energies

Article
Internal Combustion Engine Model for Combined
Heat and Power (CHP) Systems Design
Nikolaos Kalantzis, Antonios Pezouvanis and Kambiz M. Ebrahimi *
Aeronautical and Automotive Engineering, Loughborough University, Leicestershire LE11 3TU, UK;
[email protected] (N.K.); [email protected] (A.P.)
* Correspondence: [email protected]

Received: 19 August 2017; Accepted: 21 November 2017; Published: 25 November 2017

Abstract: A model based, energy focused, quasi-stationary waste heat driven, internal combustion
engine (ICE) centred design methodology for cogeneration (heat and electricity) systems is presented.
The developed parametric model could be used for system sizing, performance evaluation, and
optimization. This paper presents a systematic approach to model the behaviour of the CHP system
using heat recovery prediction methods. The modular, physics based modelling environment shows
the power flow between the system components, with a special emphasis on the ICE subsystems,
parameter identification, and model validation.

Keywords: CHP; cogeneration; IC engine; heat recovery; dynamic; model

1. Introduction
Combined heat and power (CHP) or cogeneration technology involves the generation of readily
useable heat and power from a single fuel source [1–5]. It most commonly involves the operation of
a prime mover such as an internal (spark ignited & compression ignited reciprocating engines, gas
turbines) or external (Stirling, Rankine) combustion engine, and utilizes the waste heat of the power
producing cycle for heating purposes, thus increasing the overall efficiency of the application.
Due to the potential for resource and environment conservation, as well as the desire for power
autonomy, the combined heat and power (CHP) market is expanding [6] and is starting to penetrate
the small residential sector, for the scale of which (electrical capacity lower than 10 kWe), ICE based
micro-CHP systems are currently the most marketable of the available micro-CHP types due to sharing
an established, mature, highly reliable technology, while at the same time being characterized by
low purchase and maintenance costs [1–3,7–13]. While Compression Ignition (CI) engines enjoy a
widespread proliferation in the role of larger stationary power plants due to their associated advantages,
such as a high fuel conversion efficiency and durability, the fact that Natural Gas (NG) is readily
available to households through transmission pipeline networks, the environmentally friendly nature
of the fuel, as well as the low cost and effort adaptability of Spark Ignited (SI) engines to run on
Natural Gas has made small NG fuelled SI engines the technology of choice for powering the majority
of currently available ICE based micro-CHP units. The above, when combined with the widespread
usage of simulation tools for CHP system research, design, and selection, as well as the special nature
of cogeneration technology, have created the need for fast running mathematical models of spark
ignited internal combustion engines (SI ICE ) that predict both power and waste heat components.
As a response, several researchers have developed a number of quasi-stationary ICE and CHP models.
Voorspools and D’haeseleer [14] have shown that the transient behaviour of engine fuel conversion
efficiency is fast enough to be neglected in micro-CHP simulation applications, and this is reflected
in the observation that the majority of CHP models that are encountered in literature are of the
quasi-stationary type. Caresana et al. [7], made use of a lookup table based ICE modelling layout in

Energies 2017, 10, 1948; doi:10.3390/en10121948 www.mdpi.com/journal/energies

Energies 2017, 10, 1948 2 of 14

which, one lookup table returned fuel flow, and another table returned the exhaust heat as fraction
of chemical power inlet for given combinations of engine torque and speed. On the other hand,
the parametric CHP model that is available in the ESP-r building simulation software library, and
that is used in a number of studies on micro-CHP performance [2,15,16], is based on a polynomial
fit. Under this modelling technique, the CHP system is modelled as one single unit and the model
dependent variables, such as the Brake Specific Fuel Consumption, the Electrical Efficiency, and the
Heat to Power Ratio are calculated as polynomial functions of part load ratio. Another encountered
polynomial model was used by Cho et al. [17], who calculated system electrical efficiency as a function
of the electrical output of the generator set, while they assumed coolant and exhaust heat to be constant
fractions of 0.3 of the chemical power inlet.
On the other hand, the engine thermal response is much slower, and for this reason, it is often
included in cogeneration system models. In one study, thermal transient behaviour of the engine is
described by means of polynomial functions of time elapsed after shut down [2]. Kelly et al. [15] used
an engine model with a lumped capacitance transient thermal model with one node representing the
engine and the heat exchanger masses, and another node representing the water mass. In terms of the
transient behaviour of the exhaust temperature, Zavala et al. [18], found in their experimental data
that it can be adequately described by 1st order linear transfer functions, with AFR, spark timing, and
engine speed as their main model inputs.
Since the majority of CHP modelling applications involves the study of a particular CHP system
design, or the use of an averaged map based engine model on a CHP model, the encountered layouts
as described above have been found to be very well suited for most studies. In some cases though,
especially when the model is to be used in heat recovery system design applications, separating the
engine from the heat exchanger performance, while at the same time retaining engine model scalability,
can be a desirable combination of model characteristics.
The main aim of this document is the development of an SI ICE model layout for use in CHP
sizing and design applications whose special nature may require that the engine model incorporates
the behaviour of all of the necessary power components, temperatures, and flow rates for use with a
heat recovery system model, while at the same time being characterized by an increased scalability
and connectivity to heat recovery component models, simplicity, and a low computational load.
The structure of this paper is as follows: First, Section 2 presents the structure of the system model
and the mathematical relationships that are used to describe the flow of power between the model
components. Section 3 describes the procedure that is followed to collect the necessary experimental
data and identify the required model parameters. The developed model is then simulated and validated
in Section 4, while a discussion on simulation results takes place in Section 5. Finally, Section 6 contains
the conclusions of this study.

2. System Model
The communication layout of the developed model will be power based, as in such a configuration;
power can be used to connect individual subsystems and phenomena that handle energy on multiple
domains. In addition, contrary to most existing lookup table based engine models whose main maps
include parameters that are directly linked to mechanical power and fuel consumption, the developed
model will follow a reverse logic—not unlike the general layout presented by Heß et al. [19]—by means
of which, specific sub-models and their respective lookup tables provide the rates of the generated
.
waste heat components, whose values are then subtracted from the energy input rate Qin to provide
the mechanical power output Pmech .
.
Heat inlet from the fuel is released in the combustion chamber at a rate Qin , and part of it is
converted to mechanical power output Pmech . Exhaust heat is carried by the high temperature exhaust
.
gases at a rate Q Exh . Convective heat flows from the hot cylinder content to the colder engine mass
.
at a rate Qconv and is finally released into the engine coolant as well as the engine surroundings by
.
means of conduction, convection, and radiation. Heat that is generated by friction at a rate Q f r is
Energies 2017, 10, 1948 3 of 14

Energies 2017, 10, 1948 3 of 14

dissipated by the engine oil [20,21]. This configuration gives the researcher the ability to fine-tune the
simplified
three schematic
different of the
waste heat developedtomodel
sub-models analysing its alayout
better approximate given and main
engine. A constituent subsystems
simplified schematic of
is shown in Figure 1.
the developed model analysing its layout and main constituent subsystems is shown in Figure 1.

Figure 1.
Figure 1. Block
Block diagram
diagram of
of the
the proposed
proposed engine
engine model.
model.

ThePI
The PIcontroller
controlleruses uses thethe
setset speedspeed Nset𝑁and
𝑠𝑒𝑡 and the actual
the actual speedspeed N to set 𝑁 the
to set
value theofvalue of thevariable
the control control
variable volumetric efficiency
volumetric efficiency ηvp . The physical 𝜂 𝑣𝑝 . The physical limits of the system are defined
limits of the system are defined by a saturation filter. In order for by a saturation
filter.
the In order
engine model fortothebe engine
generally model to be generally
applicable, and for the applicable, and for the
sake of simplicity, fluidsake of simplicity,
densities fluid
and specific
densities and specific volumes for
volumes for atmospheric pressure Pamb = 1 atm, rather than for atmospheric pressure 𝑃 = 1 atm, rather than for
𝑎𝑚𝑏the manifold pressure, are used due to the manifold
pressure,
the fact thataretheused
pressuredue characteristics
to the fact that themanifold
of the pressuremay characteristics
vary significantly of thethroughout
manifold the may vary
engine
operating range and from one engine type to another. The following assumptions can be made of The
significantly throughout the engine operating range and from one engine type to another. the
following
model and assumptions
these are generally can betrue made of the model available,
for commercially and thesecalibrated,
are generally healthy truerunning
for commercially
SI engines:
available, calibrated, healthy running SI engines:
• The charge air fuel ratio (AFR) is assumed to always be stoichiometric.
• Spark The charge
timingair is fuel
assumed ratio to(AFR) occur is at
assumed
maximum to always be stoichiometric.
brake torque (MBT) for all of the operating points.
 Spark timing is assumed to occur at maximum brake torque (MBT) for all of the operating
In addition, the amount of fuel that is present in the charge is assumed to have completely
points.
evaporated before it enters the cylinder, and thus, in the case of port injected engines, the volumetric
In addition,
efficiency during the theparameter
amount ofestimation fuel that phaseis present in the charge
is calculated from: is assumed to have completely
evaporated before it enters the cylinder, .
and thus,
.
in the case of port injected engines, the volumetric
. .
8 (V
1.2 × 10estimation 108 (υ f _vap + υa × AFRst )
efficiency during the V chparameter f _vap +phase
V a ) is m f × 1.2 × from:
calculated
ηvp = . = = (1)
V ×̇ N V ×N
V sw𝑉̇𝑐ℎ 1.2 × 108d(𝑉𝑓_𝑣𝑎𝑝 + 𝑉𝑎̇ ) 𝑚̇𝑓 × 1.2 × 108 (𝜐d𝑓_𝑣𝑎𝑝 + 𝜐𝑎 × 𝐴𝐹𝑅𝑠𝑡 )
𝜂𝑣𝑝 = = = (1)
. ̇
𝑉𝑠𝑤 𝑉𝑑 × 𝑁 𝑉𝑑 × 𝑁 .
where V ch , the volumetric flow rate of the experimentally acquired charge, V sw the theoretical swept
where 𝑉̇𝑐ℎ , flow ̇ the
. .
volumetric the volumetric
rate, V a , and flow Vrate of the experimentally acquired charge, 𝑉𝑠𝑤 andtheoretical swept
f _vap the volumetric flow rates of the air charge the evaporated
volumetric
fuel, flow rate,
respectively, all under ̇
𝑉𝑎 , and ̇
𝑉𝑓_𝑣𝑎𝑝 the conditions
atmospheric volumetric in flow 3 rates
m /s, Vdofthe the air charge
engine and the evaporated
displacement
.
in cm3 , m f
3 3
fuel,mass
the respectively,
flow of fuel, all υunder atmospheric conditions in m /s, 𝑉𝑑 the
f _vap , and υ a the specific volumes of the fuel
engineand
vapour displacement in cmat
air, respectively, ,
𝑚 ̇ 𝑓 the
atmospheric mass flow
conditions. of fuel, 𝜐 𝑓_𝑣𝑎𝑝 , and 𝜐𝑎 the specific volumes of the fuel vapour and air,
.
respectively, at atmospheric conditions.
Since the rate of flow of exhaust heat Qexh is a function of the exhaust temperature Texh and the
exhaust Since theflow
mass raterate
of flow.
mexh (equal of exhaust to theheat charge 𝑄̇𝑒𝑥ℎ
mass is aflow
function
rate mof
. the exhaust temperature 𝑇𝑒𝑥ℎmass and
ch ), the calculation of the exhaust
the exhaust
flow rate is anmass flow rate
important 𝑚̇𝑒𝑥ℎof(equal
element to the charge
the functionality mass
of the model.flowUnderrate 𝑚 ̇ 𝑐ℎ ),above
the the calculation
assumptions, of the
the
exhaust mass flow rate is an important element of the functionality
mass flow rate of an air-fuel mixture (charge mass flow rate) of a known ratio for a given combination of the model. Under the above
assumptions,
of the massfrom:
N, ηvp is calculated flow rate of an air-fuel mixture (charge mass flow rate) of a known ratio for a
given combination of 𝑁, 𝜂𝑣𝑝 is calculated from:
N × Vd × ηvp 11 +
+ AFR
mch = 𝑁 × 𝑉𝑑 ×8 𝜂𝑣𝑝×
.
𝐴𝐹𝑅 (2)
𝑚̇𝑐ℎ = 1.2 × 10 8 ×υ f _vap + AFR × υa (2)
1.2 × 10 𝑓_𝑣𝑎𝑝𝜐 + 𝐴𝐹𝑅 × 𝜐 𝑎

Experience
Experiencehas hasshown
shownthatthat
exhaust temperature
exhaust may vary
temperature maysignificantly throughout
vary significantly the operating
throughout the
range of anrange
operating internal combustion
of an engine. In the
internal combustion case of
engine. Inspark ignited
the case (SI) engines,
of spark exhaust
ignited (SI) temperature
engines, exhaust
temperature for higher loads lies in the region of 600 °C, and in some cases, may even reach 900 °C,
while at idle, it lies in the region of 300 °C, [22], and according Zavala et al. [18], engine speed has a
Energies 2017, 10, 1948 4 of 14

for higher loads lies in the region of 600 ◦ C, and in some cases, may even reach 900 ◦ C, while at
idle, it lies in the region of 300 ◦ C, [22], and according Zavala et al. [18], engine speed has a stronger
influence than the air mass inlet rate on exhaust temperature. At the time of this writing, the most
relevant study on the behaviour of the exhaust temperature of SI engines has been found to have been
conducted by Eriksson [23], who based on results from crank angle based combustion models assumed
a linear dependence of the exhaust temperature on the exhaust mass flow rate. As the experiments
of Section 3 indicate a strong correlation between engine speed and engine volumetric efficiency, in
the current document, the exhaust temperature will be calculated as a function of engine speed and
engine volumetric efficiency. The sequence of calculations of the power component calculator of the
developed engine model is illustrated in the diagram of Figure 2. Once the exhaust outlet temperature
Texh is provided from the exhaust temperature map for a given combination of N, ηvp , the specific
enthalpy of the exhaust gas hexh at this temperature is known. Since the specific enthalpy of the exhaust
under ambient temperature hexh_amb is also known, the rate of exhaust heat flow is calculated from:
. .
Qexh = (hexh − hexh_amb ) × mch (3)

Heat ratio Q.R. is a size-independent representation of the rate of convective heat transfer that is
.
used by the developed model. The model uses the calculated rate of exhaust heat flow Qexh and the
heat ratio Q.R. map to calculate the rate of convective heat transfer or a given combination of N, ηvp :
. .
Qconv = Q Exh × Q.R. (4)
. .
Q.R. links Qconv to Qexh , which makes model scaling a simple task, since engine displacement
. .
defines the rate of exhaust heat flow Qexh , the heat inlet rate Qin , and the rate of heat generated by
. . .
friction Q f r (discussed below). In turn, Qexh defines Qconv for a given operating point.
Using the quadratic fit of f mep as a function of engine speed N from Ferguson et al. [20], the rate
of generation of heat from friction (engine displacement Vd in cm3 ) is calculated from:
 
2.3N 2 4N 3
. Vd × 94.8N + 103
+ 106
Q f r (N) = (kW) (5)
1.2 × 108

For a fuel lower heating value, LHV f , the rate of energy that enters the engine in the form of fuel
chemical power is calculated from:
. . .
Qin = m f × LHVf = mch × LHVf /(1 + AFR) (6)

The engine mechanical power output is now calculated from:

. . . .
Pmech = Qin − Q Exh − Qconv − Q f r (7)
 𝑄̇𝑖𝑛 = 𝑚̇𝑓 × 𝐿𝐻𝑉𝑓 = 𝑚̇𝑐ℎ × 𝐿𝐻𝑉𝑓 /(1 + 𝐴𝐹𝑅) (6)
The engine mechanical power output is now calculated from:
𝑃𝑚𝑒𝑐ℎ = 𝑄̇𝑖𝑛 − 𝑄̇𝐸𝑥ℎ − 𝑄̇𝑐𝑜𝑛𝑣 − 𝑄̇𝑓𝑟 (7)
Energies 2017, 10, 1948 5 of 14

Figure 2. Block
Figure diagram
2. Block diagramofofthe
the power componentcalculator
power component calculator subsystem.
subsystem.

3. Parameter Identification
The main maps of the model that calculates the exhaust temperature Texh and the heat ratio Q.R.,
both as functions of the engine speed N and the engine load will be constructed. A series of tests
have been performed on an SI direct injection engine using a transient engine test cell. The engine
specifications and Engine Control Unit (ECU) settings used for in the tests are presented in Table 1.

Table 1. Characteristics and settings of the tested engine.

Test Engine Characteristics and Settings

Engine type SI, Naturally Aspirated, direct injection
Displacement 1.6 L
Number of Cylinders 4
Engine Control Unit (ECU) settings Constant stoich. AFR, Spark at MBT

Set speed was varied from 1500 rpm to 4000 rpm in 500 rpm increments. For each set speed
value, α was varied from 20% to 100% of maximum throttle angle in increments of 10%. The exhaust
temperature, the fuel mass flow rate, and the engine torque were directly measured. All of the tests
were carried out on the same day in order to reduce the potential for introduction of error due to
inconsistent conditions to the set of recorded data.
The coefficients of 5th order polynomial fits performed on the acquired datasets of Texh ( N, ηvp )
and Q.R.( N, ηvp ) of the tested engine, whose surface plots are shown in Figures 3 and 4, are provided
in Appendix A Table A1 to be used in a polynomial of the general form Equation (8).

f ( N, ηvp ) = p00 + p10 × N + p01 × ηvp + p20 × N 2 + p11 × N × ηvp + p02 × ηvp 2
+ p30 × N 3 + p21 × N 2 × ηvp + p12 × N × ηvp 2 + p03 × ηvp 3 + p40 × N 4
(8)
+ p31 × N 3 × ηvp + p22 × N 2 × ηvp 2 + p13 × N × ηvp 3 + p04 × ηvp 4 + p50 × N 5
+ p41 × N 4 × ηvp + p32 × N 3 × ηvp 2 + p23 × N 2 × ηvp 3 + p14 × N × ηvp 4 + p05 × ηvp 5
provided in Appendix A Table A1 to be used in a polynomial of the general form Equation (8).
𝑓(𝑁, 𝜂𝑣𝑝 ) = 𝑝00 + 𝑝10 × 𝑁 + 𝑝01 × 𝜂𝑣𝑝 + 𝑝20 × 𝑁 2 + 𝑝11 × 𝑁 × 𝜂𝑣𝑝 + 𝑝02 × 𝜂𝑣𝑝 2
+𝑝30 × 𝑁 3 + 𝑝21 × 𝑁 2 × 𝜂𝑣𝑝 + 𝑝12 × 𝑁 × 𝜂𝑣𝑝 2 + 𝑝03 × 𝜂𝑣𝑝 3 + 𝑝40 × 𝑁 4
(8)
+𝑝31 × 𝑁 3 × 𝜂𝑣𝑝 + 𝑝22 × 𝑁 2 × 𝜂𝑣𝑝 2 + 𝑝13 × 𝑁 × 𝜂𝑣𝑝 3 + 𝑝04 × 𝜂𝑣𝑝 4 + 𝑝50 × 𝑁 5
4 3 2 2 3 4 5
+𝑝
Energies 2017,
41 × 𝑁 × 𝜂𝑣𝑝 + 𝑝32 × 𝑁 × 𝜂𝑣𝑝 + 𝑝23 × 𝑁 × 𝜂𝑣𝑝 + 𝑝14 × 𝑁 × 𝜂𝑣𝑝 + 𝑝05 × 𝜂𝑣𝑝
10, 1948 6 of 14

Figure
Figure
Energies 2017, 10, 3. Map
1948 3. Map of
of the
the exhaust
exhaust temperature
temperature vs.
vs. engine
engine speed
speed and
and volumetric
volumetric efficiency.
efficiency. 6 of 14

Figure 4.
Figure 4. Map
Map of
of calculated
calculated Heat
Heat Ratio
Ratio Q.R.
Q.R. vs.
vs. engine
engine speed
speed and
and volumetric
volumetric efficiency.
efficiency.

4. Simulation and Validation

4. Simulation and Validation
In order to showcase the operation of the engine model that was developed in the previous
In order to showcase the operation of the engine model that was developed in the previous
sections within a CHP application, the engine model is connected to a higher level cogeneration
sections within a CHP application, the engine model is connected to a higher level cogeneration model
model whose block diagram is shown in Figure 5. A simple four-dimensional (4-D) map based heat
whose block diagram is shown in Figure 5. A simple four-dimensional (4-D) map based heat exchanger
exchanger model is connected to and receives input signals from the engine model, as well as from
model is connected to and receives input signals from the engine model, as well as from the primary
the primary circuit of a thermal storage tank model. The heat exchanger model input signals are the
circuit of a thermal storage tank model. The heat exchanger model input signals are the exhaust mass
exhaust mass. flow rate 𝑚̇𝑒𝑥ℎ and temperature 𝑇𝑒𝑥ℎ from the engine model, and the water . mass
flow rate mexh and temperature Texh from the engine model, and the water mass flow rate mwater and
flow rate 𝑚̇𝑤𝑎𝑡𝑒𝑟 and inlet temperature from the thermal storage tank primary circuit 𝑇𝑤𝑎𝑡𝑒𝑟 . The
inlet temperature from the thermal storage tank primary circuit Twater . The heat exchanger output
heat exchanger output signals are the two outlet temperatures.
signals are the two outlet temperatures.

Saturation
Limits

- Pmech N
Nset  vp Engine Shaft
Controller Model Dynamics

m exh Texh Pload

Electric
Machine
exchanger model is connected to and receives input signals from the engine model, as well as from
the primary circuit of a thermal storage tank model. The heat exchanger model input signals are the
exhaust mass flow rate 𝑚̇𝑒𝑥ℎ and temperature 𝑇𝑒𝑥ℎ from the engine model, and the water mass
flow rate 𝑚̇𝑤𝑎𝑡𝑒𝑟 and inlet temperature from the thermal storage tank primary circuit 𝑇𝑤𝑎𝑡𝑒𝑟 . The
Energies
heat 2017, 10, 1948
exchanger output signals are the two outlet temperatures. 7 of 14

Saturation
Limits

- Pmech N
Nset  vp Engine Shaft
Controller Model Dynamics

m exh Texh Pload

Electric
Machine
Twater Exh, Heat Model
Primary Circuit of
Heat Storage Tank
m water Exchanger
Model
Electrical
Pel
Load
Twater_ out Texh_ out Profile

Figure 5. Block diagram of the use of the engine model with a simple exhaust heat recovery system.
Figure 5. Block diagram of the use of the engine model with a simple exhaust heat recovery system.

In the simulation
In simulationstudy,
study,thethe
engine speed
engine was was
speed kept kept
constant with varying
constant electrical
with varying load demand
electrical load
for the duration
demand of a day (24
for the duration of ah)day
at a(24
simulation step of 1 s. step
h) at a simulation The electrical
of 1 s. The load profile load
electrical of a mid-terraced
profile of a
house during house
mid-terraced a typical January
during day, asJanuary
a typical found inday,
[24],as
is used
found asin
the[24],
system electrical
is used as theload. Following
system the
electrical
simulation, the plots of the different estimated power components flowing through
load. Following the simulation, the plots of the different estimated power components flowing the engine model
and the the
through system electrical
engine modelload
and profile can be
the system seen inload
electrical Figure 6. The
profile cantime intervals
be seen during
in Figure which
6. The all
time
of the components equal 0, correspond to the system that is being switched
intervals during which all of the components equal 0, correspond to the system that is being off. The distribution of
energy flow rate between the modelled energy components can be observed to vary with load as the
distance between each curve is not proportional to the magnitude of the heat input curve along the
simulation duration. This behaviour is particularly observable when the mechanical power and rate
of the exhaust heat flow plots are compared. While the rate of exhaust heat flow for low electrical
loads has a noticeably higher magnitude than the produced mechanical power, for higher loads the
rate of exhaust heat flow and power output magnitudes are in close proximity. This behaviour is
not surprising as it reflects the higher engine conversion efficiency that is usually observed at higher
engine loads. Due to a constant speed operation, the calculated rate of heat generated from friction
remains constant. Similarly, the effects of load fluctuation on the predicted exhaust temperature can be
observed on the plot of Figure 7 where the exhaust temperature predicted by the system simulation
as discussed above is plotted against time, while the electrical load profile is located on the same
graph and measured by the right y-axis. Again, the model is found to calculate exhaust temperature
values that come to a general agreement with the measured values and whose behaviour follows
the observations that were made on the experimental data showing higher loads leading to higher
predicted temperatures, and vice versa.
fluctuation on
fluctuation on the
the predicted
predicted exhaust
exhaust temperature
temperature can
can be
be observed
observed onon the
the plot
plot of
of Figure
Figure 77 where
where the
the
exhaust temperature predicted by the system simulation as discussed above is plotted
exhaust temperature predicted by the system simulation as discussed above is plotted against time, against time,
while the
while the electrical
electrical load
load profile
profile is
is located
located on
on the
the same
same graph
graph and
and measured
measured by by the
the right
right y-axis.
y-axis.
Again, the model is found to calculate exhaust temperature values that come to a general
Again, the model is found to calculate exhaust temperature values that come to a general agreement agreement
with the
with the
Energies measured
measured
2017, values and
10, 1948 values and whose
whose behaviour
behaviour follows
follows the
the observations
observations that
that were
were made
made onon the
the
8 of 14
experimental data
experimental data showing
showing higher
higher loads
loads leading
leading to
to higher
higher predicted
predicted temperatures,
temperatures, andand vice
vice versa.
versa.

55 Heat Input
Input
Heat
Exhaust Heat
Exhaust Heat
ConvectionHeat
Convection Heat
44 FrictionHeat
Friction Heat
MechanicalPow
Mechanical Power
er
ElectricalLoad
Electrical Load

33
(kW)
Power(kW)
Power

22

11

00 0
0 55 10
10 15
15 20
20
Time (hrs)
Time (hrs)

Figure 6.
Figure 6. Simulated
Simulated power
Simulated power component
power component distribution
component distribution vs.
vs. time
time as
time as predicted
as predicted by
predicted by the
the engine
engine model.
model.

800
800 11

600
600
Temperature oCC

(kW)
o

Load(kW)
ExhaustTemperature

ElectricalLoad
400
400 0.5
0.5

Electrical
Exhaust

200
200

Exhaust Temperature
Exhaust Temperature
Electrical Load
Electrical Load
00 0 00
0 55 10
10 15
15 20
20
Time (hrs)
Time (hrs)

Figure 7. Simulated
Figure Simulated system electrical
electrical load and
and engine exhaust
exhaust temperature plotted
plotted against time.
time.
Figure 7.
7. Simulated system
system electrical load
load and engine
engine exhaust temperature
temperature plotted against
against time.

By inspecting
By inspecting the
the profile
profile of
of the
the heat
heat recovery
recovery efficiency
efficiency curve
curve of
of Figure
Figure 8,
8, itit can
can be
be observed
observed that
that
the By inspecting
model behavestheasprofile of the being
expected, heat recovery
affectedefficiency
by the curveconditions
inlet of Figure 8,of it canthe be observed
heat that
exchanger
the model behaves as expected, being affected by the inlet conditions of the heat exchanger
the model behaves
subsystem. It can as seen
can be
be expected, being
that the
the affected by the
instantaneous heatinlet conditions
recovery of the tends
efficiency heat exchanger
to be subsystem.
be higher
higher for low
low
subsystem. It seen that instantaneous heat recovery efficiency tends to for
It can be seen that the instantaneous heat recovery efficiency tends to be higher for low load than
for high load conditions. A maximum heat recovery efficiency of 0.95 has been observed for a low
electrical load of 0.3 kW to 0.35 kW. On the other hand, a minimum heat recovery efficiency of 0.89
is observed for a high electrical load of 0.83 kW due to an increase in exhaust temperature and mass
flow rate. Thus, the minimum heat recovery efficiency is 6.3% lower than the observed maximum heat
recovery efficiency. Depending on model specifications, the inclusion of this level of accuracy may
be required by the heat recovery model and under these circumstances; an engine model layout as
developed in the current document can be the solution in supplying a heat exchanger model with the
necessary inputs.
mass flow rate. Thus, the minimum heat recovery efficiency is 6.3% lower than the observed
maximum heat recovery efficiency. Depending on model specifications, the inclusion of this level of
accuracy may be required by the heat recovery model and under these circumstances; an engine
model layout as developed in the current document can be the solution in supplying a heat
exchanger
Energies 2017, model
10, 1948 with the necessary inputs. 9 of 14

Figure 8. Plot of combined heat and power (CHP) Electrical Load and Exhaust Heat recovery efficiency
Figure 8. Plot of combined heat and power (CHP) Electrical Load and Exhaust Heat recovery
curves vs. time.
efficiency curves vs. time.

The engine model will be validated in terms of the rate of fuel consumption, as well as the
observed exhaust temperature for meeting a certain certain electrical
electrical demand
demand profile.
profile. As can be seen in
Figure 9, the plotted line of the simulated chemical power inlet exhibits a shape and magnitude that
is very similar to the the experimentally
experimentally obtained line (scaled down to 80 cc). cc). The plot of thethe simulated
simulated
chemical power inlet remains below the experimental experimental curve
curve for the complete
complete duration
duration of of the
the test.
test.
One may
mayobserve
observeininFigureFigure1010that thethe
that relative error
relative of the
error of model is rather
the model low, ranging
is rather between
low, ranging −0.8%
between
and
−0.8%−1.4% with the
and −1.4% with contour of the of
the contour line
thecorresponding to changes
line corresponding in engine
to changes load. load.
in engine
Similarly, the the degree
degree to to which
which the
the developed
developed model
model predicts the exhaust
exhaust temperature may be
observed in Figure 11, where the exhaust temperature profile recorded during the drive cycle phase
of the engine test, and the exhaust temperature predicted by the model for the same load are plotted
against time. The difference between the two lines ranges between 33 ◦℃ ◦ C, with the simulated
and 55 ℃,
C and
line being
being above
above the themeasured
measuredtemperature
temperaturecurve curvethroughout
throughoutthe theduration
duration of of
thethe
test. A difference
test. A difference of
◦ ◦
5℃
5of C for for
temperatures
temperatures positioned
positionedaround
around 700700
thethe ℃ mark
C mark translate to ato
translate relative error
a relative of less
error thanthan
of less 1%
◦ C is considered.
when
1% whena reference
a point
reference
Energies 2017, 10, 1948
of
point25of 25 ℃ is considered. 9 of 14

Figure 9. Plots of the simulated and tested heat inlet rate as well as the electrical load vs. time.
Figure 9. Plots of the simulated and tested heat inlet rate as well as the electrical load vs. time.

-0.7
-0.8
e Error (%)

-0.9
-1
Energies 2017, 10, 1948 10 of 14
Figure 9. Plots of the simulated and tested heat inlet rate as well as the electrical load vs. time.
Figure 9. Plots of the simulated and tested heat inlet rate as well as the electrical load vs. time.

-0.7
-0.8
(%) (%)
-0.7
-0.9
-0.8
ErrorError

-1
-0.9
Relative

-1.1
-1
Relative

-1.2
-1.1
-1.3
-1.2
-1.3
12 13 14 15 16 17 18
Time (hrs)
12 13 14 15 16 17 18
Time (hrs)
Figure 10. Plot of the relative error of the estimated fuel input by the model vs. time.
Figure 10. Plot of the relative error of the estimated fuel input by the model vs. time.

Figure 11. Plots of the simulated and the measured exhaust temperatures vs. time.
Figure 11. Plots of the simulated and the measured exhaust temperatures vs. time.
Figure 11.
The difference Plots ofthe
between theplots
simulated and the
of Figure 11measured exhaust temperatures
can be interpreted through the vs. observation
time. of the
relative error surface plot of Figure 12, in which, the relative error of the predicted
The difference between the plots of Figure 11 can be interpreted through the observation of the exhaust
temperature
Theerror
relative issurface
plotted
difference against
between
plot theengine
of plots of
Figure speed in and
Figure
12, 11 volumetric
can the
which, efficiency.
be interpreted
relative Itofcan
through
error bepredicted
the
the observed exhaust
observationthat
of the
relative error surface plot of Figure 12, in which, the relative error of the predicted exhaust
temperature is plotted against engine speed and volumetric efficiency. It can be observed that the temperature
is plotted against engine speed and volumetric efficiency. It can be observed that the relative exhaust
temperature error ranges between −2% and +2% throughout the complete operating range of the
engine. To ensure a fuel efficient operation, the engine was held at a constant synchronous speed of
3000 rpm, and the minimum electrical output of the generator was restricted to 40% of its maximum
electrical output. As the engine model was simulated over a map region that is characterized by a low
and positive relative exhaust temperature error, the relative error of the simulation remained positive
and below 1% throughout the whole duration of the tests.
 relative exhaust temperature error ranges between −2% and +2% throughout the complete operating
range of the engine. To ensure a fuel efficient operation, the engine was held at a constant
synchronous speed of 3000 rpm, and the minimum electrical output of the generator was restricted
to 40% of its maximum electrical output. As the engine model was simulated over a map region that
is2017,
Energies characterized
10, 1948 by a low and positive relative exhaust temperature error, the relative error of the
11 of 14
simulation remained positive and below 1% throughout the whole duration of the tests.

Figure
Figure 12. 12. Relative
Relative errorofofthe
error theexhaust
exhaust temperature
temperature predicted
predictedbyby
thethe
model.
model.

5. Design and Analysis

5. Design and Analysis
The observations that were made in Section 4 regarding the behaviour of the developed engine
The
modelobservations
come to anthat were made
agreement with in Sectionestablished
generally 4 regarding the behaviour
knowledge on theofbehaviour
the developed
of spark engine
modelignited
come internal
to an agreement
combustion with generally
engines, established
as well knowledge
as the engine on the
behaviour behaviour
recorded of spark
in the collectedignited
engine
internal data, the collection
combustion engines, as procedure of which
well as the engine is described
behaviourinrecorded
Section 3.in the collected engine data, the
The distribution of the different
collection procedure of which is described in Section 3. power components that are calculated by the engine model
exhibits a behaviour that is quantitatively and
The distribution of the different power components that are qualitatively very similar to that encountered
calculated by the engine in themodel
tested engine. As expected, the fraction of the inlet chemical power 𝑄̇𝑖𝑛 that is converted to
exhibits a behaviour that is quantitatively and qualitatively very similar to that encountered in the
mechanical power output (fuel conversion efficiency) at lower loads . is low when compared to the
testedfractions
engine.of As𝑄̇expected,
that end
the
up
fraction
to become
ofthe
theengine
inlet chemical
main waste
power
heat
Qin that is converted
components.
to mechanical
At higher loads, the
𝑖𝑛
power output (fuel
mechanical power conversion
accounts for efficiency) at lower loads
a greater proportion of theisinlet
lowchemical
when compared
power thantointhethe fractions
case of of
.
Qin that
lowend upand
loads, to become the engine
this translates main waste
to an increase in fuelheat components.
conversion At In
efficiency. higher loads,
addition, the the mechanical
proportion
powereachaccounts for acomponent
waste heat greater proportion
accounts for of the inlettochemical
relative power
the chemical thaninlet
power in the casewith
varies of low
load,loads,
and and
dependingtoonan
this translates the waste heat
increase recovery
in fuel configuration,
conversion this may
efficiency. cause a variation
In addition, in the system
the proportion each overall
waste heat
fuel utilization
component accounts efficiency as thetoengine
for relative speed and
the chemical load varies.
power inlet varies with load, and depending on the
The predicted exhaust temperature exhibits behaviour that is very close to the recorded profile.
waste heat recovery configuration, this may cause a variation in the system overall fuel utilization
As in the case of the tested engine, the predicted exhaust temperature is affected by both engine
efficiency as the engine speed and load varies.
speed and load with speed having a considerably stronger influence on the exhaust temperature
The predicted exhaust temperature exhibits behaviour that is very close to the recorded profile.
magnitude than engine load.
As in the case of theoftested
In terms engine,tothe
the degree predicted
which exhaust
the fitted model temperature
of Appendix is affected
A Tableby A1both engine
predicts thespeed
and load with speed having a considerably stronger influence on the exhaust
magnitude of engine measured outputs accurately, the model validation phase of Section 4 showed temperature magnitude
than engine load.
a high proximity of the simulation results to the measured data. The relative error of the exhaust
temperature
In terms of the fit prediction has been
degree to which thefound
fittedtomodel
be lessofthan 1% whenAsimulated
Appendix Table A1 under thethe
predicts engine drive of
magnitude
engine cycle. For theoutputs
measured same tests, the relative
accurately, the error
model of validation
the required chemical
phase power4inlet
of Section for the
showed sameproximity
a high load
profile remained between −0.7% and 1.4% throughout the whole simulated
of the simulation results to the measured data. The relative error of the exhaust temperature fit period. The strong
accuracy of the predicted values can be attributed to the high coefficients of covariance 𝑟 2 of the 5th
prediction has been found to be less than 1% when simulated under the engine drive cycle. For the
same tests, the relative error of the required chemical power inlet for the same load profile remained
between −0.7% and 1.4% throughout the whole simulated period. The strong accuracy of the predicted
values can be attributed to the high coefficients of covariance r2 of the 5th order fits of Texh and Q.R.,
which were carried out with the use of Matlab SFTOOL (R2013b, The MathWorks, Natick, MA, USA)
to be 0.9959 and 0.9881, respectively. Another potential factor that may have contributed to such
low relative errors is the fact that all of the tests were carried out on the same day, which assured
that the testing conditions were kept as constant and controlled as possible. While a test performed
during a different day of the year may give relative errors of a higher magnitude, the very low errors
encountered under testing indicate that the fitted model is adequate for the purpose of modelling
Energies 2017, 10, 1948 12 of 14

and simulating ICE based cogeneration systems, provided that no dramatic changes are made in the
engine running conditions, such as in the case of operating under a very low barometric pressure due
to high altitude.

6. Conclusions
A method to construct quasi-stationary SI ICE models for use as a subsystem in cogeneration
models is presented. The model includes the mechanical power and conversion efficiency aspect
of the energy flow, but also the various waste heat components and the exhaust mass flow rate
and temperature. Model scalability and connectivity were also considered as characteristics of
great importance.
In order to generate the model maps, a series of engine tests have been carried out. By inspecting
measured exhaust temperature plots for different engine loads and speeds, a strong dependence of the
exhaust temperature, on the engine speed, and the engine volumetric efficiency has been observed.
In addition, it is common knowledge that the heat exchanger waste heat recovery efficiency is heavily
dependent on fluid inlet conditions. For these reasons, an exhaust temperature model as a function of
engine speed and engine volumetric efficiency has been developed. Representing load with volumetric
efficiency and mapping the rate of flow of convective heat as a fraction of the rate of flow of the
exhaust heat ensures the scalability of the model. The rate of flow of convective heat as a fraction of
the rate of flow of the exhaust heat has been named heat ratio (Q.R.) for the ease of documentation.
Polynomial surface fits of the exhaust temperature and heat ratio as functions of engine speed and
volumetric efficiency of 5th order were performed and the resulting coefficients are included in this
document, thus providing the reader with a ready to run engine model. Although quasi-stationary
in nature, this layout, if necessary, allows for the incorporation of dynamic behaviour to some extent
into the various power components by means of using transfer functions with appropriately tuned
coefficients that are dedicated to a respective power component.
The resulting engine model was simulated in tandem with a map based heat exchanger model.
The engine model was found to be more than adequate in predicting the measured engine outputs,
and easy to connect to a higher level CHP model, thus proving the usefulness of the concept. It is
therefore safe to conclude that the developed model is successful in predicting the distribution of
all the power components for different speeds and engine loads accurately, while at the same time
being easy to scale, and to connect to peripheral components that operate in different energy domains.
The most important family of components that can easily be connected to the developed engine model
are models of heat recovery equipment whose inlet conditions affect their performance. Due to the
above, the developed engine modelling layout can be characterized as an attractive alternative for use
in cogeneration simulation applications.

Author Contributions: 60% of the work was carried out by Nikolaos Kalantzis, 20% of the work was carried out
by Antonios Pezouvanis, 20% of the work was carried out by Kambiz M. Ebrahimi.
Conflicts of Interest: The authors declare no conflicts of interest.

Nomenclature
Tamb Ambient atmospheric temperature
.
Qin Heat inlet rate from fuel to the engine
Pmech Mechanical power output
.
Qexh Exhaust heat rejection rate
.
Qconv Rate of heat rejection through convection from cylinder content to cylinder walls
.
mexh Exhaust mass flow rate
.
ma Air mass flow rate
.
mf Fuel mass flow rate
Texh Exhaust temperature
Energies 2017, 10, 1948 13 of 14

Nset Set engine speed

N Engine speed
Pload Mechanical power load acting on the shaft
ηvp Engine volumetric efficiency
Pamb Ambient pressure
AFRst Stoichiometric air to fuel ratio
.
V f _vap Volumetric flow rate of vaporized fuel at atmospheric conditions
υ f _vap Specific volume of vaporized fuel at atmospheric conditions
.
Va Volumetric flow of air at atmospheric conditions
υa Specific volume of air at atmospheric conditions
.
V sw Swept volume rate
Vd Engine displacement
.
mch Mass flow rate of charge
hexh Exhaust enthalpy under outlet temperature
hexh_amb Exhaust enthalpy under ambient temperature
Q.R. Heat ratio
LHV f Lower heating value of fuel
Twater Water inlet temperature (Heat exchanger)
.
mwater Water mass flow rate (Heat exchanger)
Twater_out Water outlet temperature (Heat exchanger)
Texh_out Exhaust outlet temperature (Heat exchanger)
Pel Electrical power load placed on the electric machine

Appendix A
Table A1. Coefficients of the 5th order polynomial surface fits on Texh and Q.R.

Texh Q.R.
p00 181 −15.26
p10 −0.01864 0.03765
p01 9717 −62.01
p20 −0.0002382 −2.308 × 10−5
p11 −2.664 −0.01654
p02 −4.732 × 104 359.6
p30 2.091 × 107 4.749 × 109
p21 0.0004299 4.275 × 10−5
p12 9.603 −0.2169
p03 1.034 × 105 −396.1
p40 −5.329 × 10−11 −1.954 × 10−13
p31 −4.319 × 10−8 −1.029 × 10−8
p22 −0.001427 1.043 × 10−6
p13 −8.814 0.3173
p04 −1.131 × 105 −19.82
p50 4.266 × 10−15 −2.644 × 10−17
p41 6.416 × 10−12 7.893 × 10−13
p32 3.797 × 10−8 1.077 × 10−9
p23 0.0008733 −5.83 × 10−6
p14 2.013 −0.1529
p05 4.992 × 104 176
R2 0.9959 0.9881

References
1. Onovwiona, H.; Urgusal, V. Residential cogeneration systems: Review of the current technology.
Renew. Sustain. Energy Rev. 2006, 10, 389–431. [CrossRef]
2. Onovwiona, H.I.; Urgusal, V.I.; Fung, A.S. Modelling of internal combustion engine based systems for
residential applications. Appl. Therm. Eng. 2007, 27, 848–861. [CrossRef]
3. Aliabadi, A.A.; Thomson, M.J.; Wallace, J.S. Efficiency analysis of natural gas residential micro-cogeneration
systems. Energy Fuels 2010, 24, 1704–1710. [CrossRef]
Energies 2017, 10, 1948 14 of 14

4. Angrisani, G.; Roselli, C.; Sasso, M. Distributed microtrigeneration systems. Prog. Energy Combust. Sci. 2012,
38, 502–521. [CrossRef]
5. Canova, A.; Cavallero, C.; Freschi, F.; Giaccone, L.; Repetto, M.; Tartaglia, M. Comparative economical
analysis of a small scale trigenerative plant: A case study. In Proceedings of the Industry Applications
Conference, 2007 42nd IAS Annual Meeting, New Orleans, LA, USA, 23–27 September 2007.
6. Sicre, B.; Buhring, A.; Platzer, B.; Hoffmann, K.H. Energy and cost assessment of Micro-CHP plants in high
performance residential buildings. In Proceedings of the ECOS 2005, the 18th International Conference
on Efficiency, Cost, Optimization, Simulation, and Environmental Impact of Energy Systems, Trondheim,
Norway, 20–22 June 2005.
7. Caresana, F.; Brandoni, C.; Feliciotti, P.; Bartolini, C.M. Energy and economic analysis of an ICE-based
variable speed-operated micro-cogenerator. Appl. Energy 2011, 88, 659–671. [CrossRef]
8. Gluesenkamp, K.; Hwang, Y.; Radermacher, R. High efficiency micro trigeneration systems. Appl. Therm.
Eng. 2013, 50, 1480–1486. [CrossRef]
9. Chamra, L.M.; Mago, P.J.; Stone, N.; Oliver, J. Micro-CHP (Cooling, Heating, and Power): Not just scaled
down CHP. In Proceedings of the ASME 2006 Power Conference, Atlanta, GA, USA, 2–4 May 2006.
10. Barbieri, E.S.; Spina, P.R.; Venturini, M. Analysis of innovative micro-CHP systems to meet household energy
demands. Appl. Energy 2012, 97, 723–733. [CrossRef]
11. De Paepe, M.; D’Herdt, P.; Mertens, D. Micro-CHP systems for residential applications. Energy Convers.
Manag. 2006, 47, 3435–3446. [CrossRef]
12. Kong, X.Q.; Wang, R.Z.; Wu, J.Y.; Huang, X.H.; Huangfu, Y.; Wu, D.W.; Xu, Y.X. Experimental investigation
of a micro-combined cooling, heating and power system driven by a gas engine. Int. J. Refrig. 2005, 28,
977–987. [CrossRef]
13. Silveiraa, J.L.; Walter, A.C.d.S.; Luengo, C.A. A case study of compact cogeneration using various fuels. Fuel
1997, 76, 447–451. [CrossRef]
14. Voorspools, K.R.; D’haeseleer, W.D. The evaluation of small cogeneration for residential heating. Int. J.
Energy Res. 2002, 26, 1175–1190. [CrossRef]
15. Kelly, N.J.; Clarke, J.A.; Ferguson, A.; Burt, A. Developing and testing a generic micro-combined heat and
power model for simulations of dwellings and highly distributed power systems. Proc. Inst. Mech. Eng. Part
A J. Power Energy Impact Factor Inf. 2008, 222, 685–695. [CrossRef]
16. Aussant, C.D.; Fung, A.S.; Urgusal, I.V.; Taherian, H. Residential application of internal combustion engine
based cogeneration in cold climate-Canada. Energy Build. 2009, 41, 1288–1298. [CrossRef]
17. Cho, H.; Luck, R.; Eskioglu, S.D.; Chamra, L.M. Cost-optimized real-time operation of CHP systems.
Energy Build. 2009, 41, 445–451. [CrossRef]
18. Zavala, J.; Sanketi, P.R.; Wilcutts, M.; Kaga, T.; Hedrick, J.K. Simplified models of engine HC emissions,
exhaust temperature and catalyst temperature for automotive coldstart. In Proceedings of the Fifth IFAC
Symposium for Advances in Automotive Control, Pajaro Dunes/Seascap, CA, USA, 20–22 August 2007.
19. Heß, T.; Seifert, J.; Schegner, P. Comparison of static and dynamic simulation for combined heat and power
micro-units. In Proceedings of the 17th Power Systems Computation Conference, Stockholm, Sweden,
22–26 August 2011.
20. Ferguson, C.R.; Kirkpatrick, A.T. Internal Combustion Engines: Applied Thermosciences; John Wiley & Sons Inc.:
New York, NY, USA, 2001.
21. Heywood, J. Internal Combustion Engine Fundamentals; Mcgraw-Hill: London, UK, 1989.
22. Pulkrabek, W. Engineering Fundamentals of the Internal Combustion Engine; Prentice Hall: Upper Saddle River,
NJ, USA, 2003.
23. Eriksson, L. Mean value models for exhaust system temperatures. In SAE 2002 World Congress & Exhibition;
SAE International: Detroit, MI, USA, 2002.
24. Cambridge Architectural Research, HES 24-Hour Chooser. 2011. Available online: https://www.hightail.
com/download/WFJWWWV0bThCSm9pR01UQw (accessed on 27 November 2015).

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).