1132 Ind. Eng. Chem. Res.

2008, 47, 1132-1139

Economic Incentive for Intermittent Operation of Air Separation Plants with

Variable Power Costs
Jason Miller and William L. Luyben*
Department of Chemical Engineering, Lehigh UniVersity, Bethlehem, PennsylVania 18015

Stephane Blouin
AdVanced Control and Operations Research, Process and Systems R&D, Praxair Inc.,
Tonawanda, New York 14151

This article presents a simplified economic analysis of the effect of hourly variations in power costs on
energy and capital costs of cryogenic air separation plants. The objective is to see what ratio of peak-to-
minimum energy costs is required to make intermittently operated air separation plants economically attractive.
The study focuses on super-staged argon-column air separation plants producing both gaseous and liquid
oxygen and nitrogen, along with high-purity liquid argon. Preliminary results indicate that power price ratios
between about 2 and 7 are required, depending on the process and economic assumptions. The analysis uses
thermodynamic ideal-work calculations to predict energy requirements for plants that produce varying amounts
of liquid and gaseous products. These products must be provided to customers even when the plant is not
running, so excess liquid oxygen and nitrogen must be produced and stored during the period when the plant
is running. The actual power consumption of the base 24-hour plant is used to calculate a thermodynamic
efficiency (about 33%), which is then used for the modified plants. Capital costs for the conventional
continuously operated 24-hour plant are assumed to be some multiple of the energy costs for the 24-hour
plant evaluated at the minimum power price (Pmin). Capital costs for the 12-hour plant and the 8-hour plant
are scaled up from the capital cost of the 24-hour plant. The analysis assumes a three 8-hour power-price-
window cost structure. The economics are evaluated over a range of power price ratios. In the most optimistic
case, when the annual capital cost for the 24-hour plant is assumed to be 30% of the total annual cost, the
equipment cost is scaled using the conventional 0.6 power factor and instantaneous start-up is assumed, the
8-hour plant becomes economically viable at a power-price ratio of 2.13. When the instantaneous start-up
assumption is removed and one assumes that only the minimum power price is observed on weekends, the
8-hour plant becomes economically viable at a power price ratio of 4.15.

1. Introduction and 8-hour plants must be twice and three-times, respectively,

that of the 24-hour plant. The production of more liquid for
The dominant operating cost in air separation is energy. The storage when the plant is running to provide products when the
cost of electric energy can vary drastically from hour to hour plant is not running increases the power requirements. However,
during the day as industrial and domestic demands change with the cost of a unit of power is smaller because the plant is only
ambient temperature and activity. The ratio of the cost at the operating during periods of less-expensive energy. The capital
peak period of demand to the cost at the lowest period of investments are increased using a scale-up factor.
demand can be quite large (greater than 3) depending on the
location of the plant.
2. Process Description
One approach to reducing energy cost is intermittent plant
operation: run at higher than normal throughputs when power A typical super-staged argon cryogenic air separation plant
is cheap (at night) and shut down when power is expensive. Of (Figure 1) includes a double distillation column with a side
course, a prerequisite for intermittent operation is an agile column to recover high-purity liquid argon. Feed air (from the
process that can start up very quickly. In this preliminary atmosphere) is compressed and passed through an adsorbent
analysis, we assume that oxygen, nitrogen, and argon product bed of molecular sieves to remove water, carbon dioxide,
purities can be achieved in either 0 or 2 h. acetylene, ethylene, butane, and other heavier hydrocarbons.
The analysis uses thermodynamic ideal-work calculations to This helps alleviate the potential dangers of hydrocarbon-
predict energy requirements for plants that produce varying oxygen mixtures and prevents the freezing of material in the
amounts of liquid and gaseous products. These products must system.
be provided to customers even when the plant is not running, The feed-air stream is split, with a good portion being
so excess liquid oxygen and nitrogen must be produced and expanded in the lower column turbine, after being cooled in
stored during the period when the plant is running. The base the primary heat exchanger against returning cold oxygen and
plant operates continuously (24 h). Two other plants are nitrogen product streams, along with the waste nitrogen stream.
considered in which the plants operate for 12 h (one-half of the The air expansion provides refrigeration for the plant. This
time) or 8 h (one-third the time). The capacities of the 12-hour stream provides vapor air feed to the high-pressure column. The
air that is not expanded is also cooled in the primary heat
* To whom correspondence should be addressed. Tel.: 610-758- exchanger and provides liquid air for both the high-pressure
4256. Fax: 610-758-5297. E-mail: [email protected]. and low-pressure columns.
10.1021/ie070593n CCC: $40.75 © 2008 American Chemical Society
Published on Web 01/23/2008
 Ind. Eng. Chem. Res., Vol. 47, No. 4, 2008 1133

Figure 1. Simplified Schematic of Cryogenic Air Separation Plant.

Table 1. Work and Efficiency Results for Base 24-Hour Plant

stream air GN2 LN2 waste N2 LO2 GO2 LAr
type feed product product product product product product
temperature (K) 295.37 310.35 80.85 309.75 128.05 308.95 96.05
pressure (MPa) 0.1 0.9 0.9 0.1 1.5 1.4 0.1
flowrate (F/F24) 1.0000 0.0155 0.1219 0.6436 0.1468 0.0628 0.0093
enthalpy (kJ/mol) -0.09 0.29 -11.84 0.33 -10.67 0.20 -10.48
entropy (kJ/mol‚K) 0.0044 -0.0174 -0.1093 -0.0004 -0.0903 -0.0208 -0.0953
Wideal (kw) Wreal (kw) η (%)
4184 12 474 33

The lower column (high-pressure column) operates at ap- provided by a liquid nitrogen-add stream, which is combined
proximately 85 psia (0.586 MPa) and separates the air into a with the shelf transfer. The liquid nitrogen-add stream is
high-purity nitrogen stream (top) and oxygen-enriched liquid provided by drawing from the liquid nitrogen storage tank or
stream (bottom). The nitrogen stream is referred to as the shelf by recycling liquid from a nitrogen liquefier. A nitrogen liquefier
transfer and the enriched oxygen stream is called the kettle includes a series of compression, expansion, and heat exchange
transfer. The upper (low-pressure) column operates at ap- equipment.1 Nitrogen vapor at the top of the lower column is
proximately 20 psia (0.138 MPa) and produces high-purity condensed against boiling liquid oxygen in the bottom of the
nitrogen (top) and oxygen (bottom) product streams. The oxygen upper column by heat exchange in a reboiler-condenser.
product from the bottom of the upper column typically contains Argon boils between oxygen and nitrogen, which results
99.9% oxygen with the remainder being argon. The nitrogen in a peak argon composition in the lower portion of the
product from the upper column typically contains ppm-level upper column. A vapor side stream is drawn from the upper
impurities of oxygen. The oxygen liquid-product stream is column near the argon peak and is fed to the argon
pumped to a higher pressure. A portion is vaporized in the column. The argon column produces a liquid argon product
primary heat exchanger and provides high-pressure gaseous that is drawn a few stages from the top of the column. This
oxygen product, and the remainder goes to the product oxygen argon product stream contains ppm-level impurities of oxygen
liquid-storage tank. Reflux for both columns is generated at the and nitrogen. The product stream is drawn several stages
top of the lower column (i.e., shelf transfer acts as reflux for from the top of the column to prevent too much nitrogen from
the upper column). Additional reflux for the upper column is entering the product argon stream. Reflux for the argon column
1134 Ind. Eng. Chem. Res., Vol. 47, No. 4, 2008

with

Wideal ) ∆H - T0∆S )
(∑ PnHn - FairHair -
)
T0
(∑ )
PnSn - FairSair (2)

where H ) enthalpy (J/mol)

S ) entropy (J/mol‚K)
T0 ) 295.37 K
Pn ) molar flow rate of product n (mol/s)
Figure 2. Simplified Flowsheet 24-Hour Plant. In the limiting case where vapor oxygen, nitrogen, and argon
are produced at atmospheric conditions, the enthalpy change
for the ideal plant is negligible and eqs 2 and 1 are equivalent.
Thus, the ideal work of mixing described by eq 1 is inherently
captured by eq 2.
Using this approach, the ideal work for the base 24-hour plant
was determined. All of the stream conditions, including pressure,
temperature, and flow rate, were obtained from steady-state data
for an actual operating plant. The enthalpies and entropies of
each stream were obtained from Aspen Plus process modeling
software using the Peng-Robinson equation of state.2 The inlet
and outlet flow rates, enthalpies, and entropies were then used
to determine the ideal work for the plant using eq 2. The power
Figure 3. Simplified Flowsheet 12-Hour Plant. consumption for the plant compressors obtained from the plant
data was used to determine the actual work for the plant. The
ratio of the ideal work to the actual work is the thermodynamic
efficiency (∼33%) for the plant. A summary of the stream
conditions that were used, along with the ideal work, actual
work, and efficiency, is given in Table 1. The flow rates given
in Table 1 have been normalized with respect to the feed-air
rate for the base 24-hour plant (F24).
Using the thermodynamic efficiency that was determined for
the base 24-hour plant, work calculations for the 8-hour and
12-hour plants were performed. As mentioned above, the 8-hour
and 12-hour plants will be producing higher fractions of liquid
oxygen and nitrogen while operating to meet vapor and liquid
Figure 4. Simplified Flowsheet 8-Hour Plant.
product demands when idle. Simplified schematics, which show
is provided by heat exchange in the argon condenser be- the flow rates and work requirements for each plant, are given
tween the vapor at the top of the argon column and the oxygen- in Figures 2 -4. The flow rates in Figures 2-4 are described
enriched liquid (kettle transfer) from the lower column. This in relation to the feed-air rate for the base 24-hour plant (F24).
stream, after expansion to the lower pressure, has a lower boiling It was assumed that the same stream pressures, temperatures,
point than the argon. The liquid from the bottom of the argon enthalpies, and entropies (Table 1) would exist in each plant.
column is fed to the upper column. The oxygen-enriched One can note that the daily energy consumption for the 8-hour
liquid and vapor from the cold (boiling) side of the argon plant is the highest, whereas that for the 24-hour plant is the
condenser are also fed to the upper column. lowest. This is caused by the 8-hour plant having the highest
fraction of liquid products while running and the 24-hour plant
3. Preliminary Calculations having the lowest. Producing higher fractions of liquid inherently
requires more work.
The basic approach is to calculate the thermodynamic ideal
work for each plant. The ideal work increases as the fraction of
4. Economic Analysis
liquid products increases. The ideal work of separation of air
into the three pure products at atmospheric conditions is 4.1. Assumptions. A. Process Assumptions.
3
1. The thermodynamic efficiency is the same for all of the

∑ xj ln xj
plants.
Widealmix ) -FairRT0 (1)
j)1
2. The capacity and size of the 12-hour plant is twice that of
the base-case 24-hour plant. The capacity and size of the 8-hour
where Fair ) air feed rate (mol/s) plant is three times that of the base-case 24-hour plant.
R ) 8.314 J/mol‚K 3. Enough liquid oxygen and nitrogen are produced and stored
xj ) mole fraction of component j in the air feed during the period when the plant is running to provide both
j ) 1 (nitrogen), j ) 2 (oxygen), and j ) 3 (argon) oxygen and nitrogen products (gas and liquid) when the plant
T0 ) 295.37 K is not running.
In a more general sense, the ideal work is computed on the 4. The nitrogen liquefier is running whenever the plant is
basis of changes in conditions between the feed and the products running.
 Ind. Eng. Chem. Res., Vol. 47, No. 4, 2008 1135

5. Enthalpy and entropy data for feed and product streams TACn ) total annual cost for plant n ($/year), n ) 24, 12, or
are obtained from Aspen Plus process modeling software using 8
the Peng-Robinson physical property package.2 R ) ratio of ACC24/AEC24 evaluated at Pmin
6. Component mole balances perfectly close, with the β ) ratio of ACC12/ACC24
composition of air assumed to be 78.11 mol % nitrogen, 0.93 κ ) ratio of ACC8/ACC24
mol % argon and 20.96 mol % oxygen. By manipulating the above equations, one can obtain a general
7. Product streams are assumed to be pure. relationship for the critical power price ratio (Pmax/Pmin) at which
8. Plants are assumed to start up instantaneously. This TAC12 becomes lower than TAC24, TAC8 becomes lower than
assumption will be relaxed in a later section of this article. TAC24, and TAC8 becomes lower than TAC12. The critical
B. Sizing and Economic Assumptions. power price ratios are functions of R, β, κ, W12, W24, and W8
1. There are three 8-hour power-price windows during each and are given in eqs 12-14. Note that if the plants have the
24 h day. The base-case 24-hour plant encounters the minimum same efficiency, which has been assumed, the critical powerprice
power price for 8 h, the average of the maximum and minimum ratio is not a function of the efficiency, and therefore the ideal
power price for 8 h, and the maximum power price for 8 h. work can be used instead of the real work in eqs 12-14.
The 12-hour plant encounters the minimum power price for 8 However, the efficiency does impact the magnitude of the TAC
h and the average of the maximum and minimum power price of each plant.

( )
for 4 h. The 8-hour plant encounters only the minimum power
price when it is operating for 8 h. 5 W12 1

( )
2. The peak power price is some multiple of the minimum R(β - 1) + -
Pmax 6 W24 2

( )
power price. The third power price is the average of the ) (12)
Pmin 12-24 1 1 W12
maximum and minimum power prices. The economics are -
evaluated over a range of Pmax/Pmin ratios. 2 6 W24

( ) [ ( ) ]
3. The annual capital cost (capital investment divided by
Pmax W8 1
payback period) for the 24-hour plant is assumed to be some ) 2 R(κ - 1) + - (13)
multiple (R) of the annual energy cost of the 24-hour plant Pmin 8-24 W24 2

( ) [( ) ( ) ]
evaluated at Pmin. The annual capital cost for the 12-hour plant
Pmax W24 W8 5
is assumed to be some multiple (β) of the annual capital cost )6 R (κ - β) + - (14)
for the 24-hour plant. The annual capital cost for the 8-hour Pmin 8-12 W12 W12 6
plant is assumed to be some multiple (κ) of the annual capital
cost for the 24-hour plant. 4.3. Results. A comparison of the TAC for the three plants
4. Total annual costs (annual capital cost plus annual energy as a function of the economic parameters R, β, and κ is given
cost) for the three plants are calculated for different peak-to- in Figure 5. The parameter values used in part a of Figure 5 are
minimum power prices to assess at what ratio the intermittent R ) 1 (i.e., ACC24 and AEC24 (at Pmin) are equal), β ) 2 (i.e.,
plants become economically attractive. ACC12 ) 2ACC24), and κ ) 3 (i.e., ACC8 ) 3ACC24). A value
4.2. Calculations. On the basis of the assumptions listed of R ) 1, which corresponds to the annual capital cost equaling
above, the following relationships can be developed to determine the annual energy cost, has been derived from an article by
the annual costs for the 24-hour, 12-hour, and 8-hour plants: Castle.3 However, it is unclear if the author uses the exact
definition for R described in this article and under what power
ACC24 ) RW24Pmin (3) price structure the value was determined. Nevertheless, a 50/
50 split between annual capital and energy costs seems like a
AEC24 ) (1/3)W24[Pmax + Pmin + 0.5(Pmax + Pmin)] (4) logical starting point. Other values of R are explored below.
Likewise, direct scaling of the annual capital cost to the plant
TAC24 ) AEC24 + ACC24 ) capacity (i.e., β ) 2 and κ ) 3), which would be the case if
(1/3)W24[Pmax + Pmin + 0.5(Pmax + Pmin)] + RW24Pmin (5) parallel units were used, seems to be a reasonable starting point.
Other scale-up relationships are explored below. With these
ACC12 ) βACC24 ) βRW24Pmin (6) economic parameters, the 8-hour plant begins to yield the lowest
TAC at a power price ratio of about 6.78. All of the TAC plots
AEC12 ) W12[(2/3)Pmin + (1/6)(Pmax + Pmin)] (7) assume a minimum power price of $0.03/kwh.
A second value for R (0.43) has been derived from a article
TAC12 ) AEC12 + ACC12 ) W12[(2/3)Pmin +
by Scharle and Wilson.4 With this value, 70% of the TAC would
(1/6)(Pmax + Pmin)] + βRW24Pmin (8) be attributed to energy. Again, it is unclear if the authors use
the exact definition for R described in this article and under
ACC8 ) κACC24 ) κRW24Pmin (9) what power price structure the value was determined. Also, the
AEC8 ) W8Pmin (10) plants described by Scharle and Wilson did not include argon
production. However, it seems quite possible that an energy-
TAC8 ) AEC8 + ACC8 ) W8Pmin + κRW24Pmin (11) intensive process such a cryogenic air separation could have
an annual energy cost that is higher than the annual capital cost.
where Pmin ) minimum power price ($/kwh) From part b of Figure 5, one can observe that changing R from
Pmax ) maximum power price ($/kwh) 1 to 0.43 reduces the critical power price ratio at which the
Wn ) yearly energy requirements for plant n (kwh/year), TAC curves for the 8-hour and 12-hour plants intersect to a
n ) 24, 12, or 8 value of 3.55.
ACCn ) annual capital cost for plant n ($/year), The economics are also improved (part c of Figure 5) if one
n ) 24, 12, or 8 assumes that capital costs scale up by the ratio of the capacities
AECn ) annual energy cost for plant n ($/year), to the power of 0.6 (i.e., β ) 20.6 and κ ) 30.6), as is
n ) 24, 12, or 8 conventionally used (see Peters, Timmerhaus, and West).5 In
1136 Ind. Eng. Chem. Res., Vol. 47, No. 4, 2008

Figure 5. TAC Analysis - Instantaneous Start Up.

Table 2. Cost Breakdown - Most Optimistic Case

ACC AEC TAC ACC AEC TAC ACC AEC TAC
8-hour 8-hour 8-hour 12-hour 12-hour 12-hour 24-hour 24-hour 24-hour
plant plant plant plant plant plant plant plant plant
Pmax/Pmin (MM$/yr) (MM$/yr) (MM$/yr) (MM$/yr) (MM$/yr) (MM$/yr) (MM$/yr) (MM$/yr) (MM$/yr)
1.0 2.73 3.53 6.26 2.14 3.47 5.61 1.41 3.28 4.69
1.5 2.73 3.53 6.26 2.14 3.76 5.90 1.41 4.10 5.51
2.0 2.73 3.53 6.26 2.14 4.05 6.19 1.41 4.92 6.33
2.5 2.73 3.53 6.26 2.14 4.34 6.47 1.41 5.74 7.15
3.0 2.73 3.53 6.26 2.14 4.63 6.76 1.41 6.56 7.97

this case, the critical power price ratio is about 3.48. Thus, a instantaneous start-up was assumed, the TAC for the 8-hour
similar change in the critical power price ratio is observed by plant becomes a function of the power price ratio (part b of
adjusting β and κ, as was observed by adjusting R. Figure 6). Using the most optimistic economic parameters (R
Under the most favorable conditions (part d of Figure 5), ) 0.43, β ) 20.6, and κ ) 30.6) and assuming a 2 h start-up
when the annual capital cost for the 24-hour plant is assumed time, the critical power ratio is about 3.25 compared to 2.13
to be 30% of the total annual cost and the equipment cost is for instantaneous start-up.
scaled using the 0.6 power factor, the critical power price ratio Alternatively, the capacities of the 8-hour and 12-hour plants
is about 2.13. A breakdown of the annual capital and energy could be increased so that the required products could be
costs for the most optimistic case is shown in Table 2. produced in 6 and 10 h of operation, respectively, during which
4.3.1. Effect of Startup Time. Thus far, we have assumed products are produced from the columns. With this mode of
that the plants start up instantaneously. However, it typically operation, the critical power price ratio is about 3.02 (part c of
takes about 2 h for the oxygen and nitrogen product streams to Figure 6). Because the 24-hour plant runs continuously with
reach the desired product purities. Thus, there is a period of no daily start up, the TAC does not change when the start-up
time for the 8- and 12-hour plants when no product is being time is included in the analysis (compare parts a, b, and c of
drawn. To meet the product demands, the 8- and 12-hour Figure 6). One should also keep in mind that it takes about 10
plants would have to have higher throughputs than those h for the product argon to reach the desired purity given the
described in Figures 3 and 4 or run for 10 and 14 h per day, current plant design. In this analysis, it has been assumed that
respectively. it takes only 2 h. In a recent paper,7 we will discuss some design
If one assumes that the plants run for 10 and 14 h, the extra and control changes that can be implemented to reduce the time
2 h of operation for both the 8-hour and 12-hour plants fall in required to achieve argon product purity to about the time
the Pavg power price period. Thus, in contrast to Figure 5 where required for that of oxygen and nitrogen.
 Ind. Eng. Chem. Res., Vol. 47, No. 4, 2008 1137

Figure 6. TAC Analysis - Effect of Start-Up Time and Weekend Prices.

4.3.2. Effect of Weekends. Thus far, it has also been assumed The second 8-hour block of time had an average power price
that three 8 h power price windows are observed each day of of $0.0626/kWh, while the last 8-hour block had an average
the week. However, this is typically not the case on weekends, power price of $0.0715/kWh (i.e., Pmax). Thus, for this set of
where only Pmin is observed. When this is taken into account, data, the power price observed in the second 8-hour block is
the critical power price ratio assuming a 2 h start-up period not the average of Pmax and Pmin.
rises to about 4.15 (part d of Figure 6). In this case, we have Given the above power price data and assuming instantaneous
assumed that the 8-hour and 12-hour plants would operate for start-up along with the most favorable economic parameters (R
an extra 2 h per day instead of increase in capacity. ) 0.43, β ) 20.6, κ ) 30.6), the lowest TAC is obtained with
4.3.3. Example Using Actual Historic Power Price Data. the 12-hour plant (7.90 × 106 $/yr), followed by the 8-hour
Throughout this article, we have assumed a power price structure plant (7.95 × 106 $/yr), and finally the 24-hour plant (8.06 ×
having three 8 h windows, with the price encountered in the 106 $/yr). In relation to the base 24-hour plant, the incremental
second 8 h window being the average of the first (Pmin) and ROI eq 15, assuming a 4-year payback period, of the 12-hour
third (Pmax). However, actual energy prices will vary from hour plant is approximately 4.4%, whereas that of the 8-hour plant
to hour and from location to location.6 In this section, we present is 1.7%. Thus, the modified plant designs with intermittent
a concrete example of how this actual variability affects the operation are not economically attractive given the 2006
economic analysis. Actual power price data was obtained from powerprice data for the Baltimore area. However, as mentioned
PJM Interconnection8 for the Baltimore and central Maryland above, power-price variability changes from location to location
region (Baltimore Gas and Electric Company). The data set and there are most likely regions where the economics are more
contains hourly power price data for the entire year of 2006. favorable. The economics have to be evaluated separately for
The average hourly power price over the entire year can be each particular location of interest.

( )
observed in Figure 7.
The average power price for the entire year, which determines TAC24 - TACN
incROI ) 100 (15)
the annual energy cost for the 24-hour plant, was $0.0574/kWh. t (ACCN - ACC24)
*
The least expensive 12 h block of time, which is when one
would operate the 12-hour plant, occurred between 10:00 p.m. where: incROI ) incremental return on investment (%)
and 10:00 a.m. During this time the average power price was TAC24 ) total annual cost of base 24-hour plant ($/yr)
$0.0448/kWh. The least expensive 8-hour block of time, which TACN ) total annual cost of modified plant with N ) 8 or
is when one would operate the 8-hour plant, occurred between 12 ($/yr)
10:00 pm and 6:00 a.m., with an average power price of ACC24 ) annual capital cost of base 24-hour plant ($/yr)
$0.0381/kWh. Thus, the annual capital cost for the 24-hour plant, ACCN ) annual capital cost of modified plant with N ) 8 or
which is based on the minimum power price (section 4.2), will 12 ($/yr)
use $0.0381/kWh for Pmin. t* ) payback period (years)
1138 Ind. Eng. Chem. Res., Vol. 47, No. 4, 2008

Figure 7. Average Hourly Power Price - Baltimore, MD 2006.

Figure 8. Effect of Fraction of Liquid Products on Critical Power Price Ratio.

4.3.4. Effect of Higher Fraction of Liquid Products. The phase, the critical power price ratio rises to 3.2 (Figure 8).
product distribution given in Table 1 shows that 70% of As above, we have assumed that all of the plants have the
the product oxygen and 89% of the product nitrogen are same thermodynamic efficiency. As one would expect, plants
sold in the liquid phase. However, depending on demand, that supply larger fractions of liquid products are more amenable
the plant may be designed to produce higher or lower frac- to intermittent operation. The results shown in Figure 8 assume
tions of liquid products. If one assumes that all of the pro- instantaneous start-up and the most favorable economic param-
duct oxygen and nitrogen is sold in the liquid phase, the cri- eters. Here, we define the critical power price ratio as the
tical power price ratio is about 2.08 (Figure 8). If one as smallest power price ratio at which the lowest TAC is that of
sumes that all of the oxygen and nitrogen is sold in the vapor the 8-hour plant.
 Ind. Eng. Chem. Res., Vol. 47, No. 4, 2008 1139

5. Conclusions argon than the continuously operated 24-hour plant. In a recent

paper,7 we discuss methods for reducing the time required to
In conclusion, this article discusses the use of thermodynamic
achieve the desired argon product purity. The methods involve
ideal-work calculations to predict energy requirements for
both design and control changes. The design changes involve
cryogenic air separation plants that produce varying amounts
the addition of storage tanks to collect argon-column liquid
of liquid and gaseous products. The work calculations were used
during shutdown and the reintroduction of the collected liquid
to perform a simplified economic analysis to determine the effect
during the subsequent start up.9,10 Thus, the extra capital required
of hourly variations in power costs on energy and capital costs.
to build and operate the agile plant will have to be included in
The objective has been to see what ratio of peak-to-minimum
the economic analysis.
energy costs is required to make intermittently operated air
separation plants economically attractive. Preliminary results
indicate that power price ratios between about 2 and 7 are Literature Cited
required, depending on the process and economic assumptions. (1) Olszewski, W. J.; Union Carbide Corporation. Gas Liquefaction
Clearly, the TAC for the given plants is highly dependent on Process and Apparatus. U.S. Patent 3,677,019, July 18, 1972.
the economic assumptions that are made. However, the general (2) Peng, D. Y.; Robinson, D. B. A New Two-Constant Equation of
approach described in this article is generic and can be applied State. Ind. Eng. Chem. Fundam. 1976, 15, 59-64.
(3) Castle, W. F. Air Separation and Liquefaction: Recent Developments
to specific design cases. For example, one needs to consider and Prospects for the Beginning of the New Millennium. Int. J. Refrig.
the plant’s desired liquid-to-vapor product splits. As shown 2002, 25, 158-172.
above, a plant producing a higher fraction of products as liquids (4) Scharle, W. J.; Wilson, K. Oxygen Facilities for Synthetic Fuel
would be more amenable to intermittent operation. One would Projects. J. Eng. Ind. 1981, 103.
also need to consider the location of the plant as the power (5) Peters, M. S.; Timmerhaus, K. D; West, R. E. Plant Design and
prices may change more frequently and to a greater extent in Economics for Chemical Engineers, 5th ed.; McGraw-Hill, Inc.: New York,
2002; p 242.
different regions. Also, intermittent operation does not neces- (6) Miller, J.; Luyben, W. L.; Belanger, P.; Blouin, S.; Megan, L.
sarily require modified plants that can start up faster. If existing Improving Agility of Cryogenic Air Separation Plants. Ind. Eng. Chem.
plants are running below capacity, production rates can be Res. 2008, 47, 394-404.
increased during periods when power prices are at a minimum (7) Li, Y.; Flynn, P. C. Power Price in Deregulated Markets. Power
and the plants shut down when power prices are at a maximum. Engineering Society General Meeting, 2003; Vol. 2, pp 874-879
(8) PJM Interconnection website, www.pjm.com.
(9) Billingham, J. F.; Bonaquist, D. P.; Dray, J. R.; Lockett, M. J.;
6. Extensions and Future Work Beddome, R. A.; Praxair Technology, Inc. Rapid Restart System for
Cryogenic Air Separation Plant. U.S. Patent 6,272,884 B1, August 14, 2001.
The cases in this article have all assumed that the plants are (10) Smith IV, O. J.; Espie, D. M. Air Products and Chemicals, Inc.
capable of rapid start-up. In the least favorable scenario, it was Recirculation of Argon Sidearm Column for Fast Response. U.S. Patent
assumed that oxygen, nitrogen, and argon all take about 2 h to 6,070,433, June 6, 2000.
reach the desired product purities. Given the existing plant
design, oxygen and nitrogen do reach product purity in about 2 ReceiVed for reView April 26, 2007
h, but argon takes approximately 10 h. Thus, intermittent ReVised manuscript receiVed November 15, 2007
Accepted November 19, 2007
operation would probably not be economically viable because
the 8-hour and 12-hour plants would produce significantly less IE070593N