Académique Documents
Professionnel Documents
Culture Documents
Shinya Obara
123
Shinya Obara, PhD
Department of Mechanical Engineering
Tomakomai National College of Technology
Nishikioka 443, Tomakomai
Hokkaido, 0591275
Japan
v
vi Contents
References......................................................................................................... 245
Act: If action
Act_FC: Each fuel cell operation
a: Time constant in a primary delay system transfer function [s]
b: Constant part in a primary delay system transfer function
C: Cost [US $]
C: Capacity [kW]
C' : Maximum load [kW]
C Boiler : Boiler fuel unit cost [US $/mol]
C Me : Methanol unit cost [US $/mol]
C Utility,E : Commercial power unit cost of electricity [US $/kJ]
C Utility,H : Purchase unit cost of thermal energy [US $/kJ]
D: Control variable of derivative action
Dc : Outside diameter of the heat insulating material [m]
Di : Inside diameter of the hot water piping [m]
Do : Outside diameter of the hot water piping [m]
Dp : Heat-insulating-mold outside diameter of hot water piping [m]
E , E : Power [kW]
EC : Capacity of power generation [kW]
EF : Power of an inverter outlet [kW]
Eg : Production of electricity [kW]
Errday : Difference of the power generation of the system and demand
on a representative day [W]
ErrN : Error of output data and training signal of NN
E sys , E SYS : Electric power of an inverter outlet [W]
xiii
xiv Notation
T: Temperature
t: Sampling time [Hour]
t ge : Operation finish time of a fuel cell [o'clock]
t gs : Startup time of a fuel cell [o'clock]
t : Sampling interval [s]
U: Capacity of facility
u: Input [kW]
u: Power demand [W]
u: Power load of a micro-grid [W]
V: The number of auxiliary machinery with electricity consumption
V , VS : Thermal storage volume [m3]
WB : Town gas calorific valve supplied to a burner [kW]
WH : Calorific valve of the hydrogen included in reformed gas [kW]
Wp : Water head of hot water circulation pump [m]
WR : Town gas calorific valve supplied to a steam reforming [kW]
w n,j, m
i : Weight of the neuron i of layer m, and the network of neuron j
of layer n
w: Weight before modification
w : The amount of modifications of weight
wo : Weight of the component of the objective function
x zy : Input to the neuron y of layer z
Yc : Equipment cost [US $]
Yd : Operation cost [US $]
Yf : Equipment installation cost [US $]
yj: Training signal
Greek Symbols
Subscripts
a: Cylinder
atm : Open air
B , cb : Burner
BE: Bioethanol
BL: Boiler
bh : Back-up heat source
bo : Boiler
bd : Number of houses
bw : Blower
CC: Commercial power
CL: Reformed gas cooler
CS: Cell stack
Day : Representative day
DEG : Diesel engine generator
E: City gas engine generator (NEG)
Er : Reforming gas
Ex : Exhaust gas of SEG
Ey : Cooling water of SEG
Ez : Heat radiation of SEG
EX: Off gas of the PEM-FC
el : Electrolyzer
F, f : Fuel cell
fca : Fuel cell generated with air
fco : Fuel cell generated with oxygen
hp : Heat pump
hw : Heat release of hot water piping
in : Input
l: Building model with PEM-FC system and NEG
m: Building model with NEG
m: Number of each building
n: Building model without PEM-FC system and NEG
Me: Methanol
xviii Notation
Equipment Symbols
B/L: Boiler
CB: Catalyst burner unit
CGS: Cogeneration system
C/O, CO: Carbon monoxide oxidization unit
DA: DA/AC converter
D C/A C: DC-AC converter
F/C: Fuel cell
FS: Fuel cell stack unit
G /T: Generator
HC: Hydrogen cylinder
H CGS: Hybrid cogeneration
Heater: Electric heater
I/T: Inverter
NEG : Gas engine generator
OU: CO oxidation unit
P/I: Interconnection device
Rad: Radiator
RE, R/M: Reforming unit
Notation xix
1.1 Introduction
The solid high-polymer-film-type fuel cell (PEM-FC) system is used as the power
supply equipment for transportation and replaces an internal combustion engine. A
reduction of the environmental load is expected through the cogeneration systems
(CGS) use of the PEM-FC system as a distributed power supply to individual
houses, apartments, and so forth [13]. The growing use of distributed power
systems, such as fuel cells, the reduction of power-transmission losses, and an
increase of waste heat recovery are expected. Therefore, the reduction of carbon-
dioxide emission is also expected as compared to conventional energy supply
methods using commercial electric power. At present, a decisive method of hy-
drogen supply and storage in houses is not proposed. Fuel cell CGS of the reform-
ing type which obtains hydrogen from natural gas and methanol is promising. In
order to provide high efficiency and to perform a catalyst reaction by methanol
steam reforming, it is necessary to manage the heat quantity [47]. Furthermore, if
approximately 3% carbon monoxide is contained in the gas after the steam reform-
ing reaction and is supplied to a fuel cell, poisoning of the electrode catalyst will
occur, and power generation will be difficult. Thus, after preparing the shift unit
and CO oxidation unit in the process after the reforming reaction, the CO concen-
tration is approximately 10 ppm. Since the shift reaction, CO oxidation reactions,
and the reforming reaction are performed using a catalyst, each reaction system is
controlled in a suitable temperature range. Moreover, the evaporation and reform-
ing reactions of methanol fuel (methanol solution) are endothermic, while other
reactions are exothermic, such as cell reactions. Under operation with unsuitable
temperature management of each reaction system, the by-product generating rates
of reacted and unreacted substances increase in the process of catalyst reaction.
The fuel cell system with a reformer is a complicated heat system; considering that
the electric energy or heat demand of individual houses change, one should pre-
pare energy buffers, such as battery and thermal storage, which should make the
chemical reaction stable.
This study considers the optimal operation when minimizing fuel consumption
(making it equal to the operation cost) into an objective function assuming the
introduction of the PEM-FC cogeneration system installation of a methanol steam
reformer for individual houses with energy load fluctuation. Furthermore, a simu-
lation that introduces this system into the energy consumption pattern of individ-
ual houses in Sapporo in Japan is performed, and the operation of the system and
the operation cost are compared with those of conventional energy systems.
Figure 1.1 shows a schematic diagram of the fuel cell CGS for the individual
houses investigated in this study. In this chapter, the evaporation unit 2, reformer
unit 3, shifter unit 4, and CO oxidization unit 5 of Fig. 1.1 are collectively called a
reformer. In the evaporator 2, methanol fuel is evaporated by the combustion gas
supplied from the catalyst burner 12. The methanol fuel reformed by reformer unit
3 and a reforming gas with much hydrogen are generated. The shifter unit 4 and
CO oxidation unit 5 are formed on removing carbon monoxide of the reforming
gas. The electric power generated by the fuel cell stack 1 is changed into an alter-
nate current by the DC/AC converter 11. The electric power supplied to a demand
side is chosen to be a commercial electric power system or the fuel cell system
with a change over switch 6. Fuel cell heat exhaust is stored in the thermal storage
tank 10. The electric heater 9 is installed in the thermal storage tank, and the heat
medium in the thermal storage tank can be heated by the heater 9 and the back-up
boiler 8. The energy generation cost by CGS is compared with commercial power
(utility) cost, the cheaper one is chosen, and power supply is changed with the
high-speed-changeover switch 6. The electric power supplied from the system is
outputted in CGS or the utility. In addition, the power generation capacity in CGS
is determined to be greater than the quantity of the maximum electricity demand
assumed. The pressure of the reforming gas is made slightly higher than the at-
mospheric pressure. The off-gas of the catalyst burner 12 is supplied to the re-
former unit 3, and the temperature of reforming reaction is controlled. The heat
supplied to a user is based on values other than an endothermic reaction by the
evaporation unit 2 and reforming unit 3 among the combustion gas of the catalyst
burner 12, waste thermal energy of the fuel cell stack 1, and the back-up boiler 8.
All the waste thermal energy is brought together in the thermal storage tank 10,
and heat exchange is carried out with tap water. If the waste thermal energy ex-
ceeds the thermal storage capacity, a part will be released to the atmosphere from
the radiator 7. The CGS operation cost in a sampling time of electric energy and
thermal energy is calculated from the quantity of methanol fuel consumed by the
reformer operation and the back-up boiler operation. The total cost of the energy
supply is added and estimated to be higher than the fuel costs and the utility en-
ergy supply costs.
Reaction equations in the reformer and fuel cell stack are shown in Fig. 1.2.
With respect to evaporation and reforming, they are endothermic reactions, and
CO shift, CO oxidization, and cell reaction are exothermic reactions. Figure 1.3
shows the target temperature region in each reaction system, and inputs and out-
puts of heat. The exothermic heat of the CO oxidation reaction device is
283 kJ/mol by combustion of carbon monoxide and hydrogen, and the exothermic
heat of the fuel cell stack is 265 kJ/mol by the fuel cell reaction. Reforming is
performed after having increased the temperature of methanol steam to about
570 K. After that, the target temperature in each reaction system falls in step. The
off-gas temperature of the fuel cell exit is set to 340360 K.
Figure 1.4 shows the energy loss and mass loss with each unit. These loss values,
however, do not include heat release from the case, duct, and pipe. The off-gas of
the fuel cell stack is given as the heat source of the evaporator and reforming
units. When requiring an additional quantity of heat, methanol burns with a cata-
lyst burner. Heat is radiated after its generation in the shifter and CO oxidization
units. The rate of conversion from the methanol in the reforming unit to hydrogen
is expressed as RE , and the back-up boiler efficiency is expressed as Boiler . A
blower for oxygen supply is installed in the catalyst burner, the CO oxidization
unit, and the fuel cell. The above-mentioned blower, methanol fuel pump, and
methanol pump for catalyst combustion, system controller, and back-up boiler
consume electric energy. An estimation as regards the loss of electric energy by
hydrogen combustion in the CO oxidization unit ( O 2 ), joule loss ( R FS ) by
PEM-FC, the concentration fault voltage in the electric pole ( Air , Me ), and the
DC/AC converter ( 1 DA ) must be made.
1.2 System Configuration 5
Fig. 1.4 Energy loss and mass loss with each unit
Figure 1.5 shows profiles of the hydrogen generation quantity, the volume rate of
carbon monoxide in the quantity of reforming gas in the shift unit exit, and the
volume rate of hydrogen conversion. The methanol inversion rate shown in the
figure is the rate at which the methanol fuel supplied to the system finishes the
steam reforming reaction. This model is applied for the actual measurement using
an experimentally manufactured cell stack. S/C expresses the molar ratio of water
to methanol and is 1.4 in this model. The average data in the trial production
PEM-FC system for 3 kW power generation is shown in Fig. 1.5. In this chapter,
the following assumptions are made regarding system operation.
(a) The temperature of the evaporation unit, reforming unit, and the shift unit can
be controlled in a suitable range as shown in Fig. 1.3.
(b) Although a catalyst reaction is dependent on the space velocity (SV) value,
CO concentration in the shifter unit exit is always less than 1.0 volume %.
(c) The temperature range and SV value in the CO oxidization unit is appropri-
ately managed so that CO concentration in the fuel cell stack entrance may be
set to 10 ppm or less. By flow rate of reforming gas, the number of CO oxidi-
zation catalyst paths where reforming gas flows is changed. Moreover, the
amount of cooling of the catalyst paths is controlled by a blower to react in the
regal temperature range.
6 1 A Small-scale Fuel Cell Cogeneration System Considering Partial Load
The output characteristic model of a single cell is shown in Fig. 1.6. For the
single cell output characteristic the model depends on the CO concentration in the
reforming gas. The sampling time t k (Eqs. (1.1) and (1.2)) divides one day into N
intervals (here N = 24 ), and is defined by the following equations by setting the
starting time to t r . Here k = 0, " N 1 . For the quantity of methanol consumption
at the system from t k to t , the amount of supply to the evaporation unit is
FVP,t k , the amount of supply to the catalyst burner unit is FCB, t k , and the amount
of supply to the back-up boiler unit is FBoiler,
*
t k . Equation (1.3) shows that the
CGS operation is cheaper than the utility energy supply. E Utility, t k expresses utili-
ties power and FBoiler,t k is a flow rate of a boiler. In terms of energy supply cost it
is cheaper to supply the same amount of electric power and thermal energy by the
system of Fig. 1.1 rather than supplying from the utility when Eq. (1.3) is needed.
t k = t r + k t (1.1)
1.2 System Configuration 7
t = 24 /N (1.2)
The following equation defines the energy supply cost of CGS in this chapter.
C CGS, t k = (C Utility,E E Utility,t k + C Me FVP, t k
(1.4)
+ C Me FCB, t k + C Boiler FBoiler,t k ) t
The C Boiler term of Eqs. (1.3) and (1.4) shows the boiler fuel unit price of
CGS, and C Utility,H shows the boiler fuel unit price of a utility. In the analysis
example described later, the kerosene fuel unit price is used for C Utility,H for
methanol or the kerosene fuel unit price for C Boiler .
The boiler used for utility and CGS backed heating has the same unit, and
therefore, FBoiler,
*
tk
is replaced by FBoiler,t k and C Utility,H is replaced with by
C Boiler of Eq. (1.4).
The boiler efficiency differs by the input and output temperatures of medium, hot
water supply quantity, and so forth. When the boiler efficiency falls below the
total efficiency of the fuel cell CGS, a part of the generated electric energy is con-
verted into heat in an electric heater.
Figure 1.7 shows the model of the relationship between the methanol fuel supply
quantity to the reformer and the energy output. The dashed line denotes electric
energy output and the solid line expresses exhaust heat output produced by the cell
reaction and the internal resistance of the fuel cell stack. In this model, the maxi-
mum output is approximately 3 kW with the methanol quantity of supply of
0.032 mol/s. The electric energy load E needs is in a smaller range than the greatest
outputting point of the fuel cell stack and sets this operating point to Pa . The heat-
ing of H Boiler is required by the boiler in order to satisfy the energy demand quan-
tity ( H needs ) at the point Pa . When changing the electric energy E into thermal
energy in an electric heater, the driving point Pa moves the maximum output point
Pb . H *Boiler shows the heating quantity of the back-up boiler, and h shows the
thermal output differential between points Pa and Pb . Therefore, ( h + E) is the
decrease in heating quantity of the back-up boiler. H Fuel is the increase in fuel
supply of the fuel cell system with the previous driving point shift. In the case of
electrical thermal exchange, a favorable condition is defined by following equation.
H Fuel + H *Boiler / * < H Boiler / (1.5)
Here, * denotes the back-up boiler efficiency with boiler heating H *Boiler .
8 1 A Small-scale Fuel Cell Cogeneration System Considering Partial Load
The following equations denote the energy balance of the system shown in
Fig. 1.1.
(a) Electric energy:
E FS, t k + E Utility,t k = E DA,t k + E Heater, t k + E DA, t k + E Boiler,t k
Pump Blower
(1.6)
+ E Controller, t k + E Pump,n,t k + E Blower,m,t k
n =1 m =1
Here,
H REFORMER, tk = H VP, t k + H RE, t k + H SH, t k + H CO, t k + H FS, t k (1.8)
H DA,t k = E FS (1 DA ) (1.14)
The left-hand sides of Eqs. (1.6) and (1.7) denote the input energy of the sys-
tem, and the right-hand sides of both equations denote the output energy.
The electric energy output of the system is given as E DA,t k in Eq. (1.6), and
thermal energy output is given as H System,t k in Eq. (1.7). Figure 1.8 shows the
CGS energy output diagram; this illustration is made up of E DA,t k and H System,t k .
The running fuel cost is estimated as shown in Fig. 1.8. F1 to F3 are operation
areas of the system stated in the section Load Fluctuation and Fuel Consump-
tion, and (a) to (d) express the routes of load fluctuation.
The objective function J C is given by Eq. (1.15) as the minimization of the en-
ergy cost for one day. A random method is used in the search analytical algorithm
of the optimum operation plan of the system which made Eq. (1.15) an objective
function.
N 1
JC = C CGS, t k
(1.15)
k =0
The maximum capacity of the electric heater H Heater,max is divided into L pieces,
and the quantity of the loads of the heater is determined based on
10 1 A Small-scale Fuel Cell Cogeneration System Considering Partial Load
Equations (1.17) and (1.18) are made to fit the curve of Fig. 1.7. Electric energy
output and thermal energy output are divided into approximately two or more
domains of the quantity of fuel supply, and the approximation equation in each
domain is created.
Electric energy output:
EFS,t k = a FS,1 (FVP,t k )2 + a FS,2 (FVP,t k ) + a FS,3 (1.17)
The fuel quantity of supply FVP, t k is calculated by giving EFS,t k ,l of Eq. (1.16)
to E FS, t k of Eq. (1.17), and the amount of heat outputs H FS, t k is further calcu-
lated from Eq. (1.18). The quantity of electric energy EFS, t k ,l and the backed
heating quantity H Boiler,t k of the boiler can be calculated by substituting the elec-
tric energy output E FS, t k of a fuel cell stack, the heat output H FS, t k , and other
factors into Eqs. (1.6), (1.7), and (1.8), and solving for H Boiler,t k . The calculation
method of the quantity of thermal storage H St, t k is described later. The fuel quan-
tity of supply FBoiler,t k is calculated by substituting H Boiler,t k in the expression of
the thermal energy output the relationship shown in Fig. 1.9. FBoiler,t k and FVP, t k
are substituted in Eq. (1.4) and the operation cost of CGS is estimated. The above
calculation is performed under the restricted conditions of Eq. (1.16), and it de-
termines the minimum operation cost for the sampling time t k .
Fig. 1.9 Relation of fuel supply and electric power consumption of the boiler
1.4 Energy Output Characteristics 11
Thermal storage will be needed if the heat output of a system exceeds the thermal
energy demand. Equations (1.20) and (1.21) are restricted conditions, and use the
quantity of the maximum thermal energy storage capacity, and the maximum
temperature of the thermal medium. When thermal energy input exceeds the quan-
tity of the maximum thermal storage capacity, heat is radiated from a radiator in
the thermal energy exceeding maximum. Equation (1.22) defines the quantity of
thermal energy storage, using specific heat Cp and thermal storage loss St includ-
ing heat release loss.
0 SSt, t k SSt, max (1.20)
Figure 1.8 shows the result of calculating the relationship of electric energy and
thermal energy output when making the calorific value of fuel supply to CGS into
a parameter. Table 1.1 lists the specifications of the fuel cell system assumed in
this chapter, and Table 1.2 lists the efficiency of each reaction unit. Each value of
Table 1.2 was decided from measurement results of each device. Figure 1.8 was
obtained by the introduction of the value of characteristic of fuel cell CGS of
Fig. 1.7 and Table 1.2 for the optimization analysis. Area A of the lower part of
Fig. 1.8 (hatching part) shows an independent energy supply operation by the fuel
cell system. Area B (solid line) shows the operation when incorporating back-up
heating by the boiler to the fuel cell system. Area C (dashed line) shows the opera-
tion when incorporating an electric heater in area B. The fuel consumption in area
A is determined only by the quantity of electric energy output; in the large region
of thermal load, the fuel consumption of the boiler is added to the fuel consump-
tion of the fuel cell, because the operation of the boiler is applied to the operation
12 1 A Small-scale Fuel Cell Cogeneration System Considering Partial Load
of the fuel cell system. In area A, the fuel supply for the same quantity of electric
energy generation is minimum compared with other cases of operation. When the
driving point at the electric heater is moved (area C), there is area in which ther-
mal energy output increases also by the same fuel quantity supply compared with
area B (this area has a dashed line above a solid line). This domain is in the small
portion of electric energy output.
Figure 1.9 shows the model of the fuel quantity of supply and electric energy
consumption of the back-up boiler (methanol fuel is assumed). The back-up boiler
output is 55 kW at the maximum, it is assumed it uses space heating and a hot-
water supply, and that the hot-water temperature is 340 K or more and the thermal
efficiency is 85%. Heat release loss of the case or duct is set to 10%, and the util-
ity electric energy supply E utility,t k and thermal storage H St, t k are taken as zero.
CO concentration at the entrance of the fuel cell is assumed to be 10 ppm or less.
In the analysis, output characteristics of the fuel cell are improved, when the CO
concentration falls below this value, and these characteristics were calculated as
being constant.
in which backed heating is possible. Load fluctuations of areas F1, F2, and F3 were
set up, respectively, as 30%, 50%, and 70% for thermal energy demand, and
60%, 60%, and 70% for electric energy demand, respectively. These load
fluctuation ranges were determined supposing power consumption of home elec-
tric appliances, and thermal energy consumption of hot-water supply and space
heating. As shown in Fig. 1.7, if the driving point passes over the maximum power
point Pb , the efficiency of the fuel cell stack will fall. Then, each of the load fluc-
tuation areas of F1F3 shall be in the range of electric energy lower than Pb .
Figure 1.10 in the system which combined PEM-FC-CGS and the methanol
boiler shows the results of measuring the quantity of fuel consumed at each point
of the F1F3 areas. (a) to (d) is the route of load fluctuations described in Fig. 1.8.
The pattern Q1 shown in the figure is the result of calculating for the operation
system which performs neither heat conversion of electric energy, nor change in
utility (commercial power) energy supply. Although Q2 performs heat conversion
of electric energy, a change in a utility is not performed. Q3 is result of the opera-
tion system which performs heat conversion of electric energy, and change in the
utility. In this Chapter, the electric energy unit price was set to 6.03 c$/MJ for the
energy supply by the commercial power, and 0.0000251 $/kJ and the price was set
to 380 $/m3 for the energy unit price when using methanol as the boiler fuel, and
the boiler fuel kerosene is set at 373 $/kJ. These unit prices are the charges in
Japan. On the other hand, the unit price of the methanol in US is approximately
250 $/m3, and commercial power unit price is approximately 2.28 c$/MJ. In the
following analyses, a methanol unit price and commercial power unit price of
Japan are used. When the results of the Q2 system and Q3 system are compared,
there is no difference in fuel consumption observed in all the areas of F1F3.
Therefore, the change to the utility energy supply is not performed. However, the
Q2 system becomes advantageous over the Q1 system in many cases. In the case
study of the following Section, the Q1 and Q2 systems use methanol fuel, and
kerosene oil is assumed as the fuel for the boiler for utilities by Q3. Therefore,
there are different operation areas from the fuel consumption in load average value
in the width of load fluctuation, the path of load fluctuation and the area of system
operation.
The analysis results of operation plan of the system of representation day in Janu-
ary and September are Figs. 1.11(a) and (b), respectively. In the system with
heater, operation cost of CGS may be less from energy supply cost by the com-
mercial power and kerosene boiler. For example, 5:00, 6:00 in January of a result,
etc. The energy cost of combination of the fuel cell and the back-up boiler is ex-
pensive compared with the energy supply by the commercial power and kerosene
boiler.
1.5 Case Study 15
Figure 1.12 shows a comparison result of the energy supply cost on a representa-
tive day for each month. This figure is the case where the boiler (methanol fuel),
or the boiler and the electric heater are combined with methanol steam reforming
type fuel cell, and the results of comparing the energy cost by the utility (commer-
cial power and kerosene boiler). The annual cost ratio of CGS with a boiler to
utility cost is 2.16. Moreover, the annual cost ratio of CGS with an electric heater
to utility cost is 1.42. If methanol unit price is 267 (=380/1.42) $/m3, the operation
cost for one year of the system with a boiler and a heater is almost the same as the
energy cost for one year of the conventional utility. If a methanol unit price is 190
(unit price of the half of 380 $/m3) $/m3, the operation cost of this system will be
less than the energy cost by the utility to all the month. On the other hand, if
methanol unit price is 176 (=380/2.16) $/m3, the operation cost for one year of a
system with a boiler is almost the same as the energy cost for one year of the con-
ventional utility.
16 1 A Small-scale Fuel Cell Cogeneration System Considering Partial Load
1.6 Conclusions
As a CGS for the individual houses, the operation cost of solid polymer membrane
fuel cell system with the methanol steam reforming was examined. The thermal
conversion of electric energy by an electric heater was introduced into the system
in order to avoid the lowering of generation efficiency of a fuel cell in low-load
operation. By the introduction of the energy demand pattern in the Sapporo district
in Japan, and performing optimization analysis with the objective of determining
the operation cost of the system, the following results were obtained.
(1) The system driven by carrying out thermal conversion of electric energy by
the electric heater may reduce the operation cost further than when the system
is not introduced, in the case of marked fluctuations in thermal load in low-
electric-power demand.
(2) The energy supply cost to individual house of methanol-steam-reforming-type
fuel cell CGS with a back-up boiler and an electric heater will be less than the
energy supply cost by the present utility, if the unit price of methanol de-
creases from the present 380 $/m3 to 267 $/m3. The methanol unit price in US
is approximately 250 $/m3, and commercial power unit price is approximately
2.28 c$/MJ. In this case, in order for the energy supply cost of propose system
to be less than the energy supply cost of the utility, the unit price of methanol
needs to be 107 $/m3.
Chapter 2
Equipment Arrangement Planning
of a Fuel Cell Energy Network Optimized
for Cost Minimization
2.1 Introduction
In recent years, uses of the distribution of fuel cells have been studied [9, 10].
Furthermore, fuel cell systems are connected by a network and the micro-grid of
the electrical power operated in cooperation under the objective function given
beforehand is investigated [1113]. Moreover, compound utilization of green
energy equipment is considered to have little environmental impact although it is
unstable, and stable power-generator equipment is investigated by the electrical
power grid of small energy equipment [11]. Construction of an electrical power
micro-grid may to develop competition with the existing energy infrastructure
other than in short-term utilization, such as a backup power supply at the time of
disaster. Installation of both a micro-grid of solid polymer membrane-type fuel
cells and a hot water piping network used to supply fuel cell exhaust heats to each
building is anticipated to cause a large reduction in environmental impact [9, 10].
In order to achieve the energy network of electrical power and heat described
above, cooperative control of the distributed energy equipment according to the
objective function given beforehand, operation planning, the arrangement design
of equipment, and capacity design of equipment are required. The fuel cell energy
network (FEN) investigated in this paper examines the micro-grid (FMG) used for
an electric power supply, and the hot water piping network (HWN) that uses the
exhaust heat of equipment for supply or collection from each house. In small-scale
FEN, since the power transmission length is short, the energy loss of electric
power transmission is small, and the loss of exhaust heat transport can also be
maintained at low levels. When loss of electric power transmission is improved
rather than heat transport due to the difference in energy unit price, the total en-
ergy cost is generally low. However, in cold regions, houses, apartment houses,
office buildings, hospitals, etc., have a great annual heat demand. It is expected
that the heat transport loss of FEN in buildings with high heat demand has a large
influence on the efficiency of the system. Therefore, when transporting a lot of
heat using HWN, route planning of piping considering the heat release loss of
HWN and an arrangement plan of equipment are required; these aspects depend
on the route planning of HWN, and the optimum arrangement plan of heat output
equipment depends on the energy demand pattern of each building linked to FEN,
and the position of each house. The heat medium (hot water) temperature that
flows through the inside of the HWN in a house outlet is decided by the heat sup-
ply and demand of each house, and the heat release of HWN is dependent on the
hot water temperature and the piping length. Thus, in this paper, a plan is made to
optimize the equipment arrangement of the fuel cell, reformer, and boiler, and an
FEN with high energy efficiency and cheap facility cost is planned by optimizing
the route of HWN used for supplying the exhaust heat of a fuel cell and a reformer
to each house. If the scale of FEN is large, the total length of HWN will increase
and heat release will also increase. From this, it is expected that the increase in the
scale of FEN will lead to a decline in the energy efficiency of the system. How-
ever, compared with FEN that is not optimized, the cost and efficiency of the
system may be improved by optimizing the equipment arrangement of each build-
ing linked to FEN, and the route of HWN. That is, compared with FEN that is not
optimized, the same facility cost as FEN that has an optimized equipment, ar-
rangement is expected and a HWN route whose scale (the number of houses
linked to FEN) can be extended. In this paper, the arrangement plan of FEN that
differs in the number of houses, and the route plan of HWN are conducted using a
genetic algorithm [1419]. From this analysis, the operation cost of the system,
facility cost, and the installation cost of a facility are investigated, and the cost per
house connected by FEN is investigated.
The model of the fuel cell energy network (FEN) assumed in this paper is shown
in Fig. 2.1. Each building is connected with the micro-grid (FMG) of electrical
power, and the hot water piping network (HWN) used for waste heat recovery and
heat supply. A fuel cell assumes a solid polymer membrane model, and a reformer
assumes the steam reforming of city gas. Each facility of the fuel cell, the re-
former, and the boiler is capacity free in each building, and it is assumed that it
can be installed freely. It is necessary to optimize and perform design arrangement
and capacity planning of this equipment based on the objective function given
beforehand for the system. Houses in which fuel cells are installed, and the houses
in which the installed reformers are connected with a reformed gas piping network
are shown in Fig. 2.1. Supposing a gas piping network reformed from an existing
city gas piping network is used, FMG decides to use the existing commercial elec-
trical power network more. Therefore, in the analysis of cost in this paper, equip-
2.2 System Scheme 19
ment cost of a reformed gas piping network and FMG is not taken into considera-
tion. Since the route of HWN is planned considering the heat release in HWN that
connects each house, they differ depending on outside air temperature. The outside
air temperature of a representative day differs in different seasons, so the heat
release of HWN is calculated using an outside air temperature model in summer
(August), winter (February), and mid-term (May), and the optimal route is ex-
plored.
Figure 2.2 shows a model in which a fuel cell is installed in house S x . The re-
formed gas produced by the reformer is supplied to the fuel cells through a re-
formed gas network (town gas piping network). The power generated by the fuel
cells is supplied to FMG through a DCAC converter, an inverter, and an inter-
connection device, and supplies the load in each house. As shown in Fig. 2.2, the
exhaust heat of a fuel cell is supplied to the heat load of house S x , but when heat
remains, heat is supplied to other houses through HWN. On the other hand, when
heat is insufficient, heat is obtained and supplied by the HWN.
20 2 Equipment Arrangement Planning of a Fuel Cell Energy Network
The heat source that supplies the heat demand in each house gives priority to the
exhaust heat of the fuel cell and the reformer installed in the same house. When
such exhaust heat is insufficient for the heat amount demanded, heat is obtained
from HWN. Although a heat storage tank is installed in HWN, it stores the heat
when the heat that flows through HWN remains. The stored heat can be used by
conducting a time shift. A boiler is operated when heat runs short, even if it sup-
plies the heat obtained from HWN to a house.
Figure 2.3 shows the model of installing a reformer in house S y . Although city
gas is supplied to the reformer, in order to remove the carbon monoxide and water
in the reformed gas, carbon monoxide oxidation equipment and a dryer are install-
ed. The exhaust heat of the reformer can be supplied to each house through HWN.
Figure 2.4 shows a model of equipment arrangement planning at the time of con-
necting the distributed fuel cell with an energy network. As shown in Fig. 2.4(a),
in order to fulfill the demand of the electric power and the heat of six houses (from
S A to S F ), a reformer is installed in houses S B and S F , and a boiler is installed
in houses S A , SC , and S D for a fuel cell at houses S A , SC , and S E .
Each house is connected with HWN and can transport the exhaust heat of the fuel
cells and reformers, and the heat output of the heat storage tank and boilers with the
heat medium (hot water) that flows through the inside of the piping. Here, the de-
mand model of the electric power and the heat of each house is made to have the
characteristics shown in Fig. 2.4(b) and (c), respectively. As shown in Fig. 2.4(e),
the exhaust heat output in each fuel cell in this case depends on the production of
electricity (Fig. 2.4(d)) of the fuel cell. Moreover, Fig. 2.4(f) shows a model of the
exhaust heat output of a reformer, and Fig. 2.4(g) shows a model of the heat output
of a boiler. The hot water quantity of heat that flows through the inside of HWN is
decided from the heat demand model shown in Fig. 2.4(c), and the model of the heat
output in each piece of equipment is shown in Fig. 2.4(g) from Fig. 2.4(e). The heat
release of HWN equipped with thermal insulation is dependent on the difference in
temperature between the hot water and the atmospheric air. Figure 2.4(h) shows
a model of the heat release of HWN that connects each house. Therefore, the
amount of heat release in HWN differ depending on which equipment is installed in
the house linked to a network, and the route of HWN.
Figure 2.5(a) shows the model of incomings and outgoings of the heat of
HWN that connects house Si , Si +1 , and Si + 2 . As shown in Fig. 2.5(b), the hot
water of temperature Tin,Si , t and the quantity of heat H in,Si , t inputs into Si
through HWN. The exhaust heat output of a fuel cell when generating the fuel
cell installed in S i is determined so that the amount E need,Si , t of the electricity
demand of sampling time t that can be supplied is H fc,Si , t . Moreover, H need,Si , t
is the heat amount demanded in Si . The hot water quantity of heat ( H out,Si , t )
outputted from Si is H in,Si , t + H fc,Si , t H need,Si , t . The cost of fuel cell exhaust
heat H fc,Si , t differs according to the production of electricity E fc,Si , t of a fuel
cell; the following section describes the details of these relations. Although
the temperature of the hot water outputted from Si is Tout,Si , t , there is heat re-
lease of H w,Si (i +1) , t for the piping as it arrives in Si +1 from Si . Therefore,
with the hot water inputted into Si +1 , the temperature falls to Tin,Si +1 , t , and
the quantity of heat is H in,Si +1 , t = (H out,Si , t H w,Si (i +1) , t ) . Furthermore, in house
Si +1 , the exhaust heat H fc,Si +1 , t is outputted by the generation operation of
the fuel cell. The analysis of the hot water quantity of heat that sets the
heat amount demanded in Si +1 to H need,Si+1 ,t , and is outputted from Si +1
is H out,Si +1 , t = H in,Si +1 , t + H fc,Si +1 , t H need,Si +1 , t is calculated for all the houses.
Figure 2.5(c) shows a model of HWN that connects Si and Si +1 . The inside di-
ameter of the hot water piping is set to D i,Si (i +1) , the outside diameter is set to
D o,Si (i +1) , and the outside diameter of the heat insulating material with which the
piping is equipped is expressed as D c,Si (i +1) . When the heat conductivity of the
piping material and the thermal insulation is set to k p and k c , respectively, the
coefficient of the overall heat transmission ( K Si (i +1) ) of the hot water and the
surface of the heat insulating materials is expressed by the following equation.
2.3 Amount of Heat Release of the Hot Water Piping Network (HWN) 23
Fig. 2.5 Heat model for the hot water piping network
The number of the houses linked to a fuel cell network shown in Fig. 2.1 is set
to N bd . Each number is set to N fc , N rm , and N bo , although a fuel cell, a re-
former, and a boiler are installed in any of the houses. In the lower part of the
figure, the energy balance of the power and the heat of the system in sampling
time t is described.
Equation (2.3) is a balance equation of power. The left-hand side of Eq. (2.3)
expresses the power outputted to FMG from the fuel cell of the number N fc that
is generated. On the other hand, the first term on the right-hand side of Eq. (2.3)
expresses the power consumption in the number of the houses N bd linked to a
network. The second term on the right-hand side of Eq. (2.3) expresses the power
consumed with the circulating pump supplying hot water to HWN. In the analysis
example of following section, the power of a hot water circulating pump is calcu-
lated as consumption according to a hot water quantity of flow.
N fc N bd
E fc,i, t = E need, j,t + E pump,t (2.3)
i =1 j=1
Equation (2.4) is a balance equation of heat. The left-hand side of Eq. (2.4) ex-
presses the exhaust heat of a fuel cell, the exhaust heat of a reformer, the heat
output of a boiler, and the heat output of a heat storage tank, respectively. On the
other hand, the right-hand side expresses heat consumption with the number of
houses N bd linked to a network, and the heat release in hot water piping that
connects each house.
Nfc Nrm Nbo Nbd Nbd
Hfc,i,t + Hrm,j,t + Hbo,l,t + Hst,t = Hneed,m,t + Hw,n,t (2.4)
i =1 j=1 l=1 m=1 n =1
2.5 Cost Calculation and Objective Function 25
Equation (2.5) expresses the operation cost of a system required between t from
sampling time t . The first term in the right-hand bracket of Eq. (2.5) expresses the
operation cost of reformers. The operation cost of reformers is calculated by mul-
tiplying by the amount of city gas consumed by the reformers (the number is
N rm ), and the unit price J s,rm of city gas. The second term in the right-hand
bracket of Eq. (2.5) expresses the operation cost of the circulating pump used for a
hot water network. The operation cost is calculated by multiplying by the power
consumption and the power unit price. The third term in the right-hand bracket of
Eq. (2.5) expresses the operation cost of boilers.
Nrm Nst
Yd,t = Qrm,i,t Js,rm + Epump,t Js,pump + Qbo,k,t Js,bo t (2.5)
i=1 k=1
Equation (2.6) calculates equipment cost from the installed capacity and the ca-
pacity unit price. The right-hand side of Eq. (2.6) expresses the equipment cost of
fuel cells, reformers, boilers, heat storage tanks, HWN, and hot water circulating
pumps, respectively.
Nfc N rm N bo
Yc = U fc,i J c,fc + U rm, j J c,rm + U bo,k J c,bo + Ust J c,st
i =1 j=1 k =1
N bd
(2.6)
+ U pih,m J cl,pih + U pump J c,pump
m =1
Equation (2.7) is a formula for the installation cost of equipment. In this paper,
cost is taken into consideration for the installation of the fuel cell, the reformer,
the boiler, and the heat storage tank, which are shown on the right-hand side of
Eq. (2.7). However, it assumes that the other equipment shown in Fig. 2.2 is in-
cluded in the installation cost of a fuel cell, and all the equipment shown in
Fig. 2.3 is included in the installation cost of a reformer.
N fc N rm N bo
Yf = J f,fc,n + J f,rm, j + J f,bo,k + J f,st (2.7)
n =1 j=1 k =1
26 2 Equipment Arrangement Planning of a Fuel Cell Energy Network
The objective function of the system is calculated by Eq. (2.8) using the value of
the cost of Eqs. (2.5), (2.6), and (2.7). The minimum arrangement planning of the
equipment, capacity planning, and operation plan in the case objective function F
is investigated. g c , g y , and g d in Eq. (2.8) express the weight of equipment cost,
the weight of the installation cost of equipment, and the weight of operation cost,
respectively. When any of the terms on the right-hand side of Eq. (30) is much
larger than the other terms, it is necessary to determine a weight so that the objec-
tive function does not depend on the term too heavily. However, in the analysis of
following section, all g c , g y , and g d are calculated as 1.0.
Day
FO = g c Yc + g y Yf + g d Yd, t (2.8)
t =0
The third term on the right-hand side of Eq. (2.8) expresses the operation cost
of the system on a representative day. In the analysis, an operation plan in the case
of the value of Eq. (2.8) being the minimum is decided as the optimal solution.
Since calculation of optimization of this paper is non-linear with many variables, it
uses a genetic algorithm (GA) [1419] for optimization calculation. The chromo-
some model showing the operating method of the system using GA decides that it
is a solution with high fitness value as an individual with a small value of the
objective function of Eq. (2.8).
Figure 2.6 shows the chromosome model used in the optimization analysis of GA.
The chromosome model composes gene models, and a gene model expresses the
arrangement and the output of fuel cells and reformers, and the route of HWN. In
the analysis example described later, the number of houses is set at 49, and the
installation of a fuel cell and a reformer is determined at random using a gene
model for each house. Furthermore, the output of a fuel cell and a reformer is also
decided at random, and is expressed by the gene model.
2.6 Analysis Method and Case Study 27
Fig. 2.6 The chromosome model used for the genetic algorithm
As the section Operation Model of the System describes, according to the route
of HWN, the amount of heat release of the whole system differs, and the effi-
ciency of the system is affected. Therefore, the arrangement planning of equip-
ment and the route plan of HWN are predicted to affect cost. In this paper, the
path-planning program of HWN is developed using the idea of the traveling
salesman problem (TSP) [19]. Herein, the view of the order expression of route by
Dewdney is installed into TSP. According to this view, it is managed by a sym-
metrical number that is different from the housing number in the gene indicating
the route of HWN. By this method, when gene manipulation such as cross-over
and mutation is added to a chromosome model, models that show unachievable
routes do not appear. From this, the analysis efficiency improves sharply.
The production of electricity E fc,i, t ( i = 1,2,..., N fc ) of the fuel cell installed in any
N bd
house at sampling time t determines the amount of power demanded E need, j,t
j=1
in time t by random distribution. The gene model group of (a) of the chromosome
models shown in Fig. 2.6 shows the number of installations of fuel cells N fc , and
the houses in which they are installed. Moreover, although the group of the gene
model of (d) expresses the generation rate in each fuel cell, the production of elec-
28 2 Equipment Arrangement Planning of a Fuel Cell Energy Network
tricity E fc,i, t (however, i = 1,2,..., N fc ) of each fuel cell is decided from these
values. The exhaust heat output of each fuel cell H fc,i, t (however, i = 1,2,..., N fc )
is decided by the characteristic model of the electric power and the exhaust heat
output of the fuel cell shown in Fig. 2.7. The characteristics of the electric power
and the heat output of a fuel cell shown in Fig. 2.7 were obtained by the experi-
mental results of the fuel cell manufactured as an experiment.
2.6 Analysis Method and Case Study 29
When the production of electricity of each fuel cell is determined, the quantity of
reformed gas required for generation will be determined from the characteristic
model of the production of electricity and the consumption of reformed gas of the
fuel cell shown in Fig. 2.8. A characteristic of the reformer of Fig. 2.8 is the model
created from the output characteristics of a fuel cell and a reformer manufactured
as an experiment. By the reformer installed in each house, the quantity of re-
formed gas produced is decoded and determines the value of the gene model of (b)
and (e) of the chromosome model shown in Fig. 2.6. If the amount of reformed
gas production of each reformer is decided; the exhaust heat output H rm,Sx , t
(however, x = 1, 2,..., N rm ) will be decided from the relation of the ratio of load
and the ratio of reformation of Fig. 2.9.
The heat of (H fc,total,t + H rm,total,t H need,total,t ) is stored when the value to which
N fc
is added exhaust heat H fc,total,t = H fc,i, t of the fuel cell and exhaust heat
i =1
N rm
H rm,total, t = H rm, j,t of the reformer compared with the heat demand
j=1
N bd
H need, total,t = H need,k,t of all the houses linked to FEN at sampling time t is
k =1
exceeded. On the other hand, when H fc,total,t + H rm,total,t + H st, t is less than
N bo
H need,total,t , the heat of (H need, total, t H fc, total,t H rm,total, t H st, t ) = H bo,l,t is
l =1
outputted to HWN from the boiler. On installing a boiler in all the houses, the heat
output of each boiler is determined so that the heat balance in each house does not
become negative. The city gas consumption of a boiler is calculated to a boiler
efficiency of 85%.
As shown in Table 2.1, the hot water piping uses a coat of polyethylene tube with
an inside diameter of 28 mm, and a wall thickness of 10 mm. The heat transfer
coefficients between the piping inner wall and the hot water is 2500 W/m2K, the
30 2 Equipment Arrangement Planning of a Fuel Cell Energy Network
heat transfer coefficient between the outer surface of the piping and the atmos-
pheric air is 20 W/m2K, and the heat conductivity of the piping is 0.043 W/mK.
The overall heat transfer coefficient between the hot water and the atmospheric air
is calculated using the value described above. Heat release H w,n, t (however,
n = 1, 2,..., N bd ) of the hot water is calculated by multiplying the overall heat
transfer coefficient by the piping length and the difference between the tempera-
ture of the hot water and the open air. The power of hot water circulating pump
Ppump is calculated by Eq. (2.9). In the analyses of this paper, the water head Wp
shall be 7 m considering pipeline friction.
Ppump = w g q w Wp (2.9)
given to the program. For the chromosome model to which gene manipulation is
added, the fitness value of each model is re-calculated. Individuals with a low
fitness value in a chromosome model are discarded, and a chromosome model that
is newly determined is supplied at random. This calculation is iterated with the
generation number first given to the program. Among the last generations chro-
32 2 Equipment Arrangement Planning of a Fuel Cell Energy Network
The city area model assumed in the analysis of this paper is shown in Fig. 2.11.
FEN to four buildings of houses S A , SC , SE , and SG of Fig. 2.11 is called the
four-houses model, and FEN to five buildings S B , S D , S F , S H , and S I is called
the five-houses model. Moreover, FEN to nine buildings S A to S I is called the
nine-houses model. Although the energy demand of each house shown in Fig. 2.11
differs due to the number of residents, composition age, lifestyle, etc., it is ana-
lyzed by giving the simplest possible energy demand pattern to an analysis pro-
gram in this analysis. Then, the average energy demand pattern of a 34 person
household in Sapporo [8] that shows the energy demand pattern of each house in
Fig. 2.12 is used. The power load pattern of Fig. 2.12(a) is consumption by house-
hold electric appliances and lighting, and electric air-conditioning equipment is
not used throughout the year. For this reason, as shown in Fig. 2.12(a), there is not
a large difference in the electricity demand pattern of each month. On the other
hand, the items of heat load are hot water supply, baths, and space heating. More-
over, the outside air temperature model of Sapporo used for calculating the heat
release in HWN is shown in Fig. 2.13 [20]. For Sapporo, a cold, snowy area, the
annual average temperature is 288 K, and the mean temperature in February and
August is 269 K and 294 K, respectively. The equipment efficiency, the energy
cost, the cost of equipment, and the installation cost of equipment are used for cost
analysis are shown in Table 2.2. The fuel cell and the reformer are calculated to be
2500 dollars/kW and 1500 dollars/kW, respectively. It is expected that these cost
values will be attained within several years. The equipment costs of other boilers,
heat storage tanks, and hot water circulating pumps are decided by a product cata-
log as reference. For equipment installation costs, such as for a fuel cell or a
boiler, the installation cost of a common home boiler or a hot water supply system
was assumed.
Figure 2.14 shows the analysis result of the electric power of a fuel cell, and the
exhaust heat output and the analysis result of the reformed gas output of a re-
former when installing a fuel cell network into the four-houses model.
Table 2.3 is the result of analysis using an energy demand pattern and the out-
side air temperature data of a representative day in February, May, and August,
and is a result of the capacity of a fuel cell, a reformer, and a boiler and the instal-
lation location of the four-houses and the five-houses models. In the analysis result
of the four-houses model shown in Fig. 2.14, although a fuel cell is installed in
two houses, SC and SG , for all months, a reformer is scheduled to be installed in
one house. The electric power and the heat output of fuel cells are different in their
output characteristics of two sets to other months, although the two sets of outputs
will be almost the same in February. The reason for this is explained in detail in
the section Quantity of Flow of a Hot Water Circulating Pump, and it is because
heat release by HWN is large, and a plan is made for representative days in Febru-
ary so that the heat transport using HWN can decrease. That is, although a fuel cell
of the same capacity will be installed in a symmetrical position on a representative
day in February, and exhaust heat is supplied to the house with the equipment
installed and a nearby house, heat is not supplied to distant houses. Furthermore,
2.7 Analysis Result 35
Fig. 2.14 Analysis results of the operational schedule of the fuel cells and reformers installed in
the four-houses model
36 2 Equipment Arrangement Planning of a Fuel Cell Energy Network
Table 2.3 Results of the arrangement plan and equipment capacity for FEN
2.7.2 Amount of Hot Water Heat Release and the Hot Water
Piping Route
Figure 2.15 shows the analysis results of the route plan of the hot water piping of
the four-building model and the five-building model, and the flow direction of hot
water. The route plan of hot water piping, the house used as the starting point of
HWN, and the flow direction of hot water show the same results every month in
the four-building model. On the other hand, although the route result of the hot
water piping of a five-building model is the shortest route by which each house is
connected in any month, each result of the house used as the starting point of
HWN, and the flow direction of hot water differs every month. However, since the
energy demand pattern of each house is the same, the house used as the starting
point of HWN and the flow direction of the heat medium do not affect the objec-
tive function equation (2.8). Figure 2.16 shows the analysis result of the time
change of the hot water piping heat release every month. Since it is dependent on
the difference in temperature of a medium and the outside air temperature, there
will be great heat release of HWN in February. Especially on representative days
in February, there is great heat release in a time zone with little heat demanded.
The heat release at each time in August is characterized by many upward slants to
the right. This is because there is little heat demanded by each house, so the heat
transported by HWN increases, and as a result, the hot water temperature rises and
the heat release increases.
2.7 Analysis Result 37
Fig. 2.15 The results of hot water supply planning of the four-houses model and the five-houses
model
38 2 Equipment Arrangement Planning of a Fuel Cell Energy Network
Figure 2.17 shows the analysis result of the heat-medium (physical properties
using the value of water) quantity of flow that flows through the inside of HWN.
The heat-medium quantity of flow (quantity of flow of a hot water circulating
pump) determined that 363 K was not exceeded at any position of HWN. There is
little heat-medium flow in the four-building model and the five-building model,
and there will be little transportation of heat between each house in February. The
difference in temperature of a heat medium and the open air is large, and the heat
release will be large in February. Then, transportation of heat between each house
through HWN is suppressed, and a plan is made so that the heat demand of each
house may be supplied with the exhaust heat of the fuel cell installed in each
house, the reformer, and the heat output of the boiler.
Figure 2.18 shows the analysis result of the amount of heat surplus that was set at
the starting time of operation of the system as 0:00, and was calculated from the
heat balance for every time. Since there is high heat demand at each time, there is
little heat surplus in February compared with other months. From the characteris-
tics of Fig. 2.18, the capacity and the operation plan of thermal storage can be
designed considering the energy demand pattern of every month. Figure 2.19
shows the analysis result of the boiler output in each time of representative days in
February. As Table 2.3 shows, on representative days in February, any model is
planned so that a boiler can be installed in all the houses.
Figure 2.20 shows the analysis results for cost operation (Fig. 2.20(a)) for every
month, equipment cost (Fig. 2.20 (b)), and the installation cost (Fig. 2.20(c)) of
equipment for the four-building model, the five-building model, and the nine-
building model. Moreover, Fig. 2.20(d) shows the total cost as a result of
Fig. 2.20(c) from Fig. 2.20(a). However, each result of Fig. 2.20 is arranged as
cost per house. Furthermore, the cost of an independent system that installs a fuel
cell, a reformer, a boiler, and a heat storage tank in an individual house, and that
performs energy supply is shown in Figs. 2.20(b)(d). The cost per house com-
pared with an independent system decreases, so that the number of the houses
connected to FEN from the result of Fig. 2.20(d) increases. In FEN planned using
the energy demand pattern in February, 18%, 22%, and 25% of the total cost is
reduced by the four-houses model, the five-houses model, and the nine-houses
model compared with an independent system, respectively. When constructing
FEN in the houses of four to nine buildings analyzed in this paper, the total cost
per building can be reduced by optimizing the arrangement planning of the
equipment as well as the system operation plan, so that there are many houses.
In order to investigate the difference in fitness values and the difference in opera-
tion plans, the solution (solution with a sufficient fitness value to the second term
of Eq. (2.8)) of the second ranking of the four-building model is shown in
Fig. 2.21 and Table 2.4. Figure 2.21 shows the cost analysis results of the optimal
solution and the second ranking solution. Figure 2.21(a) expresses the analysis
result of the operation cost of a representative day every month, Fig. 2.21(b)
shows equipment cost, and Fig. 2.21(c) shows the result of the installation cost of
Table 2.4 The results of the arrangement plan of the FEN for the four-houses model. These
analysis results are the optimal solutions of the second ranking
Fig. 2.21 Four-houses model cost comparison of a different fitness value analysis result
42 2 Equipment Arrangement Planning of a Fuel Cell Energy Network
equipment. Figure 2.21(d) shows the total cost as a result of Fig. 2.21(c) from
Fig. 2.21(a). The difference in the total cost of the optimal solution and the second
ranking solution will be 0.1% in February and May, and will be 0.4% in August.
When the results of Tables 2.3(a) and 2.4 are compared on a representative day,
there will not be a big difference in the setting position or the output of the fuel
cells, reformers, and boilers in February. However, on representative days of May
and August, there will be a difference in the setting position and the output of the
fuel cells and reformers. In the analysis of the operation plan using GA, when the
difference in the value of the objective function is 0.4% or less, as shown in Ta-
bles 2.3(a) and 2.4, it differs.
2.8 Conclusions
3.1 Introduction
at the time of load fluctuation, the risk abatement at the time of accident due to
distribution of power facilities, and the interruption of service not caused at the
time of maintenance.
The buildings in which power plants are installed (Fig. 3.3(a)) and other buildings
(Fig. 3.3(b)) are connected to each of Grid A and Grid B of Fig. 3.2. A town gas
reformer, PEM-FC, a boiler, a heat storage tank, and a system-interconnection
device are installed in the building of Fig. 3.3(a). The boiler for heat supply and
the system-interconnection device for receiving electric power from a grid are
installed in the building of Fig. 3.3(b). Figure 3.3(a) shows the model of a building
in which a power plant is installed as shown in Fig. 3.2. In this chapter, a power
plant (Fig. 3.3(a)) connected to one independent grid (Grid A and Grid B, respec-
tively) may be at one place. Buildings other than the power plant are buildings
(Fig. 3.3(b)) in which a boiler is installed for heat supply, and a system-inter-
connection device. An actual FC micro-grid requires a town gas distribution net-
work and exhaust heat distribution network other than an electric power supply
network. This chapter does not describe the management of a town gas distribu-
tion network or an exhaust heat distribution network.
Figure 3.4 shows the cell performance curve for an operating temperature of
333 K, with the gas pressure at the anode and cathode being 0.1 MPa [2629]. This
characteristic is the model to be used when setting the reformer efficiency to be
constant at 73%. Details of RM are given below. Although steam reforming of
the town gas at a flow rate of Q RM, t is conducted, and reformed gas is produced
in the reformer, the heat source for the reforming reaction is burned and obtained
using the town gas at a flow rate of Q B,t . The following equation defines the
reformer efficiency, and the maximum of this value is 73% [29]. WH is the calo-
rific value of the hydrogen included in reformed gas, and WB and WR are the
calorific values of town gas at flow rates of Q BN1 , t and Q RM, t , respectively. For
the power generation efficiency of a micro-grid it depends on the fuel cell output
characteristics of Fig. 3.4. For example, the output characteristics of a fuel cell
3.3 Installation Planning of the FC Micro-grid 47
change with the operating method of a reformer. Care should be taken that the
characteristics of Fig. 3.4 are not the result of representing a fuel cell system.
WH
RM = 100 (3.1)
(WB + WR )
Figure 3.5 shows the model indicating the relation between the load pattern of a
power demand model, and the generation efficiency of a fuel cell. As Fig. 3.4
shows, the generation efficiency of a fuel cell changes with the load ratio. Next,
the operation method of a fuel cell when there are two or more power demand
models is described.
In Fig. 3.6, each power demand model of Buildings A, B, and C is shown. The
following three methods of supplying power to these buildings from PEM-FC can
be considered.
(1) Installing a fuel cell in all the buildings (stand-alone system).
(2) Connecting all the buildings to a micro-grid, and supplying power using one
set of FC (central system).
(3) Setting two or more micro-grids considering the power demand pattern of the
buildings (partition cooperation system).
The stand-alone system determines the capacity of the PEM-FC installed in
each house so that it may exceed the maximum of each power load of Buildings
A, B, and C of Fig. 3.3. In the central system, the power load of Buildings A, B,
and C of Fig. 3.6 is added for every sampling time, and the capacity of one set of
PEM-FC is determined to exceed the maximum. In addition, in the partition coop-
eration system, Building A and Building B of Fig. 3.6 are connected, for example,
and Grid 1 operates. Building C is operated independently (Grid 2). The capacity
of the PEM-FC installed in Grid 1 and Grid 2 is determined so that it may exceed
the maximum of the power load of Building A and Building B, and the maximum
of the power load of Building C. The stand-alone system does not need a grid;
moreover, it does not need to transport exhaust heat to a neighboring building.
Therefore, the heat loss of exhaust heat is small, and installation of a grid is un-
48 3 Effective Improvement in Generation Efficiency due to Partition Cooperation Management
necessary. However, there are many fuel cells that need to be installed, and it is
necessary to install large-capacity fuel cells so that load fluctuation does not have
an impact. Unlike the stand-alone system, the central system should just install FC
in one building. Therefore, although it is advantageous with respect to equipment
cost, the transport distance for the exhaust heat is long, and heat loss is a problem.
Furthermore, the diversification of risks at the time of an accident and the extendi-
bility of equipment are problems. In the central system, the power demands of the
buildings are added, and the capacity of the fuel cell is optimized and determined
to exceed the maximum. Therefore, if the number of the buildings connected to
the grid and the pattern of the power demand model changes, the load ratio will
change and generation efficiency will vary. On the other hand, the partition coop-
eration system can partially achieve each merit of the stand-alone system and the
central system. Partition of the micro-grid in the partition cooperation system is
optimized to maximize generation efficiency. For this reason, depending on the
composition of the power demand model of the buildings in an urban area made
into an analytical object, generation efficiency higher than that with the stand-
alone system or the central system may be applicable.
3.3 Installation Planning of the FC Micro-grid 49
Figure 3.7 shows the power demand models of each building, and is the mean
power load of each sampling time on a representative day in January (winter),
May (mid-term), and August (summer) [3033]. However, the actual power de-
mand pattern is a mixture of loads that change rapidly over a short time, such as an
inrush current. In addition, a power demand estimate for the house actually shown
in Fig. 3.7(a) is difficult, and a power demand estimate for the small offices and
factories indicated in Figs. 3.7(d) and (e) is comparatively easy. Although load
fluctuation is not taken into consideration in the power demand model in the
analysis of this chapter, when accompanied by load fluctuations, system intercon-
nection between the grids is expected to occur frequently. The power demand
pattern of a house (Fig. 3.7(a)) has peaks in the morning and the afternoon. In the
hotel of Fig. 3.7(b), demand is stabilized when the midnight-to-early-morning
period is excluded, and the power demand is stable for 24 hours at convenience
stores (Fig. 3.7(c)). The difference between the time zone with a small power
demand from night to early morning and the time zone with great power demand
from morning to evening is clear in small offices (Fig. 3.7(d)), factories
(Fig. 3.7(e)), and small hospitals (Fig. 3.7(f)).
Figure 3.9 shows the urban area model used for this analysis, and shows the type
of power demand model of 23 buildings in this figure. In addition, the number
given in this figure is the number of buildings, and shows the area and the maxi-
mum load of each building assumed in Table 3.1. Figure 3.9 shows a two-person
family house and a six-person family house as well as a four-person family house
(the power demand model is shown in Fig. 3.7(a)). Each power demand model
compares and determines the number-of-persons rate of the model of Fig. 3.7(a).
Moreover, although apartment houses are shown in Fig. 3.9, these power demand
models are also determined relatively from the number-of-persons rate of the
52 3 Effective Improvement in Generation Efficiency due to Partition Cooperation Management
model of Fig. 3.7(a). In this analysis, Fig. 3.9 is used as an urban area model.
Moreover, Figs. 3.7(a)(f) is used as a power demand model of each building that
composes an urban area. The relation of the load ratio and generation efficiency of
the PEM-FC installed in the model is shown in Fig. 3.4. Since the maximum gen-
eration efficiency of the model of Fig. 3.4 is 32%, the maximum generation effi-
ciency of the micro-grid analyzed here is theoretically 32%.
Figure 3.10 shows the analysis results of the average generation efficiency of a
representative day of every month in the case of installing the stand-alone system
in the urban area model of Fig. 3.9. Although the average generation efficiency
differs each month, convenience stores with a small load fluctuation range of
power throughout the year show about 30% at maximum. The average generation
efficiency of a representative day in August, with high power consumption due to
air-conditioners, is high in hotels and hospitals. The average generation efficiency
of other buildings is less than 20%.
3.5 Analysis Results and Discussion 53
The system that distributes the PEM-FC system whose generation capacity is
1 kW or 2 kW to the urban area model of Fig. 3.9 is investigated. To install PEM-
FC (1 kW or 2 kW) using the stand-alone system, it is necessary to look for build-
ings that can install fuel cells of this capacity from among buildings that compose
an urban area model. Figure 3.11 shows the analysis results of the number of
buildings in which a fuel cell of each capacity can be installed using the stand-
alone system, and the number of the fuel cells installing PEM-FC using the parti-
tion cooperation system. If equipment cost is taken into consideration from the
difference in the installed number of PEM-FC, compared with the stand-alone
system, the partition cooperation system is more advantageous.
54 3 Effective Improvement in Generation Efficiency due to Partition Cooperation Management
Figure 3.12 shows the analysis results of the relationship between the capacity of
PEM-FC and generation efficiency in the case of installing the central system in
the urban area model. The results of the generation capacity of PEM-FC using the
central system is 110 kW (the average generation efficiency of each representative
day is 25.8% and 26.1%, respectively) on representative days in January and May,
and is 160 kW (the average generation efficiency of a representative day is 24.8%)
on a representative days in August. Therefore, it is required that a 160 kW PEM-
FC be installed in the central system in the urban area model, and be applied
throughout the year. When a 160 kW PEM-FC is installed, the average generation
efficiency of representative days in January and May is 20.6% to 20.9%. Conse-
quently, the generation efficiency of an FC micro-grid when installing the central
system is operated at 20.6% to 24.8%.
Figures 3.13(a)(c) show the analysis results of the partition cooperation system in
which a PEM-FC whose generation capacity is 2 kW is installed. However, be-
cause two or more buildings that have the same power demand model are included
as described in Fig. 3.9, when these are exchanged, it is expressed by the route
difference from Figs. 3.13(a)(c). The average generation efficiency of representa-
tive days in January, May, and August is 23%, 22.7%, and 20.4%, respectively.
3.5 Analysis Results and Discussion 55
Fig. 3.13 Analysis results of the partition cooperation micro-grid system at the time of install-
ing 2 kW PEM-FC
the whole generation efficiency of the micro-grid, where a PEM-FC with small
generation capacity is installed, is improved, and equipment cost is reduced. How-
ever, a maximum power demand greatly over 2 kW cannot be installed in this grid
in buildings.
Figures 3.15(a)(c) show each representative day in the grid route obtained in the
analysis, and show the analysis results at the time of combination, so that the gen-
eration efficiency of the whole grid is maximized. However, because two or more
buildings that have the same power demand model in an urban area model are
included as described in Fig. 3.9, if these are exchanged, the combination will
become different from that in Figs. 3.15(a)(c). Figure 3.16 shows the analysis
results of the generation efficiency when giving the route result of a representative
day in August (Fig. 3.15(c)) to the power demand model of a representative day in
January. Compared with Fig. 3.15(c), the result of the generation efficiency of
Fig. 3.16 falls due to Grid A to Grid C, and the average generation efficiency of
the whole grid falls by 23.6% to 21.1%. Therefore, as for the FC micro-grid whose
average generation efficiency in total comprises a combination of the highest
grids, generation efficiency is operated at 21.1% to 27.6%. The range of this value
exceeds the average generation efficiency (from 20.6% to 24.8%) when installing
a 160 kW PEM-FC using the central system shown in Fig. 3.12.
Figures 3.17(a)(c) show the results of the combination for which the generation
efficiency of a representative day is a grid route exceeding 25%, and the genera-
tion efficiency of the whole grid peaks every month. The grid route for which
3.5 Analysis Results and Discussion 57
Fig. 3.15 Analysis results of the partition cooperation system which supplies the power to all
the buildings
generation efficiency does not exceed 25% does not appear in Figs. 3.17(a)(c).
Consequently, there is no guarantee that all the buildings in an urban area model
can be connected to any grid. Figure 3.18 shows the analysis results of the genera-
tion efficiency when installing the power demand model of a representative day in
January into the analysis result (Fig. 3.17(c)) of the grid route of a representative
day in August. Compared with Fig. 3.17(c), the generation efficiency of Fig. 3.18
falls compared to the other grids (Grid A and Grid B), and the average generation
58 3 Effective Improvement in Generation Efficiency due to Partition Cooperation Management
Fig. 3.17 The analysis results of the partition cooperation system when supplying the power to
a building of 25% or more of average generation efficiency
efficiency of the whole grid falls by 19%, being 23.2%. Therefore, as for the FC
micro-grid whose generation efficiency is composed from grids exceeding 25%,
the generation efficiency is operated at 23.2% to 28.6%. If this system is installed,
the FC micro-grid can be operated at the highest generation efficiency, but build-
ings that lower the generation efficiency of the whole grid are separated from this
grid.
3.6 Conclusions 59
3.6 Conclusions
Even if an FC micro-grid is used for urban areas composed of buildings with vari-
ous power demand patterns and supplies power to them, generation efficiency is
generally reduced. Therefore, this chapter examined the generation efficiency of
the following three power supply methods: (1) Installing fuel cells in all the build-
ings in an urban area model (stand-alone system); (2) connecting all the buildings
to a micro-grid, and supplying power from one set of fuel cells (central system);
and (3) dividing a micro-grid into multiple grids considering the power demand
pattern of the buildings (partition cooperation system). The relationship between
the capacity of the fuel cell to be installed and the generation efficiency became
clear from the analysis results. Furthermore, in this chapter, generation efficiency
is optimized to maximize generation efficiency to the grid route planning of the
partition cooperation system. As a result, a system with a higher generation effi-
ciency than the stand-alone system and the central system could be proposed. The
average generation efficiency of the stand-alone system is 20% or less, and the
average generation efficiency of the central system is 20.6% to 24.8%. On the
other hand, the generation efficiency of the partition cooperation system proposed
herein is 21.1% to 27.6%.
Chapter 4
Fuel Cell Network System
Considering Reduction in Fuel Cell Capacity
Using Load Leveling and Heat Release Loss
4.1 Introduction
In order for installation of the fuel cell system to houses or a small-scale and mid-
dle-scale building to spread, it is necessary to reduce the equipment cost. Conse-
quently, a fuel system network (hydrogen piping and oxygen piping) and an en-
ergy network (a power transmission line and hot water piping) of distribution fuel
cells are proposed [24]. In this system, common auxiliary machinery is installed in
a machinery room. In this chapter, in order to reduce the capacity of the fuel cell
connected to the network, a method of leveling the load is proposed. By this
method, hydrogen and oxygen are generated by water electrolysis at the time of
low load with little power demand, and each gas is compressed and stored. On the
other hand, the stored gas is supplied and generated to the fuel cell in a period of
large power load. Experimental results show that the power generation characteris-
tics improve greatly compared with air supply, when supplying oxygen to the fuel
cell [34]. Therefore, if the oxygen generated when load is small can be used for a
high-load period, the installed capacity of the fuel cell can be reduced. Moreover,
the heat-energy network is hot water piping, and supplies heat to each building.
Hot water piping distributes heat via each building. When there is heat excess with
some buildings, it can also recover this heat through the hot water piping. In a
heat-energy network, the hot water temperature in a building outlet changes with
the heat consumed by each building and the fuel cell exhaust heat of each build-
ing. Therefore, the heat release of the overall network differs according to the
outside air temperature, the piping distance, the starting point of the hot water
supply, and the flow direction of the hot water. Consequently, to counteract the
piping heat release loss of the heat-energy network, the minimum piping route is
examined.
In the analysis case, the capacity reduction effect of a fuel cell when installing
load leveling using the water electrolyzer described above is investigated with
respect to a local energy network that includes houses, a hospital, a factory, an
office, and a convenience store. Furthermore, the hot water piping route and the
fuel cell capacity placed on each building when optimizing the system with the
object of minimizing the hot water piping heat release are considered.
4.2 Load Leveling and the Arrangement Plan of the Fuel Cell
The network model with the proton exchange membrane fuel cell (PEM-FC) in-
stalled that is assumed in this paper is shown in Figs. 4.1 and 4.2. As shown in
each figure, the fuel system (hydrogen piping and oxygen piping), the power sys-
tem (power transmission line), and the heat-energy network (hot water piping)
between the fuel cells installed in each building are connected. A heat transfer
medium (hot water) flows in hot water piping, the exhaust heat of a fuel cell is
recovered, and heat is distributed to each building. The route of the hot water
piping can be set up arbitrarily, and the flow direction is one way, as the arrow in
each figure shows. Figure 4.1 shows the system that supplies the power to Build-
ings A to G from one set of the fuel cell installed in the machinery room, and is
described as the R1 type below. A machinery room can be installed in an arbitrary
building (Building A in Fig. 4.1). As shown in Fig. 4.1(c), the fuel cell (1: this
number corresponds to that in Fig. 4.1), water electrolyzer (6), city gas reformer
(7), hydrogen and oxygen compressor (9 and 11), cylinders (10 and 12), geo-
thermal heat pump (13), heat storage tank (14), etc. are installed in the machinery
room. The heat output by fuel cell exhaust heat, the heat storage tank, and the geo-
thermal heat pump is distributed to each building through a heat transfer medium.
The piping route can be planned arbitrarily and it is in the order of Building
ABCFEDGA in the example of Fig. 4.1(a). As shown in Fig. 4.1(b), headers (4
and 5) are set in each building at a hot water gate. The heat of the radiator (3) and
a heat exchanger connected to the header are used for space heating and hot water
supply. Figure 4.2 shows the system that distributes a fuel cell in all the buildings,
and this system is described as the R2 type below. Although the number of fuel
cells increases and the equipment cost increases for the R2 type, heat release loss
with heat transport is small. The hot water piping route of the R2 type and the
building with a machinery room can be planned arbitrarily. In the example of
Fig. 4.2(a), hot water is supplied in the order of Building ADGBFECA. The ma-
chinery room of Fig. 4.2(c) is installed in Building A. The equipment scheme
installed in the building and machinery room in the R2 type is shown in Figs.
4.2(b) and (c).
Ambient air is usually supplied to the fuel cell installed in R1 and R2 from a
blower. However, both types can also supply oxygen through piping. Moreover, it
is assumed that reformed gas of the city gas reformer and hydrogen of the cylinder
can be supplied to the fuel cells at arbitrary times through the network.
4.2 Load Leveling and the Arrangement Plan of the Fuel Cell 63
Figure 4.3 shows the power generation characteristics when hydrogen and oxygen
are supplied, and when supplying hydrogen and air by the results of the perform-
ance measurement of a PEM-FC. The differences in these power generation char-
acteristics are considered to be due to the difference in oxygen partial pressure, the
water balance inside the cell, and the electrical receptivity change of the ion ex-
change membrane. The power generation characteristics differ in supplying re-
64 4 Fuel Cell Network System Considering Reduction in Fuel Cell Capacity
formed gas to a fuel cell, and supplying hydrogen. However, since there are few
differences in the power generation characteristics of reformed gas or hydrogen,
this difference is ignored.
Figure 4.4 shows the characteristics of the power and heat output when supply-
ing air or oxygen to a cathode using the same fuel cell (the electrode surface is
1 m2) as shown in Fig. 4.3. The maximum power output when supplying air to the
cathode is E fca, max =1.05 kW, and it is E fco, max =1.9 kW in the supply of oxygen.
In this way, if oxygen is supplied to the cathode, the power output will increase.
Therefore, if oxygen is supplied and generated to a fuel cell when there is a high
power demand, the fuel cell can be miniaturized compared with the design capac-
ity by air supply. If the fuel cell with the characteristics shown in Fig. 4.4 is used
with maximum output, the fuel cell facility capacity will decrease by the value of
( E fco, max E fca, max ).
4.2 Load Leveling and the Arrangement Plan of the Fuel Cell 65
Figure 4.5 shows the model indicating the power demand amount E need,t to which
is added the power demand amount of each building in Fig. 4.1 or Fig. 4.2 for
every sampling time t. E sep in this figure is the threshold value of the region of
low load and high load. By using this threshold value, load leveling is attempted
using the method described below. When E need,t is less than E sep , it generates
electricity by supplying reformed gas and air to the fuel cell. However, the elec-
tricity production of the fuel cell is always E sep , it supplies power the value of
which is the difference between E sep and E need,t to the water electrolyzer and
produces hydrogen and oxygen (the black area in Fig. 4.5). After compressing
these gases, they are stored in each cylinder. When E need,t exceeds E sep , it gener-
ates electricity by supplying the hydrogen and oxygen in the cylinders to the fuel
cell through the network. In the proposed method of load leveling, it is necessary
to determine E sep , where hydrogen and oxygen are produced at the time of low
load, and the amount is consumed at the time of high load balance.
Figure 4.6 shows the model of (a) the hot water piping route, (b) the fuel cell ca-
pacity of each building, (c) the change of temperature of the hot water, and (d) the
piping heat release per unit length of the R1 and R2 types. The machinery room of
both types is installed in Building A, and hot water flows in the order of Building
ABCDEFGA for the R1 type, and it flows in the order of ADGBFECA for the R2
type as shown in Fig. 4.6(a). As shown in Figs. 4.6(a) and (b), one fuel cell is
installed in Building A (FA) for the R1 type, and the fuel cell of the capacity of FA
to FG is installed in Buildings A to G for the R2 type. Hot water of temperature
TA,in, t is input into Building A in the R1 type. Heat is supplied for the hot water
from the fuel cell exhaust heat (FA), heat storage tank, and geo-thermal heat
pump, and as shown in Fig. 4.6(c), hot water of temperature TA,out, t is output from
Building A. After this, there is no heat input to the hot water, and hot water of
66 4 Fuel Cell Network System Considering Reduction in Fuel Cell Capacity
temperature TA,in, t returns to the machinery room of Building A due to the heat
consumption of Buildings B to G, and piping heat release. The temperature falls as
hot water of the R1 type progresses to Building G from Building A. Therefore,
since the difference in temperature of the outside air and the hot water is small, as
shown in Fig. 4.6(d), the piping heat release per unit length is small. On the other
hand, in the R2 type, heat is supplied to hot water from distributed fuel cells.
Therefore, the outlet hot water temperature of each building fluctuates, as shown
in Fig. 4.6(c). As a result, the heat release per unit length of piping also fluctuates,
as shown in Fig. 4.6(d).
At sampling time t , the water electrolyzer installed in the machinery room and the
fuel cells of M installed in M buildings are operating ( M = 1 in the R1 type).
The power balance equation in this case is expressed with the following equation.
M M V
E f,m,t = E need,m,t + E el,t + E hp,t + E sub,v,t (4.1)
m =1 m =1 v =1
4.2 Load Leveling and the Arrangement Plan of the Fuel Cell 67
The left-hand side of Eq. (4.1) expresses the power output in the DCAC con-
verter outlet of the fuel cells of M . Moreover, the first term on the right-hand side
is the power demand amount in each building, the second term expresses the
power consumption of the water electrolyzer, the third term expresses the power
consumption of the heat pump, and the fourth term expresses the power consump-
tion of the auxiliary machinery (the pump of the hot water network, and the com-
pressor of hydrogen and oxygen).
The heat balance of the system is expressed below.
M M M
Hf,m,t + Hst,t + Hhp,t = Hneed,m,t + Hhw,mm,t (4.2)
m=1 m=1 m=1
The first term on the left-hand side of Eq. (4.2) expresses the exhaust heat of the
fuel cell of M , and the second and third terms express the heat output from the
heat storage tank and the heat pump, respectively. The right-hand side of Eq. (4.2)
expresses heat consumption, the first term is the heat demand of each building
connected to the network, and the second term expresses the heat release of the hot
water piping that connects each building. H hw, mm, t expresses the heat release of
the hot water piping that connects Building m to Building m , and is calculated
from Eq. (4.3).
H hw, mm', t = h D p l mm' (Tm,out, t Tatm, t ) (4.3)
Equation (4.4) is the balance equation of hydrogen. The first term on the left-
hand side of Eq. (4.4) expresses the quantity of hydrogen production of the water
electrolyzer, the second term expresses the hydrogen quantity supplied to the net-
work from the cylinder, and the third term expresses the quantity of hydrogen
production of the reformer. Moreover, the right-hand side expresses the hydrogen
consumption of the fuel cell of M . Equation (4.5) is a balance equation of oxy-
gen. The first term on the left-hand side expresses the oxygen concentration of
the water electrolyzer, the second term expresses the amount of oxygen supplied
from the cylinder, and the third term expresses the amount of oxygen in the air
supply of the blower. The right-hand side is the amount of oxygen consumed with
the fuel cell.
M
Q el, H 2 , t + Q a, H 2 , t + Q r, H 2 , t = Q f,m, H ,t2 (4.4)
m =1
M
Q el,O 2 , t + Q a, O 2 , t + Q bw, O 2 , t = Q f,m,O ,t 2 (4.5)
m =1
68 4 Fuel Cell Network System Considering Reduction in Fuel Cell Capacity
The exhaust heat of each fuel cell connected to the network is used for buildings in
which a fuel cell is installed, which is given priority. The surplus heat of each
building is recovered in the hot water network. On the other hand, when the heat
of a certain building runs short, heat is received from the hot water piping. More-
over, when the heat of the overall network system runs short, heat is supplied to
the network from the heat storage tank and the heat pump. When the network has
excess heat, surplus heat is stored in the heat storage tank. The heat pump is oper-
ated when the heat of the heat storage tank is insufficient.
The load-leveling method using water electrolysis is employed in the R1 type (the
R2 type also uses the same procedure). In order to determine E sep in Fig. 4.5, an
initial value is decided at random concerning the given power demand pattern. At
this time, the amount of production of hydrogen and oxygen in a low-load period
is calculated, and hydrogen and oxygen consumption in a high-load period are also
calculated. The balance is calculated from the amount of production and consump-
tion of hydrogen and oxygen. The value of E sep is changed, and it is repeatedly
calculated until the balance of hydrogen and oxygen becomes sufficiently small.
In the analysis case in following section, the time of less than 1% of the balance
error was adopted. Balance equation (4.1) of the power, balance equation (4.4) of
hydrogen and oxygen, and Eq. (4.5) are used for the calculation of the balance of
hydrogen and oxygen. Figures 4.4 and 4.7 are used as the power generation char-
acteristics of the fuel cell and the characteristics of the water electrolyzer. When
the fuel cell capacity in the analysis exceeds that of Fig. 4.4, it is assumed that the
relation of Fig. 4.4 can be extrapolated.
Figure 4.8 shows the heat release model of the hot water piping. The fuel cell is
installed in four houses, Buildings A to D. Each building is connected with piping,
and hot water returns to Building A. The machinery room is set in Building A, and
the heat outputs of the heat storage tank and the heat pump installed in this ma-
chinery room are H st, t and H hp, t . There is heat demand of H need,A,t to H need,D, t
in Buildings A to D, respectively. In the fuel cell installed in each building, there
is exhaust heat power output of H f,A,t to H f,D, t . Therefore, the heat balance of
Buildings A to D is calculable from Eq. (4.2). Moreover, the heat release (from
H hw,AB,t to H hw,DA,t ) of the hot water piping that connects each building is
calculated using Eq. (4.3). Tatm,t in Eq. (4.3) employs the outside air temperature
in Tokyo, as shown in Fig. 4.9 [35].
(3) Route Planning of Hot Water Piping Considering Heat Release Loss
Since a fuel cell is placed in each building, for the R2 type it is necessary to de-
termine the capacity of each fuel cell. The outlet hot water temperature of a certain
building is decided by the heat balance in the building, and the heat release of the
hot water piping is calculated from the difference between the outlet hot water
temperature and the outside air temperature. Therefore, the heat release of the
overall network differs according to the capacity of the fuel cell installed in each
building. In this chapter, as shown in Fig. 4.10, information on the capacity of the
fuel cell installed in each building and the piping route is expressed with genes,
and these are installed into a genetic algorithm. It is evaluated as a solution with
high fitness, so that there are few values in Eq. (4.6) showing heat release from the
hot water piping. The calculation is iterated, chromosomes are evolved, and a
solution with high fitness is sought. In the last generations chromosomes, a solu-
tion with the highest fitness is determined as an optimal solution. From the infor-
mation on the optimum chromosome, the capacity of the fuel cell installed in each
building and the piping route are determined.
Period M
F= Hhw,l,t (4.6)
t =1 l =1
When threshold value E sep of a low-load region and a high-load region is calcu-
lated according to the procedure of the previous section, a representative day in
January is 109 kW and a representative day in May and August is 125 kW. By
installing E sep into the load leveling described in previous section, the fuel cell is
made to follow the power load pattern of Fig. 4.12. Figure 4.14 shows the fuel cell
exhaust heat in this case. In the heat balance on a representative day in January
shown in Fig. 4.14, since heat runs short in the 7:00 to 17:00 period, the heat
pump is operated. On the other hand, the heat supply and demand on representa-
tive days in May and August show much heat surplus. Moreover, Fig. 4.15 shows
the calculation result of the electrode surface of the fuel cell at the time of install-
ing Esep and performing load leveling. The fuel cell electrode surface in Fig. 4.15
expresses the fuel cell capacity. The data enclosed within the broken line in
Fig. 4.15 are power generation using hydrogen and oxygen. These gases are pro-
duced using the power generated with reformed gas and air. The fuel cell gener-
ated with reformed gas and air is operated at times other than the broken-line re-
gion in Fig. 4.15. At the time of high load from 12:00 to 16:00 on a representative
day in August, about 180 m2 electrode surface was conventionally taken. If load
leveling using water electrolysis is employed, the fuel cell can be reduced to a
120 m2 electrode surface, which is equivalent to 2/3 at the peak at 20:00.
4.4 Case Study 73
The result of the outlet hot water temperature of each building that composes the
network is described. The outlet hot water temperature differs according to the
R1 type or the R2 type. Moreover, since the heat release of the hot water piping
differs according to the outside air temperature, the sampling time is different. The
results of 4:00 and 16:00 on representative days in January and August are shown
in Fig. 4.16. As Fig. 4.12 shows, the sum of the power demand of each building
connected to the network at 4:00 on representative days in January and August is
small. On the other hand, this value is large at 16:00. The horizontal axis in
Fig. 4.16 is the route order (Nos. 17) of the hot water piping. Letters A to G in
Fig. 4.16 correspond to the building number shown in Fig. 4.13. For example, in
the analysis results at 4:00 and 16:00 for the R1 type on a representative day in
January, hot water flows in the order of GFDCABE. The optimal path on a repre-
sentative day in January for the R1 type is GFDCABE, and the optimal path on
a representative day in August is BEGFDCA. This way, the starting points of the
hot water differ according to each month. Moreover, the route of the representa-
74 4 Fuel Cell Network System Considering Reduction in Fuel Cell Capacity
Fig. 4.16 Hot water temperature of outlet piping of buildings in January and August
tive days of both months for the R2 type is GEBACDF. The routes GFDCABE,
BEGFDCA, and GEBACDF are the same as a result of the shortest route shown in
Fig. 4.13. However, GFDCABE and BEGFDCA are clockwise rotations, and
GEBACDF is counterclockwise. The outlet hot water temperature of each building
differs in the starting point of the hot water piping, route and the flow direction, as
shown in Fig. 4.17. Figure 4.17 shows the result of the hot water temperature
when setting the starting point of the hot water piping as B, E, or G. Figure 4.18
shows the result of the hot water piping heat release relevant to the piping routes
in Fig. 4.17. Figure 4.19 shows the result of the piping heat release in the network
on a representative day every month. Under these analysis conditions, the differ-
ence in the heat release for the R1 type and R2 type on a representative day is less
than 3% every month. Considering the analysis error of the genetic algorithm,
these can be estimated as the same value. Therefore, if the heat release of the R1
type and R2 type is optimized, it will converge at almost the same value. How-
ever, since the R1 type in this case assumes that the starting point of the hot water
piping is movable to arbitrary buildings according to the month, it is not realistic.
4.4 Case Study 75
Figure 4.20 shows the result of the fuel cell arrangement plan for the R2 type. The
fuel cell capacity installed in each building is a circle of the broken line in
Fig. 4.20. When the electrode surface of each building shown in Fig. 4.20 is added,
it is 97 m2. The electrode surface when installing the capacity reduction by load
leveling is 120 m2. Furthermore, if the optimum arrangement plan of a fuel cell is
installed, the electrode surface will be reduced to 97 m2. When load leveling using
water electrolysis and optimization of fuel cell distribution are installed, the fuel
cell electrode surface is reduced by 46% compared to the conventional system.
76 4 Fuel Cell Network System Considering Reduction in Fuel Cell Capacity
4.5 Conclusions
A load-leveling method that supplies air and water electrolysis oxygen to the cath-
ode of the fuel cell has been proposed for the fuel cell energy network system.
Furthermore, the optimum operation plan of the hot water network has been pro-
posed, and the fuel cell capacity of each building, position of the machinery room,
piping route, and hot water flow direction have been investigated. The fuel cell
energy network composed of individual houses, a hospital, a hotel, a convenience
store, an office building, and a factory has been analyzed, and the following con-
clusions obtained.
(1) If the load-leveling method is used, the installed capacity of by fuel cell will
be reduced by 34% compared with the conventional system.
(2) Moreover, when fuel cell distribution is optimized, in accordance with the
effectiveness of (1), there is a 46% reduction compared with the conventional
system.
Chapter 5
Equipment Plan of a Compound
Interconnection Micro-grid Composed
of Diesel Power Plants and a Fuel Cell
5.1 Introduction
Installation of the fuel cell micro-grid in an urban area is the technology of spread-
ing the utilization of hydrogen energy. For example, hydrogen production using
green energy and reforming technology of natural gas can be introduced. Gener-
ally, as for the introduction of a micro-grid in a city area, the following points are
expected: (a) The distance for the heat supply is short and effective use of exhaust
heat is possible [38, 39]; (b) load leveling of existing large-sized power generation
equipment is possible; and (c) since a facility suitable for the energy demand char-
acteristics of a community can be installed, it is expected to be a technology in
which energy efficiency is high and environmental impact is low [21, 22, 40].
However, the proton exchange membrane-type fuel cell (PEM-FC) has expensive
electrode material (catalyst material and a solid polymer membrane). Furthermore,
since the system is complex, it is difficult to commercialize it immediately. Thus,
reduction of the number of installations of the expensive fuel cell by connecting
PEM-FC to a micro-grid, and supplying power to two or more buildings is consid-
ered in this chapter. However, the subject of this system is the frequent partial-
load operation with low efficiency, when power is supplied to two or more build-
ings using a large-capacity fuel cell (FC). As technology to solve this issue, a fuel
cell is divided into small-capacity units, and as the method of increasing the load
factor of each unit is used [25]. However, by this method the number of fuel cell
units increases greatly, and facility costs increase. Consequently, the base load of a
micro-grid is supplied using a diesel engine power generator (DEG), and how to
install and interconnect two or more PEM-FC grids is examined. The compound
grid of DEG and PEM-FC is interconnected in this chapter. This micro-grid is
termed CIM (compound interconnection micro-grid). In CIM there are an inter-
connection system and an independent system. The interconnection system is
connected with other grids, such as commercial power, and operated. Although
achieving an independent system is predicted to be difficult compared with the
interconnection system, it is considered that the effect of (b) and (c) described
above will be large. Thus, this chapter examines the independent micro-grid sys-
tem built by two or more FC grids in which system interconnection is possible.
To date there have been many cases of DEG being installed as cogeneration,
and the characteristics, such as the power generation efficiency, facility costs, and
power cost due to a number of achievements, can be estimated. Although it is
expected that the micro-grid using DEG has high potential, it is accompanied by
the problem of carbon dioxide discharge. Consequently, DEG is introduced as
generating equipment corresponding to the base load of the whole CIM, and op-
eration near the maximum efficiency point is examined. On the other hand, the
dynamic characteristics at the time of load fluctuation and carbon dioxide emis-
sions of PEM-FC are good compared with DEG [41]. PEM-FC linked to CIM is
controlled to operate correspondingly to the load fluctuation of the grid. However,
the load factor of PEM-FC changes with the power demand patterns of each build-
ing linked to CIM in this system. As a result, the power generation efficiency of
the whole micro-grid may improve by dividing a micro-grid into two or more
parts, and determining the buildings connected to each grid (grid route) with the
object of maximizing the power generation efficiency. In this chapter, an inde-
pendent micro-grid with a high power generation efficiency is planned by optimiz-
ing the capacity of DEG and PEM-FC, and selecting the grid routes.
This chapter examines two or more FC grids and the independent micro-grid built
by DEG operated according to a base load. However, herein only the power sys-
tem is investigated, and the exhaust heat of the generating equipment is not
touched upon. Figure 5.1 shows the CIM model that introduced seven FC grids
into 61 buildings. DEG is installed in the buildings of any grid and outputs con-
stant power corresponding to the base load of all the FC grids. The power demand
patterns of each building in a city area differ. Therefore, the power load pattern of
each FC grid changes with the route of the building linked to the grid. Conse-
quently, as shown in Fig. 5.1(a), the building linked to each FC grid is selected
and arranged with the object of maximizing the power generation efficiency. Fig-
ure 5.1(b) shows the model of the FC grid (FC grids A to G) in part (a), and power
supply-and-demand is possible for each grid through a system interconnection
device (CP1CP7). The system interconnection between the grids is effective when
supplying power from another system in the case of an accident, maintenance,
etc., and when there is large load exceeding a certain grid capacity.
5.2 Compound Interconnection Micro-grid 79
Figure 5.1(c) shows the model of the FC grid; an interconnection device is shown
in parts (a) and (b). In CIM, DEG of with a capacity of Pc,DEG is installed and
PEM-FC with a capacity from Pc,FC,A to Pc,FC,G is installed from FC Grids A to
G, respectively. Each grid can change over and interconnect the system intercon-
nection device of CP1 to CP7. DEG is operated correspondingly to the base load of
the city area model shown in Fig. 5.1(a). DEG is operated by a constant load. The
load fluctuation, power is supplied from FC Grids A to G. As shown in Fig. 5.1(c),
all FC Grids A to G are connectable with DEG.
Figure 5.2(a)(c) shows the facility scheme installed in the building linked to CIM
shown in Fig. 5.1. Figure 5.2(a) shows the facility scheme of a building where
DEG is installed, and Fig. 5.2(b) shows the facility scheme of a building where
PEM-FC is installed. A building with the installed DEG facility shown in
Fig. 5.2(a) is connected to any one grid, and a building with the installed facility
shown in Fig. 5.2(b) is connected to all the FC grids. Figure 5.2(c) shows the facil-
ity scheme of a building in which DEG or PEM-FC is not installed. Generating
equipment composed from a diesel engine, a power generator, a boiler, a heat
storage tank, a system interconnection device, etc., is installed in Fig. 5.2(a).
Moreover, the generating equipment composed from a town gas reformer, PEM-
FC, a boiler, a heat storage tank, a system interconnection device, etc., is installed
in Fig. 5.2(b). In a reformer, reformed gas is produced on a catalyst by making the
combustion gas of town gas into a heat source. Since there is a large amount of
water in reformed gas generated by steam reforming, reformed gas is cooled by
the air supply of a blower with a dryer, and the water is condensed and separated.
In order for the carbon monoxide concentration in the reformed gas in a fuel cell
stack entrance to be several ppm, the carbon monoxide oxidization part is pre-
pared. In the carbon monoxide oxidization part, carbon monoxide is burned on a
catalyst and it changes into carbon dioxide. Reformed gas is supplied to a fuel cell
stack from the carbon monoxide oxidization part, and it generates electricity. The
generated DC power is changed into an alternating current of constant frequency
through an inverter, and is supplied to a system interconnection device. Moreover,
the boiler for heat supply and the system interconnection device for obtaining
power from a grid are installed in Fig. 5.2(c).
5.2 Compound Interconnection Micro-grid 81
The model of the operating method of CIM is shown in Fig. 5.2(d). The power
load of a representative day is divided into the base load of the constant load, and
other loads as shown in the figure. In operating and generating DEG at the base
load, other loads correspond by the power generation of FC. Although DEG is one
set, the FC grid consists of two or more sets. Since the FC grid corresponds to load
fluctuation, it may operate at partial load with low efficiency, but DEG can be
operated by the constant load of the maximum efficiency point.
The output characteristics result of the cogeneration system using DEG is shown
in Fig. 5.3(a). This result is the relation among the calorific heat of the kerosene
fuel supplied to DEG, the engine-cooling-water heating value and the engine
exhaust gas heating value, and the production of electricity. The engine speci-
fications of the cogeneration system of Fig. 5.3(a) are shown in Table 5.1(a).
Moreover, the specifications of a synchronous power generator are shown in
Table 5.2(b). The fuel of a diesel engine is kerosene and uses two cylinders and
four cycles. A power generator is of a single-phase synchronous type, and power
is transmitted through a belt from the power shaft of the diesel engine. If the
amount of kerosene fuel is increased, the production of electricity and the exhaust
gas heating value increase, but the engine-cooling water heating value decreases.
The maximum power generation output is 3 kW, and the kerosene supply heating
value at this time is 9.8 kW. Figure 5.3(b) shows the production of electricity of
DEG and the relation of power generation efficiency obtained by the test. Al-
though the power generation efficiency changes with the number of engine rota-
tions, since this difference is small, the approximated curve shown in Fig. 5.3(b) is
used in the analysis of this research. Moreover, the relation between the load fac-
tor and power generation efficiency shown in Fig. 5.3(b) should be maintained
even if the capacity of DEG changes.
5.3 Equipment Characteristics 83
The output characteristics of the fuel cell stack used in the analysis of this chapter
are shown in Fig. 5.3(c). The maximum power generation efficiency of PEM-FC
shown in Fig. 5.3(c) is 32%. The supply of town gas has a system of changing it
into reformed gas, and a system of supplying it to the burner used as a heat source
of the reforming reaction. The power generation efficiency is calculated by divid-
ing the production of electricity of PEM-FC by the calorific power adding these
two town gas systems [2629]. The electrode area of an anode and cathode of the
examined PEM-FC is 1.0 m2, respectively. Moreover, the cost calculated by divid-
ing the calorific power of hydrogen in the reformed gas by the calorific power of
the two town gas systems described above is defined as the reformer efficiency.
When there is the same pattern among the power-demand patterns of each
building introduced into a city area model, two or more grid routes considered to
be optimal appear.
The analysis flow that searches for the optimal solution of CIM is shown in
Fig. 5.4. First, the power demand data of each building that composes a city area
model is given to the analysis program. Next, the base load of a micro-grid is
calculated from the power demand data, and the capacity of DEG is determined.
Regarding all the FC grid routes, the power generation efficiency for every sam-
pling time is calculated. However, the FC grid route is given to a program widely
assuming the number of grids and the capacity of PEM-FC installed in each grid,
and calculates the power generation efficiency for all the FC grid routes. By add-
ing and equalizing these results, the average generation efficiency on a representa-
tive day is obtained. The power generation efficiency of DEG and PEM-FC is
used to calculate the load ratio from the power demand of all the buildings con-
nected to a grid and the capacity of DEG and PEM-FC that were set up before-
hand, and these are calculated by inserting them into Figs. 5.3(b) and (c). The
power generation efficiency of DEG and each route of FC grid is given to
Eq. (5.1), and the route of FC grid in case FO, Day of Eq. (5.2) is the largest value,
and the capacity of DEG and PEM-FC are decided on as the optimal solutions.
Figure 5.5 shows the power demand model of each building in Tokyo used in the
analysis, and is the mean power load of each sampling time of the representative
day in January (winter), May (mid-term), and August (summer) [3032]. How-
ever, the actual power demand pattern is a meeting of the load that changes rapidly
in a short time, such as an inrush current. In Tokyo, the annual average tempera-
ture for the past five years was 289 K. The average temperature in January is
25.8 K, and the highest and the lowest temperatures on a representative day in
January are 283 K and 25.4 K, respectively. The average temperature in May is
292 K, and the highest and the lowest temperatures on a representative day in May
are 296 K and 288 K, respectively. The highest and the lowest temperatures on a
representative day in August for the past five years have been 302 K and 296 K,
respectively, and the average temperature is 298 K [20]. There is a high power
demand on a representative day in August compared with other months, including
the space-cooling load. The power demand estimate of the family household
shown in Fig. 5.5(a)(d) is difficult, and the power demand estimate of the small
offices and factories indicated in Figs. 5.5(g) and (h) is regular, and comparatively
easy to estimate. Although load fluctuation in a short time is not taken into consid-
eration for the power demand model in the analysis of this chapter, when accom-
panied by load fluctuation, it is necessary to investigate the dynamic characteris-
tics of the grid. The power demand pattern of a family household (Fig. 5.5(a)(d))
shows peaks in the morning and the afternoon. The demand of hotels (Fig. 5.5(e))
stabilized when midnight to early morning was excluded, and there is continuous
power demand at convenience stores (Fig. 5.5(f)) that are open around the clock.
The difference in the time zone of night to early morning with little power demand
and the time zone from morning to evening with a high power demand is clear in
offices (Fig. 5.5(g)), factories (Fig. 5.5(h)), and hospitals (Fig. 5.5(i)).
5.4 Analysis Method 87
Figure 5.6 shows the urban area model used for analysis. The number shown in
the figure is the building number, and also shows the type of each building. The
urban area model is composed from 20 buildings, and Table 5.3 shows the type of
each building. The analysis investigates each case of Table 5.3(a) and (b). Case A
is a model (complex community) assuming an urban area that consists of various
buildings, and Case B is a model (residential area) assuming a residential street. In
addition, each power demand pattern uses Fig. 5.5. For example, building numbers
1 and 2 are family households (two persons), and each power demand model used
for analysis is shown in Fig. 5.5(b). Therefore, the grid route of building numbers
1 and 2 is exchangeable. The power demand model of Fig. 5.5 and the urban area
model of Fig. 5.6 are installed into the analysis program described in Fig. 5.4, and
the efficiency of the CIM system is verified in analysis. However, in the analysis,
the power demand model of a representative day in May of Fig. 5.5 is used. The
analysis using the power demand model of the representative days of other months
is the same as that of the example of a representative day in May; other months are
not analyzed in this chapter.
Figure 5.7(a) shows the rate of the power demand of a representative day in May
in the urban area model of Case A of Table 5.3(a). A representative day shows the
greatest power demand to be for convenience stores (two buildings), followed by
hotels, factories, and small hospitals, in that order. As Fig. 5.5 describes, the dif-
ference in the power demand for day and night is comparatively small with con-
venience stores, hotels, and small hospitals, and is large for small offices and fac-
tories. There is a difference between family households and apartments in the
amount demanded from midnight to early morning, and daytime. In order to main-
tain the high power generation efficiency of the whole micro-grid, it is necessary
to plan the path of the FC grid containing convenience stores, hotels, factories,
small hospitals, etc., with a large power demand. The grid route shown in
Fig. 5.7(b) was obtained from the results of analyzing the power generation effi-
ciency of each grid route. Figure 5.7(b) consists of three FC grids: Grid A, Grid B,
and Grid C. Figure 5.7(c) shows the analysis results of the generation capacity of
FC installed in each FC grid, and the generation capacity of DEG corresponding to
the base load. PEM-FC of 10 kW, 15 kW, and 47 kW is connected to each of Grid
A, Grid B, and Grid C, respectively. In addition, 57 kW DEG is installed and it
corresponds to the base load of the whole grid.
Figure 5.7(d) shows the analysis results of the electric energy to be outputted on a
representative day in May by each FC grid. In each grid, there is base power sup-
plied from DEG and power corresponding to the load fluctuation supplied from
FC. The base load of each grid distributes the power outputted by DEG. The out-
put of DEG is larger than FC, removing Grid B by the building composition of
Case A. In addition, the power supply of a representative day has more DEG than
the sum total of each PEM-FC. Figure 5.7(e) shows the analysis results of the
power generation efficiency of FC of each FC grid, and the power generation
efficiency of the whole grid. Although the power generation efficiency of Grid A,
Grid B, and Grid C is 19.2%, 15.3%, and 18.6%, respectively, a base load opera-
tion is added due to DEG, and the power generation efficiency of the whole grid
improves to 27.1%. Because there are two or more buildings with the same power
demand model in the urban area model, the grid routes shown in Fig. 5.7(b) differ,
but there is a case where Figs. 5.7(c), (d), and (e) show the same results. More-
over, one set of DEG or one set of PEM-FC is installed into the conditions of the
urban area model of Case A, and the analysis result of the power generation effi-
ciency of the system that supplies the power demand of all the buildings (central
system) is shown in Table 5.4(a). The power generation efficiency of the DEG
central system and the FC central system is 22.4% and 26.2%, respectively. There-
fore, the CIM system of power generation efficiency (27.1%) proposed in this
chapter is larger. The result of the load distribution of the whole micro-grid of
Case A is shown in Fig. 5.8(a). In this figure, allocation of the load of DEG and
the load of the FC grid (Grid A, Grid B, and Grid C) is shown. The magnitude of
the load during the time zone from midnight to early morning, and others, differs
greatly, and the power generation efficiency of the FC grid and the total efficiency
of the micro-grid (equal to CIM efficiency) are influenced by this.
5.5 Case Study 91
Figure 5.9(a) shows the rate of the power demand of a representative day in May
in the urban area model of Case B in Table 5.3(b). In Case B, family households
account for 18 buildings and convenience stores account for 2 buildings. However,
the power demand rate of convenience stores is 84%, and the power demand rate
of family households is 16%. The grid route shown in Fig. 5.9(b) was obtained
from the analysis result of the power generation efficiency of each grid route.
Figure 5.9(b) consists of two FC grids: Grid A and Grid B. Figure 5.9(c) shows
the analysis result of the generation capacity of FC installed in each FC grid, and
the generation capacity of DEG corresponding to the base load. PEM-FC of 8 kW
is connected to each of Grid A and Grid B, and DEG of 33 kW is operated as a
base load of the whole grid.
92 5 Equipment Plan of a Compound Interconnection Micro-grid
Figure 5.9(d) shows the analysis result of the electric energy to be outputted on a
representative day in May by each FC grid. From the analysis result of Fig. 5.9(d),
with the composition of the buildings of Case B, the output of DEG is overwhelm-
ingly larger than FC, and the power supply of a representative day has more DEG
than the sum total of each PEM-FC. Figure 5.9(e) shows the analysis result of the
power generation efficiency of FC connected to each FC grid, and the power gen-
eration efficiency of the whole grid. Although the power generation efficiency of
Grid A and Grid B is 19.5% and 14.5%, respectively, a base load operation is
added due to DEG, and the power generation efficiency of the whole grid im-
proves to 29.9%. In addition, one set of DEG, or one set of PEM-FC, is installed
into the conditions of the urban area model of Case B, and the analysis result of
the power generation efficiency of the system (central system) that supplies the
power demand of all the buildings is shown in Table 5.4(b). The power generation
5.6 Conclusions 93
efficiency of the DEG central system and the FC central system is 23.2% and
29.1%, respectively. Therefore, the CIM system (29.9%) of power generation
efficiency proposed in this chapter is larger. The result of the load distribution of
the whole micro-grid of Case B is shown in Fig. 5.8(b). There is little load distri-
bution of the FC grid compared with Case A, and the difference in the load at each
sampling time is small. This is why the power generation efficiency (equal to the
power generation efficiency of CIM) of the whole micro-grid is high.
5.6 Conclusions
CIM, which divides the grid of an independent micro-grid into multiple units and
interconnects between grids, was proposed in this chapter. Although there are
many examples of introducing DEG (diesel power plant generator) as cogenera-
tion until now, there has been the problem of carbon dioxide emission. Therefore,
in this chapter, DEG was installed as generating equipment corresponding to the
base load of whole CIM, and the method of operating a proton exchange mem-
brane-type fuel cell (PEM-FC) so that it may correspond to the load fluctuation of
the grid was investigated by numerical analysis. The generation capacity and the
grid route of DEG and PEM-FC in the case of maximum power generation effi-
ciency of the whole micro-grid were sought. From the results of this analysis, an
independent micro-grid with high power generation efficiency was planned. As a
result, compared with the method of installing DEG or PEM-FC into a micro-grid
independently (central system), the power generation efficiency is confirmed to
have improved. Moreover, it has been verified that the power generation effi-
ciency of a micro-grid, the number of FC grids introduced, and the capacity of
DEG and PEM-FC change with the types of building that compose an urban area.
In the complex community model and the residential area model investigated in
this chapter, the power generation efficiency of the whole micro-grid (CIM) was
27.1% and 29.9%, respectively. Although the micro-grid that combines DEG and
PEM-FC is advantageous with respect to power generation efficiency and carbon
dioxide emission, studies including those on the increase in equipment costs are
now required.
Chapter 6
The Effective-use Method of Exhaust Heat
for Distributed Fuel Cells
6.1 Introduction
When a proton exchange membrane fuel cell cogeneration system with a town gas
reformer is distributed to each building and energy is supplied, the transportation
loss of the heat supply is small. However, if the fuel cell system is installed in
each building, a drop in energy efficiency is predicted due to the imbalance be-
tween the rate of heat supply and electric power supply. Although the installation
of a battery and a heat storage tank is effective, using these methods, the system
and the operational plan become complicated, and serve to raise the equipment
costs. In this study, the fuel system (the hydrogen piping network), the electric
power system (the power line network), and the heat-power system (the hot water
piping network) of the fuel cells installed in each building are connected, and the
best method of satisfying two or more of electric power and heat loads of the
buildings in cooperation is examined. When two or more fuel cells are connected
in a network, then the method of cooperation and control of the electric power
output and the heat power output is called a fuel-cell-energy-network (FEN) in
this chapter. In an FEN, it is possible to stop a fuel cell, which represents the op-
erational status of a partial load with low efficiency in the cooperative operational
control by FEN, to utilize the electric power generated by other fuel cells and to
improve the efficiency of the system as a whole. Since being charged, from now
on according to the amount of discharge of carbon dioxide gas is also considered,
connecting renewable energy equipment and unused energy equipment to the
FEN, which is composed of tens of buildings, is also considered. Compared to a
conventional system, which is installed as an independent system in each building,
the FEN energy supply system, which is comprised of common auxiliary machin-
ery (a heat storage tank, an auxiliary heat source, a reformer, etc.) confined to one
machinery room, reduces the overall facility costs. A reformer installed in the
machinery room produces reformed gas with a high hydrogen concentration from
town gas. The existing town gas network can be used as the hydrogen piping net-
work for supplying reformed gas to the fuel cell in each building. Using the exist-
ing power line network, the exchange of electric power, based on supply and de-
mand, produced by the fuel cells in each building between each other is possible
between the buildings connected to the FEN. However, a hot water piping network
must be newly installed for the construction of an FEN, and piping needs to be
connected to all the buildings; therefore the heat loss from the piping is an impor-
tant issue. It is expected that the total amount of heat released in a hot water piping
network varies between networks depending on the path of the piping.
In this chapter, a scheme for determining and predicting the optimal piping path
for minimizing the heat release from a hot water piping network has been devel-
oped. For planning a hot water piping network, the program uses a genetic algo-
rithm (GA) [14, 16], and a detailed analysis has been performed in this study.
With respect to optimization using GA, previous reports have explored the piping
path and the layout of equipment [4244]. However, there have been no studies on
the analysis of an energy network using fuel cells and other energy equipment.
Sapporo City in Japan is used as an urban area model for the path planning pro-
gram of the hot water network in this study, and the optimal path and the amount
of heat release of the hot water piping network in FEN are investigated. It is ex-
pected that the actual electric power loads and heat loads of buildings deviate
sharply from the average energy demand patterns. Therefore, in this chapter, the
optimal path and the amount of heat released in a hot water piping network in the
presence of random load fluctuations for an average electricity demand pattern are
also investigated.
An example of applying the hot water piping network by FEN to two or more
buildings is shown in Fig. 6.1. S1 to S7 in Fig. 6.1 indicate the buildings in the
network, and L Si (i +1) is the length of the hot water piping that connects Si and
Si +1 . The hot water piping path shown in Fig. 6.1 is in an order that connects each
building by the shortest distance. When the paths of the hot water piping differ,
the amounts of heat released from the network also differ. Therefore, effective use
of the fuel cell exhaust heat requires a path plan that minimizes the amount of heat
released from the hot water piping network.
Figure 6.2(a) presents a machinery room scheme showing the installation of com-
mon facilities of an FEN. It is assumed that the machinery room is installed in the
building S m linked to the FEN. This machinery room is the starting point of the
hot water piping network. A town gas reformer, a heat storage tank, a radiator, and
backup heat source equipment are installed in this machinery room. Town gas,
with the quantity of flow Q f,rm,t , and water are supplied to a reformer, and re-
formed gas with high hydrogen concentration is produced from the town gas by a
steam reforming reaction using a catalyst. Steam reforming is an endoergic reac-
tion and the heat source of this reaction is a supply of town gas of the quantity of
flow Q f ,cb,t to a burner. The composition of hydrogen and steam in reformed gas
varies considerably. The reformed gas is then cooled with atmospheric air and a
dryer is used to condense and remove moisture. Carbon monoxide with a concen-
tration of several percent is typically present in the reformed gas when it comes
out of the dryer. If this reformed gas containing carbon monoxide is supplied to a
fuel cell, the reaction in an electrode catalyst will be blocked due to poisoning, and
the power generation performance will fall. Therefore, CO oxidization is carried
out using oxidation equipment so that the concentration of the carbon monoxide
gas in the reformed gas can be reduced to 10 ppm or less.
The exhaust heat output H rm,t from the reformer is stored in the heat storage
tank using a heat medium. The heat in the heat storage tank can be supplied to
each building through the hot water piping network. However, when the quantity
of heat, including H rm,t , H in,Sm , t of the return of the hot water piping network,
and H st, t of thermal storage, exceeds the overall heat storage capacity, the excess
heat H ra,t is emitted out of the system via a radiator. On the other hand, when the
quantity of heat obtained by adding H rm,t , H in,Sm , t , and H st, t is less than the heat
demand in each building, a quantity of heat H bh, t is supplied to the heat storage
tank using auxiliary heat sources, such as a boiler and a heat pump.
At first, after supplying the heat output from a fuel cell installed in a building
to the heat demand of the building, excess heat (deficiency) is output (input) from
the hot water piping network. As shown in Fig. 6.2(b), a fuel cell, an inverter, and
a heat exchanger, which can output or input heat from or to the hot water piping
network, are installed in each building linked to the FEN. When the building in
Fig. 6.2(b) is set to Si , it has a hot water input with quantity of heat H in,Si 1 , t
at temperature Tin,Si , t from the hot water piping network. Moreover, when
the heat demand in Si is set to H n,Si , t and the exhaust heat output of a fuel
cell is set to H f,Si , t , the hot water output from Si is the quantity of heat
H out,Si , t = H in,Si 1 , t + (H f,out,Si , t H n,Si , t ) at temperature Tout,Si , t . The electric
power generated by a fuel cell is E out,Si , t , and the electric power after changing
into the regulation frequency of exchange with an inverter is E sys,Si , t . Although
the electric power, excluding the electricity demand E n,Si , t from E sys,Si , t , can be
supplied to any building in the FEN through the power line network, this is not
taken into consideration in the case analysis described later.
A system that includes only one fuel cell installed in a machinery room and that
supplies electric power in each building is called a centralized system. Fig-
ures 6.3(a)(e) show the model of the FEN and a centralized system. Figure 6.3
6.2 Outline of the Fuel Cell Energy Network System 99
Fig. 6.3 The model of the FEN and a centralized system showing the relationship between the
allocation of distributed fuel cells and heat release
shows (a) the installation of the hot water piping, (b) the path of the hot water
piping, (c) the fuel cell capacity in each building, (d) the change in hot water tem-
perature, and (e) the amount of heat release from the piping per unit length in the
FEN. In the model shown in Fig. 6.3, the machinery room in the FEN and a cen-
tralized system is installed in Building S1 , and hot water for heat supply flows in
the order of the Buildings S1 , S 2 , S3 , S 4 , S5 , S6 and S7 , as shown in
Fig. 6.3(b). However, as shown in Fig. 6.3(a), the hot water piping is equipped
with heat insulating material, and all the piping is exposed to the open air and
connects each building. Moreover, as shown in Fig. 6.3(c), in a centralized system,
100 6 The Effective-use Method of Exhaust Heat for Distributed Fuel Cells
only one fuel cell is installed in the machinery room of Building S1 , and a fuel cell
is installed in the FEN based on the capacity required for supplying the maximum
electric load for all buildings. Figure 6.3(d) shows the model for the hot water
temperature as a function of the building location. In a centralized system, the hot
water input into S1 at temperature Tin,S1 , t is output at temperature Tout,S1 , t at the
exhaust of a fuel cell. In this system, there is no heat source heating the hot water
after this point, and due to the heat demand in each building and the heat released
from the piping connecting each building, the hot water temperature falls to
Tin,S1 , t , and the hot water returns to S1 . With a centralized system, since the hot
water temperature falls sequentially as the hot water progresses from Building S1
to S7 , the difference in temperature between the outside temperature and the hot
water gradually becomes smaller, as shown in Fig. 6.3(e), and it is assumed that
the amount of heat release per hot water piping unit length becomes sequentially
smaller as well. On the other hand, the fuel cells are installed in all the buildings
by the FEN, and the outlet hot water temperature for each building is determined
by the balance relation of the quantity of heat of the hot water input into each
building, the amount of exhaust heat of the fuel cell installed in the building, and
the heat demand in the building. Therefore, the outlet hot water temperature of the
buildings will be changed upwards or downwards as shown in Fig. 6.3(d), and the
hot water goes to each building. As shown in Fig. 6.3(e), as a result, the heat re-
lease per unit length of piping changes in an upwards or downwards direction. The
amounts of heat released in the whole piping will differ based on the course plan
of the hot water piping of the FEN, and the piping path is likely to affect the sys-
tem-wide energy efficiency.
Assuming that the buildings with fuel cells M are connected to the FEN, and that
the fuel cells M generate electricity at a sampling time t, the heat balance of the
FEN is then expressed using Eq. (6.1).
M M M
Hf,S ,t + Hrm,t + Hbh,t + Hst,t = Hn,S ,t + Hr,S
i i i (i +1) , t
+ H ra,t (6.1)
i =1 i =1 i =1
The left-hand side of Eq. (6.1) shows the heat outputs, and the right-hand side
shows the consumption of heat. The first term of the left-hand side is the exhaust
heat output by the fuel cell M , the second term expresses the exhaust heat of a
reformer, the third term expresses heat supply from an auxiliary heat source, and
the fourth term expresses the heat output from the heat storage tank. The first term
of the right-hand side is the heat demand in each building, the second term is the
amount of heat released by the hot water piping that connects Si and Si +1 , and the
third term expresses the amount of heat released in the radiator installed in the
machinery room.
6.2 Outline of the Fuel Cell Energy Network System 101
H r,Si (i +1), t in Eq. (6.1) is calculated by the method described below. Figure 6.4(a)
shows the model for heat input and output from the hot water piping that connects
Buildings Si and Si +1 , and the amount of heat released in the piping is set to
H r,Si (i +1) , t . The hot water temperature is Tin,Si , t and the quantity of heat H in,Si , t
is input into building Si through the network. When carrying out the power gen-
eration operation of the fuel cell installed in Si so that the electricity demand
E n,Si , t in sampling time t may be satisfied, the exhaust heat output of the fuel cell
is H f,Si , t . Moreover, H n,Si , t is the heat demand in Si . The quantity of hot water
heat H out,Si , t output from Si is H in,Si , t + H f,Si , t H n,Si , t . Here, the values of
H f,Si , t vary based on the value of E n,Si , t , as described in the next section. Al-
though the temperature of the hot water output from Si is Tout,Si , t , by the time the
hot water reaches Si +1 , there will be heat release H r,Si (i +1) , t from the piping.
Therefore, for the hot water input into Si +1 , the temperature falls to Tin,Si +1 , t , and
the quantity of heat is H in,Si +1 , t = (H out,Si , t H r,Si (i +1) , t ) . Furthermore, in Building
Si +1 , since power generation operation of the fuel cell is performed so that the
electricity demand E n,Si +1 , t is satisfied, the exhaust heat H f,Si +1 , t is output. This
calculation is repeated for the following condition for all buildings: For a heat
demand of H n,Si +1 , t in Si +1 , the quantity of heat Si +1 of the hot water output from
H out,Si +1 , t is H in,Si +1 , t + H f,Si +1 , t H n,Si +1 , t . Figure 6.4(b) shows the model of the
hot water piping connecting Si and Si +1 . The bore diameter of the hot water pip-
ing is expressed as D i,Si (i +1) , the outside diameter is expressed as D o,Si (i +1) , and
the outside diameter of the heat insulating material, with which the piping is
Fig. 6.4 Heat model for the hot water piping network
102 6 The Effective-use Method of Exhaust Heat for Distributed Fuel Cells
equipped, is set to D c,Si (i +1) . The heat conductivity of the piping material and
thermal insulation are set to k p and k c , respectively. The coefficient of overall
heat transmission K Si (i +1) between the hot water and the surface of heat insulat-
ing material is determined as follows.
1 1 Do,Si(i+1)
K Si(i+1) = 1 + ln
h w,Si( i+1) Di,Si(i+1) 2 k p Di,Si(i+1)
(6.2)
1 Dc,Si(i+1) 1
+ ln +
2 kc Do,Si( i+1) h ,Si( i+1) Dc,Si(i+1)
In the analysis example in this chapter, although the hot water piping is
installed above the ground, K Si (i +1) can also be obtained by similar calculations
for piping buried underground. In the analysis example herein, the hot water
piping used is assumed to be a hard vinyl pipe with an outside diameter
D o,Si (i +1) = 50 mm, and a bore diameter D i,Si (i +1) = 40 mm, and it is assumed that
the piping is covered with a 30 mm thick polystyrene-foam system heat insulating
material ( D c,Si (i +1) = 110 mm). Moreover, the heat transfer coefficient h w of the
hot water in the piping is set to 3.5 kW/m2K, and the heat transfer coefficient h
between the heat-insulating material surface and atmospheric air is 11.6 W/m2K.
These values are assigned to Eq. (6.2), and as a result, K S, t is calculated to be
0.05 W/m2K. The outlet hot water temperature for Building Si is Tout,Si , t , and
the outdoor air temperature in sampling time t is T, t . The amount of heat re-
leased H r,Si (i +1), t in the piping of length L Si (i +1) that connects Si and Si +1 for
this time is calculated by the following equation.
H r,Si (i +1), t = K Si (i +1) D o,Si (i +1) L Si (i +1) ( Tout,Si , t T , t ) (6.3)
The relationship between the ratio of electric power load to the power generation
capacity of a fuel cell, defined as the load factor, and the ratio of heat output to the
electric power output, calculated from the results of the power generation of a
proton exchange membrane fuel cell stack, is shown in Fig. 2.7 [45]. This essen-
tially represents the performance of the fuel cell stack. The characteristics of
Fig. 2.7 are the measurement results of the electric power output at an ACDC
converter exit, and the heat output in a fuel cell stack exit. The load factor for the
electric power load E n,Si , t of the fuel cell in Building Si is calculated, and the
heat output H f,Si , t of the fuel cell can be calculated using the plot in Fig. 2.7.
6.3 Model of the Fuel Cell 103
The urban area considered in the model is assumed to be Sapporo city in Japan,
where the FEN is introduced as shown in Fig. 6.5. The total number of buildings
in the urban area model is 15. The use of the buildings is shown in Fig. 6.5. The
buildings include an individual house of a single-person household (SF in
Fig. 6.5), a two-person household (DF), a three-person or four-person household
(F), two households living together (DH) consisting of five or more persons, a
small-scale office (SO), and an apartment house (AP1). Also, in Fig. 6.5, the
shortest path of the hot water piping going via all the buildings is also shown. The
electric power and thermal energy demand patterns for each building in winter
(February), summer (August), and mid-term (May) as used in the analysis in the
following section is shown in Fig. 6.6 [31]. The power generation capacity of the
fuel cell linked to FEN makes the maximum value of the annual power consump-
tion in each building nearly 1.2 times. The power generation capacity of the fuel
cell used in the analysis in the following section is shown in Table 6.1.
In Sapporo city in Japan, which is in a cold and snowy area, the annual mean tem-
perature for the past five years has been 282 K. The average temperature in Febru-
6.4 Case Analysis 105
ary is 270 K, and the highest and lowest temperatures on a representative day in
February are 273 K and 266 K, respectively. Moreover, the average number of
snowy days for February is 25 days. On the other hand, the average temperature
for a representative day in July day for the past five years has been 293 K, and the
highest and the lowest temperatures 298 K and 290 K, respectively. The average
temperatures in Sapporo for the sampling time on representative days in February,
May, and August are shown in Fig. 6.7 [46]. There is no summer cooling load in
Sapporo. Electricity demand includes household appliances and electric light, and
heat demand comes from heating, the hot water supply, and baths. Thermoelectric-
ity ratios for a representative day in February and August are 0.90:0.1 and 0.5:0.5,
respectively. The area of an average individual house (three-person or four-person
household) in Sapporo is 140 m2, the number of stories is two, and the houses are
made of wood.
In this chapter, a program for the hot water piping path using the traveling sales-
man problem (TSP) [19] has been developed. The path of the hot water piping is
described by the chromosome model used by the genetic algorithm. Crossovers
and mutations are added to many chromosome models, and an analysis program
evaluates the value of the objective function of each chromosome model. When
the value of the objective function of a certain chromosome model fills the objec-
tive in a better fashion, this indicates that the adaptive value is high. In an analy-
sis program, a chromosome model with a high adaptive value is made to survive
with high probability, and other chromosome models become extinct. The model
with the highest adaptive value is the optimal solution for the chromosome mod-
els, which are repeated and calculated for gene manipulation by crossovers and
mutations of the chromosome models, and survive with the last generation number
decided beforehand. However, if the problem in this chapter is analyzed by a gen-
eral GA, many chromosome models showing a path passing through the same
building two or more times will be generated. For this reason, it is necessary
to extinguish many chromosome models, and the calculation efficiency will then
106 6 The Effective-use Method of Exhaust Heat for Distributed Fuel Cells
fall significantly. Therefore, in this chapter, the view of order expression of the
path by Dewdney is introduced [9]. The gene model in a chromosome model does
not express the number of an actual building with the view of order expression of
the path of Dewdney. A gene model expresses the number of the new buildings
list devised so that the same building is not passed two or more times. Since the
chromosome model, which must be canceled by devising the order expression of
a path does not appear, the efficiency of the path planning analysis program
improves. A chromosome model with a smaller amount of heat release from
the hot water piping on a representative day is evaluated as an individual model
with a high adaptive-value. The objective function FO is shown in the following
equation.
Day M
FO = H r,Si (i +1) , t (6.4)
t =1 i =1
The analysis procedure of planning the path of a hot water piping network is de-
scribed below. First, the coordinates of each ridge of the urban area model shown
in Fig. 6.5, the energy demand pattern of the building shown in Fig. 6.6, and the
outside temperature data of the representative day for each month in Sapporo
shown in Fig. 6.7, are input into the program developed in this chapter. Next, N cr
chromosome models showing the path order of the hot water piping are prepared
at random. For each of these chromosome models, the objective function shown
by Eq. (6.4) is evaluated. The top R cr % of the high chromosome model of the
adaptive value is made to survive, and the remaining chromosome models are
excluded. Crossover and mutations are added to the surviving chromosome mod-
els using the probabilities rcs and rmu .
Furthermore, N ge generation numbers decided beforehand repeat the calcula-
tion for evaluating the adaptive value of the chromosome models and selecting the
individuals with a low adaptive value. The hot water piping path obtained for the
highest adaptive value individual, is made into an optimal path in the previous
generations chromosome models. In the calculation of Eq. (6.4), the heat
H r,Si (i +1) , t released in the piping, which connects each building in the equation, is
calculated using the following procedures. From the electricity demand pattern in
Building i , the electricity demand E n,Si , t in the sampling time t is found. The
power generation capacities of the fuel cells installed in each building are shown
in Table 6.1, and the characteristics of electric power and heat output are obtained
from Fig. 2.7. The load factor of the electric power can then be calculated by the
power generation capacity of a fuel cell, and E n,Si , t and H f,Si , t can be obtained by
applying the load factor and E n,Si , t to the curve in Fig. 2.7.
The heat storage capacity is decided beforehand, and when the heat input ex-
ceeds this capacity, it is assumed that H ra,t of the quantity concerned is emitted
6.5 Analysis Results 107
from a radiator. The amount of heat released from the hot water piping
( H r,Si (i +1) , t ) in the sampling time t can be calculated by the ability to give
H f,Si , t , H rm,t , H bh, t , H st, t , H n,Si , t , and H ra,t to Eq. (6.1). However, the exhaust
heat temperature of a fuel cell is set to 353 K, and the temperature of the hot water
returning from the hot water piping network to the machinery room is assumed to
be nearly 333 K.
Figure 6.8 shows the results of analyzing the planned route of the hot water piping
using the urban area model shown in Fig. 6.5. Although the path of the shortest
distance, which goes via all the buildings, was also shown in Fig. 6.5, the path
analysis results in Figs. 6.8(a), (b), and (c) are the same as the path of the shortest
distance in Fig. 6.5. In Figs. 6.8(a), (b), and (c), the total distance of the network,
the hot water flow direction, and the starting point of the hot water network are
shown. Figure 6.8(d) shows the calculation results of the heat demand of the FEN
on a representative day of each month, and the amount of heat release in a hot
water piping network. Also, Fig. 6.8(e) shows the ratio of the heat released in the
hot water piping network to the heat demand on a representative day of each
month.
As shown in Fig. 6.8(d), the amount of heat release decreases in the order of
winter, mid-term, and summer. The reason for this is that the difference in tem-
perature between the hot water, which flows through the inside of piping, and the
open air becomes smaller, and the amount of piping heat release becomes less,
when the outdoor air temperature becomes higher. On the other hand, the ratio of
the amount of heat release in the hot water piping network to the heat demand for
108 6 The Effective-use Method of Exhaust Heat for Distributed Fuel Cells
Fig. 6.8 The path planning results for minimum heat. The arrangement of the buildings is the
pattern shown in Fig. 6.5.
every monthly representative day shown in Fig. 6.8(e) increases in the order of
winter, mid-term, and summer. Although the quantity of heat transported by the
hot water in a season or time in a hot water piping network differs, this means that
the amount of change of network heat release is smaller than the amount of change
of the quantity of heat in the hot water. Therefore, in an FEN with various heat
demands, the rate of the amount of the heat demand in hot water piping is small,
and fuel cell exhaust heat is used effectively. In the analysis in Fig. 6.8, calcula-
tions were performed using several different energy demand patterns. In the fol-
lowing section, using the arrangement of the buildings shown in Fig. 6.5, the rela-
tionship between the energy demand pattern and the hot water path is investigated.
Figure 6.9(a) shows the urban area model of Fig. 6.5, and the path analysis results
for the hot water piping assuming that the energy demand pattern of all the build-
ings is for a two-person household (DF). The path in the analysis results for repre-
sentative days in February and May is the same as the shortest path shown in
Fig. 6.5, and the starting point building is the same as that in Fig. 6.8. However, in
contrast, the path of the hot water piping for a representative day in August is not
the shortest one. Moreover, although the starting point building was also S15 on
representative days in February and May, the starting point for a representative
day in August is S12 . The heat demand of a building connected to the FEN for a
6.5 Analysis Results 109
Fig. 6.9 Path planning results for minimum heat release. The case which installed a different
pattern into the same energy demand pattern
110 6 The Effective-use Method of Exhaust Heat for Distributed Fuel Cells
Figure 6.10(a) shows the results for the hot water piping path for the energy de-
mand pattern of Building S8 when it is a store (CV), for a time period of 24 hours,
and by setting the energy demand pattern of other buildings to DF. For the energy
demand pattern of CV, since there is a 24-hour electricity demand, the exhaust
heat output of the fuel cell is large, and the path planning results of the hot water
piping for each monthly representative day differs from the path of the shortest
distance. The path of the hot water piping, the starting point building, and the flow
direction of the hot water are also the same for a representative day for each month
in Fig. 6.10(a). The path planning results of hot water piping in the presence of
30% and 60% random load fluctuations to the electricity demand pattern in
6.6 Conclusions 111
Fig. 6.10 The energy demand pattern of all the buildings other than S8, which is a building for
a two-person household (DF). The energy demand pattern of S8 is of a store of day-long business
(CV)
each ridge of the system shown in Fig. 6.10(a) are shown in Figs. 6.10(b) and (c).
The path planning results of the hot water piping shown in Fig. 6.10(b) is the same
as the results when there is no load fluctuation (Fig. 6.10(a)). However, when load
fluctuations of electric power is 60%, as shown in Fig. 6.10(c), the path planning
results of the hot water piping differs from Figs. 6.10(a) and (b). These results
demonstrate that when planning the hot water piping network in an FEN, it is
necessary to take the load fluctuations of electricity demand into consideration.
6.6 Conclusions
For FEN, cooperative control of each operation of the fuel cell installed in two or
more buildings was carried out, and an effective utilization approach of the ex-
haust heat of the fuel cells was considered. In this chapter, path planning at the
time of supplying exhaust heat of the fuel cell is linked through the FEN to each
building through hot water piping was investigated. In consideration of the differ-
ence in temperature of the open air and the hot water in the hot water piping net-
work, a program that searches for the path of hot water piping for minimal piping
heat release was developed. As a result of conducting a case analysis, the optimal
path of the hot water piping network in an FEN confirmed the energy demand
pattern in each building, and the difference in the power generation capacity of the
fuel cell installed in each building. Furthermore, if the fluctuation of the electricity
demand in each ridge linked to the FEN becomes large, the optimal path of the hot
water piping network will be influenced.
Chapter 7
Load Response Characteristics of the Fuel Cell
for Individual Cold-region Houses
7.1 Introduction
characteristics of a system change greatly with the value of the control variables.
When introducing a fuel cell system into a house, it is necessary to consider the
two transient response characteristics: electric power and heat power. However,
there have been no reports on the details of the transient response characteristics of
the fuel cell system composed of a reformer, a fuel cell, an inverter, a system in-
terconnection device, etc. Similarly, there are examples reporting the transient
response characteristics of the heat supply of a fuel cell system composed of fuel
cell exhaust heat, an auxiliary heat source, a heat storage tank, etc. So, in this
chapter, the transient response characteristics of the power in the proton exchange
membrane type (PEM) fuel cell cogeneration with a town gas reformer and heat
output are investigated. However, town gas is natural gas of a methane principal
component. An overall transient response characteristic is investigated by numeri-
cal analysis using the response characteristic of each instrument obtained in the
experiment. Furthermore, the dynamic characteristics of the system when intro-
ducing the energy demand pattern of an individual cold-region house into the fuel
cell system in which a geo-thermal heat pump is installed as an auxiliary heat
source are clarified.
The system outline figure of the PEM fuel cell examined in this chapter is shown
in Fig. 7.1. Further, the assumed specifications of the system are shown in Ta-
ble 7.1. The main equipment of the system is a town gas reformer, a PEM fuel
cell, a dryer, a carbon monoxide oxidization system, an inverter including a
DC/AC inverter, a heat pump, a heat storage tank, and a radiator. The dynamic
characteristics of the heat storage tank are not taken into consideration in the
analysis described below. The dynamic characteristics of a heat output examine
only the geo-thermal heat pump. If a complex system is installed in an individual
house, the time to recover facility costs becomes long. However, the cost of a
PEM fuel cell system may be greatly reduced due to advances in material devel-
opment and manufacturing techniques. The output characteristics of the fuel cell
stack used in this chapter are shown in Fig. 7.2. The maximum power generation
efficiency of the fuel cell stack shown in Fig. 7.2 is 48% [28, 29]. The town gas
supplied to the system is of the quantity reformed by the reformer ( Q f,rm,t , t is
sampling time), and the quantity consumed by the burner as a heat source of the
reformer ( Q f,cb,t ). The relationship to the supply adding these two amounts of
town gas ( Q f,rm,t + Q f,cb,t ) and the power outputted by a fuel cell system
( E Inverter,t ) is shown in Fig. 7.3 [28, 29]. However, Fig. 7.3 shows the characteris-
tics when the electrode area of the anode cathode is 0.5 m2. Here, the value divid-
ing the calorific power of the hydrogen in the reforming gas generated by the
7.2 System Configuration 115
reformer by the calorific power of town gas supply ( Q f,rm,t + Q f,cb,t ) is defined as
reformer efficiency RM, t . In this analysis RM, t is 73% [23, 26]. The power that
can be supplied to the demand side from a PEM fuel cell is a maximum of 1.0 kW,
and the maximum heat power of the heat pump is 15 kW.
In a reformer, reforming gas is produced in the catalyst using the combustion gas
of town gas as a heat source. There is a large amount of water in the reforming gas
generated by steam reforming. Therefore, a dryer is prepared, reforming gas is
cooled by air supply from a blower, and the water is condensed and dissociated. In
order for the carbon monoxide concentration in the reforming gas in a fuel cell
stack entrance to be several ppm, carbon monoxide oxidization equipment is pre-
pared. With carbon monoxide oxidization equipment, carbon monoxide is burned
in a catalyst and changes into carbon dioxide. In this chapter, about 2% of hydro-
gen in reforming gas burns with carbon monoxide oxidization equipment. There-
fore, the efficiency of carbon monoxide oxidization equipment is 98%. These
values refer to past experiments [52]. Reforming gas is supplied to a fuel cell stack
from carbon monoxide oxidization equipment, and after changing the DC power
generated by the fuel cell into AC power of a fixed frequency through an inverter,
the demand side is supplied from a system interconnection device. It is possible to
switch and supply the power generated by the fuel cell system and commercial
power to the demand side. Moreover, commercial power is used for the operation
of the heat pump.
7.2 System Configuration 117
The exhaust heat of a PEM fuel cell and a reformer is stored in a heat storage tank
(ST). Tap water is supplied to the heat exchanger installed in the heat storage tank,
exhaust heat and the heat of the tap water are exchanged, and hot water is supplied
to the demand side. When the exhaust heat exceeds the heat demand, the surplus
heat is stored. Moreover, a heat pump is operated when there is high heat demand
and exhaust heat and thermal storage output are insufficient. However, the re-
sponse time of the exhaust heat output in the reformer and fuel cell stack is long
compared with the response time of the heat output in the thermal storage and heat
pump. Priority is given to the heat output of the thermal storage and heat pump
when the heat demand requires a quick response. When there is a large amount of
exhaust heat and it exceeds the capacity of the heat storage tank, valves VA2 and
VA3 in Fig. 7.1 are operated, and excessive heat is released out of the system.
It is assumed that this system will be used in a cold region. A heat exchanger is
buried in the ground, and the heat source of the heat pump is obtained from the
soil. With the circulation pump PP2 in Fig. 7.1, it circulates through a heat transfer
medium between a subterranean heat exchanger and the low-temperature-side heat
exchanger of the heat pump, and heat is obtained from soil. The heat power in a
trial production geo-thermal heat pump and the test result of the coefficient of
performance (COP) introduced into this system are shown in Fig. 7.4.
Two controllers are installed in the proposed system. As shown in Fig. 7.1, Con-
troller 1 controls the power generation system and Controller 2 controls the heat
supply system. In each Controller, each control variable of a proportional action
(P), an integral action (I), and a derivative action (D) can be set up, and the output
is put close to the target value by feedback control. In Controller 1, the data of the
temperature ( T2 ) of a fuel cell stack and the amount of electricity demand
( E need,t ) is inputted, and in Controller 2, the data of the heat-transfer-medium
118 7 Load Response Characteristics of the Fuel Cell for Individual Cold-region Houses
temperature in the heat storage tank ( T1 ) and the heat amount demanded ( H need,t )
are inputted, and operation commands are outputted to each blower, valve, and
pump.
The power generated by the fuel cell system and commercial power can be sup-
plied to the demand side through a system interconnection device. However, the
auxiliary machinery in Fig. 7.1 described below is operated with commercial
power. The auxiliary machinery operated with commercial power is a blower for a
reformer burner (BW1), a blower for a dryer (BW2), a blower for a fuel cell cath-
ode (BW3), a exhaust heat extraction pump (PP1) of a fuel cell stack, a circulating
pump for a geo-thermal heat pump (PP2), and an electric motor for a heat pump
(MT). By the command of Controller 1 and Controller 2, the switch of SW1, SW2,
SW3, SW4, and SW5 is operated, and each of these auxiliary machines are started
and stopped. Moreover, the number of rotations of MT is controlled by the com-
mand of Controller 2.
Figure 7.5 shows the operation model of the fuel cell system of Fig. 7.1. System
operation when giving the power load shown in Fig. 7.5(a) and the heat load
shown in Fig. 7.5(b) to a fuel cell system is considered. Figure 7.5(c) shows the
model of operation of the switch of a system and valves that are controlled by the
controller. The symbols in this figure correspond to the symbols shown in Fig. 7.1.
In starting up a system, in order to supply air to the burner for the heat source and
the dryer of a reformer, switches SW1 and SW2 are turned ON by the command of
Controller 1, and blowers BW1 and BW2 are operated. At time t1 , in order to
supply air to the cathode of the fuel cell, a command is given to SW3 from Con-
troller 1, and BW3 is operated. A control command is sent so that Controller 1 to
valve VA1 may be opened simultaneously. Figure 7.5(d) shows a fuel cell in op-
eration following the power load. The transient response characteristics, such as
the reformer, fuel cell stack, and inverter, affect the response characteristics of the
system. Therefore, compared with the input characteristics shown in Fig. 7.5(a),
time delay, overshooting, etc., produce the response characteristics of Fig. 7.5(d).
Immediately after start-up, a heat pump with large heat output is the main heat
source. By the control command of Controller 2, the electric motor MT of heat
pump, SW5, and circulating pump PP2 are operated at time t1 . If the operating
time of a system passes and temperature T2 of the fuel cell stack increases, SW4
will be turned ON by the command of Controller 1 exhaust heat-taking pump PP1
of the fuel cell stack will be operated. If operating time passes and the exhaust
heat output of the fuel cell stack and reformer increases, the heat output of the heat
pump can be reduced. When the amount of exhaust heat of the fuel cell stack and
reformer satisfies the heat amount demanded, the heat pump is stopped by the
command of Controller 2. Figure 7.5(e) shows a model of these operations. Fur-
thermore, when the two amounts of exhaust heat exceed the heat amount de-
manded, as shown in Fig. 7.5(f), excessive heat is stored in the heat storage tank.
Moreover, as shown in Fig. 7.5(g), when the exhaust heat exceeds the capacity of
the heat storage tank, valves VA2 and VA3 are operated by the command of Con-
troller 2, and surplus heat is released from the system.
Figure 7.5(h) shows a model of the amount of consumption of town gas in the
system. The consumption of town gas has the same characteristics as shown in
Fig. 7.5(d) depending on the production of electricity of the fuel cell. The output
characteristics of Fig. 7.5(d) differ greatly from the control variables set as Con-
troller 1. The characteristics of the town gas consumption of the system also
change with the value of the control variable set up by Controller 1.
In order to investigate the transient response characteristics of the power and the
heat of the system shown in Fig. 7.1, the transient response characteristics of each
piece of equipment of the fuel cell, reformer, heat pump, inverter, and system
interconnection device are expressed by a primary delay system. The time con-
stant of each piece of equipment is determined as described below.
120 7 Load Response Characteristics of the Fuel Cell for Individual Cold-region Houses
The heat output characteristics of a fuel cell stack are calculated by the enthalpy
transportation of the off-gas discharged from the cathode obtained using the test
system of Fig. 7.6. In the test system of Fig. 7.6, the initial operation temperature
of the fuel cell stack installed in a case that is thermally insulated can be set up
freely. From the examination results of this test system, the power generation
efficiency of a fuel cell is considered depending on the operation temperature.
According to an experimental result, when the operating temperature of the fuel
cell changes from 288 K to 358 K, there is about 30% difference in the power
generation efficiency. In this chapter, the temperature of the fuel cell in operation
is a constant 333 K. Figure 7.7 shows the transient response characteristics of the
heat output of a fuel cell stack. This figure was prepared from the experimental
results of the test system using the power and the heat output of the fuel cell stack.
As shown in Fig. 7.7, the strong production of electricity depends on the transient
response characteristics of the heat output of the fuel cell stack. In this chapter, as
shown in Fig. 7.7, approximated curves are prepared for each production of elec-
tricity, and transfer functions are also prepared. Time constants a and b of the
transfer function of the fuel cell shown in Table 7.2 are arranged using the produc-
tion of electricity of the fuel cell ( E f, t ), as shown in Figs. 7.8 and 7.9. Therefore,
the values of a and b are determined by giving the amount of power outputs of the
fuel cell to the approximate expression in Figs. 7.8 and 7.9. The transfer function
of the heat output in the fuel cell stack is determined by giving a and b to the
transfer function of Table 7.2.
Figure 7.10(a) shows the examination result of an experiment on a trial produc-
tion fuel cell stack using the test system of Fig. 7.16. This fuel cell stack is 100 W
and capacity differs from the fuel cell stack (1.1k W) shown in Fig. 7.11. There-
fore, in this chapter, it is assumed that the capacity factor of a fuel cell is the
same characteristic. In the examination of Fig. 7.10(a), a load of 70% of the capac-
ity factor was inputted into the fuel cell stack in a step-wise manner, and the tran-
sient response characteristics were obtained. Here, the value that divided the
output of the equipment by the equipment capacity is defined as the capacity
Fig. 7.8 The approximate equation for the time constant of the transfer function
Fig. 7.9 The approximate equation for the constant of the transfer function
122 7 Load Response Characteristics of the Fuel Cell for Individual Cold-region Houses
factor. As a result, the transient response characteristics shown in Fig. 7.10(b) are
obtained, and the formula expressing these characteristics with the transfer func-
tion of a primary delay system is shown in Table 7.3. To be exact, the transfer
function depends on the capacity factor. However, according to the examination
results of this chapter, the influence of the capacity factor on the transfer function
is small and is not taken into consideration.
7.3 The Time Constant of Each Piece of Equipment 123
Figure 7.11(a) shows the response characteristic when giving the step input to a
town gas reformer. The result shown in this figure changed the capacity factor
between 100%80%, and between 100%50%. Figure 7.11(b) shows the transient
response characteristics acquired from this result. This characteristic is expressed
with the transfer function of a primary order system and is described in Table 7.3.
A transfer function influences a capacity factor exactly like a fuel cell stack. In
this chapter, within the range described above, there is no large difference, and
this influence is not taken into consideration.
Figure 7.12 shows the response characteristic model of the exhaust heat of the
heat-source burner installed in the reformer when inputting a capacity factor of the
reformer from 50% to 100% in a step-wise manner [23, 26, 27]. As shown in this
figure, unlike the heat output characteristics of the fuel cell stack and heat pump
described later, the response characteristics of the exhaust heat of the heat-source
burner installed in the reformer show an S shape. Then, an approximated curve is
independently prepared for the transfer function of the response characteristics of
Fig. 7.12 for each range of A1 and A2. The transfer function of the reformer
shown in Table 7.2 is a formula for the A2 range.
124 7 Load Response Characteristics of the Fuel Cell for Individual Cold-region Houses
Fig. 7.12 Model of the burner exhaust gas heat at the time of a load going up from 50% to
100%
A cheap voltage control type inverter was used. Although the power inputted into
the inverter is changed and outputted to the voltage and the frequency of the re-
quirement (less than 5% of the regulation in this chapter), 120 ms is taken [53].
When the transfer function of an inverter is expressed with a primary delay sys-
tem, it is the formula shown in Table 7.3. System interconnection equipment is
used to change the power of single-phase 100 V to a network. It takes about 10 s
to change the system interconnection equipment [53]. The operation time of the
model of the system interconnection equipment assumed in this chapter is set at
12 ms [53]. The transfer function of the system interconnection equipment by a
primary delay system is set up using the formula shown in Table 7.3.
Figure 7.13 shows the experimental result of the heat output characteristics of a
geo-thermal heat pump when inputting a step-wise load so that it operates at
maximum output in an instant from the unloaded condition [54]. The high-
temperature-side heat exchanger of the trial production geo-thermal heat pump
was installed in the water tank, and this experimental result was obtained from the
water temperature change. On the other hand, the black dot in the figure is an
approximation of the examination result of the transient response characteristics of
the heat output of the heat pump. From the approximated curve, the transfer func-
tion of the heat pump shown in Table 7.2 was prepared.
7.4 Analysis Method 125
Figure 7.14 is a control block diagram of the system shown in Fig. 7.1. If the load
of the power and heat is given to Fig. 7.14, a transient response characteristic can
be investigated. Random fluctuation is added for every sampling time supposing
the actual power load. There are two controllers: Controller 1, which controls a
production-of-electricity network, and Controller 2, which controls the heat supply
network in this system. Controller 2 is contained in Subsystem 0 in Fig. 7.14, and
Fig. 7.15 shows the details of Subsystem 0. Immediately after Controller 1 and
Controller 2, a limiter is provided so that extremely large overshooting does not
occur.
Subsystem 18 is a transfer function of the response characteristics of the ex-
haust heat output in the fuel cell stack, and Fig. 7.16 shows this information. a and
b of the transfer function of Fig. 7.16 express the time constant and the constant
part in a primary delay system transfer function, respectively. a and b can decide
to give the production of electricity E f, t of a fuel cell to Fig. 7.8 and Fig. 7.9, as
the section The Time Constant of the Fuel Cell Stack described. The exhaust
heat of a fuel cell stack is a non-linear output to the power load, as Fig. 7.7 shows.
So, in this chapter, the relation between the power load and the exhaust heat out-
put of a fuel cell is divided into eight ranges for the power load. The relation be-
tween the power load and the exhaust heat output of each range is approximated
by the transfer function formula shown by Subsystems 18. The dynamic charac-
teristics of a heat storage tank are not taken into consideration in analysis for sim-
plification. Moreover, operation of SW4 and PP1 shown in Fig. 7.1 and the time
delay of the heat pump are not taken into consideration, either.
The transient response characteristics of the power and heat output of the sys-
tem of Fig. 7.14 is analyzed by MATLAB (Ver.7.0)/Simulink (Ver.6.0) of The
MathWorks Corporation. In the solver to be used, the RungeKutta method is
installed, and the sampling time of analysis is calculated automatically and deter-
mined so that the error may be less than 0.1%.
126 7 Load Response Characteristics of the Fuel Cell for Individual Cold-region Houses
The analysis results of the response characteristics of the control variables set up
by Controller 1 and Controller 2, the power output of the system, and the heat
output of the geo-thermal heat pump are shown in Figs. 7.17 and 7.18, respec-
tively. Figures 7.17(a), (b), and (c) inputted the power load (0.2 kW, 0.6 kW, and
7.5 Results and Discussion 127
Fig. 7.18 Characteristics of heat output of the geo-thermal heat pump system
128 7 Load Response Characteristics of the Fuel Cell for Individual Cold-region Houses
1.0 kW) at t = 5 s , respectively. The control variables of action P and action I set
up by Controller 1 were calculated by changing them arbitrarily, as shown in
Fig. 7.17. When action P is used independently in the control of the system by
Controller 1, steady-state error is large. Moreover, with the use of action D, such
as D control, PD control, and PID control, the response characteristics of the sys-
tem become extremely unstable, and we decided not to use action D for this analy-
sis. The optimal values of the control variables installed into Controller 1, which
controls the production-of-electricity network of a fuel cell system, differ in the
volume of the power load from each result of Figs. 7.17(a), (b), and (c). However,
the response characteristics do not depend on the control variables in the range of
the control variables of the P action and the I action shown in Fig. 7.17(c) by
1.0 kW of the maximum production of electricity of the system. On the other hand,
the response characteristics of the heat pump of Fig. 7.18 are the result of input-
ting a heat load (5.0 kW and 15.0 kW) in a step-wise manner at t = 5 s . The opti-
mal value of the control variables set up by Controller 2 from this result changes
with the volume of the heat load. Therefore, in order to improve the transient re-
sponse characteristics of the heat pump, it is necessary to change the control vari-
ables of the controller in the magnitude of the heat load.
Figure 7.19 shows the analysis results of the transient response when inputting a
power load of 0.2 kW, 0.5 kW, and 1.0 kW into the system of Fig. 7.14 in a step-
wise manner. The step input is performed at t = 5 s . Figure 7.19(a) shows the
result of calculating the response of each power load. In these calculations, the
control variables of the P action and the I action control were set as the values
shown in this figure. Control variables for the P action and the I action that con-
sidered that a transient response is good for every power load were chosen (offset
and overshooting is small, settling time is short), and the result shown in
Fig. 7.19(a) was obtained. However, as Fig. 7.17(c) shows, the response character-
istics of a 1.0 kW power load do not depend on the control variables of P and I.
The response characteristics of 1.0 kW with the largest power load has a long
settling time compared with other results in Fig. 7.19(a). When the response of the
system has a large power load, the response speed is low, and with the maximum
power output, it is the latest response. When a power load of 1.0 kW of maximum
output is put into the system of unloaded condition, the settling time is about 10 s
In order to shorten the settling time of the system, it is necessary to improve the
reformer with the slowest response speed.
7.5 Results and Discussion 129
Fig. 7.19 Analysis results of the fuel cell system with a geo-thermal heat pump
130 7 Load Response Characteristics of the Fuel Cell for Individual Cold-region Houses
Figure 7.19 (b) shows the analysis result of the heat load response of the system.
Since the response speed of the exhaust heat in the fuel cell stack and reformer is
lower than the heat pump, as shown in the figure, it starts the heat supply using the
heat pump. Gradually, the exhaust heat of the fuel cell stack and the reformer
increases, and the heat output of the heat pump are controlled to reduce the quan-
tity concerned. Figure 7.19(c) shows the analysis result of the power consumption
of the heat pump when the heat loads are 5 kW, 10 kW, and 15 kW in a 1.0 kW
power load. If the operation time passes as Fig. 7.19(b) describes, the exhaust heat
output of the fuel cell stack and reformer will increase, and the electric power
consumption of the heat pump will decrease with time.
The cold start of a reformer takes about 10 minutes to 1 hour [8, 27]. It is assumed
in this analysis that the reformer operates when having already finished the warm-
up. In this case, the analysis result of the amount of town gas consumed by the
reformer is shown in Fig. 7.19(d). The analysis result in the case of a power load
0.2 kW, 0.5 kW, and 1.0 kW is shown in this figure. For the town gas consumption
amount it depends on the response characteristic of power load. Moreover, the
settling time of 1.0kW of power loads is the longest (Fig. 7.19(a)). The response
characteristics of a reformer will be good from now on due to modification of the
control method and the improvement of the heat transfer performance.
Figure 7.20 shows the analysis result of the response characteristics of the system
when considering load fluctuations to the step input of the power load. Fig-
ure 7.20(a) shows the load pattern that added 10% of fluctuation to a 1.0 kW
power load. As shown in Fig. 7.20(a), a step load was inputted at t = 5 s , and load
fluctuations were given at random within a range given beforehand for every sam-
pling time. Figures 7.20(b) and (c) show the error analysis result of the step re-
sponse when adding 10% of load fluctuation to the base power load (0.5 kW and
1.0 kW). The main causes of the error shown in Figs. 7.20(b) and (c) are a time
delay of the system and the influence of overshooting.
Figure 7.21 shows the analysis result of the difference in the step load and sys-
tem response with load fluctuation when operating the system for 200 s. There are
fewer response results of the system than loading when the load fluctuation is 0%,
and if load fluctuation is added, there is an increase compared to loading. In
7.5 Results and Discussion 131
Fig. 7.22 Percentage of the integrated value of a response result to a load integrated value if a
sampling time period is operating for 200 s
Fig. 7.22, the response result at the time of 0% load fluctuation is a negative value.
This main reason is the time delay of the response. On the other hand, when the
response result is a positive value, an overshoot follows the response.
Figure 7.22 shows the analysis result of the difference in the amount of heat load
and the heat output of the heat pump when inputting a step load with a load fluc-
tuation of power for an operating time of 200 s. When load fluctuation increases,
there are fewer heat outputs of the heat pump than the heat load. As Fig. 7.21
describes, when load fluctuation becomes large, the power output of the system
will increase. From this, when load fluctuation becomes large, the amount of ex-
haust heat of the fuel cell stack and the reformer will increase, and the heat output
of heat pump will decrease. Moreover, if the heat load becomes large, the heat
output of the heat pump is approximated to the amount of heat load. This is be-
cause the ratio of the exhaust heat output of the fuel cell stack and reformer to the
heat output of the heat pump is small when the heat load is large. When power
load increases, the amount of exhaust heat of the fuel cell stack and the reformer
will increase. Therefore, the rate of heat output of the heat pump to heat load falls.
As a result, the power consumption of the heat pump decreases, so that the power
load is large and the load fluctuation is larger.
Figure 7.23 shows the analysis results of the response of the heat pump output
when changing the PI control parameter set up with the controller (Controller 2) of
the heat pump shown in Fig. 7.15. When the analysis results of the transient re-
sponse characteristics are compared in the difference in control parameters, Ph
=200.0 and Ih =0.05, which has the best evaluation of settling time, overshooting,
and steady-state error. So, the control parameters of the controller of the heat
pump are set at Ph =200.0 and Ih =0.05 in the following analyses.
7.5 Results and Discussion 133
In Sapporo in Japan, which is a cold and snowy area, the annual mean temperature
for the past five years has been 282 K. The average temperature in February is
270 K, and the highest and lowest temperatures on a representative day in Febru-
ary are 273 K and 266 K, respectively [14]. Moreover, the average number of
snowy days in February is 25. On the other hand, the average temperature for a
representative July day has been 293 K for the past five years, and the highest and
the lowest temperatures are 298 K and 290 K, respectively. There is no cooling
load in the summer in Sapporo. Electricity demand includes household appliances
and electric lights, and heat demand comes from heating, hot water supply, and
baths.
134 7 Load Response Characteristics of the Fuel Cell for Individual Cold-region Houses
Figures 7.24 and 7.25 show the analysis results for the heat output and power
output in the system at the time of installing a PEM fuel cell under the load pattern
of an average individual house in Sapporo [8]. The horizontal axis of Figs. 7.24
and 7.25 is the sampling time of the analysis. Real time is also displayed on the
horizontal axis of Fig. 7.24(a). Since the computation time is enormously long,
real-time analysis is performed by shortening the real time to 1/180 in this chapter.
A model of the heat and power load pattern of the representative time of winter
(February), mid-term (May), and summer (August) of an individual house in Sap-
poro is shown in Figs. 7.24(a) and 7.25(a). The heat load for hot water supply and
baths is included in the heat load pattern. Figure 7.24(b) shows the analysis result
of the heat response of the heat pump. From this result, the response speed of the
heat output of the heat pump is sufficiently rapid, and there is no problem in fol-
lowing the heat load. Figure 7.24(c) shows the analysis result of the heat response
that added the exhaust heat of the heat-source burner of the reformer to the ex-
haust heat of the fuel cell stack. The time constant of the exhaust heat output of the
7.6 Conclusions 135
fuel cell and the exhaust heat output of the heat-source burner of the reformer are
large. As shown in Fig. 7.24(c), the heat output does not converge within this
sampling time. Figure 7.24(d) shows the analysis result of the heat dissipation
from the system through a radiator. However, this heat is recoverable if the heat-
ing storage capacity is extended. On the other hand, Fig. 7.25(b) shows the analy-
sis results of the response characteristics of the power when controlling the PEM
fuel cell to follow the power load pattern of Fig. 7.25(a). The electric power needs
model of Fig. 7.25(a) is an average in each time. It consists of many dynamic
peaks of transient power demand. The electric power demand of a time average
was used for analysis. This way, the response speed of the power of the system is
very fast compared with the response speed of the exhaust heat of the fuel cell and
reformer. From this result, in order to supply heat demand only with the exhaust
heat of the fuel cell and the reformer, improvement of the heat response is re-
quired. In order to supply early morning heat demand, to recover the exhaust heat
of the fuel cell and the reformer, waste heat recovery of a long period is required.
Therefore, the heat dissipation loss of the heat storage tank is predicted to have a
large influence on the overall efficiency of the system.
7.6 Conclusions
The dynamic characteristics of a fuel cell system for individual houses in cold
regions have been investigated using numerical analysis. This chapter examined
the transient response characteristics of the power and heat output of proton ex-
change membrane type (PEM) fuel cell cogeneration with a town reformer. Fur-
thermore, from experiments, etc., the transient response characteristics of each
piece of equipment of the fuel cell, the reformer, the heat pump, the inverter, and
136 7 Load Response Characteristics of the Fuel Cell for Individual Cold-region Houses
8.1 Introduction
The micro-grid is expected to reduce the discharge of carbon dioxide gas, to cut
the peak of an electric power plant, and to supply backup power in an emergency
[2123]. A micro-grid technique connects energy equipment, such as an engine
generator and a fuel cell, and power is supplied by each cooperating piece of
equipment. In forming a micro-grid, the coordinated grid system with commercial
power, etc., and the independent grid system should be considered. In the coordi-
nated grid type, the supply and demand of power with a commercial system are
possible, and the peak cut of an electric power plant and the buying and selling of
power are possible. If the exhaust heat of the generating equipment linked to a
micro-grid needs to be conveyed only a short distance, it can be supplied to the
consumer with small radiation loss. On the other hand, in the exhaust heat of the
large-scale conventional power plant, long distance transport has many heat
losses, and utilization of exhaust heat is limited. One of the problems predicted by
the construction of a micro-grid is that power quality deteriorates when the power
demand and supply balance of the grid do not balance. The deterioration of the
power quality described in this chapter means that fluctuation of voltage and fre-
quency because the dynamic characteristics of an electric power supply do not
meet the demand. Power quality can be maintained comparatively easily by con-
trolling the voltage and frequency of a grid by a coordinated grid type to synchro-
nize with another network, such as commercial power [51]. On the other hand, in
an independent grid type, the reference of the power is determined to be any
power generator linked to a grid. In addition, other voltages and frequencies of
generators are controlled to synchronize with this reference electrode. Therefore,
if the power quality of the power generator made into a reference is not stabilized,
the power quality of the whole grid may deteriorate. In order to stabilize the power
quality of a micro-grid, the method of connecting a battery to a micro-grid can be
used. However, when facility cost and the maintenance cost of a facility are taken
The energy network and micro-grid that are installed in an urban area are shown in
Fig. 8.1. Figure 8.1(a) shows the micro-grid of the system linked with a commer-
cial power network, and Fig. 8.1(b) shows the micro-grid of the independent sys-
tem that does not connect with other grids. An energy network consists of a hot
water piping network used for the waste heat recovery of energy equipment, and
8.2 The Micro-grid Model 139
the heat supply to each house, a city gas network that supplies city gas to the re-
former for fuel cells, and a gas-engine power generator, and a micro-grid that
supplies electric power to each house. The power transported by the independent
micro-grid shown in Fig. 8.1(b) needs to control the voltage and frequency of each
power generator on the basis of the power of any one power generator. On the
other hand, it is controllable by the micro-grid connected with the other power
grids shown in Fig. 8.1(a) to synchronize with the network voltage and the fre-
quency that were connected. Even if a rapid load fluctuation is added to a grid, and
a difference occurs in the production of electricity and the amount demanded, the
power quality of the coordinated grid type is stabilized by synchronizing with the
connected network. With the independent grid type, if rapid load fluctuation is
added to a grid, and a difference occurs in the production of electricity and the
amount demanded of a power generator, it is considered that a long time is re-
quired to stabilize the power quality. If the load fluctuation is not appropriately
predicted, in an independent grid type, the power quality of the grid is expected to
be low over a long time.
140 8 Load Response Characteristics of a Fuel Cell Micro-grid
System Scheme
number of fuel cells installed in a distributed system is determined from the elec-
tricity demand pattern shown in Fig. 8.3 used in the analysis. The electricity de-
mand pattern of Fig. 8.3 is a measurement result of an individual house for repre-
sentative days in February in Sapporo, Japan [6]. On the horizontal axis of this
figure, the sampling time of analysis and the time assumed (assumed time) are
written together. In this electricity demand pattern, the minimum load is deter-
mined as a value of the base load (in the example in Fig. 8.3, it could be 3 kW). In
the case of a centralization system, there is one set of installed fuel cells, and, in
the case of a distributed system, a fuel cell of 1 kW maximum output is installed in
each house. Moreover, let the fuel cell capacity of the centralization system be a
value where the maximum load added to a grid is satisfied. In the distributed sys-
tem, the fuel cells of F/C (1) to F/C (16) all have the same capacity and dynamic
characteristics. The electric transmission loss of a micro-grid is considered in this
analysis.
Figure 8.4(b) shows the test results of the response characteristics when inputting
a step load of 70% of a load factor into the PEM fuel cell trial production as an
experiment [57]. The response characteristics shown in Fig. 8.4(c) are obtained
from the results of Fig. 8.4(b), and require the transfer function of a primary-
delay-system so that these characteristics may be approximated. The equations for
the transfer function are shown in Fig. 8.4(c). To be exact, although the transfer
function is considered to depend on the load factor, it is not taken into considera-
tion because this difference is small as a result of examination. The settling time of
a fuel cell when generating 1 kW maximum output is about 3 seconds [57].
144 8 Load Response Characteristics of a Fuel Cell Micro-grid
Figure 8.4(d) shows the model of the step input given to the city gas reformer of a
load factor when it is between 100% and 80%, and between 100% and 50% [57,
58]. Figure 8.4(e) shows the response result of the load input to Fig. 8.4(d), and
the transfer function of a primary-delay-system shows this response characteristic
in this figure. In addition to fuel cells, although it is considered that a transfer
function influences the load factor directly, since there is no big difference, the
result of Fig. 8.4(f) is used in the range managed in this chapter.
An inverter uses a cheap voltage control form, and converts and outputs input
power to regular voltage and frequency. An inverter requires 120 ms to stabilize
voltage and frequency within 95% of a regular value [58]. Figure 8.4(f) expresses
the transfer function of this inverter with the primary-delay-system. When switch-
ing the power of the single phase 100 V by an interconnection device, the duration
of the change is about 10 s. However, since it is necessary to synchronize the
frequency by control, the interconnection device assumed in this chapter sets the
change time to 12 s. As a result, the transfer function of the interconnection de-
vice by the primary-delay- system is the value of Fig. 8.4(f).
Figure 8.5 shows the output characteristics of a fuel cell and a city gas reformer
obtained in the experiments. Figure 8.5(a) shows the relation between the load
factor of a fuel cell and the generation efficiency assumed in this chapter, and
Fig. 8.5(b) shows the characteristics of the amount of hydrogen supplied to a fuel
cell, and generation and exhaust heat output. Figure 8.5(c) shows the relation be-
tween the load factor of a city gas reformer and reformer efficiency. Here, the
value that divides the calorific power of the hydrogen contained in the reforming
gas by the calorific power of the city gas supplied to a reformer is defined as the
reformer efficiency. The city gas supplied to a reformer has the object of produc-
ing reforming gas and the object of being the heat source of a reformer. In analy-
sis, a load factor is calculated from the capacity of a fuel cell, and the quantity of
the load, and the generation efficiency is calculated using the relation shown in
Fig. 8.5(a). Moreover, the amount of exhaust heat of a fuel cell is obtained by
giving a power load to Fig. 8.5(b). Because the load factor of a reformer is calcu-
lable from the load and capacity, reformer efficiency is determined if this value is
given to Fig. 8.5(c).
8.4 Control Variables and Analysis Method 145
Fig. 8.5 Output characteristics of a fuel cell system with city gas reformer
The response characteristics that inputted step load into the independent micro-
grid shown in Fig. 8.2(a) for 0.2, 0.6, and 1.0 kW loads are shown in Figs. 8.6(a)
(c) [57, 58]. The response characteristics of a fuel cell changes with the control
parameter set up with a controller, and changes and analyzes the parameters of PI
control in Fig. 8.6. As shown in Fig. 8.6(c), the result of the 1.0 kW loads does not
depend on the rise time and the settling time of the control parameter. In the result
of the 0.2 kW loads, although the rise time of P = 12.0, I = 1.0 is short; as for the
146 8 Load Response Characteristics of a Fuel Cell Micro-grid
settling time, P = 1.0 and I = 1.0 are short. P = 12.0 and I = 1.0 is compared with
P = 1.0 and I = 1.0, as for the rise time, P = 12.0 and I = 1.0 is shorter, and the
settling time is almost the same, and the overshooting is large. Moreover, in the
result of P = 5.0 and I = 1.0, the steady-state error of low load is large, and is
not suitable as a control variable. Therefore, in the analysis example of the follow-
ing section, the control parameter of a fuel cell is analyzed as P = 1.0, I = 1.0 and
P = 12.0, I = 1.0.
The dynamic characteristics of a micro-grid are analyzed using MATLAB
(Ver.7.0) and Simulink (Ver.6.0) of The MathWorks Corporation. However, in
the analysis example of the following section, the solver to be used is made into
the RungeKutta method; it is calculated, and the sampling time of the analysis is
determined so that error is less than 0.01%.
8.5 Load Response Characteristics of the Micro-grid 147
The response characteristics when inputting a step load into the micro-grid is
shown in Fig. 8.7(a). The parameter of the PI controller of all the fuel cell systems
was set to P = 1.0 and I = 1.0. Figure 8.7(b) shows the result of a response of a
micro-grid, Fig. 8.7(c) shows the result of a response of an engine generator, Figs.
8.7(d)(f) show the results of a response of fuel cell Nos. 1, 8, and 12, respec-
tively. When the response results of an engine generator (Fig. 8.7(c)) is compared
with the response results of a fuel cell (Figs. 8.7(d)(f)), there is little output vibra-
tion and the settling time is short in a fuel cell. If the load added to a grid in-
creases, the number of fuel cells to be operated is increased corresponding to the
load. According to the load of a grid, a larger number of fuel cells is operated such
as Nos. 1, 8, 12 (Fig. 8.7(d)(f)). How the dynamic characteristics of a micro-grid
would change according to the difference in the control parameter of the PI con-
troller of a fuel cell system was investigated.
The results of investigating the response characteristics of the grid for the con-
trol parameter with P = 12.0 and I = 1.0 are shown in Fig. 8.8. Figure 8.8(a) shows
P = 12.0 and I = 1.0 and is a response result when inputting the step load of
Fig. 8.7(a) into a micro-grid. When Fig. 8.7(b) is compared with Fig. 8.8(a), it
shows that the overshooting in Fig. 8.8(a) is larger. The analysis results of the
power load and settling time when the control parameters of a fuel cell system are
Fig. 8.7 Step response characteristics of the system. The control parameter of the fuel cell
system is P = 1.0 and I = 1.0
148 8 Load Response Characteristics of a Fuel Cell Micro-grid
Fig. 8.8 Analysis results of the step input for micro-grid system
P = 1.0, I = 1.0 and P = 12.0, I = 1.0 are shown is Fig. 8.8(b). The period until
convergence of less than 5% of the load by the power supplied to a grid is de-
fined as the settling time. From the result of Fig. 8.8(b), a load does not depend on
the settling time of a control parameter of 15 kW or less. However, when a load
exceeds 16 kW, control parameter P = 12.0, I = 1.0 of a fuel cell system is about
5 seconds short compared with P = 1.0, I = 1.0. Figure 8.8(c) shows the analysis
result concerning the difference in the power load of a micro-grid, and the power
supplied to a grid from EG and PEMFC. The difference in the power load and
electric power supply is due to a time delay in the response of EG and the fuel
cell. The result of Fig. 8.8(c) shows that the difference in the control parameter of
a fuel cell does not affect the difference in the power load or the electric power
supply. However, the control parameter of a fuel cell system strongly affects the
settling time and overshooting.
Fig. 8.9 Response results analyzed using an electricity demand pattern of a house in Sapporo in
February. The control parameter of the fuel cell system are P = 1.0 and I = 1.0
Fig. 8.10 Response results analyzed using an electricity demand pattern of a house in Sapporo
in February. The control parameters of the fuel cell system are P = 12.0 and I = 1.0
of Fig. 8.3 into the micro-grid are shown in Figs. 8.9(a) and 8.10(a). In the analy-
sis of Figs. 8.9 and 8.10, the control parameter of the PI control device of a fuel
cell system was set up with P = 1.0, I = 1.0 and P =12.0, I = 1.0, respectively. As
for overshooting, in these analysis results, P = 12.0, I = 1.0 are larger than P = 1.0,
I = 1.0. Moreover, immediately after startup (past [0:00]), the reason for the many
fluctuation elements is based on the starting characteristics of an engine generator,
as shown in Figs 8.9(b) and 8.10(b).
The response results of fuel cell Nos. 1, 3, 5, and 9 are shown in (f) from (c) of
Figs. 8.9 and 8.10. The analysis results of Figs. 8.9 to 8.10 have large overshoot-
ing when load fluctuation is added to a fuel cell and an engine generator. As for
the analysis results of the sufficiency ratio of power supply to the power demand
150 8 Load Response Characteristics of a Fuel Cell Micro-grid
The analysis results of generation efficiency when supplying power by the fuel
cell micro-grid of a centralization system and a distributed system to 17 houses is
shown in Fig. 8.12. The average value of the generation efficiency of each fuel
cell under operation defines the generation efficiency of the fuel cell in a distrib-
8.5 Load Response Characteristics of the Micro-grid 151
Fig. 8.12 Fuel cell efficiency results analyzed using an electricity demand pattern of 17 houses
in February in Sapporo. The control parameters of the fuel cell system are P = 12.0 and I = 1.0
8.6 Conclusions
9.1 Introduction
If the micro-grid is introduced into an urban area, it will be expected that the en-
ergy cost of a distributed power supply and emission of greenhouse gas can be
reduced. To date, authors have investigated the operating method that connects
distributed proton exchange membrane fuel cells (PEM-FC) in a power network
and cooperates with it [41, 59]. Although the generation efficiency of a PEM is
high, greenhouse gas discharges by the reforming reaction of city gas. On the
other hand, micro-combined heat and power (micro-CHP) using a small-scale
Stirling engine generator (SEG) has been examined in the UK as an energy system
for individual houses [60, 61]. By using woody biomass so that carbon dioxide
may circulate, the greenhouse gas amount of emission of a power generation sys-
tem can be decreased. Therefore, the introduction of SEG using woody biomass is
effective in emission control of greenhouse gas [6264]. However, compared with
an internal combustion engine or a fuel cell, the generation efficiency, volume
efficiency, equipment cost, etc. of the conventional SEG are small. The energy
supply system using the micro-grid can reduce equipment cost compared with the
method of introducing generating equipment into each house. Moreover, the
energy equipment linked to the micro-grid can help, for example to minimize the
amount of greenhouse gas emission. Thus, in this chapter, the dynamic charac-
teristics of the power of the independent micro-grid using hybrid cogeneration
(PWHC) of PEM-FC and SEG using woody biomass are investigated. The con-
trol response characteristic of SEG depends on the engine structure, the configura-
tion of the combustion chamber, the heat transmission characteristic of the heat
source, etc. Until now, optimization of the combustion chamber configuration and
the heat transmission characteristic of combustion gas have been investigated
[65, 66]. Commonly, the power demand pattern of a house or an apartment house
consists of many peaks changed for a short time. Since such a power load is fol-
lowed, a rapid control response characteristic is required of the generating equip-
Figure 9.1 shows the model of the independent micro-grid that introduces two-set
PWHC (PWHC (1) in House (1), and PWHC (2) in House (5)). The micro-grid of
this model consists of eight buildings of House (1) to House (8). The heat supply
of the exhaust heat of PWHC, a heat storage tank, and a boiler is separated into the
group of House (1) to House (4), and the group of House (5) to House (8). The
power of two-set PWHC is supplied to each building through the power grid. The
system interconnection device is installed in the contact point of PWHC and a
power grid. Moreover, the power of PWHC is changed into 100 V and 50 Hz with
an inverter. On the other hand, the exhaust heat of PWHC, the heat of a heat stor-
age tank, and a boiler is supplied to each building through hot water piping (1) and
(2). However, this study is limited to the dynamic characteristics of the power for
the micro-grid.
Figure 9.2 shows the energy flow and chemical reaction of each component of
the proposed system. Chip fuel is supplied to a woody biomass engine (SEG), and
power is transmitted to an alternating current synchronous power generator. The
heat output of SEG is the high-temperature exhaust gas of the combustion cham-
Fig. 9.1 Independent hybrid micro-grid model with PEM-FC and a woody biomass engine
9.2 System Scheme 155
ber, and engine-cooling hot water. Moreover, as the heat output of PEM-FC, there
is fuel cell stack exhaust heat and reformer exhaust heat. In the proposed system,
the combustion chamber high-temperature exhaust gas of SEG is supplied to the
heat exchanger of the reformer. With a catalyst in the reformer, city gas is changed
into reformed gas with a high hydrogen concentration with a reaction temperature
of 970 K to 1070 K using this exhaust heat. Reformer exhaust heat is the remain-
ing heat after providing heat to the catalyst through the heat exchanger. In the case
study, exhaust heat that can be supplied to the demand side is taken as the re-
former exhaust heat and SEG cooling water. Moreover, the demand side is sup-
plied after changing the power of SEG and PEM-FC into an alternating current of
constant frequency.
Tables 9.1 and 9.2 show the operating conditions and specifications of SEG and
the power generator that are examined in this chapter. Although the maximum
output of SEG is 3.7 kW, the maximum power load examined according to restric-
tions of the combustion chamber capacity, etc., is 1.6 kW. Figure 9.3 shows a
general view of the test equipment. Chip fuel (woody biomass) is fed into the
hopper of the combustion chamber. Chips are mixed with preheated air before
entering the combustion chamber. The rate of feed of chip fuel is controllable by
the fuel feed system installed in the lower part of the hopper. Power is transmitted
to the power generator shown in Table 9.2 by a belt from the power shaft of SEG.
Since the test SEG is a single cylinder, its vibration is large. Consequently, the
combustion chamber is connected with the engine by a buffer duct so that the
vibration of the engine does not spread to the combustion chamber. The exhaust
gas of the combustion chamber is discharged from the system through a duct. The
quantity of heat of the exhaust gas Q Ex and cooling water Q Ey is obtained from
the value of the temperature sensor and the flow meter by calculating the transport
volume of enthalpy. Moreover, the amount of heat radiation on the combustion
chamber surface ( Q Ez ) is measured by heat flow rate sensor q , and the heat-
medium pressure is measured using sensor Pg .
Figure 9.4 shows the experimental results of the energy flow of the test SEG.
The energy flow is separated into auxiliary machinery loss, cooling water quantity
of heat, exhaust gas quantity of heat, production of electricity, and other losses.
Other losses of the energy flow are mechanical loss of radiation of heat and fric-
tion of SEG, vibration, etc. Other losses decrease, so that the production of elec-
tricity of SEG is large. The power generation efficiency of SEG improves by re-
ducing other losses that hold a large part of the energy flow at the time of low
load. The quantity of exhaust gas heat holds the largest part in the energy flow,
and it is always large compared to the cooling water quantity of heat. Since there
is large exhaust gas heat, the development of a compound cycle of operating a
steam turbine using the high-temperature exhaust gas of SEG, for example, is
possible. Auxiliary machinery loss holds very few parts in the whole energy flow.
Figure 9.5 shows the PWHC operation model on a representation day. In this op-
eration pattern, SEG is operated in a range smaller than the base load set up be-
forehand. In addition to SEG, PEM-FC is operated in a larger load range than the
base load. When a load exceeds the base load, SEG can be operated at a maximum
efficiency point. However, when a load is less than the base load, a load following
operation is required of SEG.
Figures 9.6(a) and (b) are the block diagram of the feedback control on the micro-
grid by SEG and PEM-FC, respectively. Proportional-plus-integral control (PI
control) is introduced into the control of each system. PEM-FC and SEG are con-
trolled by the controller. Each controller is controlled based on the PI control pa-
rameters ( P and I ) set up beforehand. The power generated by SEG and PEM-
FC is supplied to the demand side through an inverter and a system interconnec-
tion device. The transfer functions of each equipment shown in Figs. 9.6(a) and
158 9 Dynamic Characteristics of a PEM-FC/Woody Biomass Engine Hybrid Micro-grid
base load set up beforehand. PEM-FC is also operated when the load of the micro-
grid exceeds the base load. The control block diagram in the case of one-set PEM-
FC operating corresponding to the base load, and corresponding to the load ex-
ceeding the base load by one-set SEG is shown in Fig. 9.6(d).
The control block diagram of the PWHC micro-grid in the case of one-set SEG
operation corresponding to the base load, and corresponding to the load exceeding
the base load with multiple generators is shown in Fig. 9.6(e). In Fig. 9.6(e), SEG
(1) operates corresponding to the base load, and operates SEG (2), PEM-FC (1),
and PEM-FC (2) according to the magnitude of a load. When supplying power to
the micro-grid from the combined cycle system, the dynamic characteristics of the
micro-grid are determined with the transfer functions of each piece of equipment.
So, in this chapter, the transfer function and control parameters of PEM-FC, SEG,
an inverter, and a system interconnection device in Fig. 9.6 are determined by the
method described in the sections Response Characteristics of PEM-FC and
Response Characteristics of SEG.
The transient response characteristics of the power output of the SEG, PEM-FC,
the auxiliary machine, and the PWHC micro-grid have been analyzed by
MATLAB (Ver.7.0)/Simulink (Ver.6.0) of The MathWorks Corporation. In
the solver to be used, the RungeKutta method is installed, and the sampling time
of analysis is calculated automatically and determined so that error may be less
than 0.1%.
Table 9.3 shows the result of the investigation of the transfer function in the previ-
ous study about the fuel cell stack, the reformer, the inverter, and the system inter-
connection device [57, 64]. The transfer function of the fuel cell stack was deter-
mined from the experimental result, and the transfer function of other equipment
was decided from references [27, 4751, 67, 68]. In the further last study, the
optimal value of the parameters of the PI control introduced into the controller of
PEM-FC was also investigated. The transfer function and control parameters on
the PEM-FC of the control block diagram shown in Fig. 9.6 introduce each value
of Table 9.3.
Figure 9.7 shows the results of the step response of 0.2 kW, 0.4 kW, 0.6 kW,
0.8 kW, 1.0 kW of the PEM-FC with a reformer [57, 64]. In the analysis of
Fig. 9.7, the control block diagram of Fig. 9.6(b) was used. The control parameter
with a short settling time and small overshooting was investigated by numerical
analysis, and P = 12.0 and I = 1.0 were determined. The response time of a sys-
tem converging on 5% of a target value is defined as the settling time.
Figure 9.8 shows the experimental result of the step response of 0.2 kW, 0.4 kW,
0.6 kW, 0.8 kW, and 1.0 kW of the testing SEG. As shown in Fig. 9.8, the step
response of the testing SEG has large overshooting, and its settling time is long
compared with PEM-FC. The heat transmission characteristics between the com-
bustion gas of a chip and the heat exchanger of SEG is considered to influence the
settling time greatly. However, it is difficult to improve the rate of heat transfer of
the combustion gas of a chip, so that the load fluctuation of the power can be fol-
lowed. So, in order to shorten the settling time of SEG as much as possible and to
reduce overshooting, PI control is added to the operation of SEG.
Figure 9.9 shows the example as a result of a step response obtained in the op-
erating experiment of SEG (Fig. 9.4). The model of the transfer function that
simulated this step response is shown in Fig. 9.9. The settling time of the testing
SEG exceeds 10 s. Therefore, when SEG is operated so that the fluctuating load
Fig. 9.9 Step response result of the test SEG, and response model
may be followed, the unstable time of voltage and frequency is long. Figure 9.10
shows the analysis results of a step response when adding PI control to the system
using the transfer function in Fig. 9.9. The control block diagram used in this
analysis is Fig. 9.6(a), and the control parameters of SEG introduced P = 0.1 and
I = 0.001 . Moreover, a response result in the case without PI control is also shown
in Fig. 9.10. The settling time becomes short by adding PI control to SEG, and an
overshoot does not appear. For example, the settling time of the 2 kW step re-
sponse that does not use PI control is about 16 s. However, if PI control is added, it
will improve at about 6 s.
The 1 kW PWHC micro-grid consists of 0.5 kW SEG and 0.5 kW PEM-FC. Fig-
ure 9.11 shows the analysis results of the step response of 0.2 kW, 0.4 kW, 0.6 kW,
162 9 Dynamic Characteristics of a PEM-FC/Woody Biomass Engine Hybrid Micro-grid
0.8kW, and 1.0kW of this system. Results in the case when SEG corresponds to
the base load and PEM-FC follows the fluctuating load exceeding the base load
are shown in Figs. 9.11(a) and (b). The control block diagram used in the analysis
in Figs. 9.11(a) and (b) is shown in Fig. 9.6(c). However, the values in Figs. 9.7
( P = 12.0 and I = 1.0 ) and 9.10 ( P = 0.1 and I = 0.001 ) were used for the control
parameter of the analysis in Fig. 9.11(a). The speed of response of the PEM-FC
shown in Fig. 9.7 is quick compared with the speed of response of SEG shown in
Fig. 9.10. From the difference in this speed of response, as shown in the step re-
sponse of 0.8 kW and 1.0 kW in Fig. 9.11(a), the response of a quick response part
and a late response part appears. Consequently, the control parameters of PEM-FC
with a quick speed of response are changed, and an improvement of the response
characteristics of the PWHC micro-grid is tried. Figure 9.11(b) shows the response
characteristics at the time of changing the control parameters of PEM-FC into
P = 0.95 and I = 1.1 . These control parameters were decided by trial and error.
Two response parts, 0.8 kW and 1.0 kW in Fig. 9.11(a), have improved.
Step response results in the case when PEM-FC corresponds to the base load
and SEG follows the fluctuating load exceeding the base load are Figs. 9.11(c) and
(d). In the analysis in Figs. 9.11(c) and (d), the control block diagram shown in
9.4 Results of Dynamic Characteristics Analysis of the PWHC Micro-grid 163
Fig. 9.6(d) was used. In Fig. 9.11(c), the control parameters of PEM-FC are
P = 12.0 and I = 1.0 , and the control parameters of SEG are P = 0.1 and
I = 0.001 . Since the overshoot of the response shown in Fig. 9.11(c) is large, the
control parameters of PEM-FC are changed and an improvement is tried. Fig-
ure 9.11(d) shows the response characteristics at the time of changing the control
parameters of PEM-FC into P = 0.95 and I = 1.1 . These control parameters were
determined by trial and error. Compared with the response of Fig. 9.11(c), the
response of Fig. 9.11(d) has small overshooting, and its settling time is short.
The response characteristics in the case of power supplied to the micro-grid from
SEG or PEM-FC are investigated. However, the power load pattern added to the
micro-grid assumes two houses on a representative day in February in Sapporo.
The power load pattern consists of time average values of the load consumed by
the household appliances and electric lights [8]. Space cooling and heating loads
are not included in this power load pattern. Therefore, the power load pattern does
not have a large difference every month.
Figure 9.12 shows the analysis results of a load response at the time of supply-
ing the power to the micro-grid using 2 kW SEG. The control block diagram used
in the analysis of Fig. 9.12 is that of Fig. 9.6(a). Moreover, the control parameters
of SEG are P = 0.1 and I = 0.001 as well as Fig. 9.10. The horizontal axis of
Fig. 9.12 is the representative time of analysis. Real time is also displayed on the
horizontal axis of Fig. 9.12. Since the calculation time is enormously long, the
real-time analysis is performed by shortening the real time to 1/180 in this chapter.
Figure 9.12(a) shows the results of a load input and the system response, and
Fig. 9.12(b) shows the results of the error of a load input and a response. As for
the broken-line part shown in Fig. 9.12(b), the error of the load and the response is
over 5%. A large rising error occurs immediately after 0:00 in Fig. 9.12(b). Ac-
tually, since the system is operated continuously, this rising error does not exist.
Figure 9.12(c) shows the analysis results of the time period for the error of the
load and the response to exceed 5%. Accordingly, the results of Fig. 9.12(c)
express settling times. The settling time when installing SEG into the micro-grid
from the result of Fig. 9.12(c) is 10.2 s at the maximum. When the micro-grid is
composed from SEG, the unstable period of voltage and a frequency is 10.2 s at
the maximum.
164 9 Dynamic Characteristics of a PEM-FC/Woody Biomass Engine Hybrid Micro-grid
Figure 9.13 shows the analysis result of a load response at the time of installing
2 kW PEM-FC into the micro-grid. The control block diagram used in the analysis
of Fig. 9.13 is that of Fig. 9.6(b). The control parameters set up with the controller
are P = 12.0 and I = 1.0, as well as Fig. 9.7. The settling time in the case when
PEM-FC composes the micro-grid from the result of Fig. 9.13(c) is 1.6 s or less.
However, rising parts are excluded. The power supply due to PEM-FC has a short
settling time compared with SEG. Therefore, the dynamic characteristic of the
power of the PEM-FC micro-grid is good compared with SEG micro-grid.
9.4 Results of Dynamic Characteristics Analysis of the PWHC Micro-grid 165
Figure 9.14 shows the analysis results of a load response of the micro-grid com-
posed from 8 kW PWHC. Eight houses are connected to the micro-grid. Woody
biomass engine generators installed into the micro-grid are 2 kW SEG (1) and
2 kW SEG (2), in addition 2 kW PEM-FC (1) and 2 kW PEM-FC (2) are installed.
Moreover, the control block diagram used in the analysis of Fig. 9.14 is that of
Fig. 9.6(e). The control parameters set up with the controller of PEM-FC are
P = 0.95 and I = 1.1 as well as Fig. 9.11(d), and the SEG parameters are P = 0.1
and I = 0.001 . Since the speed of response of SEG is slow, the dynamic character-
istics of SEG (2) have a large influence on the micro-grid. It is because SEG (2) is
166 9 Dynamic Characteristics of a PEM-FC/Woody Biomass Engine Hybrid Micro-grid
followed and operated. As a result, the settling time becomes long as shown in
Fig. 9.14(c). Consequently, installation of SEG shall be one set corresponding to
the base load. With respect to the load exceeding the base load, it corresponds by
installing two-set PEM-FC. Figure 9.15 shows the analysis results of the load
response of the micro-grid composed from a one-set of 2 kW SEG and a two-
set of 2.5 kW PEM-FC. Eight houses are connected to the micro-grid. This system
was analyzed by modifying the control block shown in Fig. 9.6(e). The control
parameters set up with the controller of PEM-FC are P = 12.0 and I = 1.0 , and
the SEG parameters are P = 0.1 and I = 0.001 . The error analysis results in
Figs. 9.14(b) and 9.15(b) are similar. However, as shown in Fig. 9.15(c), the
settling time of the micro-grid becomes very short compared with that shown
9.5 Conclusions 167
in Fig. 9.14(c). The system of Fig. 9.15 is the PWHC micro-grid stabilized
dynamically.
9.5 Conclusions
The load response characteristics were investigated using the testing Stirling en-
gine power generator (SEG) that uses woody biomass as a fuel. The transfer func-
tion was determined from these results, and the dynamic characteristics of the
power of the micro-grid composed from SEG were investigated. Moreover, hybrid
cogeneration (PWHC) that uses the combustion exhaust heat of SEG for the heat
168 9 Dynamic Characteristics of a PEM-FC/Woody Biomass Engine Hybrid Micro-grid
10.1 Introduction
in the engine has an advantageous mean effective pressure of 0.8 MPa or less.
When the mean effective pressure exceeds this value, the hydrogen supply in the
engine has low efficiency. As hybrid operations of PEM-FC and NEG, there is (a)
the method of operating NEG in a low-load region and operating PEM-FC in a
high-load region. Furthermore, since the maximum generation efficiency of PEM-
FC is higher than NEG, there is (b) the operating method of the base load response
of PEM-FC, corresponding to the fluctuating load of NEG.
In this chapter, a system operation map was prepared from the test data of PEM-
FC and NEG that was evaluated in the past [71, 28, 29]. A system operation map
expresses the city gas calorific power supplied to HCGS (it consists of PEM-FC,
NEG, and a boiler), and the relation between power generation and the heat output
of HCGS. In order to compare systems, this chapter investigates: (a) The hydro-
genation NEG individual-operation system (OM-A); (b) the PEM-FC individual-
operation system (OM-B), (c) the hybrid system (OM-C) that combines the opera-
tion of the low-load region of NEG, and operation of the high-load region of
PEM-FC, and (d) the hybrid system (OM-D) that combines the base load opera-
tion of PEM-FC and the fluctuating load following operation of NEG. These sys-
tems were installed into the load pattern of an apartment building composed of ten
family apartment houses in Tokyo, and fuel consumption, carbon dioxide emis-
sions, amount of heat storage, generation efficiency, and total efficiency were
analyzed. As a result, OM-D regarding fuel consumption, generation efficiency,
and total efficiency was understood to be the best. On the other hand, OM-A has
the least carbon dioxide emissions. A load pattern changes greatly according to
different times of the day. Therefore, fuel cell cogeneration that includes a NEG
hydrogenation method with sufficient low-load characteristics is dramatically
advantageous. The total efficiency of OM-A on a representative day in January,
May, and August was 66%, 75%, and 61%, respectively, and it was 65%, 68%,
and 63% in OM-B. On the other hand, the total efficiency of OM-C was 70%,
79%, and 71%, respectively, and the total efficiency of OM-D was 88%, 94%, and
73%, respectively.
A block diagram of HCGS proposed in this chapter is shown in Fig. 10.1. City gas
(CH4) is supplied to a reformer and a heat-source burner, and reformed gas is
produced. Equations (10.1) and (10.2) express the reactions of natural gas reform-
ing. Since Eq. (10.2) is an endothermic reaction, the heat of city gas burning
shown in Eq. (10.3) is supplied. After removing some of the water in the reformed
gas with a gas cooler, the fuel is supplied to the PEM-FC and NEG system. The
hydrogen in reformed gas after removing water is about 80% of the volume ratio.
10.2 System Scheme 171
CO + H 2 O CO 2 + H 2 + 41 [kJ/mol] (10.2)
Figure 10.2 shows the operation model using the HCGS shown in Fig. 10.1. OM-
A, OM-B, OM-C and OM-D are the operation models shown in Figs. 10.2(a), (b),
(c) and (d), respectively, where C 'N and C 'F express the maximum load of NEG
and PEM-FC, respectively. It is necessary to decide the capacity ( C N C F ) of
NEG and PEM-FC installed for each operation model with a value exceeding C 'N
and C 'F , which are shown in the figures. OM-A follows the load pattern by a NEG
independently, and OM-B follows the load pattern by a PEM-FC independently.
172 10 A Fuel Cell and Hydrogenation Engine Hybrid System
Figure 10.3 illustrates the model showing the relation between the load factor and
the generation efficiency of hydrogenation NEG and PEM-FC. In the low-load
operation region of NEG, an improvement in emission cleanup and brake thermal
efficiency is expected by increasing the hydrogen concentration of the fuel [71].
However, there is no advantage of hydrogenation in the high-load operation region
of NEG. Moreover, as shown in Fig. 10.3, compared with PEM-FC, the maximum
efficiency of NEG is low. Therefore, operation in a low-load region of NEG is
advantageous, and operation in a high-load region of PEM-FC is advantageous.
Thus, at OM-C, PEM-FC is changed to hydrogenation NEG according to the
magnitude of the load. However, compared with the capacity of NEG and PEM-
FC installed by OM-A and OM-B, although NEG decreases, PEM-FC does not
change. From Figs. 10.2(a) and (c), the capacity of NEG installed into OM-C
10.3 Equipment Characteristics 173
becomes less rather than OM-A. However, the expensive fuel cell installed into
OM-C is the same capacity, as seen in Figs. 10.2(b) and (c). Therefore, the equip-
ment cost of OM-C increases compared with OM-A and OM-B. Thus, in order to
use OM-C, the energy unit cost needs to decrease significantly compared to OM-A
and OM-B. On the other hand, when using PEM-FC for a base load, FC can al-
ways be operated at a maximum efficiency point. So, with OM-D, the load fluc-
tuation region where low load frequently occurs is followed by hydrogenation
NEG. The capacity of NEG and PEM-FC installed in OM-D decreases compared
with OM-A and OM-B.
The output characteristics of NEG with hydrogenation take into account past
evaluation results. The specifications of the engine examined and those of the power
generator assumed in this chapter are shown in Table 10.1. Figure 10.4(a) shows the
examination results of the hydrogenation rate and the brake thermal efficiency of
the gas engine in Table 10.1(a). This examination results are data from [71]. The
effective range of hydrogenation is 0.8 MPa of the mean engine effective pressure.
On the other hand, in the range in which the mean effective pressure exceeds
0.8 MPa, even if hydrogen is not added, high thermal efficiency can be obtained.
Figure 10.4(b) shows the relation between the hydrogenation rate, the mean effec-
tive pressure, and the brake thermal efficiency [71]. The broken line in the figure is
the hydrogenation rate to express maximum thermal efficiency. Figure 10.5 shows
the relation of the city gas (CH4) consumption, the amount of hydrogenation, and
the electricity production of NEG at the time of maximum thermal efficiency. The
characteristics of Fig. 10.5 were calculated from the hydrogenation rate based on
the examination results of Figs. 10.4(a) and (b). In Fig. 10.5, when the production of
electricity exceeds 14 kW, the amount of hydrogenation is zero. This is because
high thermal efficiency can be obtained even if there is no hydrogenation in the
large range of engine power, as Fig. 10.4(a) describes.
(2) Efficiency
Figure 10.6 shows the relation between the load factor (load added to NEG/the
capacity of NEG), and the generation efficiency. Reformer efficiency is included
in the generation efficiency shown in Fig. 10.6. Here, Eq. (10.4) defines the re-
former efficiency.
q
R = H2 100 [%] (10.4)
q CH 4
10.3 Equipment Characteristics 175
of Fig. 10.2(c) and OM-D of Fig. 10.2(d). Generally, the engine thermal efficiency
increases, so this capacity grows. The NEG generation efficiency characteristics of
5 kW and 10 kW are shown in Fig. 10.6. The difference in the maximum genera-
tion efficiency shown in Fig. 10.6 is about 3%. Figure 10.7 shows the result of the
power of NEG (5 kW and 10 kW), the heat output, and the total efficiency. The
heat outputs are engine exhaust, cooling hot water, and reformer exhaust heat.
However, it is assumed that the total efficiency of Fig. 10.7 consumes all power
and heat produced by NEG. The difference in the total efficiency of NEG of
10 kW and 5 kW is less than 1%. Under conditions where all exhaust heat is con-
sumed, the difference in total efficiency by the capacity of NEG is small.
The hydrogen quantity supplied to NEG can be determined from Fig. 10.5. CH4
supplied to a reformer, and the amounts of carbon dioxide discharged by a reform-
ing reaction are calculable using Eqs. (10.1), (10.2), and (10.3). The sampling time
is expressed as t . When adding hydrogen to NEG, the amount of carbon dioxide
discharged by the reforming reaction, and the heat-source burner is expressed as
g N,R, t and g N,B,t , respectively. Moreover, the amount of carbon dioxide dis-
charged by city gas burning in NEG is expressed as g N,I,t . The amount g N, t of
carbon dioxide discharged by the power generation of NEG is calculated by
Eq. (10.5).
g N, t = g N,R, t + g N,B,t + g N,I,t (10.5)
Figure 10.8 shows the carbon dioxide emission characteristics of NEG of Ta-
ble 10.1. This figure shows the relation between the load factor, and CO2 emis-
sions are also shown. The characteristics of the total CO2 emissions differ by
about 60% of the load factor. The fuel supply rate to NEG is due to the fact that
there is a large quantity of hydrogen in a low-load region, and there is much city
gas in a high-load region. In a low-load region, a large amount of CO2 is dis-
charged by reforming of the city gas reforming, which produces hydrogen, as well
as reformer burners. Moreover, in a high-load region, there is a large amount of
CO2 where city gas is burned in engines. The characteristics of Fig. 10.8 change
with the capacity of NEG, as well as with the generation efficiency. In the calcula-
tion of the carbon dioxide emissions of the analysis case, 5 kW NEG was set up
compared with 10 kW NEG, with a maximum increase of 3%.
10.3 Equipment Characteristics 177
Figure 10.9 shows the output characteristics of 5 kW PEM-FC and the model of
total efficiency with a city gas reformer [26, 28, 29, 67, 75]. The model of
Fig. 10.9 was prepared from the analysis results in this chapter, and the literature
on PEM-FC [26, 28, 29, 67, 75]. The heat outputs are the exhaust heat of the re-
former and the cell stack. Moreover, the power output is the value of the inverter
outlet. Total efficiency is assumed when all power and heat is consumed. Fig-
ure 10.10 shows a model of the carbon dioxide emissions and generation effi-
ciency of PEM-FC of Fig. 10.9. The carbon dioxide discharged by the operation of
PEM-FC is the city gas burning of a reformer burner (Eq. (10.3)), and the reform-
ing reaction (Eqs. (10.1) and (10.2)). The generation efficiency F of Fig. 10.10
was calculated using Eq. (10.6). E F, t on the right-hand side of Eq. (10.6) is the
power of the inverter outlet of PEM-FC. q R,CH 4 , t is the calorific power of CH4
supplied to a reformer, and q B,CH 4 , t expresses the calorific power of CH4 supplied
to the heat-source burner of the reformer. The maximum generation efficiency of
the fuel cell shown in Fig. 10.10 is 32%.
178 10 A Fuel Cell and Hydrogenation Engine Hybrid System
Fig. 10.10 Characteristics model of the load factor of a PEM-FC with reformer, and power
generation efficiency. The area of the electrode including the anode and cathode of the fuel cell
stack is 1m2, respectively, and the reformer efficiency is 73%
E
F, t = F, t
(q R, CH 4
+ q B, CH 4 , t ) 100
(10.6)
The amount of CO2 discharged by the reforming reaction and a heat-source burner
is expressed as g F,R, t and g F,B, t , respectively. The amount g F, t of CO2 dis-
charged by the power generation of PEM-FC is calculated from Eq. (10.7).
g F, t = g F,R, t + g F,B, t (10.7)
10.4 Power and Heat Output Characteristics of HCGS 179
The changing ratio of the CO2 emissions of NEG is large in a high-load region,
as Fig. 10.8 shows. On the other hand, the CO2 emissions of PEM-FC are large in
a low-load region, as Fig. 10.10 shows. From the difference among these CO2
emission characteristics, the operation of NEG is advantageous in a low-load re-
gion, and operation of PEM-FC is advantageous in a high-load region.
The chart showing the relation between the calorific power of the city gas supplied
to HCGS, the production of electricity, and the heat output is described as the
operation map of HCGS. The city gas calorific power of an operation map ex-
presses the fuel consumption of HCGS. Figure 10.11 shows the operation map of
OM-A (10 kW hydrogenation NEG). Operation Area A of this figure is a region of
the production of electricity and exhaust heat power when operating NEG. Area B
is a region of the production of electricity and the heat output (the amount of ex-
haust heat, and the boiler power) when operating NEG with a boiler. In the figure,
the city gas calorific power (described as city gas consumption in the following)
supplied to a system is also shown. Furthermore, the boiler efficiency of Area B is
90%. Figure 10.12 shows the operation map of OM-B (10 kW PEM-FC). When
OM-A is compared with OM-B, Area A with OM-A is wider. Furthermore, in
Area A with OM-A, if the production of electricity increases, the exhaust heat
Fig. 10.11 Relation of fuel supply and output of 10 kW NEG with a boiler (OM-A)
180 10 A Fuel Cell and Hydrogenation Engine Hybrid System
Fig. 10.12 Relation of fuel supply and output of 10 kW PEM-FC with a boiler (OM-B)
power will also increase. However, with OM-B, even if the production of electric-
ity increases, the increase in heat output is lost. PEM-FC has the feature that the
rate of heat output to production of electricity becomes low when the production
of electricity increases.
The city gas consumption of OM-A and OM-B in the same heat output range
(050 kW) as the same power generation range (210 kW) are 1173 kW, and
1769 kW, respectively, as shown in Figs. 10.11 and 10.12. The difference of the
city gas consumption by Area A of OM-A and OM-B is large. For example, in the
power generation range of 210 kW, city gas consumption of OM-A is 1144 kW.
On the other hand, city gas consumption of OM-B is 1731 kW.
Figure 10.13 shows the operation map of OM-C. In OM-C, NEG is operated in a
low-load region, and PEM-FC is operated in a high-load region. The efficiency of
exhaust cleanup and improvement at partial load by the hydrogenation of NEG is
known. So, in OM-C, NEG is operated in a low-load region. The operation map of
5 kW NEG and 10 kW PEM-FC is sketched in Fig. 10.13. Areas A and B are the
operation maps of NEG, and Areas C and D are the operation maps of PEM-FC.
10.4 Power and Heat Output Characteristics of HCGS 181
Fig. 10.13 Relation of fuel supply and output of 5 kW NEG and 10 kW PEM-FC hybrid opera-
tion (OM-C)
In OM-C, a system with lower city gas consumption per production of electricity
is selected, and NEG or PEM-FC is operated. The operation of switching NEG
and PEM-FC makes a threshold value of 3.9 kW production of electricity. By
switching the operation of NEC and PEM-FC, OM-C shows the load characteris-
tics of OM-A and OM-B.
Figure 10.14 shows the operation map of 5 kW NEG and 5 kW PEM-FC operated
by OM-D. In OM-D, 5 kW PEM-FC is operated corresponding to a base load, and
5 kW-NEG is operated according to the fluctuating load. Since the base load does
not change, PEM-FC can always be operated at maximum efficiency. The maxi-
mum efficiency range of PEM-FC is near the 5 kW production of electricity of
Fig. 10.14(b). On the other hand, NEG operates corresponding to the load fluctua-
tion, and low-load operations occur frequently. The low-load operation of NEG
has a high generation efficiency compared with PEM-FC. Therefore, it is expected
that the generation efficiency of OM-D is high compared with OM-A and OM-B.
182 10 A Fuel Cell and Hydrogenation Engine Hybrid System
Figure 10.15 shows the power and heat demand model of ten family apartment
houses in Tokyo, and is used for analysis. The power and heat demand model are
the average loads of each sampling time of a representative day in January (win-
ter), May (mid-term), and August (summer) [3033]. However, the actual power
demand pattern is an assembly of the load that changes rapidly in a short time,
such as an inrush current. Space cooling, space heating, and household appliances
are power loads, and hot water supply and baths are heat loads. In Tokyo, the
10.5 Case Study 183
Fig. 10.15 Power and heat demand model for ten family apartment houses
annual average temperature for the past five years has been 289 K. The average
temperature in January is 279 K, and the highest and the lowest temperatures on a
representative day in January are 283 K and 275 K, respectively. The average
temperature in May is 292 K, and the highest and the lowest temperatures on a
representative day in May are 296 K and 288 K, respectively. The highest and the
lowest temperature on a representative in day August for the past five years have
been 302 K and 296 K, respectively, and the average temperature is 298 K [46].
Since the cooling load of air-conditioners will be included in the power demand
amount of a representative day in August, the power demand is high compared
with other months. The capacity of NEG and PEM-FC of OM-A to OM-D was
determined so that the power demand of Fig. 10.15(a) may be satisfied. The maxi-
mum value of the power demand of Fig. 10.15 is 9 kW. Therefore, OM-A and
OM-B that operate NEG or PEM-FC independently set the power generation ca-
pacity to 10 kW. Moreover, OM-C set NEG to 5 kW, and set PEM-FC to 10 kW.
OM-D set NEG and PEM-FC to 5 kW, respectively. The capacity of NEG of OM-
C, NEG of OM-D, and PEM-FC can be changed into other values. In this chapter,
in order to simplify the analysis, NEG and PEM-FC may be 5 kW and 10 kW.
The operation map of HCGS shown in Figs. 10.1110.12 is installed into a power
and heat demand model (Fig. 10.15), and the city gas consumption and the amount
of exhaust heat power are obtained. When the exhaust heat power exceeds the heat
demand, surplus heat is stored in a heat storage tank. On the other hand, when the
power is lower than the heat demand, heat is supplied from a heat storage tank. If
heat still runs short, it is output by a boiler. City gas consumption and the amount
of heat storage are calculated using all the sampling times of a representative day
every month. The generation efficiency of a representative day is calculated by
dividing the power demand amount by the consumption calorific power of city
gas. The total efficiency is calculated by dividing the value adding production of
184 10 A Fuel Cell and Hydrogenation Engine Hybrid System
electricity and the amount of heat output by the consumption calorific power of
city gas. The load factor is given in Figs. 10.810.10, and the carbon dioxide
emissions for every sampling time are obtained. The amount of emission on a
representative day is calculated by adding these carbon dioxide emissions.
Figures 10.16(a)(c) show the analysis results of city gas consumption using each
operation method. With OM-B the change in fuel consumption in every month is
small compared with other operation methods. The reason for this can be ex-
plained from the operation map of OM-B shown in Fig. 10.12. In Area A of OM-
B, it is because the city gas consumption with a fluctuating production of electric-
ity has a change smaller than other operation methods. On the other hand, a
change in the fuel consumption of each month with OM-A is large compared with
the other operation methods. This is because the change from city gas consump-
tion to the production of electricity is large in Area A of the NEG operation map
(Fig. 10.11). With OM-D the least amount of fuel consumption will be in January
and May. Figure 10.16(d) shows the analysis results of city gas consumption for
every month on a representative day. There are few operation methods of OM-D
for fuel consumption. Since August has a high power demand due to the cooling
load, there is more city gas consumption with OM-A and OM-B than in other
months. However, in January there will be more city gas consumption with OM-C
than in August. This is because in August the generation efficiency of PEM-FC
operated in the high-load region of OM-C is higher than in January.
Figure 10.17 shows the analysis results of the generation efficiency and the total
efficiency of each operation method. OM-D has the highest generation efficiency
and total efficiency every month. The total efficiency on a representative day in
May with OM-D is the highest. This is because there is little fuel consumption, as
Fig. 10.16(d) shows. However, a representative day in August has a high load
factor of NEG due to the cooling load, and the generation efficiency and total
efficiency of OM-D are lower than in other months. The generation efficiency and
total efficiency of OM-C are high following OM-D. OM-C has particularly high
generation during a load peak in August when the power demand is high. This is
because the hours of operation of 10 kW PEM-FC are long, and the load factor at
the time of operation is higher in August than in other months. Furthermore, the
generation efficiency of OM-B is strongly influenced by the power demand
amount compared with other operation methods. As described in the section Ca-
pacity of the Heat Storage Tank, heat output characteristics differ according to
each operation method of OM-A to OM-D. However, the heat demand considering
the heat output characteristics of each operation method is required for the total
efficiency shown in Fig. 10.17(b).
Figures 10.18(a)(c) are the analysis results of the carbon dioxide emissions for
every sampling time in each operating method. Compared with the relation be-
tween each operation method of Fig. 10.16(a)(c) and city gas consumption, the
result of Fig. 10.18 shows a great difference according to the operation method.
The characteristics of the carbon dioxide emissions of NEG and PEM-FC shown
in Figs. 10.8 and 10.10 are due to the fast that the difference is large between low
load and high loads. For this reason, fuel consumption (Fig. 10.16(d)), generation
efficiency (Fig. 10.17(a)) and total efficiency (Fig. 10.17(b)), and the relation of
carbon dioxide emissions are not in a simple proportional relationship. This rela-
tionship is remarkable when power demand with load fluctuation and low load
operation is used. The carbon dioxide emissions of each operation method are
dependent on the operation map of each operation method shown from Figs.
10.1110.14. When the operating point of the system has been determined, the
fuel consumption will be obtained from the operation map. Generation efficiency
and carbon dioxide emissions are calculable from this value. Therefore, the rela-
tionship of fuel consumption, generation efficiency, and carbon dioxide emission
can be predicted by preparing the operation map of the system. OM-B has the
largest amount of carbon dioxide emission on a representative day every month.
Figure 10.8(d) shows the analysis results of the carbon dioxide emission on a
representative day. OM-A has the lowest carbon dioxide emission, followed by
OM-C, OM-D, and OM-B. From this, it is considered that the operation of NEG
with hydrogenation is most suitable with respect to the carbon dioxide emission of
the power demand pattern with the large load fluctuation shown in Fig. 10.15(a).
Figures 10.19(a)(d) show the analysis results of the heat balance of the heat stor-
age tank for every sampling time of each operation. VS,1 VS,4 shown in
Fig. 10.19(a) express the heat output in a heat storage tank. The operation method
with the greatest excess heat balance is OM-D, and OM-C shows the least excess
heat balance. When OM-A is compared with OM-B, there is a particular differ-
ence in the heat output in January and May, but this is due to the difference in the
exhaust heat characteristics of NEG and PEM-FC. According to Figs. 10.19(a)
(d), VS determines the largest value to be the heat storage tank capacity. The result
of the heating storage capacity obtained in the analysis is shown in Table 10.2.
The heat storage capacity of the equipment with a lower rate of heat output to
power output such as PEM-FC is larger, and it is so large that the peak of the heat
load is high. The capacity of the heat storage tank installed in OM-D from the
analysis results of Table 10.2 is the smallest.
10.7 Conclusions
The city gas-engine cogeneration facility is cheap compared with PEM-FC. How-
ever, it is inferior with respect to maximum generation efficiency, exhaust gas
composition, and efficiency lowering at partial load and noise. Therefore, attention
has been focused on the gas-engine generator (NEG) with hydrogenation for im-
proving efficiency and exhaust gas emission. In hydrogenation NEG, emission
cleanup and brake thermal efficiency improvement is confirmed by increasing the
hydrogen rate of a fuel at the time of low load operation. Therefore, this study
investigated the output characteristics of the hybrid cogeneration (HCGS) of
PEM-FC and NEG by numerical analysis. These output characteristics were in-
corporated into the power and heat demand model of ten family apartment houses
in Tokyo, and the following conclusions were obtained.
(1) Compared with the conventional method (the power supply at one set of
PEM-FC, or one set of NEG), an operation method where generation effi-
ciency and total efficiency are improved was proposed. These methods are the
operation method that switches hydrogenation NEG to PEM-FC in the magni-
tude of load, and the method of combining the base load operation of PEM-FC
and the load fluctuation operation of NEG.
(2) The carbon dioxide emissions characteristics of NEG with hydrogenation and
PEM-FC show a large difference between low load operation and high load
operation. For this reason, the carbon dioxide emissions of a system and its
fuel consumption, generation efficiency, and total efficiency are not in a sim-
ple proportional relationship.
(3) The exhaust heat characteristics of HCGS and the capacity of the heat storage
tank were investigated. Since exhaust heat characteristics differ depending on
the operation method, in order to maintain total efficiency, the heat demand
corresponding to the heat output characteristics for every operation method is
used.
(4) The total efficiency of the method of combining the base load operation of
PEM-FC and the load fluctuation operation of NEG is the highest. Moreover,
under a load pattern with large fluctuation of power demand, individual NEG
operation with hydrogenation has the lowest carbon dioxide emissions.
Chapter 11
CO2 Discharged from a Compound Micro-grid
of a Hydrogenation City Gas Engine
and a Fuel Cell
11.1 Introduction
generation efficiency and carbon dioxide emission when connecting NEG and the
proton exchange membrane fuel cell (PEM-FC) to an independent micro-grid. In
this chapter, the load of an independent micro-grid is divided into a base load
region and a fluctuating load region, and the system that compounds NEG and
PEM-FC is examined. Below, the independent micro-grid that compounds the
NEG and PEM-FC systems is described as IMPE.
PEM-FC system is defined as an IMPE system. By the IMPE system, the kinds of
generating equipment of the base-load operation and the fluctuating-load operation
differ.
Figure 11.2 shows the power load pattern of the independent micro-grid shown in
Fig. 11.1(b). The load pattern of Fig. 11.1(b) is separated into a base load region
and a fluctuating load region in systems other than the centralized system. As
Fig. 11.1(b) shows, the PEM-FC system of capacity PF,l is installed in building 5
linked to Grid A, and NEG of capacity PE,m is installed in building 19 linked to
Grid B. Grids A and B can deliver and receive the power by system interconnec-
tion equipment CP. Therefore, the PEM-FC system of building 5 is made to corre-
spond to the base load region of Fig. 11.2, and NEG of building 19 is made to
correspond to a fluctuating load region.
The hydrogen produced by the reformer is supplied to NEG at the time of low
load, and PEM-FC stack, NEG, PEM-FC stack, a city gas reformer, a boiler, a
heat storage tank, an interconnection device, etc., are installed in the building
shown in Fig. 11.3(b). City gas ( Q S ) is a heat source, and city gas ( Q R ) produces
reformed gas with the fuel for reforming. Furthermore, in order to reduce the CO
concentration in the reformed gas in a fuel cell stack entrance to several ppm, a
carbon monoxide oxidization section is provided. In the carbon monoxide oxida-
tion section, carbon monoxide is burned on a catalyst and it changes into carbon
dioxide. The direct current power generated by the fuel cell stack is changed into
an alternating current of fixed frequency through an inverter, and is supplied to an
interconnection device. Figure 11.3(c) shows the equipment scheme of Building
n in which NEG or PEM-FC system is not installed. The power demand of Build-
ing n is received from a micro-grid through an interconnection device. Moreover,
heat supply is obtained by city gas ( Q B ) burning of a boiler. Carbon dioxide
emissions are calculated from the city gas supplied to a reformer ( Q R and Q S )
and NEG ( Q E ).
Figure 11.4(a) shows the examination results of the hydrogenation rate and brake
thermal efficiency of a one-cylinder city gas engine [71]. The engine mean effec-
tive pressure of hydrogenation is effective in a range that is less than 0.8 MPa.
Thermal efficiency with a large mean effective pressure without hydrogenation in
a range exceeding 0.8 MPa can be obtained. Figure 11.4(b) shows the relation
between the mean effective pressure and the brake thermal efficiency, and the
hydrogenation rate [71]. The broken line in the figure is the hydrogenation rate
indicating the maximum thermal efficiency. Figure 11.4(c) shows the analysis
results of the production of electricity of NEG, city gas consumption, and the
amount of hydrogenation calculated from the model of Figs 11.4(a) and (b). The
amount of hydrogenation of Fig. 11.4(c) is the result when obtaining the maxi-
mum thermal efficiency. The specifications of a city gas engine and a power gen-
erator are shown in Tables 11.1(a) and (b). Hydrogen consumption is zero when
the production of electricity exceeds 14 kW, as shown in Fig. 11.4(c). This is be-
cause high thermal efficiency can be obtained even if there is no hydrogenation in
the large range of engine power, as Fig. 11.4(a) describes. Figure 11.4(d) shows
the relation of the production of electricity and the generation efficiency of NEG.
Although reformed gas is supplied to NEG, the generation efficiency of
Fig. 11.4(d) includes reformer efficiency. The reformer is as common as the PEM-
FC system described below. Details of reformer efficiency are given in the section
The PEM-FC System.
Figure 11.5(a) shows the relation between carbon dioxide emission and the pro-
duction of electricity of NEG and engine hydrogenation. This model was calcu-
lated from the characteristics of the thermal efficiency described in Figs. 11.4 and
11.3 Equipment Characteristics 195
the equations (Eq. (11.1)-(11.3)) described below. The fuel supplied to NEG has
many hydrogen rates in a low-load region, and there are many rates of city gas in a
high-load region. Therefore, there are many rates of carbon dioxide discharged
with a reforming reaction and a reformer burner in a low-load region, and there are
many rates of carbon dioxide discharged by the engine burning of city gas in a
high-load region. Figure 11.5(b) shows the model of a load factor and CO2 emis-
sions calculated from Fig. 11.5(a). In Region A in this figure, NEG is mainly op-
erated using reforming gas. In this region, CO2 emissions decrease slightly with
the rise of a load factor. It is because reformer efficiency will improve when a load
factor rises as described in the section The PEM-FC System and Fig. 11.6.
PEM-FC stack is determined, the amount of city gas supplied to a reformer and
the carbon dioxide to be discharged are calculable using Eqs. (11.1)(11.3). The
CH4 quantity supplied to an engine is calculable using Eqs. (11.2), and Figs.
11.4(a) and (b).
CH 4 + H 2 O CO+ 3 H 2 206 [kJ/mol] (11.1)
CO + H 2 O CO 2 + H 2 + 41 [kJ/mol] (11.3)
Figure 11.6 shows the model of the generation efficiency of PEM-FC system and
the city gas reformer efficiency [77]. Moreover, generation efficiency F of
Fig. 11.6(a) was calculated using Eq. (11.5). When the sampling time is expressed
with t , E F, t of Eq. (11.5), the right-hand side is the power in the inverter outlet of
a PEM-FC system. Q R,F, t expresses the calorific power of CH4 for reforming, and
Q S,F, t expresses the calorific power of CH4 supplied to a heat-source burner. The
maximum generation efficiency of the fuel cell shown in Fig. 11.6 is 31%. More-
over, the reformer efficiency in Fig. 11.6(a) improves with the increase in a load
factor. Figure 11.6(b) shows the CO2 emissions of the PEM-FC system. Fig-
ure 11.6(b) shows the result of calculating based on the power-generation effi-
ciency and reformer efficiency in Fig. 11.6(a). At the time of the hydrogen supply
to the PEM-FC stack, the amount of CO2 discharged by a reforming reaction is
expressed with G R, F,t , and the quantity discharged by a heat-source burner is
expressed with G S, F, t . Therefore, the amount G F,t of CO2 discharged by the
generation of the PEM-FC system is calculated by Eq. (11.6).
E
F, t = F, t
(Q R,F,t + QS,F,t ) 100 (11.5)
The urban area model analyzed in this chapter is shown in Fig. 11.7. The number
of buildings is shown in this figure and the application for each building is shown
in Table 11.2. The number of buildings of an urban area model is 20. The urban
area model can consider various patterns. This chapter examines the characteris-
tics of the carbon dioxide emission of the compound grid of NEG and the PEM-
FC system from the case of Fig. 11.7.
Figure 11.8 shows the power demand model of each building in Tokyo, and is the
mean power load of each sampling time of a representative day in January (win-
ter), May (mid-term), and August (summer) [3033]. However, the actual power
demand pattern is an assembly of the load that changes rapidly in a short time,
such as an inrush current. A power demand estimate of the houses shown in Figs.
11.8(a)(d) is difficult. On the other hand, the power demand of the small offices
of Fig. 11.8(g) and the factories of Fig. 11.8(h) is regular and easy to predict. The
power demand pattern of a house has a peak in the morning and in the afternoon.
When midnight to early morning is excluded, hotels (Fig. 11.8(e)) have a stable
demand and convenience stores (Fig. 11.8(f)) have a 24-hour power demand. In
small offices (Fig. 11.8(g)), factories (Fig. 11.8(h)), and small hospitals
(Fig. 11.8(i)), there is small power demand at night to early morning, and there is
much power demand from morning until evening. In the case study, the CO2 emis-
sion of August representation days, which have the largest power demand, are
calculated. Figure 11.9 shows the heat demand model in August of each building
described at the top of the figure [3033]. However, in a convenience store, an
office, and a factory, because the heat pump is introduced, heat demand is not
taken into consideration.
The analysis flow of the centralized system, the base load-sharing system, and the
IMPE system is shown in Fig. 11.10. First, the power demand data of each build-
ing are given to the analysis program, and the base load of the whole micro-grid is
calculated. Next, the power plant capacity installed into a micro-grid is given, and
the power generation efficiency and carbon dioxide emission are calculated for
every sampling time concerning all the grid routes of an urban area model. By
adding all these, the total power generation efficiency and the total carbon dioxide
emission in the operation period, and the capacity of a power plant are determined.
The load factor is calculated from the capacity and power load of a power plant.
A load factor is given to the approximation of Fig. 11.4(d) or Fig. 11.6, and the
power generation efficiency is determined. The carbon dioxide emission of a
system are calculated by giving a load factor to the approximation of Figs. 11.5
or 11.6.
11.4 Case Study 199
Figure 11.11(a) shows the result of the power load pattern of a representative day
in August of the urban area model. As the result of the time change of the power
11.5 Results and Discussion 201
demand, Fig. 11.11(a) shows that the power plant capacity of a fluctuating load is
100 kW, and the power plant capacity of a base load is 66 kW. Figure 11.11(b)
shows the result of the rate of a base load and a fluctuating load. The base load is
1.32 times larger. Figure 11.11(c) shows the composition of the power demand
added to a micro-grid. The largest load component is convenience stores (two
buildings), which takes 36% of the whole load. Because there is a 24-hour power
demand in convenience stores (Fig. 11.8(f)) and hotels (Fig. 11.8(e)), it is a com-
ponent that smoothes the whole load added to the micro-grid. Factories and small
offices of the ratio of the whole load are large. However, the demand difference
during the day and at night is large, and is a component for which the fluctuating
load region of a micro-grid is made to increase.
The analysis results of a representative day in August are shown in Fig. 11.12.
Figure 11.12(a) shows the results of the capacity of the power plant installed in
each micro-grid system. (A)(E) in Fig. 11.12 expresses the power supply method
described in the figure. In a centralized system ((A) and (B)), one set of the
11.5 kW power plant is connected to a micro-grid. On the other hand, in the base
load-sharing system ((C) and (D)) and the IMPE system ((E) and (F)), the power
plant capacity corresponding to a base load and load fluctuation is 66 kW and
100 kW, respectively.
Figure 11.12(b) shows the analysis results of the total power generation efficiency
of the system of (A)(F). Total power generation efficiency is high at (A), (C), (E)
and (F). Most these systems are a method of corresponding to base-load operation
by FC. Figure 11.12(c) shows the distribution of the power-generation efficiency
of the system of (C)(F), except for the centralized system. In FC base-load opera-
tion, it can operate at the maximum power-generation efficiency shown in
Fig. 11.6(a). The maximum power-generation efficiency of PEM-FC system is
higher than the efficiency of NEG shown in Fig. 11.4(d). Therefore, the total
power generation efficiency of the system of (A), (C), and (E) using FC to base-
load operation is high.
Figure 11.12(d) shows the analysis result of the amount of carbon dioxide dis-
charged from each system. (C) and (E) have small CO2 emission, and these are
PEM-FC system base load operations. Moreover, (F) (NEG base load and FC load
fluctuation operation) also has small CO2 emission. When Fig. 11.5(b) is com-
pared with Fig. 11.6(b), the change of NEG CO2 emissions is larger than the
PEM-FC system. As Figs. 11.5(b) and Fig. 11.6(b) show, when the load factor of
PEM-FC system is large and the load factor of NEG is small, CO2 emission will
decrease. Therefore, there is little CO2 emission of (C) and (E). Although (F) is the
NEG base load operation, because it corresponds to load fluctuation by the large
capacity PEM-FC system (100kW), there is less CO2 emission than in the system
composed only from NEG ((B) and (D)). (E) in Fig. 11.12(d) is CO2 emission of
NEG without hydrogenation. When the hydrogenation of NEG is introduced,
compared with the method that does not add hydrogen, about 15% of CO2 emis-
sion will reduce. Finally, the order with little CO2 emission is (C), (F), (E), (E),
11.5 Results and Discussion 203
(D), (A), and (B). The order ((A) and (C), (E), (F), (B) and (D)) of the power-
generation efficiency described above differs from this order. Furthermore, when
facility cost is taken into consideration, the smallest possible system of fuel cell
capacity is advantageous. Power-generation efficiency is high, there is little CO2
emission, and a system with cheap facility cost is the best. Therefore, system (E) is
proposed in this chapter.
204 11 CO2 Discharged from a Compound Micro-grid of a Hydrogenation City Gas Engine
The energy supply by commercial power and a kerosene boiler is defined as the
conventional method. The amount of greenhouse gas discharge of the conven-
tional system is calculable based on the investigative commission report of the
calculation method of the amount of greenhouse gas discharge (the Ministry
of Environment, Japan, August, 2003). The commercial power of a greenhouse
gas discharge factor is 0.331 kg-CO2/kWh, and a kerosene boiler is set up at
0.0685 kg-CO2/MJ, 9.5 kg-CH4/TJ, and 0.57 kg-N2O/TJ. As a result, as shown in
Fig. 11.1(d), the CO2 emission of system (E) decreases slightly more than with the
conventional method. It depends on the amount of discharge of CO2 on the magni-
tude of a power load factor. Therefore, the amounts of discharge of the system (A)
differ greatly compared with the system (C).
Figure 11.13 shows the analysis results of the exhaust heat output of the base load
operation and the load fluctuation operation of each system for representative days
in August. The exhaust heat of each system of (A) to (F) exceeds the heat amount
demanded of the urban area model in Fig. 11.7, as shown in Fig. 11.13. When an
exhaust heat network is introduced into a micro-grid, and exhaust heat is distrib-
uted to each building, the boiler shown in Fig. 11.3 will become unnecessary.
Fig. 11.13 Heat demand and exhaust heat output on representative days in August
11.6 Conclusions
the PEM-FC system, and are operated. The following conclusions were obtained
as a result of installing this system into an urban area model composed from the
power load patterns of 20 buildings, such as houses, offices, and hospitals.
(1) The total power generation efficiency of the centralization system, base-load
sharing system, the IMPE system using PEM-FC and NEG was in the range
of 19%30%. Especially, power-generation efficiency has a high introduction
of the PEM-FC system base load operation.
(2) The load factor of the PEM-FC system is large, and the system with a small
load factor of NEG has little CO2 emission. There is little CO2 emission with
the PEM-FC base-load sharing system as a result of analysis. Further, there is
little CO2 emission with the IMPE system of PEM-FC and NEG. Moreover,
when hydrogen is added to NEG, CO2 emission will be reduced by 15%.
(3) A PEM fuel cell base load and the system operating a hydrogenation city gas
engine in a load fluctuation region are the most advantageous under the over-
all evaluation of facility cost, power generation efficiency, and CO2 emis-
sions. When the urban area model was analyzed using the highest system of
an overall evaluation, 25% of power-generation efficiency and CO2 emissions
were 1,106 kg/day. The amount of CO2 emission is influenced by the magni-
tude of the power-demand fluctuation (load factor) of the micro-grid.
The maximum effect is expected by reflecting making an energy-demand char-
acteristic in the planning of a micro-grid. The relation between the locality of an
energy demand characteristic and the optimal design of a micro-grid will be inves-
tigated hereafter.
Chapter 12
Development of a Fast Operation Algorithm
of a Fuel Cell System with Solar Reforming
12.1 Introduction
There are many methods to produce the hydrogen gas supplied to a fuel cell sys-
tem [7881]. Moreover, the production method for hydrogen fuel depends on the
emission characteristic of the fuel cell greenhouse gas. For this reason, a reform-
ing method using sunlight as the heat source is examined [82, 83]. If solar reform-
ing equipment can be developed, the distribution of the small-scale fuel cell sys-
tem will be accelerated. To date, there have been cases involving study of the solar
reforming system using methane gas [82, 83]. On the other hand, the authors are
investigating steam reforming of bioethanol [84]. Within the proposed system,
hydrogen is produced by steam reforming of the bioethanol and used to operate
the proton exchange membrane fuel cell (PEM-FC). The heat of the solar collector
with condensing is then used for vaporization and steam reforming of the bioetha-
nol fuel. Bioethanol is predominantly carbon-free and the heat source of this sys-
tem uses renewable energy. The bioethanol solar reforming system proposed in
this chapter is described as FBSR. Although FBSR is clean compared with hydro-
gen production methods using fossil fuel, the amount of hydrogen production
fluctuates depending on the weather, time, season and installation location, etc.
Previous weather patterns (outside air temperature and global solar radiation) are
put into the neural network (NN), and the training signal is analyzed via the input
of the genetic algorithm (GA). This method was used to investigate optimization
of the collecting area of the solar collector with condensing, and the capacity of
the reforming gas cylinder and the heat storage tank [84]. In this chapter, the algo-
rithm of NN, which analyzes the dynamic operation of the arbitrary days of the
FBSR at high speed, is developed. If the dynamic operation method of the system
is determined within a short time, it can be optimized by installing this algorithm
in the control device. As for the dynamic operation plan in the non-linear system
of many variables, the method of approximating a linear problem [85] and that of
using the GA [86] are proposed. By the approximation method of a linear prob-
Figure 12.1 is a block diagram of the fuel cell system with the bioethanol solar
reforming system (referred to as FBSR) [84]. In FBSR, two paraboloid rotating
mirrors (described as condensing solar collectors) that have a solar tracking sys-
tem are used. The high-density solar energy in condensing solar collector A is
used for vaporization of the bioethanol fuel. The solar energy in condensing solar
collector B is used as the heat source to reform the fuel steam. When solar energy
cannot be used, it generates electricity using the reformed gas stored in the cylin-
der, or commercial power is supplied. In order to reduce CO, a shifter and CO
oxidation equipment are installed. Reformed gas is supplied to the fuel cell and the
power is converted into regular voltage and a regular frequency with a DCDC
converter and an inverter. This power is supplied to an individual house or a
Fig. 12.1 Fuel cell system with bioethanol solar reforming system (FBSR)
12.2 System Configuration 209
power grid through an interconnection device. The exhaust heat of the cooler, the
fuel cell, and the CO oxidization equipment shown, respectively, in Fig. 12.1 is
supplied to the heat storage tank, and this heat is supplied to the demand side.
The S/C (mole ratio of steam to ethanol) of the steady-state value of the reformed
fuel supplied to the vaporizer equipment is 3.0. Moreover, the SV value (space
velocity) in the catalyst layer in the solar reformer unit is 3000 hour1, the conver-
sion ratio of the reformed gas is 95%, the methanization rate is 5%, and the CO
generation percentage is 10% [88]. When there is little solar radiation, the SV
value is controlled to maintain the reaction temperature of the reforming unit at
673 K or more.
210 12 Development of a Fast Operation Algorithm of a Fuel Cell System
Equations (12.1) and (12.2) express the balance of the power and heat, respec-
tively. The left-hand side of each equation expresses inputs, and the right-hand
side expresses outputs. IT and CT in Eq. (12.1) are the efficiency of the in-
verter and the DCDC converter, respectively. Moreover, E AUX expresses the
demand of auxiliary machinery. H Loss on the right-hand side of Eq. (12.2) is the
heat loss of the system.
Power balance
Heat balance
CL + H
H CS + H
OU + H
ST + H
BL = H
SYS + H
Loss (12.2)
Equation (12.3) expresses the mass balance of hydrogen, and Eq. (12.4) expresses
the reaction formula of the steam reforming of ethanol. Hydrogen is calculable
) of bioethanol based on Eq. (12.4). However, the
from the supply flow rate ( Q BE
rate converted into hydrogen among Q
BE is decided by the methane conversion
ratio of bioethanol.
Hydrogen balance
+Q
Q SR HC + Q EX = Q OU + Q OU Q SU (12.3)
Figure 12.3 shows the preparation procedure of the operation prediction algorithm
of the SRF developed in this chapter. In the operation prediction using the NN, as
12.4 Dynamic Operation Prediction of SRF 211
shown in Fig. 12.3, it is necessary to perform the learning process first. The train-
ing signal used for the learning calculation is previously calculated by the GA.
The training signal is the optimal solution of the dynamic operation plan on a
representative day. When predicting operation, the input data given to the NN is
the meteorological data (the amount of global solar radiation, and the outside air
temperature) and the energy-demand data (power and heat demand), both of which
were previously measured. The results obtained by the proposed NN are the
amount of hydrogen production and the amount of exhaust heat output for every
sampling time in the representative day to be predicted.
Based on these results, the quantity of hydrogen and exhaust heat to be stored
for every sampling time in the representative day to predict can be known.
The structure of the layered neural network introduced in this chapter is shown in
Fig. 12.4. This neural network consists of three layers, the input, the medium, and
the output. All the neurons between each layer of the NN are connected with net-
works. Each neuron is outputted to output layer So,k, j according to the magnitude
of the input. Inputoutput between the neurons of n 1 layer and n layer is
shown in Fig. 12.5. Input x nj of neuron j in n layer is calculable using output
out nk 1 and weight w n,j, nk 1 of neuron k of the n 1 layer, as shown in Eq. (12.5):
where j = 1, " , L n and k = 1, " , L n 1 .
L n 1
x nj = w n,j, nk1 out nk 1 (12.5)
k =1
212 12 Development of a Fast Operation Algorithm of a Fuel Cell System
Output out nj of neuron j in n layer is given with the sigmoid function of in-
put x nj , as shown in Eq. (12.6).
1
out nj =
(
1 + exp x nj ) (12.6)
1
All weights w n,n
j,k in the NN shown in Fig. 12.4 using the error-correction learn-
ing method are determined. So, the past weather pattern and the energy-demand
12.4 Dynamic Operation Prediction of SRF 213
pattern in a building are given to the NN. Training signals are solutions of the
operation plan on the representative day previously calculated by the GA. When
w n,j, nk 1 is decided by learning of the NN, the error of the training signal and the
data of the output layer can be evaluated from Eq. (12.7). This error is expressed
as ErrN , and the NN is made to learn, changing weight w n,j, nk1 so that ErrN may
approach zero.
( )
1 Ln 2
ErrN = y j out Nj (12.7)
2 j=1
Corrected weight w n,j, nk1 is expressed with Eq. (12.8) using the weight before mo-
dification w n,j, kn 1 and the amount of modifications w n,j, nk 1 . Amount of modifica-
tions w n,j, nk1 in Eq. (12.9) is expressed with Eq. (12.10). The term of the partial
differential of the right-hand side of Eq. (12.9) is calculable by Eqs. (12.10) and
(12.11) [12].
ErrN ErrN
w n,j, nk 1 = = out nj 1 (12.9)
w n,j, nk 1 x nj
In the case of n = N
ErrN
x nj
( )
= y j x Nj out Nj 1 out Nj( ) (12.10)
ErrN L n ErrN
x j
= +
(
w l,n +j 1, n out nj 1 out nj ) (12.11)
n
l =1 x j
n 1
Figure 12.6 shows the analysis flow of the learning process of the NN. All weights
w n,j, nk1 are first determined at random, and learning rate in Eq. (12.9) is given
to the program. Next, input data x nj described in the section Analysis Procedure
of the Operation Prediction Algorithm and training signal y j described at the
bottom section are input into the program, and input x nj and output out nj of each
214 12 Development of a Fast Operation Algorithm of a Fuel Cell System
neuron are calculated. Equation (12.10) is calculated using out nj and y j , and
Eq. (12.11) is calculated further These results are introduced into Eq. (12.9), and
w n,j, nk 1 is calculated. When this value is given to Eq. (12.8), the weight of each
neuron can be updated. The analysis error is calculated using Eq. (12.7), and cal-
culation will finish if this value is smaller than the value set up previously. On the
other hand, when the analysis error is larger than the value set up previously, as
shown in Fig. 12.6, the process is returned and calculated repeatedly.
When the current time is t11 , as shown in Fig. 12.7 for example, the operation of
the system in time t14 is predicted. In this case, by time t 0 to t10 , the weather
measurement data and energy demand data of the same day are input into the
learned NN described previously. Moreover, the meteorological data measured on
the previous day is used for the input value of t12 , and t15
and t13 to t 23 . The
input data given to the NN at the prediction time t14 should give the predictor of
the best possible accuracy. In this chapter, the monthly average value at the same
time of the previous year is input.
12.5 Preparation of the Training Signal Using a GA 215
The amount of hydrogen production of the FBSR is expressed with the chromo-
some code shown in Fig. 12.8. The chromosome code is a 13-bit genetic code of 0
or 1. Figure 12.9 shows the analysis flow for obtaining training signals using the
GA. Many chromosome codes are generated, and the adaptive values (objective
function) described in the following section for all the chromosomes are calcu-
lated. Chromosome groups with a low adaptive value are exchanged for chromo-
some groups newly generated at random. Moreover, handling of crossover and
mutation are added to chromosomes, and diversity is maintained. These calcula-
tions are repeated by the number of generation numbers set up previously. In the
last generation, it is decided that the chromosome with the highest adaptive value
is the optimal solution. The amount of the hydrogen production is decided by
decoding this chromosome code. The amount of hydrogen storage, the output of
exhaust heat and thermal storage, etc., are calculated by introducing this solution
into Eqs. (12.1)(12.6).
Figure 12.10 shows the case of the power-demand pattern in a building and the
operation pattern of the system. In this figure, solar radiation is obtained with a
condensing solar collector in period R dh from 06:00 to 12:00. In this case, start
time t st of the system operation is 06:00. In insufficient time zones, solar radia-
tion generates electricity using the reformed gas stored during the day. In period
R of time t gs to time t ge shown in Fig. 12.7, energy is supplied from the fuel
cell. Although t gs and t ge can be determined arbitrarily, in the case study in the
section Case Study, it is set at 23:00 and 07:00. This time zone is set as mid-
night power in many domestic electric power companies.
Equation (12.12) expresses the sum total of the difference in the power of the
system and demand on a representative day. When the cylinder capacity of the
reformed gas and the capacity of the heat storage tank are expressed as VH 2 and
Vhw , respectively, objective function f o of the system is given in Eq. (12.13).
w o,1 , w o,2 , and w o,3 in Eq. (12.13) express the weight of each term. In this
chapter, it is decided that the adaptive value in the GA is a high solution, so that
218 12 Development of a Fast Operation Algorithm of a Fuel Cell System
the value of f o is small. Moreover, the reformed gas cylinder and the heat storage
tank are designed to be as small as possible.
t ge
Errday = E N E SYS (12.12)
t gs
Operation of the system when the FBSR shown in Fig. 12.1 is introduced into an
individual house is predicted. The energy demanded of the individual house used
as input data introduces the data of the reference [8]. Moreover, the database of
the Japanese Meteorological Agency is used concerning outside air temperature
and global solar radiation. Operation of the system is taken as the pattern shown in
Fig. 12.10 where, t gs = 08:00 and t ge = 21:00. The data input into the NN by the
learning calculation are the weather pattern in 2006 and the energy-demand pat-
tern of a house [90]. The operation prediction of the FBSR on arbitrary days is
calculable by giving the learning NN the weather pattern and the energy-demand
pattern.
The single-cell performance of the PEM-FC is shown in Fig. 12.11. This perform-
ance is obtained from the experiments in the references [28, 29]. The maximum
output point of the single cell is decided as 100% of a load factor in this chapter.
Figure 12.12 shows the relation between the load factor and the power density
prepared from single-cell performance. The characteristics shown in Fig. 12.12 are
divided into three areas, and approximate expressions are shown in this figure.
When the production of electricity of the PEM-FC is decided, the generation effi-
ciency, the power density, and the amount of exhaust heat will be decided from
Figs. 12.11 and 12.12. The conditions of the system components used in analysis
are shown in Table 12.1.
The parameters introduced into the analysis of the GA and the NN are shown in
Tables 12.2 and 12.3, respectively. These values were determined by applying a
trial-and-error method for many analyses. When operation of the FBSR is ana-
lyzed by the NN using the conditions in Table 12.3, the computation time is 240
times faster compared with the GA.
The analytical object is an individual house in Sapporo with the weather pattern
and the energy-demand pattern in 2006 shown in Fig. 12.13. There is no cooling
load in the summer season, and hot water supply, baths, and the space heating load
220 12 Development of a Fast Operation Algorithm of a Fuel Cell System
Fig. 12.13 Meteorological pattern and energy demand pattern in Sapporo in 2006
are included in the heat demand. Moreover, the load of electric lights and house-
hold appliances is included in the power demand, and there is no large monthly
difference. Figure 12.14 shows the measurement data of the outside air tempera-
ture for each day in January and August, 2007. The NN is made to learn using the
weather pattern in 2006. Operation on each representative day in January and
August, 2007 (J1, J2, A1, A2) is predicted using this NN. In the operation plan in
this chapter, the priority of design variables is in order of the power balance, the
miniaturization of the cylinder capacity of the reformed gas, and the miniaturiza-
tion of the capacity of the heat storage tank. Then, weights w o,1 , w o,2 and w o,3 ,
of the objective function shown in Eq. (12.13) are set at 0.7, 0.2, and 0.1, respec-
tively.
12.7 Results and Discussion 221
Figure 12.15 shows the result of the optimizing operation of the FBSR using the
GA based on the weather pattern in 2006. Figure 12.15(a) shows the analysis re-
sults of the collecting area of the solar collector with condensing on a representa-
tive day every month. Moreover, Fig. 12.15(b) shows the result of the rate,
whereby the energy supplied by FBSR takes to the amount demanded. The power
supplied from the FBSR is 67% to 76% of the demand. However, the heat sup-
plied from the FBSR is 3%47%, differing greatly according to the season. Sup-
plying energy for the power demand from t gs to t ge is included in the objective
function (Eqs. (12.12) and (12.13)) of the FBSR. Therefore, the power supply rate
from the FBSR is high compared to demand.
Figure 12.16 shows the meteorological data measured on January 4 and 5, 2007
(representative day J1 in Fig. 12.14(a)). When the present is 3:00 on January 5,
operation of the system of 3 hours after (6:00) is predicted. In this case, the meas-
urement data on January 5 in Figs. 12.16(a) and (b) is used for the input data from
0:00 to 3:00 to the NN. Moreover, the measurement data on January 4 in Figs.
12.16(a) and (b) are used for the input data 4:00 and 5:00, and 7:00 to 23:00.
Moreover, concerning the input data to the NN at 6:00 for prediction, the average
data of the sampling time for every month of 2006 is used.
Figure 12.17 shows the operation plan showing the amount of hydrogen produc-
tion by the NN, and the amount of exhaust heat thermal storage. In the same fig-
ure, the operation result calculated by the GA, which is considered to be the opti-
mal solution, is also shown. Figures 12.17(a)(c) represent the prediction results
of the current time, up to 1 hour after, 3 hours after, and 6 hours after, respec-
12.7 Results and Discussion 223
tively. When the amount of hydrogen production and the amount of exhaust heat
thermal storage are negative, the purchased power and operation of the boiler are
required. Figures 12.18(a)(c) show errors of the analysis result of the NN and
GA. The integrated value is defined for the error in the forecast for one day shown
in Fig. 12.18 as the integration error in the forecast in this chapter. Compared with
the result of Fig. 12.17(a), the integration error in the forecast of Figs. 12.17(b)
and 12.17(c) is 1% and 7% of the increase in hydrogen production, and 2% and
11% of the increase in thermal storage respectively. Accordingly, analytic accu-
racy deteriorates, so that the time to predict operation is the future.
Figures 12.1912.21 show the analysis results of the meteorological data and the
operation prediction concerning the representative day J2 (January 12, 2007) in
Fig. 12.14(a). Figure 12.20 shows the operation plan of the amount of hydrogen
production, and the amount of exhaust heat storage, while Fig. 12.21 shows the
error of the operation result in the NN and the GA. In this chapter, the monthly
average outside-air-temperature data in 2006 (Fig. 12.19) are used for the learning
process of the NN. Therefore, when the outside-air-temperature characteristic
(Fig. 12.19) of a prediction day differs from the outside-air-temperature data in the
monthly average in 2006, the analysis error of operation prediction will be consid-
ered to increase. When the characteristics of the monthly average and representa-
tive day in Figs. 12.16(a) and 12.19(a) are compared, Fig. 12.16(a) is larger and
when the difference of monthly average and representative day, as shown in
Figs. 12.16(a) and 12.19(a), is compared, Fig. 12.16(a) is larger. Consequently, the
Figures 12.22 and 12.23 show the analysis results of the meteorological data and
the operation prediction on the representative day A1 (August 15, 2007) and the
representative day A2 (August 25, 2007) in Fig. 12.14(b), respectively. Fig-
ures 12.22(c) and 12.23(c) show the operation plan of the amount of hydrogen
production, and the amount of exhaust heat storage, while Figs. 12.22(d) and
12.23(d) show the error of the operation result in the NN and the GA. Compared
with Fig. 12.23(a), the difference between the outside-air-temperature character-
istic in the monthly average in 2006 and the measurement outside-air-
temperature (representative day) input into the NN is larger in Fig. 12.22(a).
However, the difference in the global-solar-radiation characteristic on these repre-
sentative days is small compared with that in the outside-air-temperature charac-
teristic. In this chapter, the monthly average outside-air-temperature data in 2006,
as shown in Figs. 12.22(a) and 12.23(a), are used for the learning process of the
NN. According to the difference of the monthly average outside-air-temperature
characteristic in 2006, and the temperature characteristic used for prediction, as
shown in Figs. 12.22(d) and 12.23(d), the magnitude of the analysis error changes.
228 12 Development of a Fast Operation Algorithm of a Fuel Cell System
12.8 Conclusions
The operation prediction algorithm of a fuel cell system with a bioethanol solar
reforming system using a neural network (NN) was developed. In this algorithm,
the operation plan of a building analyzed using a genetic algorithm (GA) is intro-
duced as a training signal. When input data (weather pattern and energy-demand
pattern) are given to the learning program, the operation method of the system (the
amount of hydrogen production, and the amount of exhaust heat thermal storage)
on arbitrary days can be obtained. A proposal algorithm was formed to learn about
use of the meteorological data in Sapporo in 2006, and a dynamic operation plan
was analyzed for two or more representative days in January and August, 2007.
These results were then compared with the optimal solution by the GA. Any error
in the operational analysis depends on the difference between the outside-air-
temperature pattern used for the learning process of the NN and the outside-air-
temperature pattern of the prediction day. Moreover, the analysis error of opera-
tion prediction increases, meaning there is previous prediction time. The operation
prediction of the system, based on past weather patterns, is rapidly analyzable over
a constant error range with the proposal algorithm.
Chapter 13
Power Characteristics of a Fuel Cell Micro-grid
with Wind Power Generation
13.1 Introduction
Figure 13.1 shows the fuel cell independent micro-grid model investigated in this
chapter. There is a network of power and city gas in this micro-grid. Although
Fig. 13.1 Fuel cell micro-grid system with a wind power generator
13.3 Response Characteristics of System Configuration Equipment 233
a power network connects all houses, a city gas network connects houses in which
a fuel cell is installed. The fuel cell installed in each house is a proton exchange
membrane type (PEM-FC). The output of a 2.5 kW fuel cell is determined to be
a base power of the micro-grid. Moreover, PEM-FC of 1 kW power is installed in
seven houses. However, the fundamental dynamic characteristics of all the fuel
cells are the same, and a fuel cell and a city gas reformer are installed as a pair. One
set of wind power generators is installed, and the power produced by wind force is
supplied to a micro-grid through an inverter and an interconnection device. The
power supply of the micro-grid assumes 50 Hz of the single-phase 200 V.
Figure 13.2(b) shows the output model that step-wise inputted a load of a 100%
load factor into the city gas reformer [4, 26, 27, 67, 92]. An approximated curve
was prepared from the result of the measurement, and the transfer function of the
primary delay of the city gas reformer was obtained. Although the transfer func-
tion of a city gas reformer influences the magnitude of the load significantly, since
there is no large difference, the result of Fig. 13.2(b) was used as a fuel cell. Com-
pared with the condition of the steady operation of the reformer, the characteristics
of startup and shutdown differ greatly. Cold start and shutdown operations require
about 20 minutes each. In the analysis of this chapter, it is assumed that the startup
of the methanol reformer is always a hot start.
234 13 Power Characteristics of a Fuel Cell Micro-grid with Wind Power Generation
Figure 13.4 shows a model of the relation between the load factor of a fuel cell
and the generation efficiency [52, 93]. Power generation efficiency is obtained by
dividing the power output of the fuel cell system by the city gas calorific power
supplied to the system. This model was prepared from the results of the power
output when attaching the fuel cell show in Fig. 13.2(a) to the city gas reformer
236 13 Power Characteristics of a Fuel Cell Micro-grid with Wind Power Generation
show in Fig. 13.2(b). If the load of a fuel cell is given to Fig. 13.4, power genera-
tion efficiency is calculable. The maximum efficiency of one set of a fuel cell
system is 32%.
It is assumed that an inverter of a voltage control type is used, and 120 ms is re-
quired to output power on regular voltage and frequency (in this chapter, it is less
than 95%) [53]. Figure 13.5(a) expresses the transfer function of such an inverter
with primary delay.
When changing the power with a system interconnection device, the change
takes about 10 s [8]. However, there is the operation of taking the synchronism
of the frequency between systems, and the model of the system interconnection
device sets the change time to 12 ms. The transfer function of the system intercon-
nection device by primary delay is shown Fig. 13.5(b).
The response characteristics of the 1 kW fuel cell system when inputting 0.13.10,
and a 1.0 kW load step-wise is shown in Fig. 13.6. The response characteristics
of a fuel cell system changes by the control parameters set up with the controller.
As shown in Fig. 13.6(c), with 1 kW step input, the rising time and settling time
(time to converge on 5% of the target output) are not based on control parame-
ters. With a 0.2 kW step input, the rise time of P = 12.0, I = 1.0 is short, and the
settling time of P = 1.0, I = 1.0 is short. With a 0.6 kW step input, P = 12.0, I = 1.0
and P = 1.0, I = 1.0 have almost the same settling time. Moreover, overshooting
is large although the rise time of P = 12.0, I = 1.0 is short. Considering the follow-
ing load fluctuations, the control parameters of the fuel cell are analyzed by
P = 12.0, I = 1.0. The dynamic characteristics of a micro-grid are analyzed using
MATLAB (Ver.7.0) and Simulink (Ver.6.0) of The MathWorks Corporation.
However, in the analysis, the solver used is the positive RungeKutta system, and
this determines the sampling time from calculation converged to less than 0.01%
by error.
of the target output) are described in the left-hand part of Fig. 13.7(a). Moreover,
the maximum power due to overshooting is described in the right-hand part of the
figure. The settling time when not installing a wind power generator has the long-
est period of step input of 6 kW and 8 kW for 3.9 seconds. If a wind power genera-
tor is connected to the micro-grid, many fluctuations in the system response char-
acteristics will occur in a short period. If the power produced by wind power
generation is supplied to the micro-grid, the dynamic characteristics of power of
the micro-grid will be influenced. Figure 13.7(b) shows the analysis results of the
response error corresponding to Fig. 13.7(a). If a wind power generator is con-
nected to the grid, the response error will become large as the load of the grid
becomes small. It is expected that the power range of the fluctuation of the micro-
grid will increase as the output of the wind power generation grows. Therefore,
when the load of a micro-grid is small compared with the output of a wind power
generator, and the power supply of the independent micro-grid becomes unstable.
Fig. 13.8(a) shows the power demand pattern of a micro-grid formed from ten
individual houses in Sapporo, Japan, and assumes a representative day in Febru-
ary. This power demand pattern is the average value of each hour, and the sam-
pling time of analyses and the assumption time are written together on the hori-
zontal axis. As a base load of the power demand pattern shown in Fig. 13.8(a), F/C
(0) is considered as operation of 2.5 kW constant load. Figures 13.8(b) and (c) are
the power demand patterns when adding load fluctuations (1 kW and 3 kW) to
Fig. 13.8(a) at random. The variation of the load was determined at random within
the limits of the range of fluctuation for every sampling time.
Figure 13.9 shows the response results of F/C (0) to F/C (6) when wind power
generation is connected to the micro-grid and the power load has 1 kW fluctua-
tions. F/C (0) assumed operation with 2.5 kW constant output, with the result that
the response of F/C (0) is much less than 2.5 kW in less than the sampling time of
100 s, as shown in Fig. 13.9(a). This is because F/C (0) was less than 2.5kW with
the power of wind power generation. Although the micro-grid assumed in this
chapter controlled the number of operations of F/C (1) to F/C (7) depending on the
magnitude of the load, since the power supply of wind power generation existed,
there was no operating time of F/C (7).
240 13 Power Characteristics of a Fuel Cell Micro-grid with Wind Power Generation
Figure 13.10 shows the analysis results of the average power generation efficiency
of fuel cell systems for every sampling time. The average efficiency of a fuel cell
system is the value averaging the efficiency of F/C (0) to F/C (7) operated at each
sampling time. However, the fuel cell system to stop is not included in average
power generation efficiency. The average power generation efficiency of
Fig. 13.10(a) is 13.4%, and that of Fig. 13.10(b) is 14.3%. The difference in aver-
age efficiency occurs in the operating point of a fuel cell system shifting to the
efficient side, when load fluctuations are added to the micro-grid. Thus, if load
fluctuations are added to the micro-grid, compared with no load fluctuations, the
load factor of the fuel cell system shown in Fig. 13.4 will increase.
Figure 13.11 shows the power generation efficiency of each fuel cell in the case
of connecting wind power generation to the micro-grid of 1.0 kW of the load
fluctuation. F/C (0) operated corresponding to a base load has maximum power
generation efficiency at all sampling times. Since the number of operations of a
242 13 Power Characteristics of a Fuel Cell Micro-grid with Wind Power Generation
fuel cell is controlled by the magnitude of the load added to the micro-grid, the
operating time falls in the order of F/C (1) to F/C (6). Moreover, there is no time
to operate F/C (7) in this operating condition.
13.6 Conclusions
A 2.5 kW fuel cell was installed in a house linked to a micro-grid, operation corre-
sponding to a base load was conducted, and the dynamic characteristics of the grid
when installing a 1 kW fuel cell system in seven houses were investigated by
13.6 Conclusions 243
[1] Krist, K, Gleason, K, Wright, J (1999) SOFC-based residential cogeneration systems. In:
Stimming U et al. (eds) In: Proceedings 6th international symposium on solid oxide fuel
cells VI. Electrochemical Society, Pennington, NJ, p 107
[2] Nagata, Y, Hori, M, Sonai, A, Shimotori, S (1999) In: Uchida et a (eds) Development of
PEMFC systems and components at Toshiba. Proceedings 3rd International fuel cell confer-
ence. Nagoya, Japan, p 69
[3] Dhathathreyan, S, Sridhar, P, Sasikumar, G, Ghosh, K, Velayutham, G, Rajalakshimi, N,
Subramaniam, K, Raja, M, Ramya, K (1999) Development of polymer electrolyte mem-
brane fuel cell stack. Int J Hydrogen Energy 24(11): 11071115
[4] Nagano S, Yamamoto S, Asano, T, Ohsawa, K (2002) Plate-type methanol steam reformer
using new catalytic combustion for a fuel cell. In: Proceedings SAE Technical Paper Series,
Society of Automotive Engineers, p 10
[5] Changrani S, Samms, S, Rogers S, Tasic S, Gervasio D, Koripella R (2001) Reactor design
and analysis of a miniature methanol steam reformer. Prepr Pap-Am Chem Soc Div Fuel
Chem 46(2): 476478
[6] Amphlett, C, Mann F, Peppley A (1996) Performance and operating characteristics of
methanol steam-reforming catalysts for on-board fuel-cell hydrogen production. Hydrog
Energy Prog 11(2): 17371743
[7] Amphlett, C, Creber M, Davis M, Mann F, Peppley A, Stokes, M (1992) Hydrogen produc-
tion by steam reforming methanol for polymer electrolyte fuel cells. Hydrog Energy Prog
9(3): 15411550
[8] Narita K (1996) The research on unused energy of the cold region city and utilization for the
district heat and cooling. PhD thesis, Hokkaido University (in Japanese)
[9] Alibhai Z, Lum R, Huster A, Gruver WA, Kotak DB (2004) Coordination of distributed
energy resources. In proceedings of the annual meeting of the North American Fuzzy In-
formation Processing Society, Alberta, vol 2, p 913
[10] Lasseter RH (2004) Microgrid. A conceptual solution, In: Proceedings 35th IEEE annual
power electronics specialists conference, vol 6, p 4285
[11] Obara S, Kudo K (2005) Study on improvement in efficiency of partial load driving of
installing fuel cell network with geo-thermal heat pump and water electrolysis operation. In:
Proceedings 8th international energy agency heat pump conference (CD)
[12] Aki H, Murata A, Yamamoto S, Kondoh J, Maeda T, Yamaguchi H, Ishii I (2005) Penetra-
tion of residential fuel cells and CO2 mitigationcase studies in Japan by multi-objective
models. Int J Hydrogen Energy 30(9): 943952
[13] Hernandez-Aramburo CA, Green TC, Mugniot N (2005) Fuel consumption minimization of
a microgrid. IEEE Trans Ind Appl 41(3): 673681
245
246 References
[14] Goldberg DE (1989) Genetic algorithms in search, optimization and machine learning,
Addison Wesley, Reading MA
[15] Srinivas M, Patnaik LM (1994) Genetic algorithms: A survey. IEEE Computer 27(6): 1726
[16] Baker JE (1985) Adaptive selection methods for genetic algorithms. In: ICGA85, Proceed-
ings 1st international joint conference on genetic algorithms, p 101
[17] Yu H, Fang H, Yao P, Yuan Y (2000) A combined genetic algorithm/simulated annealing
algorithm for large scale system energy integration. Comput Chem Eng 24(8): 20232035
[18] Chang YC, Lin JK, Chuang MH (2005) Optimal chiller loading by genetic algorithm for
reducing energy consumption. Energy Build 37(2): 147155
[19] Dewdney AK (1985) Exploring the field of genetic algorithms in a primordial computer sea
full of flobs. Sci Amer 85(5) 1621
[20] Maruzen KK (2003) Chronological scientific tables. National Astronomical Observatory of
Japan
[21] Robert H, Paigi P (2004) Microgrid: A conceptual solution, In: IEEE 35th annual power
electronics specialists conference, Aachen, vol 6, p 4285
[22] Carlos A, Hernandez A (2005) Fuel consumption minimization of a microgrid. IEEE Trans
Indus Appl 4(3): 673681
[23] Takuma Y, Goda, T (2005) Microgrid for urban energy. Trans Soc Heat Air Cond Sanit Eng
Jpn 79(7): 573 579 (in Japanese)
[24] Obara S, Kudo K (2005) Study on improvement in efficiency of partial load driving of
installing fuel cell network with water electrolysis operation. Trans Jpn Soc Mech Eng, Se-
ries B, 71(701): 237244 (in Japanese)
[25] Obara S, Kudo K (2005) Study of efficiency improvements in a fuel-cell-powered vehicle
using water electrolysis by recovering regeneration energy and avoiding partial load opera-
tion. Trans ASME J Fuel Cell Sci Tech 2: 202207
[26] Takeda Y, Iwasaki Y, Imada N, Miyata T (2004) Development of fuel processor for rapid
start-up. In: Proceedings 20th energy system economic and environment conference, Tokyo,
p 343 (in Japanese)
[27] Ibe S, Shinke N, Takami S, Yasuda Y, Asatsu H, Echigo M (2002) Development of fuel
processor for residential fuel cell cogeneration system. In: Proceedings 21st annual meeting
of the Japanese Society of Energy and Resources, Osaka, p 493 (in Japanese)
[28] Mikkola M (2001) Experimental studies on polymer electrolyte membrane fuel cell stacks.
Masters thesis submitted in partial fulfillment of the requirements for the degree of Master
of Science in Technology, Helsinki University of Technology
[29] Ibaraki Prefecture Government Office of Education (2002) Modeling of hydrogen energy
system. High school active science project research report, Ibaraki, Japan (in Japanese)
[30] Architectural Institute of Japan (2002) The nationwide research study concerning the energy
consumption in the house in the 2001 fiscal year, vol 3, p 36 (in Japanese)
[31] Nagase O, Ikaga T, Chikamoto T (2003) Study on the effect of energy saving methods and
global warming prevention in Tokyo, 19th Energy System-Economics-Environment Con-
ference, p 461466. (in Japanese)
[32] Hatano Y et al. (2003) Investigation on the energy use characteristics in the apartment house
of hot water supply and heating of an area. In: Proceedings of the meeting of the Society of
Heating, Air-Conditioning and Sanitary Engineering, p 17451748 (in Japanese)
[33] Yoshinaga M et al. (2003) Investigation on the energy consumption of a house, and a resi-
dents consciousness. In: Proceedings of the meeting of the Society of Heating, Air-
Conditioning and Sanitary Engineering, p 17291732 (in Japanese)
[34] Badami M, Caldera C (2002) Dynamic model of a load-following fuel cell vehicle: impact
of the air system. SAE2002-01-100, p 110
[35] Inoue U (1996) Air-conditioning handbook. Maruzen, Tokyo, Japan, p 14
[36] Yamano Y (2002) Development of a load-levelling technique. Yaskawa Electric Corpora-
tion, Kita-Kyushu, Japan, Denki 629: 5661
References 247
[37] Ozaki S, Tuziki I (1990) Trial calculation of the quantity of public electric power and city
gas to be replaced by a distributed energy system. In: Proceedings of the 9th energy-
resources seminar, vol 9, p 174 (in Japanese)
[38] Obara S, Kudo K (2006) Route planning of heat supply piping in a fuel cell energy network.
J Environ Eng 1(1): 1728
[39] Obara S (2006) The Hot Water Piping Route of a Fuel Cell Energy Network with a Concen-
tration Installing Method. Int J Hydrogen Energy 31(13): 18071818
[40] Takuma T, Goda T (2005) Microgrid for urban energy. Trans Soc Heat Air Cond Sanit Eng
Jpn 79(7): 573 579 (in Japanese)
[41] Obara S (2006) Load response characteristics of a fuel cell micro-grid with control of num-
ber of units. Int J Hydrogen Energy 31(13): 18191830
[42] Hongmei Y, Haipeng F, Pingjing Y, Yi Y (2000) A combined genetic algorithm/simulated
annealing algorithm for large scale system energy integration. Comput Chem Eng 24:
20232035
[43] Zhu D, Latombe JC (1991) Pipe Routing, Proceedings of the 1991. In: IEEE international
conference on robotics and automation, Sacramento, CA, 19401947
[44] Sang K. Sehyun Myung, Soon Hung Han (1999) A Design Expert System for Auto-Routing
of Ship Pipes. J Ship Prod 15(1):1
[45] Obara S (2004) Route Planning and Estimate of Heat Loss of Hot Water Transportation
Piping for Fuel Cell Local Energy Network. In: Proceedings of International Symposium on
New Refrigerants and Environmental Technology. The Japan Refrigeration and Air Condi-
tioning Industry Association, Kobe, Japan, p 225227
[46] National Astronomical Observatory (2003) Rika Nenpyo chronological scientific tables,
Maruzen, Tokyo, Japan (CD-Rom)
[47] Zhang Y, Ouyang M, Lu Q, Luo J, Li Z (2004) A model predicting performance of proton
exchange membrane fuel cell stack thermal systems. Appl Thermal Eng 24: 501513
[48] Sedghisigarchi K (2004) Dynamic and transient analysis of power distribution system with
fuel cellsPart 1: Fuel-cell dynamic model. IEEE Trans Energ Convers 19(2): 423428
[49] Hamelin J, Agbossou K, Laperriere A, Laurencelle F, Bose TK (2001) Dynamic behavior
of a PEM fuel cell stack for stationary applications. Int J Hydrogen Energy 26: 625629
[50] Yerramalla S, Davar, A, Feliach A, Biswas T (2003) Modeling and simulation of the dy-
namic behavior of a polymer electrolyte membrane fuel cell. J Power Sources 124: 104113
[51] Koike S, Inaka H, Ohmura T, Kirai K, Kishi H (1999) Demonstration program of 1kW-class
PEMFC system for residential use by the Japan Gas Association. In: Proceedings 3rd inter-
national fuel cell conference, p 497498
[52] Obara S, Kudo K (2005) Study of a small-scale fuel cell cogeneration system with methanol
steam reforming considering partial load and load fluctuation. Trans ASME J Energy Res
Tech 127(4): 265271
[53] Kyoto Denkiki Co, Ltd (2001) A system connection inverter catalog and an examination
data sheet
[54] HC-TECH Inc (1997) HC12a and HC22a properties and performance tests data sheets,
p 13
[55] Rakopoulos CD, Glakoumis EG (2006) Comparative First- and Second-law Parametric
Study of Transient Diesel Engine Operation, Energy 31(12): 19271942
[56] Katrasnik T et al. (2005) Analysis of the dynamic response improvement of a turbocharged
diesel engine driven alternating current generating set. Energ Convers Manage 46(1819):
28382855
[57] Obara S, Kudo K (2007) Installation planning of small-scale fuel cell cogeneration consid-
ering transient response characteristics (load response characteristics of electric power out-
put). J Environ Eng 2(3): 621631
[58] Obara S (2006) Dynamic characteristics of a PEM fuel cell system for individual houses. Int
J Energy Res 30(15): 12781294
[59] Obara S (2006) Effective improvement in generation efficiency due to partition cooperation
management of a fuel cell micro-grid. J Thermal Sci Technol 1(2): 4253
248 References
[60] Peacock AD, Newborough M (2007) Controlling micro-CHP systems to modulate electrical
load profiles. Energy 32(7): 10931103
[61] Hawkes AD, Leach MA (2007) Cost-effective operating strategy for residential micro-
combined heat and power. Energy 32(5): 711723
[62] De Ruyck J et al. (2007) Co-utilization of biomass and natural gas in combined cycles
through primary steam reforming of the natural gas. Energy 32(4): 371377
[63] Bonnet S, Alaphilippe M, Stouffs P (2005) Energy, exergy and cost analysis of a micro-
cogeneration system based on an Ericsson engine. Int J Therm Sci 44: 11611168
[64] Obara S (2007) Load response characteristics of the pem fuel cell for individual cold-region
houses. J Fuel Cell Sci Tech 5(1): 011005-1
[65] Ozeki h (1991) Research on a combustor for Stirling engines. 4. Effects of a combustor
engine operation parameter on exhaust characteristics. In: Proceedings of the 29th Japanese
symposium on combustion, p 208
[66] Eguchi K, Kurosawa Y (1988) Study on high load swirl burners for stirling engines. Reduc-
tion in NOx and heat adsorption characteristics in exhaust gas recirculation. In: Proceedings
of the 26th Japanese symposium on combustion, p 131
[67] Oda K, Sakamoto S, Ueda M, Fuji A, Ouki T (1999) A small-scale reformer for fuel cell
application. Sanyo Tech Rev 31(2): 99106(in Japanese)
[68] Takeda Y, Iwasaki Y, Imada N, Miyata T (2004) Development of fuel processor for rapid
start-up. In: Kimura K (ed) Proceedings 20th energy system economic and environment
conference, Tokyo, January 2930, p 343I (in Japanese)
[69] Obara S, Kito S, Hoshi A, Sasaki S (2007) Energy balance of Stirling engine cogeneration
and installation potentiality in cold region houses. J Environ Eng 2(4): 696707
[70] Ogawa T et al. (2005) Study on energy saving of the cogeneration for houses the 8th report:
The survey and fuel cell installation feasibility study of power load in an apartment house.
In: Proceedings of the meeting of the Society of Heating, Air-Conditioning and Sanitary
Engineering, vol 3, p 2037 (in Japanese)
[71] Mohammad A et al. (2005) Development of highly efficient and clean engine system using
natural-gas and hydrogen mixture fuel obtained from onboard reforming. NEDO report
ID:03B71006c (in Japanese)
[72] Yap D et al. (2004) Effect of hydrogen addition on natural gas HCCI combustion. SAE
Tech Pap Ser, SAE-200401-1972
[73] Larsen JF, Wallace JS (1995) Comparison of emissions and efficiency of a turbocharged
lean-burn natural gas and hythane-fuelled engine. ASME ICE 24: 3140
[74] Swain MR, Usuf MJ (1993) the effects of hydrogen addition on natural gas engine opera-
tion. SAE Tech Pap Ser, SAE-932775
[75] Nakamura T, Sei M (2002) Energy related technology. high-efficiency fuel processor for
fuel cell system. Technical Report, Matsushita Electric Works, Ltd, No. 77, p 4
[76] Society of Living Amenity (2006) Family-size fuel cell cogeneration systems development
of Tokyo gas, ALIA News vol. 89. http://www.alianet.org/homedock/15kinen/5-2.html (in
Japanese)
[77] Yasuda I (2005) Development of hydrogen production technology for fuel cell. Energy
Synthesis Engineering 28(2) (in Japanese)
[78] Dalle Nogare D, Baggio P, Tomasi C, Mutri L, Canu P (2007) A thermodynamic analysis
of natural gas reforming processes for fuel cell application. Chem Eng Sci 62(1820):
54185424
[79] Radulescu M, Lottin O, Feidt M, Lombard C, Le Noc D, Le Doze S (2006) Experimental
results with a natural gas cogeneration system using a polymer exchange membrane fuel
cell. J Power Sources 159(2): 11421146
[80] Schenk NJ, Moll HC, Potting J, Benders RMJ (2007) Wind energy, electricity, and hydro-
gen in the Netherlands. Energy 32(10): 19601971
[81] Lv P, Yuan Z, Ma L, Wu C, Chen Y, Zhu J (2007) Hydrogen-rich gas production from
biomass air and oxygen/steam gasification in a downdraft gasifier. Renew Energ 32(13):
21732185
References 249
[82] Kodama T (2003) High-temperature solar chemistry for converting solar heat to chemical
fuels. Prog Energy Combust Sci 29(6): 567597
[83] Petrasch J, Steinfeld A (2007) Dynamics of a solar thermochemical reactor for steam-
reforming of methane. Chem Eng Sci 62(16): 42144228
[84] Obara S, Tanno I (2007) Development of distributed energy system due to bio-ethanol pem
fuel cell with solar reforming. Part 1Evaluation of basic performance. Trans Soc Heat Air
Cond Sanit Eng Jpn 123: 2332
[85] Ito K, Shibata T, Yokoyama R (2002):_Optimal operation of a cogeneration plant in combi-
nation with electric heat pumps. J Energ Resour- ASME 116: 5664
[86] Obara S (2006) Operating schedule of a combined energy network system with fuel cell. Int
J Energy Res 30(13): 10551073
[87] Obara S (2007) Improvement of power generation efficiency of an independent micro grid
composed of distributed engine generators. J Energ Resour-ASME 129(3): 190199
[88] Urasaki K (2005) Application of lattice oxygen in metal oxides to catalysts for hydrogen
production. PhD thesis, Waseda University
[89] Homepage of Murakami and Izumida Lab, http://ipr20.cs.ehime-u.ac.jp/column/neural/
index.html, 2007 (in Japanese)
[90] Homepage of Japan Meteorological Agency, http://www.data.jma.go.jp/obd/stats/etrn/
index.php, 2007
[91] Abu-Sharkh S, Arnold RJ, Kohler J, Li R, Markvart T, Ross JN, Steemers K, Wilson P, Yao
R (2006) Can microgrids make a major contribution to UK energy supply? Renew Sustain
Energy Rev 10(2): 78127
[92] Lindstrom B, Petterson L (2003) Development of a methanol fuelled reformer for fuel cell
applications. J. Power Source.118: 7178
[93] Obara S, Kudo K (2005) Installation planning of small-scale fuel cell cogeneration in con-
sideration of load response characteristics (load response characteristics of electric power
output). Trans Jpn Soc Mech Eng, Series B, 71(706): 16781685 (in Japanese)
Index
251