assuming a
platen size equal to that of the hot pressing operation, the total number of cavities is 6 (2 widthwise by 3
lengthwise). The outer cell perimeters will require a total length of :
Outer perimeter length = 2 (2 378.4) + 2 (3 305) = 3343.6 mm
The inner perimeters are shared and will require a total length of:
Inner perimeter length = 2 (2 305) + 1 (3 378.4) = 2355.1 mm
Internal features are unique to each cell cavity and include the fluid and gas openings, and the tie rod
holes, which require a total die length of:
Feature length = 4 (2 (100 + 20)) + 2 (2 (245 + 20)) + 6 (π 10) = 2208.5 mm
Therefore the total die cutting length is:
ଵ
Annual tooling cost = (Tooling cost Number of tools purchased)
ହ
where: Number of tools purchased = Roundup(Total production / Tool life)
Total production = Annual production 5
Setup
The total number of setup operations will be dictated by the length of the shortest roll being
processed, which is the membrane at 100 meters. As shown above, the number of roll setups is
1,071. Assuming 0.5 hours per setup, the total setup time is:
Die Cutting
The primary energy input to run the press is hydraulic pump motor power. The total force required to cut
the material is the total shear area (cutting length material thickness) multiplied by the material shear
strength. Shear strength data for Nafion is not readily available, but polymer based materials typically
range from 8000 – 11,000 psi (55-76 N/mm2). Assuming the worst case shear strength, and using the
material thickness of 0.7 mm, the total required press force per part is calculated as:
Press force = Die cutting length (mm) Material thickness (mm) Shear strength (N/mm2)
Press force = 19,020 mm/die 0.7 mm 76 N/mm2 = 1,012 kN
A survey of 15 to 100 ton (150 – 1000 kN) fast-acting die cutting presses found that the motor power
required to operate the press fell in the range of 0.015 – 0.025 kW/kN. Assuming a 50% capacity margin
and using the upper end of the motor power rating, the maximum required press energy input is:
Press energy = 1,012 kN 1.5 0.025 kW/kN = 37.95 kW
Typical die cutting press speed ranges from 30 – 60 cycles/min (1800 – 3600 cycles/hour). Assuming the
slower speed, the time to process a batch of parts is calculated as
Processing time = 566,000 parts / 6 parts/cycle / 1800 cycles/hour = 52.41 hours
The total machine time for processing and setup is:
Machine processing time = 52.41 + 535.5 = 587.9 hours
Given an availability of 6000 hours per year per machine, the number of presses required is:
Roundup(587.9 / 6,000) = 1 machines
Machine utilization is:
587.9 / 6,000 = 9.80%
Total machine labor time for processing and setup:
Machine labor time = 1 operator/machine 587.9 hours = 587.9 hours
Model Approach
Process Flow
Background
The BDI software provides pre-programmed cost models for the injection molding process used to
manufacture the fuel cell stack coolant gaskets. The process selection was liquid silicon injection
molding.
Preliminary Analysis
The stack contains 3 gaskets (cathode, anode and cooling) per cell plus 2 cathode gaskets on each end of
the stack. To manufacture 1,000 100 kWnet systems consisting of two 60 kW stacks of 283 cells each
requires a total of 566,000 each anode and cooling gaskets, and 570,000 cathode gaskets. The gasket
features and dimensions are shown below:
TITLE
20mm. 30mm.
Cathode Gasket
100mm. 30mm.
Ø 10.0mm.
typ
Anode Cooling
Gas Fluid
365mm.
5mm.
245mm.
305mm.
Cathode Gas
Cooling Anode
Fluid Gas
377mm.
TITLE
Anode Gasket 5mm. 25mm.
100mm. 30mm.
Ø 10.0mm.
typ
Cooling
Fluid
5mm.
5mm.
317mm.
Cathode Gas
Cathode Gas
245mm.
318mm.
Anode Gas
Cooling
Fluid
377mm.
TITLE
Cooling Gasket
30mm. 100mm.
20mm. 25mm. 5mm.
Ø 10.0mm.
typ
Anode
Gas
5mm.
5mm.
317mm.
Cathode Gas
Cathode Gas
305mm.
245mm.
Cooling Fluid
Anode 5mm.
Gas
377mm.
Cathode Gasket
The BDI estimate for the cathode gasket is a 1.5 hour machine setup time, and calculates the total
manufacturing cycle time as 9.16 sec for a 4 cavity mold, making the total machine time for annual
production:
Machine time = (9.16 sec/cycle / 4 parts/cycle / 3600) 570,000 parts + 1.5 = 364.1 hours
Machine utilization is:
Utilization = 364.1 / 6000 = 10.1%
Assuming 1 full-time operator per 2 molding machines, the total machine labor time is equal to half the
machine time = 182.1 hours.
The BDI estimate for material weight per part is 0.024 kg, making total annual material usage:
Material usage = 0.020 kg/part 570,000 parts = 11,400 kg
Material cost was determined in accordance with Appendix A2 as:
Material cost = $15.33/kg
Tooling cost is $93,864 and is assumed to be capable of producing 1,000,000 parts. Amortizing over a 5
year production life, the total annual tooling cost is:
ଵ
Annual tooling cost = (Tooling cost Number of tools purchased)
ହ
where: Number of tools purchased = Roundup(Total production / Tool life)
Total production = Annual production 5
ଵ
Annual tooling cost = ($93,864 Roundup((570,000 parts/year 5 years) /
ହ
1,000,000 parts/tool)) = $56,318.40
Anode/Cooling Gasket
Note that the Anode and Cooling gaskets are the same design, but are installed by flipping along the
vertical center axis, and are therefore analyzed by the DFM software as the same gasket, as shown below:
The BDI estimate for the anode/cooling gasket is a 2.5 hour machine setup time, and calculates the total
manufacturing cycle time as 10.28sec for a 4 cavity mold, making the total machine time for annual
production:
Machine time = (10.28 sec/cycle / 4 parts/cycle / 3600) 1,132,000 parts + 2.5 = 810.6 hours
The BDI estimate for material weight per part is 0.054 kg, making total annual material usage:
Material usage = 0.054 kg/part 1,132,000 parts = 61,128 kg
Material cost was determined in accordance with Appendix A2 as:
Material cost = $15.33/kg
Tooling cost is $135,508 and is assumed to be capable of producing 1,000,000 parts. Amortizing over a
5 year production life, the total annual tooling cost is:
ଵ
Annual tooling cost = (Tooling cost Number of tools purchased)
ହ
where: Number of tools purchased = Roundup(Total production / Tool life)
Total production = Annual production 5
ଵ
Annual tooling cost = (($135,508) Roundup((1,132,000 parts/year 5 years) /
ହ
1,000,000 parts/tool)) = $162,609.60
Model Approach
Setup operation
o Machine setup labor time based on user input
o Tooling cost based on input insert and platen cost and life
Pre-form operation
o Measure and pre-form labor time based on user input labor time
o Part material unit cost based on usage
Compression mold
o Part handling time based on part size per BDI formula; 4 second minimum
o Press processing time based on part size and cycle time
o Compute machine utilization
Post bake
o Part handling time based on part size per BDI formula and throughput; 4 second
minimum
Process Flow
BMC940
Graphite
Composite
Compression Mold
Pre-form Post Bake Bipolar
(800 tons for 180
(12 sec) (15 min @ 175°C) Plate
sec @ 160°C)
Background
A supplier of composite bipolar plates for PEM fuel cell stacks provided the following
information regarding their process:
Process requires a special press
o High speed – 30 inches per second (ips)
o High tonnage – 800 ton capacity to produce 1 part per cycle
o Cure time in the press is 120-230 seconds
Preliminary Analysis
Unlike injection molding, compression molding requires that a pre-measured, usually pre-
formed, and generally preheated amount of material be loaded into the mold insert prior to
pressing. Given the stated consistency of the material, we will assume a manual weighing
process followed by a manual packing process to get the material into the rough rectangular
shape of the plate. No material pre-heating was mentioned by the manufacturer or the material
spec sheet.
The bipolar plates for this analysis will be:
305 mm width 378.4 mm length = 1,154 cm2
Process values will be calculated based on annual production of 2,000 stacks per year to supply
1,000 systems producing 100 kWnet. The 60 kW stack requires 284 anode bipolar plates and 284
cathode bipolar plates, requiring annual production of 568,000 of each type of plate.
Setup
We will assume one full setup per run of parts. This would include such things as platen and die
installation, die alignment, work station setup and maintenance and operational checks. An
analogous setup operation in the Boothroyd Dewhurst DFMA® software is for a powder
metallurgy compaction press, for which the default value is 4 hours.
Material cost
Flow channels cut into the plates are generally 1 mm deep. The cathode bipolar plate has flow
channels cut into one side of the plate, indicating a plate thickness of around 2 mm. The anode
bipolar plate has flow channels cut into both sides of the plate to accommodate anode gas flow
on one side, and cooling fluid flow on the other, indicating a plate depth of around 3 mm. Given
a material density of 1.9 g/cm3 (BMC940 spec sheet) and 5% overage allowance, the total annual
material required before scrap is:
Cathode plate material required = 1.9 g/ cm3 0.001 kg/g (1154 .2) cm3 1.05
568,000 parts = 261,533 kg
Anode plate material required = 1.9 g/ cm3 0.001 kg/g (1154 .3) cm3 1.05
568,000 parts = 392,300 kg
Based on quotes from BMC, the material cost can be estimated in accordance with Appendix A2
as:
Material cost = $2.066/kg
The material specification recommends molding pressure in excess of 40 MPa (0.4 tons/cm2) on
the projected part area::
Tonnage = 0.4 tons/cm2 1,154 cm2 = 461.6 tons
Discussions with a bipolar plate manufacturer indicate the use of a special fast-acting 800 ton
press. Moving the capacity up to 1,000 tons, and allowing for some capacity margin, it is
feasible to mold two plates per cycle (923.2 tons).
The primary energy input to run the press is hydraulic motor power. Surveying press
manufacturers Wabash, Beckwood, and Karunanand, the hydraulic motor size for 800 ton
presses appears as either 30 or 50 HP, but lists pressing speeds of only 20 ipm (0.3 ips).
Cylinder bore sizes are listed as 26”-30” diameter. To move a 30” diameter cylinder at 30 ips
requires a pump delivery of:
Flow rate = (30”)2 (ᴨ / 4) 30”/sec 60 sec/min 0.004 gal/in3 = 5,089 gpm
This is beyond the practical limit of most high performance hydraulic gear pumps, which tend to
have maximum flow rates of 90 gpm at 100 HP input power and 2500 psi working output
pressure (reference Commercial Intertech P365 series hydraulic pumps).
To supply 1,000 tons of force using a 30” cylinder requires a delivery pressure of:
Pressure = 1,000 tons 2,240 lbs/long ton / ((30”)2 (ᴨ / 4)) = 3,169 psi
For this analysis, we will assume two 100 HP (75 kW) pumps feeding a set of staged cylinders;
e.g. two smaller diameter cylinders to provide the necessary pressing speed, and one larger
cylinder to develop the required pressure. To provide some limited scalability, we assume that
150 kW of input power is required to mold two 1,154 cm2 bipolar plates, giving a factor of
approximately 0.130 kW/cm2 of plate area.
Total press cycle time is the sum of part handling time, press actuation time, and press dwell
time. An empirical formula developed by Boothroyd Dewhurst calculates a quantity called part
girth, then calculates a theoretical total handling time (both load and unload) with a minimum
value of 4 seconds, as follows:
Part girth = Part length + Part width + part depth
Handling time = Max((0.60714 (Part girth / 25.4) - 4.57143), 4)
Cathode plate handling time = Max((0.60714 ((305 + 378.4 + 3) / 25.4) - 4.57143), 4) =
11.8 sec
Anode plate handling time = Max((0.60714 ((305 + 378.4 + 6) / 25.4) - 4.57143), 4) =
11.9 sec
For an actuation time of 10 sec, dwell time of 230 sec, and handling times shown above, the total
cycle time is:
Cathode plate cycle time = ((2 11.8) + 230 + 10) = 263.6 sec/cycle = 0.0732
hours/cycle
Anode plate cycle time = ((2 11.9) + 230 + 10) = 263.8 sec/cycle = 0.0732 hours/cycle
Throughput is calculated as:
Parts per hour = 2 parts/cycle / 0.0732 hours/cycle = 27.32 parts/hour
Since throughput for each type of plate is essentially the same, we can calculate the total time
required to process both sets of plates (1,136,000 parts) as:
Press machine time = 1,136,000 parts / 27.32 parts/hour + (2 4) hour setup = 41,589
hours
Given an availability of 6000 hours per year per machine, the number of presses required is:
Roundup(41,589 / 6,000) = 7 machines
Machine utilization across the 7 machines is:
41,589 / (7 6,000) = 99.0%
Tooling cost
Tooling consists of the mold inserts and the heated platens. Contact with Custom Engineering
Co. (platens.com) indicates that platens in the size range required will generally consist of 2”-
2.5” thick aluminum plates loaded with electric cartridge heaters spaced 3” (7.6 cm) apart. Costs
will be in the range of $10,000 for a 7,500 cm2 platen ($1.333/ cm2), and $3500 for the
controller. No life was provided for the platens. An engineering estimate based on heater life
would be around 500,000 cycles.
Assuming 2 plates per cycle with 50 mm margin between and around each plate, the total platen
area is:
ଵ
Annual insert tooling cost = (($150,000) Roundup((1,136,000 parts/year / 2
ହ
parts/cycle 5 years) / 100,000 parts/tool)) = $870,000
ଵ
T Annual insert tooling cost = (($12,668) Roundup((1,136,000 parts/year / 2
ହ
parts/cycle 5 years) / 500,000 parts/tool)) = $15,202
The BMC940 material spec sheet calls for a post bake at 350°F (177°C) for 15 minutes after the
part reaches temperature. For a batch type oven, the strategy is to rack parts in quantities that
Model Approach
Test and condition fuel cell stack
Process Flow
Background
Following assembly, the PEM stack is tested and conditioned to determine its fitness for
installation into the system. The total test time is assumed to be 2.5 hours. Total H2
consumption at full power is determined from the equation:
For a 60 kW stack current of 312 A and cell count of 283 cells we have:
Converting to L/min:
Air is supplied in a stoichiometric ratio of 1.2:2, resulting in required air flow of:
Preliminary Analysis
Assuming setup and teardown of the stack test stand requires 1/2 hour for 1 operator per run, the setup
time per production run of 2,000 stacks is:
The Fuel Cell and Hydrogen Energy Association placed the 2010 nation-wide average cost of
hydrogen in bulk liquid form at about $7.83/kg for usage levels of 700-1,400 kg/month. Internet
quotes indicate a price of about $5.93/kg for bulk purchases of 30,000 kg or more. The mass of
1 mole hydrogen gas (H2) = 2 grams, so the mass of 22.4 liters (stp) of H2 is 2 g.
At 100% rated power, the total material usage of the hydrogen is:
Full power material usage = ((738.8 l/min / 1000 l/m3) / 11.2 m3/kg) 60 min/hr =
3.958 kg/hr
During the 2.5 hour test, we assume a conditioning and test regimen as follows:
H2 usage = 3.958 kg/hr ((0.25 1.0 hr) + (1.0 0.5 hr) + (0.25 1.0 hr)) 2,000 stacks =
7,916 kg
The material cost before scrap can be estimated in accordance with Appendix A2 as:
We will assume that 1 test station (150 kW load bank) is capable of supporting 2 stacks during
testing, making the total machine time for setup and test:
Testing machine time = ((2.5 hrs/stack / 2) + (0.5 hrs/stack)) 2,000 stacks = 3,500 hrs
We will assume that 1 operator can cover 3 testing stations, making the total labor time:
The testing process is subject to a failure rate estimated at around 5%. Stacks failing test are
reworked by disassembling the stack, replacing the defective part (assumed to be an MEA), and
reassembling the stack. Using the Boothroyd-Dewhurst Design for Assembly software, the 60
kW stack assembly labor time was estimated to be 3.1 hours.
The formula for scrap value is based on the total amount of additional production necessary to
make up for the value of the scrapped items as:
Assuming a scrap rate of 5%, the total loss associated with disassembly and reassembly labor is:
Scrap labor time = (((2 3.1 hrs/stack) / (1-0.05)) - (2 3.1 hrs/stack)) 2,000 stacks =
652.6 hours
Assuming that the part requiring replacement is a MEA, which is the worst case, the total loss
associated with replacement parts is:
Model Approach
Ceramic slurry preparation operation
o Machine setup labor time based on user input
o Compute required batch size based on part batch size and ceramic layer thickness
o Compute ceramic slurry material unit cost based on usage
o Compute ceramic slurry processing time and machine utilization
Process Flow
Background
The composition of typical SOFC ceramic slurries used in industry is not directly reported, and
fundamental work seems to be continuing in the area of ceramic powder characterization.
In the book Modern Ceramics Engineering (Richardson, 2006) lists the components of a typical
solvent based slurry as:
70 wt% ceramic powder
14 wt% organic solvent (MEK/EtOH)
9 wt% binder (ethyl methacrylate)
1 wt% dispersant (fish oil)
6 wt% plasticizer (BBP/PEG)
In their study of sintering and deformation, Cologna (2010), et al. report using a water-based
slurry in tape casting experiments as follows:
Anode: blade gap = 500 µm; dry thickness = 270 ± 5 µm; 46% reduction
26 wt% YSZ (8% mol)
37 wt% NiO
12 wt% water
24 wt% binder
1 wt% dispersant
Cologna’s values are consistent with general “rule-of-thumb” thickness reduction of 50% seen
on several web sites and used on some technical papers. Therefore, for cost purposes, we will
assume that wet ceramic deposition will be twice the thickness of the required final ceramic layer
thickness.
Preliminary Analysis
Anode Batch Volume
Slurry batch volume depends on the part size, casting width, and ceramic layer thickness.
The cells for this analysis are for a 100 kW system at a production rate of 4,000 stacks per year (1,000
systems with 4 stacks per system). The deposition area for the anode will be:
Based on the slurry composition as specified above, 100 grams of wet slurry has a volume of:
The required dried depth of 500 microns requires a deposited wet depth of 1,000 microns. (Note that final
anode depth will be achieved by casting two 250 micron tapes and pressing the tapes together to achieve
the final desired thickness) The weight of slurry material required per part is:
Wet slurry weight = 2.2 g/cm3 (501.16 0.10) cm3 0.001 kg/g = 0.220 kg/part
Batch sizes will be calculated based on a production schedule producing 4,000 stacks per year. The 30
kW stack requires 259 cells, requiring total slurry production of:
Annual Slurry Production: 259 parts/stack 4,000 stacks 0.220 kg/part = 228,201 kg
Material cost of the slurry is calculated using the weight percents of the slurry constituents
multiplied by the raw material cost to determine a cost per kilogram. Ceramic material pricing
was obtained from Inframat Advanced Materials in December, 2013. Bulk cost for the dispersant
was obtained from web quotes at around $2.50 for 2500 kg. The cost of DI water is based on
amortized distillation costs obtained from www.apswater.com. Summarizing:
YSZ = $36.00
NiO = $28.80
Water = $0.084
Binder = $2.125
Dispersant = $1.268
Raw material cost = (0.26 $36.00) + (0.37 $28.80) + (0.12 $0.084) + (0.24 $2.125) +
(0.01 $1.268)
Raw material cost = $20.55/kg
The first step is to weigh the materials out and place them in the mill. We will assume a manual process
consisting of a measurement step and a material handling step. The DFMA® software contains an
analogous operation for off-line precision measurement with a default value of 17.4 seconds for the
measurement, and a minimum of 4 seconds for material handling. The slurry is made up of 5 materials,
so that total handling time for material preparation can be estimated as:
Material prep time = 5 21.4 sec = 107 sec = 1.8 minutes = 0.03 hours
The primary cost for operating the ball mill is the energy input to the motor running the mill. Some
studies have looked into the cost of operating large ball mills used for cement and powder metallurgy
material processing, where the target parameter is the amount of energy required to process a given
amount of material, usually expressed in kW-hr/ton. The calculations are complex owing to the large
number of inputs into the calculations.
We note that production levels of 1,000 systems (4,000 stacks) per year will require a total of 264,900 kg
of catalyst production per year across all layers, or a little over 1 MT per working day. A volume of
50,000 (200,000 stacks) systems per year will require about 13,245 MT per year of slurry, or about 4 MT
per day.
A web search identified a batch ball mill capable of 5 MT loading weight measuring 2.8 m diameter and
2.6 m long and consuming 37 kW of power. For 4,000 30 kW stacks, the total number of batches
processed will be:
For other SOFC layers, the slurry volumes required will be smaller due to thinner deposition layers,
ranging from 5 to 30 microns. The total depth of all subsequent layers is around 80 microns across 6
additional layers. Assuming similar density and composition profiles, we can expect slurry production
for the remaining layers to be around 370 MT per year, or about 1.5 MT per day.
A web search identified a batch ball mill capable of 1.5 MT loading weight measuring 2.0 m diameter and
1.8 m long and consuming 15 kW of power.
Once the milling process is complete, the slurry will need to be separated from the milling balls and
transferred to the coating machine. While we presently have no information about this part of the
process, one approach would be the use of a vacuum sieve (e.g., Farleygreene, Ltd. SM950 Sievmaster
Vacu-siev) to remove and separate the slurry from the mill, and transfer the slurry to a transport container
or directly to the coater reservoir.
ShopVac reports a sealed suction of 54 in-H2O (13.4 kPa) for their 2 HP (1.5 kW) unit. Using an
equivalent vacuum sieve with a 1.5” (0.038 m) diameter hose and 80% transfer efficiency, the flow rate
is:
Since the slurry forms 90% of the charge volume, the total charge volume is:
Charge volume (m3) = 1.11 (Slurry weight (kg) / Slurry density (kg/ m3))
Charge volume (m3) = 0.0013 Slurry weight
Therefore, the optimal time required to remove the charge volume is:
Material removal time (sec) = Charge volume / Flow rate = 8.1 Slurry weight
We will estimate the total transfer time to remove the slurry from the mill and transfer it to the coater as
twice the slurry removal time. The total time required to remove the slurry from the mill would be:
(Setup time + Material prep time + Milling time + Material removal time) Number of bathes =
(0.5 + 0.03 + 10.0 + 22.5) 46 = 1,519 hours
For computing machine labor time, we assume 1 dedicated operator for the setup, material prep, and
material removal operations, but only minimal labor time required during milling. The total machine
labor time is then:
((0.5 + 0.03 + 22.5) 1 operator + (10 0.1 operator)) 46 batches = 1105.4 hours
Model Approach
Process Flow
Background
We will assume that the pre-fired anode tape has similar physical properties to those of elastomeric
materials. The primary method for blanking elastomeric materials with standard features and tolerances
is steel rule die cutting. The outline of the gasket is laid out and cut into a board. Strip steel is embedded
into the board at a uniform height and mounted on a small stroke, fast acting press. The anode material is
fed into the press where the steel rule die shears the material. The cutout areas of the blank are pushed
out of the bulk material and the blanks stacked.
Preliminary Analysis
The cells for this analysis are for a 100 kWnet system at a production rate of 4,000 stacks per year
(1,000 systems with 4 stacks per system). The deposition area for the anode will be:
187 mm width 268 mm length = 501.16 cm2
The 30 kW stack requires 259 cells, requiring total annual production of:
Annual production = 259 parts/stack 4,000 stacks = 1,036,000 parts
Setup
There are 194 setups for both the anode blanking and the anode tape pressing processes.
Tooling Cost
The primary factor contributing to steel rule die cost is the total cutting length of the die. For the anode,
the cutting length (mm) is:
Cutting length = 2 (268 + 187) = 910 mm
The total number of cavities is 4 (1 widthwise by 4 lengthwise). A steel rule die will have (2 (1 + 5)) =
10 outer edges, leaving ((5 4) -16) = 4 common sides, yielding 4 inner cutting edges of 268 mm.
Therefore the die cutting length is:
Die cutting length (mm) = ((4 – 1) 268) + ((4 2) 187) + (2 268) = 2,836 mm
A rough quote of approximately $230 was obtained from steel-rule-dies.com for a two cavity die with a
similar configuration.
ଵ
Annual tooling cost = ((2,836 mm/die $0.04/mm) Roundup((1,036,000 parts/year /
ହ
4 parts/cycle 5 years) / 20,000 parts/tool)) = $1,474.72
Die Cutting
The primary energy input to run the press is hydraulic pump motor power. The total force
required to cut the material is the total shear area (cutting length material thickness) multiplied
by the material shear strength. Assuming that the unfired anode material has the approximate
consistency of high-density polyethylene (HDPE), we will use 23 N/mm2 as the shear strength,
giving the total required press force as:
Press force = Number of cavities Cutting length (mm) Material thickness (mm)
Shear strength (N/mm2)
Press force = 4 dies 2836 mm/die 0.25 mm 23 N/mm2 = 65.23 kN = 6.55 tons
A survey of 15 to 100 ton (150 – 1000 kN) fast-acting die cutting presses found that the motor
power required to operate the press fell in the range of 0.015 – 0.025 kW/kN. Assuming a 50%
capacity margin and using the upper end of the motor power rating, the required press energy
input is:
Press energy = 65.23 kN 1.5 0.025 kW/kN = 2.45 kW
Typical die cutting press speed ranges from 30 – 60 cycles/min (1800 – 3600 cycles/hour).
Assuming the slower speed, part throughput is calculated as
Throughput = 4 parts per cycle 1800 cycles per hour = 7200 parts/hour
The total machine time required to produce 4,000 30 kW stacks is:
Machine time = Setup time + Machine time =
(0.5 hours/setup 194 setups) + (1,036,000 parts / 7200 parts/hour) = 241 hours
Machine utilization is:
Machine utilization = 194 / 6,000 = 3.2%
Assuming 1 operator per casting machine for both setup and operation, the machine labor time is
equal to the total machine time of 144 hours.
Model Approach
Screen Preparation
o Compute tooling cost
o Compute labor time for screen cleaning
o Compute labor time and material cost for emulsion coating based on required
ceramic layer thickness
o Compute energy, machine time and labor time for masking and emulsion
exposure
o Compute energy, machine time and labor time for emulsion rinse and post-cure
Screen Printing
o Compute time for machine setup
o Compute labor time for substrate load/unload
o Compute machine time for screen printing operation
Oven Drying
o Compute required heater area based on drying time and required conveyor speed
o Compute heater energy on energy watt density and heater area
Process Flow
Ceramic
Slurry
Coated
Screen Print Oven Dry
Cell
Background
The mechanics of the screen preparation and printing process are described in several on-line
sources, as well as a series of instructional videos produced by Cat Spit Productions found on
YouTube. The calculations used for the screen preparation process were based on material and
process specifications for Ulano QT-THIX emulsion and the article “Screen Coating
Techniques” available from emulsion manufacturer Kiwo at http://www.kiwo.com/Articles.
Technical details of the printing process were based on the article “Screen and Stencil Printing”
Preliminary Analysis
The cells for this analysis are for a 100 kWnet system comprised of four 30 kW stacks. The
deposition area for the anode will be:
Screen size is determined based on pattern area. DuPont recommends a squeegee length of 10-
20 mm beyond the pattern area (part width) on both sides, and squeegee travel of 50-80 mm
beyond the pattern area (part length) on both ends. Bopp, a printing mesh manufacturer,
recommends a screen width of 3 times the squeegee width and screen length of 2 times the
squeegee travel. The minimum screen size can be calculated as:
AMI indicates that polymer squeegees may be changed daily in high volume production
applications, indicating a useful life of around 5000 - 6000 parts. The squeegee tooling cost is:
Web quotes for fine mesh precision metal screens in 24” x 30” size ranged from $50 to $100,
equating to about $0.02/cm2, giving estimated screen costs of:
ଵ
Annual tooling cost = (($91.41) Roundup((1,036,000 parts/year 5 years) / 5,000
ହ
parts/screen)) = $18,940.15
Screen Preparation
Screen preparation is a manual process that consists of cleaning, emulsion coating, emulsion
masking and exposure to high intensity light, emulsion rinsing and post cure using high intensity
light. The primary cost component will be the labor involved in handling and coating the screen.
An empirical formula developed by Boothroyd-Dewhurst calculates a quantity called part girth,
then calculates a theoretical total handling time (both load and unload) with a minimum value of
4 seconds. Adapting the formula for dimensions in millimeters and handling of large, light-
weight parts, the handling time is calculated as follows:
applied to a 1 meter length in approximately 5 seconds. Using 3 seconds for tool acquisition, the
time to apply a single coat can be estimated as:
100 kW
Labor time
(sec)
Cleaning 58.91
Coating 55.22
Exposure 10.95
Rinsing 29.46
Post-cure 10.95
Total 165.49
Screen Printing
The screen printing operation consists of a part load/unload, which may be manual or robotic,
but will be driven by overall part size. Using the handling time formula developed previously,
the load/unload time is:
Following deposition, the ceramic slurry is dried, usually by means of a tunnel dryer positioned
directly after the deposition step. The drying can be done by either radiant or convective heating.
For the cost analysis, we will assume radiant (infrared) heating is used and compute the cost of
drying by determining the required heater area based on throughput and the drying time.
Drying time is a function of the evaporation rate of the solvent and is inversely and exponentially
proportional to the coating thickness. Experiments conducted by Mistler (Tape Casting of
Ceramics, Ceramic Processing Before Firing, 1978) indicate drying rates of 1.3510-5 g/cm2-sec
at room temperature for an air flow rate of 2 l/min, and 2.2210-5 g/cm2-sec at room temperature
for an air flow rate of 75 l/min.
Previous analysis assumed that the screen printed slurry material was formulated with aqueous
components as follows:
12 wt% water
24 wt% binder (Dow Duramax B-1000/B-1014)
1 wt% dispersant
The binder consists of approximately 45% solids. Roughly estimating the volume of liquid per
gram of slurry by multiplying the material density by the material weight percent:
Liquid density = (0.12 1.0) + ((0.24 0.55) 1.05) + (0.01 1.16) = 0.270 g/cm3
The weight of liquid to be removed per unit area is a function of slurry thickness. As with tape
casting, we assume a 50% thickness reduction after drying. Using the anode active layer (15
micron green thickness; 30 micron wet thickness) as an example:
At a rate of 2.010-5 g/cm2-sec drying rate, the estimated drying time is:
Drying time = 0.0008 g/cm2 / 2.010-5 g/cm2-sec = 40.5 sec = 0.675 min
The conveyor speed is a function of part throughput and belt length required to transport the part.
Using the results above, the throughput is:
Assuming a 50 mm gap between parts on the belt, the conveyor speed can be calculated as:
Conveyor speed = (187 + 50) * (160.7 / 60) = 634.77 mm/min = 0.635 m/min
Infrared heating panels are generally sold with various energy watt densities and in standard
sized units and assembled to provide the necessary heating area. Using the Casso Solar Type FB
as an example, standard watt densities are 15 and 25 W/in2 (23 and 39 kW/m2) with standard
width of 12” (0.305 m) and lengths in 12” increments up to 60” (1.524 m). They note that 25
W/in2 corresponds to an emitter temperature of 880°C, and that the conversion efficiency of
electrical power to usable radiant energy is up to 80%.
For a drying time of 0.675 minutes, the required heater area is:
Heater area = Drying time (min) Conveyor speed (m/min) (Belt length per part (mm)
/ 1000)
Heater area = 0.675 0.635 (237 / 1000) = 0.10156 m2
While the heater energy density will be taken as an input, the drying temperatures for the
ceramic slurry are fairly moderate (150°C or less), such that 23 kW/m2 should be sufficient to
maintain the drying area temperature. Using an energy cost of $0.07/kW-hr, the hourly energy
cost to power the heaters will be:
Model Approach
Kiln Firing
o Compute part handling time and labor time
o Compute machine time
Process Flow
Background
Preliminary Analysis
The cells for this analysis are for a 100 kWnet system at a production rate of 4,000 stacks per year
(1,000 systems with 4 stacks per system). The deposition area for the anode will be:
187 mm width 268 mm length = 501.16 cm2
The 30 kW stack requires 259 cells, requiring total annual production of:
Annual production = 259 parts/stack 4,000 stacks = 1,036,000 parts
Kiln Firing
The SOFC process calls for kiln firing at 1000°C (1832°F) for 1 hour after the part reaches
temperature. The moderate temperatures required allow for the use of a mesh belt furnace to
accomplish the kiln firing process.
Large mesh belt furnaces manufactured by AFC-Holcroft feature a 66” (167 cm) wide belt, 6”
(15.24 cm) workspace clearance and effective load length of 456” (1158 cm) for a total load
volume of 2.95 106 cm3. The maximum part thickness following the last screen printing
operation is 328 microns (0.0328 cm). Adding 1 cm on all sides for part spacing in the furnace
and assuming optimal racking, the total part envelope is:
Part Envelope = (18.7 + (2 1.0)) (26.8 + (2 1.0)) (0.0328 + (2 1.0)) = 1212 cm3
The maximum furnace loading is then:
Furnace loading = 2.95 106 cm3 / 1212 cm3 = 2435 parts
The firing schedule is based on the firing schedule suggested in PNNL-22732i of:
Segment Rate/Time Temp.
Ramp 3C/min 1000C
Hold 1 hour 1000C
Ramp 5C/min Ambient
Using 25C as the ambient temperature, the required preheat time is:
(1000 – 25)C / (3/min) = 325 min = 5.42 hours
The required cooling time is:
(1000 – 25)C / (5/min) = 195 min = 3.25 hours
Therefore, total furnace time is 9.67 hours, making furnace throughput:
Throughput = 2435 parts / 9.67 hours = 251 parts/hour
Total machine time to fire the cells is:
Machine time = 1,036,000 parts / 251 parts/hour = 4127.5 hours
Machine utilization is:
Utilization = 4127.5 / 6000 = 68.8%
Part load/unload, which may be manual or robotic, will be driven by overall part size. To
determine if a single operator can load and unload parts while the machine is operating, we use
the handling time formula adapted from the BDI DFMA® software to determine the total
load/unload time required per hour or machine time as:
Part handling time = Max((0.3 ((187 + 268 + 1) / 25.4) - 4.6), 4) 251 parts/hour =
0.243 hours per hour of machine time.
Given that part handling time is less than 25% of total firing time, we assume one operator is
capable of covering 3 machines, making the labor time:
Machine labor time = 4127.5 / 3 = 1375.8 hours
i
Case Study for Manufacturing of Solid Oxide Fuel Cell Power Systems (PNNL-22732), M. Weimer, L. Chick, D.
Gotthold and G. Whyatt, Pacific Northwest National Laboratory, August, 2013.
Model Approach
Sintering
o Part handling time labor cost based on part size per BDI formula and throughput; 4
second minimum
o Process cost based on furnace energy cost plus standard machine rate
Process Flow
Background
Preliminary Analysis
The cells for this analysis are for a 100 kWnet system at a production rate of 4,000 stacks per year
(1,000 systems with 4 stacks per system). The deposition area for the anode will be:
187 mm width 268 mm length = 501.16 cm2
The 30 kW stack requires 259 cells, requiring total annual production of:
Annual production = 259 parts/stack 4,000 stacks = 1,036,000 parts
Sintering Process
The sintering process schedule is based on the bi-layer sintering schedule suggested in PNNL-
22732i:
Segment Rate/Time Temp.
Ramp 0.5C/min 190C
Sintering Process 1
Manufacturing Cost Analysis of 100 and 250 kW Fuel Cell Systems for Primary Power and Combined Heat and Power
Applications / DOE Contract No. DE-EE0005250
Generally, the high sintering temperatures require the use of either pusher, walking beam, or
rotary hearth continuous furnaces. A used Surface Combustion rotary hearth furnace supports 60
5” 21” (12.7 cm 53.3 cm) trays with a workspace clearance of 35” (88.9 cm) for a total load
volume of 3.61 106 cm3. The maximum part thickness following the last screen printing
operation is 328 microns (0.0328 cm). Adding 1 cm on all sides for part spacing in the furnace
and assuming optimal racking, the total part envelope is:
Part Envelope = (18.7 + (2 1.0)) (26.8 + (2 1.0)) (0.0328 + (2 1.0)) = 1212 cm3
The maximum furnace loading is then:
Furnace loading = 3.61 106 cm3 / 1212 cm3 = 2980 parts
Using 25C as the ambient temperature, the required heating times are:
(190 – 25)C / (0.5/min) = 330 min. = 5.5 hours
(450 – 190)C / (0.5/min) = 520 min. = 8.67 hours
(1375 – 450)C / (3/min) = 308 min. = 5.14 hours
The required cooling time is:
(1375 – 25)C / (5/min) = 270 min. = 4.5 hours
Adding the 4 hours of hold time, the total furnace time is 27.81 hours, making furnace
throughput:
Throughput = 2980 parts / 27.81 hours = 107 parts/hour
Total machine time to fire the cells is:
Machine time = 1,036,000 parts / 107 parts/hour = 9,682 hours
Total number of furnaces required is:
Roundup(9682 / 6,000) = 2
Machine utilization is:
Utilization = 9682 / (2 6000) = 80.7%
Sintering Process 2
Manufacturing Cost Analysis of 100 and 250 kW Fuel Cell Systems for Primary Power and Combined Heat and Power
Applications / DOE Contract No. DE-EE0005250
Part load/unload, which may be manual or robotic, will be driven by overall part size. To
determine if a single operator can load and unload parts while the machine is operating, we use
the handling time formula developed previously, to determine the total load/unload time required
per hour of machine time as:
Part handling time = Max((0.3 ((187 + 268 + 1) / 25.4) - 4.6), 4) 107 parts/hour =
0.044 hours per hour of machine time.
Given that part handling time is a small percentage of total firing time, we assume one operator
is capable of covering 3 machines, making the labor time:
Machine labor time = 9682 / 3 = 3227.3 hours
i
Case Study for Manufacturing of Solid Oxide Fuel Cell Power Systems (PNNL-22732), M. Weimer, L. Chick, D.
Gotthold and G. Whyatt, Pacific Northwest National Laboratory, August, 2013.
Sintering Process 3
Manufacturing Cost Analysis of 100 and 250 kW Fuel Cell Systems for Primary Power and Combined Heat and Power Applications /
DOE Contract No. DE-EE0005250
Model Approach
Laser cut final cell shape
Process Flow
Background
Following sintering, the SOFC cells are laser cut to final dimensions as shown:
Preliminary Analysis
The cells for this analysis are for a 100 kWnet system at a production rate of 4,000 stacks per year
(1,000 systems with 4 stacks per system). The deposition area for the anode will be:
187 mm width 268 mm length = 501.16 cm2
The 30 kW stack requires 259 cells, requiring total annual production of:
Annual production = 259 parts/stack 4,000 stacks = 1,036,000 parts
Assuming a single setup operation requiring 1 operator per batch of parts, the final trim setup
time will be 0.5 hours.
Part load/unload, which may be manual or robotic, will be driven by overall part size. Using the
handling time formula adapted from the BDI DFMA® software, the total handling time is:
Part handling time = Max((0.3 ((187 + 268 + 3.5) / 25.4) - 4.6), 4) = 4.0 sec/part
Linde suggests that laser cutting of 1 mm thick stainless steel be performed using a 1,500 W
YAG laser under pure nitrogen flow of 8.0 m3/hr at a maximum speed of 7.0 m/min (0.117
m/sec). Assuming that the sintered ceramic has similar properties, the time to cut the cells is:
Total machine time = (7.8 + 4) sec/part 1,036,000 parts / 3600 = 3,396 hours
Given that the operator required time for load/unload represents less than 20% of the total part
processing time, we will assume one operator is capable of operating 3 machines, making the
machine labor hours:
Cutting material cost = 8.0 m3/hr 7.8 sec/part / 3600 sec/hr 1,036,000 parts = 17,957
m3
Material cost before scrap was determined in accordance with Appendix A2 as:
Model Approach
Ferritic stainless steel stamping operation
Laser etching operation
Perovskite coating operation
Heat treating operation
Process Flow
Background
The interconnect plates are dimensioned as shown:
304mm.
248mm.
28mm.
36mm.
10mm.
10mm.
Anode Gas Path Cutout
167mm.
10mm.
10mm.
28mm.
36mm.
Preliminary Analysis
The interconnects for this analysis will be working in a 30 kW stack for which the part size is:
The BDI DFMA® software provides pre-programmed cost models for the transfer stamping
operations used to manufacture the interconnect plate blanks, as shown below
The BDI DFMA® estimates a 2 hour machine setup time, and calculates the total manufacturing
time for the stamped part as 1 sec, making the total machine time for annual production:
Machine time = (1.0 sec/part / 3600) 1,072,000 parts + 2.0 = 299.8 hours
Assuming 1 full-time operator per machine, the total machine labor time is equal to the machine
time of 299.8 hours.
Tooling cost is $49,322 and the tooling is assumed to be capable of producing 400,000 parts.
Amortizing over a 5 year life, the total annual tooling cost:
ଵ
Annual tooling cost = (Tooling cost Number of tools purchased)
ହ
where: Number of tools purchased = Roundup(Total production / Tool life)
Assuming a single setup operation requiring 1 operator per production run of parts, the setup
time will be 0.5 hours.
Part load/unload, which may be manual or robotic, will be driven by overall part size. Because
the part will be turned in order to coat both sides, additional time equal to half of the load/unload
time will be added. Using the handling time formula developed previously, the total handling
time is:
Part handling time = 1.5 Max((0.3 ((239 + 304 + 0.25) / 25.4) - 4.6), 4) = 6.0 sec/part
The perovskite coating is deposited via aerosol spray to a depth of 3 microns (0.003 mm) and has
a material density of approximately 6.1 g/cm3. Assuming a 90% spray efficiency, and allowing
for 5 mm over spray on the 4 edges, the total deposited material per coated side is:
Coating material usage = 6.1 g/cm3 0.452 cm3/part 1,072,000 parts = 2,955 kg
The material cost before scrap was determined in accordance with Appendix A2 as:
Deposited depth is a function of flow rate, spray width and nozzle speed:
Coating depth = Flow rate (mm3/sec) / (Spray width (mm) Nozzle speed (mm/sec))
Spray nozzle manufacturers will generally specify a maximum flow rate associated with a
particular nozzle. Therefore, given a flow rate, coated width and coating depth, the nozzle speed
is calculated as:
Nozzle speed (mm/sec) = Flow rate (mm3/sec) / (Spray width (mm) Coating depth
(mm))
Using the SonoTek Flexicoat Impact nozzle system as an example, the maximum precision spray
width is approximately 50 mm and maximum nozzle speed of 400 mm/sec. Assuming a
maximum coating flow rate of 333 mm3/sec (20 mL/min), the nozzle speed is:
The time to coat both sides of the interconnect plate, allowing for 25 mm over spray on the 4
edges is:
Coating time per part = 2 ((239 + 50) mm (304 + 50) mm / (50 mm 400 mm/sec)) =
10.23 sec/part
Machine time = (10.23 + 6.0) sec/part / 3600 1,072,000 parts = 4,833 hours
Given that the operator required time for load/unload compared to the total part processing time
would allow sufficient transition to another machine, we will assume one operator per two coating
machines, making the machine labor time equal to the part handling time plus one half of the
machine time:
Labor time = (6.0 + (10.23 / 2)) sec/part / 3600 1,072,000 parts = 3.309 hours
The interconnect coating process calls for heat treatment at 1000°C (1472°F) for 4 hours after the
part reaches temperature. Large mesh belt furnaces manufactured by AFC-Holcroft feature a 66”
(167 cm) wide belt, 6” (15.24 cm) workspace clearance and effective load length of 456” (1158
cm) for a total load volume of 2.95 106 cm3. The maximum part thickness following the last
screen printing operation is 328 microns (0.0328 cm). Adding 1 cm on all sides for part spacing
in the furnace and assuming optimal racking, the total part envelope is:
Part Envelope = (23.9 + (2 1.0)) (30.4 + (2 1.0)) (0.025 + (2 1.0)) = 1699.3 cm3
The firing schedule is based on the metallization schedule suggested in PNNL-22732i of:
Using 25C as the ambient temperature, the required preheat time is:
Part load/unload, which may be manual or robotic, will be driven by overall part size. To
determine if a single operator can load and unload parts while the machine is operating, we use
the handling time formula developed previously, to determine the total load/unload time required
per hour or machine time as:
Part handling time = Max((0.3 ((239 + 304 + 0.25) / 25.4) - 4.6), 4) = 4 sec/part
Total part handling time = (4 sec/part / 3600) 137.1 parts/hour = 0.152 hours per hour
of machine time.
i
Case Study for Manufacturing of Solid Oxide Fuel Cell Power Systems (PNNL-22732), M. Weimer, L. Chick, D.
Gotthold and G. Whyatt, Pacific Northwest National Laboratory, August, 2013.
Model Approach
Ferritic stainless steel stamping operation
Process Flow
Background
The SOFC cell contains 3 frames:
The anode frame supports the interconnect on the anode side and provides space for the
anode mesh
The picture frame provides space for the cathode side of the cell
The cathode frame supports the interconnect on the cathode side and provides space for
the cathode mesh
The BDI DFMA® software provides pre-programmed cost models for the transfer stamping
operations used to manufacture the frames. Labor and machine times will be aggregated to
determine number of presses required and utilization. Material usage will be aggregated to
compute material cost.
Preliminary Analysis
The frames for this analysis will be working in a 30 kW stack for which the overall part size is:
Anode Frame
Frame Dimensions
304mm.
28mm.
28mm.
10mm.
10mm.
284mm.
239mm.
183mm.
10mm.
28mm.
28mm.
The resulting BDI DFMA® software analysis is shown in the following screen shot:
The BDI estimates a 2 hour machine setup time, and calculates the total manufacturing time for
the stamped part as 1.5 sec, making the total machine time for annual production:
Machine time = (1.5 sec/part / 3600) 1,036,000 parts + 2.0 = 431.7 hours
Assuming 1 full-time operator per machine, the total machine labor time is equal to the machine
time of 431.7 hours.
Tooling cost is $52,346 and assumed to be capable of producing 400,000 parts. Amortizing over
a 5 year life, the total annual tooling cost is:
ଵ
Annual tooling cost = (Tooling cost Number of tools purchased)
ହ
where: Number of tools purchased = Roundup(Total production / Tool life)
Total production = Annual production 5
ଵ
Annual tooling cost = (($52,346) Roundup((1,036,000 parts/year 5 years) / 400,000
ହ
parts/tool)) = $136,099.60
Picture Frame
Frame Dimensions
36mm.
10mm.
10mm.
Anode Gas Path Cutout
167mm.
36mm.
10mm.
The resulting BDI DFMA® software analysis is shown in the following screenshot:
The BDI estimates a 2 hour machine setup time, and calculates the total manufacturing time for
the stamped part as 1.0 sec, making the total machine time for annual production:
Machine time = (1.0 sec/part / 3600) 1,036,000 parts + 2.0 = 289.8 hours
Assuming 1 full-time operator per machine, the total machine labor time is equal to the machine
time of 289.8 hours.
Tooling cost is $62,062 and assumed to be capable of producing 400,000 parts. Amortizing over
a 5 year life, the total annual tooling cost is:
ଵ
Annual tooling cost = (Tooling cost Number of tools purchased)
ହ
where: Number of tools purchased = Roundup(Total production / Tool life)
Total production = Annual production 5
ଵ
Annual tooling cost = (($62,062) Roundup((1,036,000 parts/year 5 years) / 400,000
ହ
parts/tool)) = $161,361.20
Cathode Frame
Frame Dimensions
36mm.
10mm.
10mm.
Anode Gas Path Cutout
219mm.
36mm.
10mm.
The resulting BDI DFMA® software analysis is shown in the following screen shot:
The BDI estimates a 2 hour machine setup time, and calculates the total manufacturing time for
the stamped part as 1.0 sec, making the total machine time for annual production:
Machine time = (1.0 sec/part / 3600) 1,036,000 parts + 2.0 = 289.8 hours
Assuming 1 full-time operator per machine, the total machine labor time is equal to the machine
time of 289.8 hours.
Tooling cost is $47,377 and assumed to be capable of producing 400,000 parts. Amortizing over
a 5 year life, the total annual tooling cost is:
ଵ
Annual tooling cost = (Tooling cost Number of tools purchased)
ହ
where: Number of tools purchased = Roundup(Total production / Tool life)
Total production = Annual production 5
ଵ
Annual tooling cost = (($47,377) Roundup((1,036,000 parts/year 5 years) / 400,000
ହ
parts/tool)) = $123,180.20
Machine Utilization
Total machine time to produce the 3 frames is:
Material Cost
The BDI DFMA® software estimated the part weights as:
Annual material usage = (0.524 + 0.128 + 0.124) kg/cell 259 cells/stack 4,000 stacks
= 803,936 kg
Model Approach
Calculate glass-ceramic sealant batch size
Calculate glass-ceramic sealant production time
Calculate glass-ceramic sealant application time
Process Flow
Background
The sealant bead (dashed lines) is applied to the picture frame and cathode frame as shown:
Picture Frame
5mm. 304mm.
5mm.
248mm.
5mm.
Anode Gas Path Cutout
167mm.
5mm.
Cathode Frame
304mm.
5mm.
248mm.
5mm.
Anode Gas Path Cutout
219mm.
5mm.
Preliminary Analysis
Machine and labor time will be calculated based on an annual production of 1,000 100 kWnet
systems per year, with each system requiring 4 stacks. The 30 kW stack requires 259 of each
type of frame
Sealant Material
The sealant will be applied to areas that are 10 mm wide, and needs to fill a gap of about 148
microns (on both sides of the interconnect plates and one side of the picture frame with the same
seal length as the cathode side. Assuming a maximum finished seal width of 8 mm, the total seal
cross-sectional area is
Assuming application in a round bead, the required bead diameter that will yield the same cross
sectional area is:
The total seal length per side based on the above drawings is:
Picture frame seal length: (2 (304 – 10)) + (4 (239 – 10)) + (2 (304 – 50)) = 2012
mm
Cathode frame seal length: (2 (304 – 10)) + (4 (239 – 10)) = 1504 mm
The total volume of seal material required per cell (sealing both parts) is:
The total sealant batch size (cm3) for 4,000 stacks is:
A typical sealant is the Ceredyne VIOX V1649 glass ceramic sealant, consisting of 50/50
borosilicate glass/lanthanum oxide by volume. The density of the mixture is:
Material costs before scrap were determined in accordance with Appendix A2 as:
To create the paste, the borosilicate glass/lanthanum oxide mixture is fired at 1050C in a pot
furnace for 3 hours. The Trent PF-1000 17 kW pot furnace with a total pot size of 44,480 cm3 is
capable of holding 145 kg of sealant material at 75% fill volume. At the highest anticipated
volume, the 25 kW stack will require annual production 13.4 million cells, requiring about 12.5
times as much material, so this furnace should be sufficient for all anticipated levels of
production.
Assuming that 1 operator can setup a sealant batch in 0.5 hours results in a total furnace machine
time of:
Given that the operator required time for batch setup represents 14% of the total part processing
time, we will assume one operator is capable of operating 3 machines, making the machine labor
hours:
Machine labor hours = 130 batches (0.5 + (3.0 / 3)) = 195 hours
Assuming a single setup operation requiring 1 operator per batch of parts, the sealant application
station setup time is 0.5 hours.
Part load/unload (2 total part movements), which may be manual or robotic, will be driven by
overall part size. Using the handling time formula developed based on the BDI DFMA®
software, the total handling time for each part is:
Part handling time: Max((0.3 ((304 + 239 + .25) / 25.4) - 4.6), 4) = 4.0 sec/part
The Boothroyd-Dewhurst DFMA® software tool estimate for the bead application rate of viscous
sealants is 2 in/sec (51 mm/sec) with an applicator positioning time of 0.4 seconds. For the
picture frame, we assume that the bead is applied to the part perimeter in a single bead, followed
by the two beads between the anode gas path and the ceramic cell, followed by the two beads
between the anode gas path and the ceramic cell. Consequently, there will be 6 total re-
positionings: move applicator to start of perimeter bead, move applicator to start of each of the
four gas path beads, move applicator to second gas path bead, move applicator to home position.
The total application time is:
Picture frame application time = 4.0 sec + (6 0.4) sec + (2012 mm / 51 mm/sec) = 45.85
sec/part
For the cathode frame, we assume that the bead is applied to the part perimeter in a single bead,
followed by the two beads between the anode gas path and the ceramic cell for 4 total re-
positionings. The total application time is:
Cathode frame application time = 4.0 sec + (4 0.4) + (1504 mm / 51 mm/sec) = 35.09
sec/part
Machine time = 0.5 hours + ((80.94 sec / 3600) hours/cell 259 cells/stack 4,000 stacks))
= 23,293 hours
Given that the operator required time for load/unload represents 12% of the total part processing
time, we will assume one operator is capable of operating 3 sealant machines, making the
machine labor hours:
Model Approach
Stack Brazing
o Part handling time labor cost based on part size per BDI formula and throughput;
4 second minimum
o Process cost based on oven energy cost plus standard machine rate
Process Flow
Background
Preliminary Analysis
The cells for this analysis are for a 100 kWnet system at a production rate of 4,000 stacks per year
(1,000 systems with 4 stacks per system).
The overall part envelope is bounded by the end plate length and width and the stack height. The
30 kW stack height is estimated based on a thickness of about 2.1 mm per repeat cell, plus 12
mm each for the two end plates. The total stack envelope estimated as:
36.4 cm width 29.9 cm length ((0.21*259) + (1.2*2)) cm high = 61,808 cm3
Production time will be calculated based on an annual production of 4,000 stacks.
Stack Brazing
The brazing schedule is based on the metallization schedule suggested in PNNL-22732i of:
Segment Rate/Time Temp.
The 30 kW stack dimensions require that the furnace have a workspace clearance of at least 40
cm (15.7”), an unusually large size for conveyor-type continuous furnaces. A used Surface
Combustion rotary hearth furnace supports 60 5” 21” (12.7 cm 53.3 cm) trays with a
workspace clearance of 35” (88.9 cm) for a total load volume of 3.61 106 cm3. Allowing 5 cm
separation between stacks creates an effective part volume of:
41.4 cm width 34.9 cm length 61.8 cm high = 89,292 cm3
The maximum furnace loading is then:
Furnace loading = Rounddown(3.61 106 cm3 / 89,292 cm3) = 40 parts
Using 25C as the ambient temperature, the required heating times are:
(600 – 25)C / (3.0/min) = 192 min. = 3.2 hours
(875 – 600)C / (5.0/min) = 55 min. = 0.917 hours
The required cooling times are:
(875 – 750)C / (5.0/min) = 25 min. = 0.417 hours
(750 – 25)C / (5.0/min) = 145 min. = 2.417 hours
Adding the 7 hour hold times, total furnace time is 13.95 hours, making furnace throughput:
Throughput = 40 stacks / 13.95 hours = 2.867 stacks/hour
Total machine time to fire the cells is:
Machine time = 4,000 stacks / 2.867 stacks/hour = 1395.2 hours
Machine utilization is:
Utilization = 1395.2 / 6000 = 23.2%
Part load/unload, which may be manual or robotic, will be driven by overall part size. To
determine if a single operator can load and unload parts while the machine is operating, we use
the handling time formula adapted from the BDI DFMA® software to determine the total
load/unload time required per hour or machine time as:
Part handling time = Max((0.3 ((299 + 364 + 567.9) / 25.4) – 4.6), 4) = 9.94 seconds
Given that part handling time is a small fraction of total firing time, we assume one operator is
capable of covering 3 machines, making the labor time:
Machine labor time = 1395.2 / 3 = 465.1 hours
During the final hold at 750C, reducing gas (2% hydrogen in nitrogen) is flowed through the
anode cavities to reduce the NiO to Ni metal. Using the required testing flow rate for 30 kW
stacks of 358.86 l/min:
i
Case Study for Manufacturing of Solid Oxide Fuel Cell Power Systems (PNNL-22732), M. Weimer, L. Chick, D.
Gotthold and G. Whyatt, Pacific Northwest National Laboratory, August, 2013.
Model Approach
Tooling Cost
o Compute tooling cost
Tape Casting
o Compute machine time for machine setup
o Compute material cost for tape casting substrate
o Compute casting speed/throughput
o Compute machine and labor time for tape casting operation
Oven Drying
o Compute drying time and dryer length
o Compute radiant heater area
o Compute heater energy based on energy watt density and dryer area
Process Flow
Background
The total required anode thickness is 500 microns, which will be achieved via hot roller pressing
of two 250 micron tapes.
Preliminary Analysis
The cells for this analysis are for a 100 kW system at a production rate of 4,000 stacks per year
(1,000 systems with 4 stacks per system). The deposition area for the anode will be:
Tooling Cost
In a personal communication with Richard Mistler, author of Tape Casting: Theory and Practice
(Wiley – American Ceramic Society, 2000), estimates that a doctor blade for this application
would cost approximately $2,050 and would “last for years”. Using 100,000 parts as a life
approximation and amortizing over a 5 year production life, the total annual tooling cost is:
ଵ
Annual tooling cost = (Tooling cost Number of tools purchased)
ହ
where: Number of tools purchased = Roundup(Total production / Tool life)
Total production = Annual production 5
ଵ
Annual tooling cost = (($2,050) Roundup((1,036,000 parts/year 2 tapes/part 5 years) /
ହ
100,000 parts/tool)) = $42,640
Tape Casting
Since the slurry cost is calculated separately, the material cost will consist of the cost of the tape
casting carrier film. The carrier film is usually Mylar or polyethylene. For roll stock in 2 mil
thickness, these materials cost approximately $2.00/m2 in bulk. Assuming that the casting width
will be equal to the longest part dimension (i.e., the part length), the required casting length is
determined by the part width as:
Tape casting machine setup consists of loading and threading the casting substrate, and loading
the ceramic slurry into the reservoir. Bulk roll stock is available in 1000 meter lengths. Thus,
the number of setups required to run a batch of parts is:
Allowing 25 mm casting margin on each side, the required minimum roll widths are:
Rolls commonly appear in 6 inch (152.4 mm) incremental widths, requiring a 18” (427.2 mm)
roll width. The total material required is:
For roll stock in 2 mil thickness, these materials cost approximately $0.32/m2 in bulk. Material
cost was determined in accordance with Appendix A2 as:
Casting speed is limited by the slurry material properties, since running too fast can result in
non-uniform deposition. In “Tape casting of high dielectric ceramic composite substrates for
microelectronics applications,” Tok, et.al. plotted experimental data relating maximum green
tape thickness to casting speed, which shows a roughly exponential shape. Using the Excel
function LOGEST for estimating an exponential curve fit produced the following relationship
with maximum 3% error in the range of 150 – 300 microns:
For a green tape thickness of 250 microns, the resulting casting speed is:
Assuming 1 operator per casting machine for both setup and operation, the machine labor time is
equal to the total machine time of 18,040 hours:
Casting speed is also a function of required drying time and available dryer length. HED®
International’s PRO-CAST® series features systems ranging in length from 12 to 100 feet (3.66
to 30.5 meters).
Following deposition, the ceramic slurry is dried, usually by means of a tunnel dryer positioned
directly after the deposition step. The drying can be done by either radiant or convective heating.
For the cost analysis, we will assume radiant (infrared) heating and compute the cost of drying
by determining the required heater area.
Drying time is a function of the evaporation rate of the solvent and is inversely and exponentially
proportional to the coating thickness. Experiments conducted by Mistler (Tape casting of
ceramics, Ceramic Processing Before Firing, 1978) indicate drying rates of 1.3510-5 g/cm2-sec
at room temperature for an air flow rate of 2 l/min, and 2.2210-5 g/cm2-sec at room temperature
for an air flow rate of 75 l/min.
Previous analysis assumed that the anode slurry material was formulated as follows:
26 wt% YSZ (8% mol)
37 wt% NiO
12 wt% water
24 wt% binder (Dow Duramax B-1000/B-1014)
1 wt% dispersant
The binder consists of approximately 45% solids. Roughly estimating the volume of liquid per
gram of slurry by multiplying the material density by the material weight percent:
Liquid density = (0.12 1.0) + ((0.24 0.55) 1.05) + (0.01 1.16) = 0.270 g/cm3
The weight of liquid to be removed per unit area is a function of slurry thickness:
At a rate of 2.010-5 g/cm2-sec drying rate, the estimated drying time is:
Drying time = 0.0135 g/cm2 / 2.010-5 g/cm2-sec = 675 sec = 11.25 min
Infrared heating panels are generally sold with various energy watt densities and in standard
sized units and assembled to provide the necessary heating area. Using the Casso Solar Type FB
as an example, standard watt densities are 15 and 25 W/in2 (23 and 39 kW/m2) with standard
width of 12” (0.305 m) and lengths in 12” increments up to 60” (1.524 m). They note that 25
W/in2 corresponds to an emitter temperature of 880°C, and that the conversion efficiency of
electrical power to usable radiant energy is up to 80%.
While the heater energy density will be taken as an input, the drying temperatures for the green
tape are fairly moderate (150°C or less), so that the 23 kW/m2 should be sufficient to maintain
the drying area temperature. While researching the tape casting process, the manufacturing
specifications for the 1 kW parts were provided to HED International, a manufacturer of coaters,
dryers, kilns and furnaces. They recommended their TCM-251M tape casting machine with 12”
(300 mm) casting width and 25 foot (7.7 meter) casting length with counter-flow heated-air
dryer. The total machine power rating is 24 kW, the bulk of which would be consumed by the
drying system. This is consistent with our estimate of 25 kW/m2 0.955 m2 = 23.9 kW.
The roll-to-roll pressing operation can be either a semi-continuous process where the material is
indexed into a standard heated platen press (see DTI, Mass Production Cost Estimation for
Direct H2 PEM Fuel Cell Systems for Automotive Applications: 2010 Update, Section 4.4.6.1)
or completed by pre-heating and passing through heated rollers in a calendaring process. For the
preliminary analysis we will assume a calendaring process.
Setup
Setup consists of loading, threading and aligning the anode tapes into the calendaring rollers.
For costing purposes, we will take the setup time as a user input and assume a value of 0.5 hours.
Tapes were cast in 1000 meter lengths totaling 387,464 m, so that the number of setups required
is:
The calendaring process consists of 2 main steps: preheating and rolling. We will assume that
the anode layers are brought together and passed through an infrared tunnel oven for preheating.
Assuming that the two layers need to reach 500°C, we can estimate the oven dwell time as
(noting that 1 W = 1 J/sec):
Oven dwell time = Part weight (g) Part specific heat (J/g-°C) Temperature rise (°C) /
Energy input (W)
If we assume that the same infrared heaters used for drying are used for preheating, the energy
rate impinging on the part is:
The cast anode material has a density of approximately 4.6 g/cm3 and specific heat capacity of
0.590 J/g-°C.
Anode oven dwell time = (4.6 g/cm3 25.058cm3 0.590 J/g-°C) 475°C / 922 W =
35.0 sec
For the calendaring process, the layers will be moving together, so a heating time of 35 seconds
(0.583 min) is used to determine the required oven length. At a substrate speed of 5 m/min the
required heating length is about 2.92 meters, which can be accomplished using eight 12” by 36”
IR panels (four for each layer).
Once the material layers are preheated, they are compressed between steel rollers that bond the
anode layers together. Total machine time to setup and produce 1,036,000 parts is:
Machine time = (194 setups 0.5 hours/setup) + (1,036,000 parts / 1604.3 parts/hour) =
742.77 hours
Given an availability of 6000 hours per year per machine, the number of coating systems
required is:
Model Approach
Test and condition fuel cell stack
Process Flow
Background
Following assembly, the SOFC stack is tested and conditioned to determine its fitness for
installation into a CHP system. The total test time is assumed to be 6 hours, consisting of a 2
hour warm-up at 5% hydrogen (H2)/95% nitrogen (N2), a 2 hour test at 50% H2/50% N2, and 2
hour cool-down at 100% N2. Total H2 consumption for a 30 kW stack at full power is 359 l/min.
Machine and labor time and material usage will be calculated for production of 1,000 100 kWnet
systems, each system requiring four 30 kW stacks.
Preliminary Analysis
The mass of 1 mole hydrogen gas (H2) = 2 grams, so the mass of 22.4 liters (stp) of H2 is 2 g.
During the 2 hour full power test, the total material usage is:
Hydrogen: Full power material usage = (21.54 0.50) m3/hr 2 hrs = 21.54 m3
Nitrogen: Full power material usage = (21.54 0.50) m3/hr 2 hrs = 21.54 m3
Total material cost for a full test and conditioning cycle of 4,000 stacks is:
We will assume that 1 operator takes 0.5 hours for setup and tear-down of each stack on the test
stand, and can cover 3 testing stations, making the total labor time:
Testing labor cost = (0.5 + (0.33 6) hrs/stack) 4,000 stacks = 10,000 hours
We will assume that each test stand can support 3 stacks simultaneously, making the total
machine time:
Testing machine time = (0.5 + 6) hrs/stack 4,000 stacks / 3 stacks/test stand = 8,667
hours
Model Approach
Process Flow
Background
The BDI DFMA® software provides pre-programmed cost models for the casting and cell machining
operations used to manufacture the fuel cell stack end plates. The end plates need to be rigid in order to
apply even pressure across the face of the stack. The process selection for the SOFC end plate was
automatic sand casting of Hastelloy X to near net shape, followed by finish machining of the stack contact
face on a Haas VF-3B vertical machining center, and reaming and tapping the holes for fuel and exhaust.
Preliminary Analysis
12mm.
6mm.
66mm.
Ø 10.0mm.
typ 12mm.
40mm.
15mm.
15mm.
249mm. 58mm.
6mm.
10mm.
10mm.
130mm.
149.5mm.
80mm.
15mm.
167mm.
299mm.
M24 x 1.5
M16 x 1.5
130mm.
80mm.
364mm.
End Plate
The DFMA® estimates a 18.35 hour machine setup time, and calculates the total manufacturing time for
the end plates as 3441 sec, making the total machine time for annual production of 1,000 100 kW
systems:
Machine time = (3441 sec/part / 3600) 8,000 parts + 18.35 = 7,665 hours
Because total time exceeds available single machine time, machine utilization is:
Utilization = 7,665 / (Roundup(7,665 / 6000) 6000) = 63.9%
Assuming 2 full-time operators (1 for casting, 1 for machining) per station, the total machine labor time is
equal to twice the machine time = 15,330 hours.
Tooling cost is $19,718 and is assumed to be capable of producing 180,000 parts. Amortizing over a 5
year production life, the total annual tooling cost is:
ଵ
Annual tooling cost = (Tooling cost Number of tools purchased)
ହ
where: Number of tools purchased = Roundup(Total production / Tool life)
Total production = Annual production 5
ଵ
Annual tooling cost = (($19,718) Roundup((8,000 parts/year 5 years) / 180,000 parts/tool))
ହ
= $3,943.60
Model Approach
Stainless steel mesh laser cutting operation
Process Flow
Background
The SOFC cell contains 2 meshes:
The anode mesh is 0.5 mm thick and provides directed anode gas flow and electrical
contact between the anode contact layer and interconnect
The cathode mesh is 0.75 mm thick and provides directed cathode gas flow and electrical
contact between the cathode contact layer and interconnect
Preliminary Analysis
Machine and labor time will be calculated based on an annual production of 1,000 100 kWnet
systems per year, with each system requiring 4 stacks. The 30 kW stack requires 259 of each
type of mesh, requiring production of:
Assuming a single setup operation requiring 1 operator per batch of parts, the laser cutting setup time will
be 0.5 hours.
Part load/unload, which may be manual or robotic, will be driven by overall part size. Using the
handling time formula adapted from the BDI DFMA® software, the total handling time is:
Anode mesh handling time = Max((0.3 ((241.3 + 158.7 + 0.5) / 25.4) - 4.6), 4) = 4.0
sec/part
Cathode mesh handling time = Max((0.3 ((241.3 + 162.6 + 0.75) / 25.4) - 4.6), 4) = 4.0
sec/part
Research suggests that laser cutting of 0.75 mm thick stainless steel can be performed using a 1,500 W
YAG laser under pure nitrogen flow of 8.0 m3/hr at a maximum speed of approximately 15.0 m/min (250
mm/sec). Assuming that the mesh has similar cutting properties as stainless steel plate, the time to cut the
mesh is:
Total machine time = (((3.20 + 4) + (3.23 + 4)) sec/part / 3600) 259 cells/stack 4,000
stacks = 4,153 hours
Given that the operator required time for load/unload is about the same as the total part processing
time, we will assume one operator per machine, making the machine labor hours equal to 4,153
hours.
At a consumption rate of 8.0 m3/hr, the nitrogen (inert gas used for laser cutting) material usage
is:
Material usage = 8.0 m3/hr ((3.20 + 3.23) sec / 3600) 259 cells/stack 4,000 stacks =
14,803 m3
Model Approach
Laser weld anode frame to interconnect
Laser weld picture fame to cathode frame
Process Flow
Background
Prior to final assembly and sealing of each SOFC cell, the SOFC frames are laser welded as
shown by the dotted lines:
5mm.
5mm.
284mm.
239mm.
183mm.
5mm.
5mm.
248mm.
5mm.
Anode Gas Path Cutout
219mm.
5mm.
Preliminary Analysis
The cells for this analysis are for a 100 kWnet system at a production rate of 4,000 stacks per year
(1,000 systems with 4 stacks per system).
The 30 kW stack requires 259 cells, requiring total annual production of:
Annual production = 259 parts/stack 4,000 stacks = 1,036,000 parts
Assuming a single setup operation requiring 1 operator per batch of parts, the final trim setup
time will be 0.5 hours.
Part load/unload, which may be manual or robotic, will be driven by overall part size. Using the
handling time formula adapted from the BDI DFMA® software, the total handling time is:
Part handling time = Max((0.3 ((239 + 304 + 0.25) / 25.4) - 4.6), 4) = 4.0 sec/part
Anode frame weld length = (2 ((239 – 10) + (304 – 10))) + (2 (304 – 10)) = 1,634
mm
Cathode frame weld length = (2 ((239 – 10) + (304 – 10))) + (2 (239 – 10)) = 1,504
mm
Dilas (http://www.dilas.com/) suggests that laser welding to a depth of 0.5 mm in stainless steel
can be achieved using a 500 W YAG laser at a maximum speed of 6.0 m/min (0.10 m/sec). The
time to weld the frames is:
The total machine time required for welding both sets of frames is:
Total machine time = ((16.34 + 4) + (15.04 + 4)) sec/part 1,036,000 parts / 3600 =
11,333 hours
Given that the operator required time for load/unload represents about 25% of the total part
processing time, we will assume one operator is capable of operating 3 machines, making the
machine labor hours: