Вы находитесь на странице: 1из 11

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 2 ( 2 0 1 7 ) 7 8 3 6 e7 8 4 6

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.elsevier.com/locate/he

Multi-objective optimization of the hybrid wind/


solar/fuel cell distributed generation system using
Hammersley Sequence Sampling

Hao Chen a,b, Chen Yang a,b,*, Kangjie Deng a,b, Nana Zhou a,b,
Haochuang Wu a,b
a
Key Laboratory of Low-grade Energy Utilization Technologies and Systems, Chongqing University,
Ministry of Education, Chongqing 400030, PR China
b
College of Power Engineering, Chongqing University, Chongqing 400030, PR China

article info abstract

Article history: As the development of China's economy, environmental problems in China become more
Received 26 October 2016 and more serious. Solar energy and wind energy are considered as ones of the best choices
Received in revised form to solve the environmental problems in China and the hybrid wind/solar distributed
7 January 2017 generation (DG) system has received increasing attention recently. However, the instability
Accepted 30 January 2017 and intermittency of the wind and solar energy throw a huge challenge on designing of the
Available online 24 February 2017 hybrid system. In order to ensure the continuous and stable power supply, optimal unit
sizing of the hybrid wind/solar DG system should be taken into consideration in the design
Keywords: of the hybrid system. This paper establishes a multi-objective optimization framework
Distributed generation system based on cost, electricity efficiency and energy supply reliability models of the hybrid DG
Wind/solar/fuel cell hybrid system system, which is composed of wind, solar and fuel cell generation systems. Detailed
Multi-objective optimization models of each unit for the hybrid wind/solar/fuel cell system were established. Advanced
Hammersley Sequence Sampling ε-constraints method based on Hammersley Sequence Sampling was employed in the
multi-objective optimization of the hybrid DG system. The approximate Pareto surface of
the multi-objective optimization problems with a range of possible design solutions and a
logical procedure for searching the global optimum solution for decision makers were
presented. In this way, this work provided an efficient method for decision makers in the
design of the hybrid wind/solar/fuel cell system.
© 2017 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

Company in 2015 shows that the total energy consumption of


Introduction China in 2014 was 2972.1 Mtoe, of which 66% came from the
coal [1]. This inevitably leads to the increase of the emissions.
The fast-growing economy of China gives a huge challenge for To achieve the goals of energy-saving and emission-
China's energy industry in the early 21st century. The Statis- reduction, the reform of Chinese energy consumption struc-
tical Review of World Energy published by British Petroleum (BP) ture becomes imperative. The proportion of the renewable

* Corresponding author. College of Power Engineering, Chongqing University, 174 Shazhengjie, Shapingba, Chongqing 400044, PR China.
Fax: þ86 23 6511 1297.
E-mail address: yxtyc@cqu.edu.cn (C. Yang).
http://dx.doi.org/10.1016/j.ijhydene.2017.01.202
0360-3199/© 2017 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 2 ( 2 0 1 7 ) 7 8 3 6 e7 8 4 6 7837

energy should be increased greatly [2]. Solar energy and wind HOMER and GAMBIT software [20]. An improving particle
energy, as two of the most efficient renewable energy, have swarm optimization was employed in the optimal capacity
been developed rapidly in recent years. However, both the allocation of standalone wind/solar/battery hybrid power
solar energy and the wind energy have inherent disadvan- system by Wang et al. [21]. However, most of the researches
tages: instability and intermittency. The integration of the focused on the optimization algorithm, using some simplified
above two sources could overcome the problem partially [3,4] models of each unit for the hybrid system. The complicated
and is considered as one of the best alternatives to the fossil processes of each energy conversion in each unit were not
fuel generated energy [5]. In order to ensure the continuous involved.
and stable power supply for a long period of time, a power In this paper, a fuel cell system comprising fuel cell stacks
storage system is generally considered to be added into the and an electrolyzer was added into the hybrid wind/solar DG
hybrid system as well. system for a long period of time power storage. Detailed
In recent years, the hybrid systems combined with the models of each unit in the hybrid wind/solar/fuel cell DG
solar energy, the wind energy and the power storage systems system were built. A multi-objective optimization framework
have been built and in operation [6,7]. The hybrid wind/solar based on cost, electricity efficiency and the reliability of en-
distributed generating (DG) system has the advantages of ergy supply models of the hybrid wind/solar/fuel cell DG
clean and renewable, however, the cost, the electricity effi- system was established. Because of the high calculation effi-
ciency and the reliability of energy supply are still the ciency and the great uniformity properties, Hammersley
important issues that need to be considered [8e10]. The unit Sequence Sampling was employed in the multi-objective
sizing of the hybrid wind/solar DG system has a huge impact optimization [22,23]. On the base of the traditional
on the performance of the whole system. Therefore, it should ε-constraint multi-objective optimization method, the multi-
be taken into consideration during the design phase of the objective problem is converted into a certain amount of
whole hybrid system. And the integration of the wind and single-objective problems by using Hammersley Sequence
solar energy in the hybrid system becomes complex and re- Sampling. The Pareto optimal solution set could be formed by
quires consideration with regard to grid stabilisation [11]. For the solution set of those single-objective optimization prob-
the design of a Zero Energy Building, the problems caused by lems. The approximate Pareto surface of the multi-objective
the mismatch between the demand and the production and optimization problem is given in the end aimed to search
the challenge of coordinating fluctuating and intermittent the global optimum solution. In this way, this paper provides
renewable energy production should be addressed well when an efficient evaluation method for design makers.
considering integrating the DG system into the electricity
grids [12].
Many researches have been published on the optimization Hybrid models
of the unit sizing for the hybrid wind/solar DG system. A fuel
cell vehicle powered by photovoltaic and wind energy system The hybrid wind/solar/fuel cell DG system is combined with
was modeled by Huang et al. and the influence of temperature wind energy system, the photovoltaic power generation sys-
on hydrogen flow and efficiency of the electrolyzer was tem and the fuel cell system comprising the fuel cell stacks
studied in their paper using MATLAB Simulink software [13]. and an electrolyzer, which is used for a long period of time
Coupled operation of electrolyzer with wind turbine and four power storage. Fig. 1 shows the energy management strategy
different electrolyzer models are presented and evaluated for for this hybrid DG system from Wang [24]. During the opera-
the renewable energy system by Mena et al. [14]. Chen et al. tion, the excess electricity power generated by wind and PV is
have investigated the optimal contract and installed capac- used to generate hydrogen by the electrolyzer. Then, the
ities of the wind and PV generation system for a time of use hydrogen is stored in hydrogen reservoir tanks. When the
rate for industrial users [15]. Fazelpour et al. performed an electricity power generated by wind and PV is not enough for
economic analysis of a standalone solar and wind energy the power demand, the fuel cell stacks begins to produce
system which produces energy for a household in Tehran Iran electricity fueled from the hydrogen reservoir tanks. Thus, the
[16]. Wang and Singh use the particle swarm optimization existence of the electrolyzer and the fuel cell system, ac-
(PSO) algorithm to balance the cost, the reliability and the cording to the power demand, play a role of the energy storage
pollutant emissions of a grid-connected hybrid DG system, and the energy generation respectively.
which is composed of wind turbine generators, photovoltaic
panels and storage batteries, and a set of non-dominate so- Electrolyzer model
lutions is given in their paper [17]. Chedid et al. try to help
decision makers in designing and sizing of hybrid power To simplify the model, the following hypotheses are made:
systems with solar and wind power energy by using the An-
alytic Hierarchy Process, and a trade-off surface in 3D space is (1) The water vapor in the cathode and the anode of the
presented in their works [18]. PSO, Hopfield neural network electrolytic cells is fully saturated;
(HNN) and HNN-PSO algorithm have been proposed to opti- (2) The water in the electrolytic cells is incompressible;
mize unit size for a hybrid wind/photovoltaic DG system (3) The gas phase, liquid phase can be separated;
respectively and the comparative studies and analysis of the (4) The pressures and the temperature of the electrolytic
three algorithms were carried out in Ref. [19]. Vikas et al. cells in gas flow channels are constants
presented the optimization of hydrogen based hybrid renew- (5) The water vapor enthalpy is constant within the work-
able energy system using BIG BANG CRUNCH algorithm, ing temperature;
7838 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 2 ( 2 0 1 7 ) 7 8 3 6 e7 8 4 6

Fig. 1 e Block diagram of the overall energy management strategy for the hybrid energy system.

(6) The energy used for hydrogen compression and water open-circuit voltage of the electrolytic cell is greater than E.
supply is ignored. The concentration overpotential is very small in electrolytic
(7) The work condition of each electrolytic cell is under cells according to Yang et al. [26], and it is not to be considered
100  C. in this paper. Therefore:

Vel ¼ E þ Vel;act þ Vel;ohm (4)


Due to the parasitic current losses, the theoretical
maximum value cannot be accessed in fact. According to the where Vel is the open-circuit voltage of an electrolytic cell (V).
Faraday Law [25], the actual hydrogen rate can be obtained as: Vel,act is the activation polarization loss of an electrolytic cell
! (V). Vel,ohm is the ohmic polarization loss of a single cell (V).
0:09 75:5 Activation polarization loss can be calculated by Tafel
hF ¼ 96:5 exp  2 (1)
ie ie equation [27]:
where hF is the current efficiency or Faraday efficiency; ie is the Vel;act ¼ a þ blgi (5)
electrolyzer current density (A/cm2).
The ideal hydrogen production rate can be calculated by: where a is the Tafel constant, which value is 0.05 in this paper;
b is the Tafel slope, which value is 0.11 in this paper.
hF NC Ie Ohmic Voltage Drop can be obtained as:
qH2 ¼ (2)
2F
 
where qH2 is the hydrogen production rate (mol/s). Nc is the 2870 1 1
1273Tel

number of electrolytic cells. F is the Faraday constant, which Vel;ohm ¼ 0:126Ie e (6)
value is 96487 C/mol. Ie is the electrolyzer current(A).
According to the assumption (7), the reversible potential of
Solid oxide fuel cell model
each electrolytic cell can be calculated by the corresponding
Because of the losses across the fuel cell, the open-circuit
Nernst equation:
voltage is less than the reversible potential. Activation and
" !#
DG0 RTel PH2 P0:5
O2
ohmic polarization loss can also be obtained by equations (5)
E¼ þ ln (3) and (6) above. The reversible potential of the fuel cell can be
2F 2F PH2 O
calculated by the Nernst equation:
where E is the reversible potential of each electrolytic cell " !#
(V);DG0 is the difference of Gibbs free energy for the standard DG0 RT PH2 P0:5
O2
Ecell ¼ þ ln (7)
hydrogeneoxygen reaction; R is the universal gas constant; Tel 2F 2F PH2 O
is the temperature of the electrolytic cells.
Due to the inevitable losses across the electrolytic cell,
such as activation, ohmic and concentration polarization loss, Vcell ¼ Ncell *Ecell  hact  hohm (8)
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 2 ( 2 0 1 7 ) 7 8 3 6 e7 8 4 6 7839

where Ecell is the reversible potential of the fuel cell (V). Vcell is
the fuel cell output voltage (V). PH2 , PO2 , PH2 O is the partial
pressure of the H2, O2, H2O in the fuel cell. hact, hohm is the
activation and ohmic polarization loss in the fuel cell
respectively.
The molar flow rate of the consumed H2 and O2 in the
electrochemical reaction qrH2 , qrO2 can be calculated by:

NI
qrH2 ¼ ¼ 2Kr I (9)
2F

1
qrO2 ¼ qrH2 ¼ Kr I (10)
2 Fig. 2 e Simplified four parameters equivalent circuit
The molar flow rate of the generated H2O in the electro- model of PV.
chemical reaction qrH2 O can be obtained by:

qrH2 O ¼ qrH2 ¼ 2Kr I (11)

where N is the number of the single cells; I represents the fuel where IL is the light current (A). I0 is the saturation current (A).
cell current (A); Kr ¼ N/4F. I is the load current (A). U is the output voltage (V). Rs is the
The molar flow rate of the H2, O2, H2O in the outlet of the series resistance (r). a is the thermal voltage timing comple-
fuel cell qout out out tion factor (V).
H2 , qO2 , qH2 O can be obtained by:
Light Current IL can be obtained from equation (16):
8 out
< qH2 ¼ KH2 PH2
qout ¼ KO2 PO2 f  
: Oout2
(12) IL ¼ IL;ref  mI;sc Tc  Tc;ref (16)
qH2 O ¼ KH2 O PH2 O fref

where KH2 , KO2 , KH2 O is the semi-empirical constant, which are where f is the solar radiation (W/m2). fref is the reference solar
0.843, 2.52 and 0.281 respectively obtained from Ref. [28]. radiation (which value is 1000 W/cm2). IL,ref represents the
According to the law of mass conservation and the Ideal light current at reference condition. Tc is the temperature of a
gas law, the following equations can be obtained: PV cell ( C). Tc,ref is the reference temperature (which value is
25  C). mI,sc is the temperature coefficient of short-circuit cur-
8
>
> d RT  in  rent (which value is A/25  C).
>
> PH2 ¼ qH2  KH2 PH2  qrH2
>
> dt Va Temperature has an enormous impact on the PV perfor-
>
>
<
d RT  in  mance. The thermal model is established to model the heat
PO2 ¼ qO2  KO2 PO2  qrO2 (13)
>
> dt Vca transfer in the PV module. The temperature of the PV module
>
>
>
>
>
> d RT  in  is influenced greatly by the ambient temperature, solar radi-
: PH2 O ¼ qH2 O  KH2 O PH2 O  qrH2 O
dt Vca ation, the output voltage and the current. Equation (17) shows
the relationship between the temperature of the PV module
where Va and Vca represent the volume of the anode and
and those variables:
cathode of the fuel cell respectively.
Thus, the partial pressure of the H2, O2, H2O can be calcu- dTc U*I
lated by: CPV ¼ kin;PV f   kloss ðTc  Ta Þ (17)
dt As
8   . where CPV is the overall heat capacity per unit area of the PV
>
> P ðtÞ ¼ qin  2Kr I 1  et=tH2 KH2 þ PH2 ð0Þet=tH2
< H2  H2 .  . cell/module [J/( C m2)]. Kin,PV is the absorption rate of PV cells
> PO2 ðtÞ ¼ qinH2 rHO  Kr I 1  et=tO2 KO2 þ PO2 ð0Þet=tO2 for the solar radiation. Kloss is the heat transfer coefficient [W/
>
:  
PH2 O ðtÞ ¼ 2Kr I 1  et=tH2 O KH2 O þ PH2 O ð0Þet=tH2 O ( C m2)]. Ta is the ambient temperature ( C). As is the absorp-
(14) tion area of a PV cell (m2).

where tH2 ¼ Va =KH2 RT, tO2 ¼ Vca =KO2 RT, tH2 O ¼ Vca =KH2 O RT, the
Wind energy conversion system model
values are 26.1, 2.91 and 78.3 respectively, obtained from Ref.
[28].
A simplified wind turbine generator (WTG) model is used in
this study. Equation (18) is used to calculate the output power
PV cell/module model
of WTG [18]:
8
The simplified four parameters equivalent-circuit model is > 0 v < vci
>
<
applied in this work, shown in Fig. 2: a  v3  b  Pr vci  v  vr
PWTG ¼ (18)
Equation (15) shows the relationship between output > Pr
> vr  v  vco
:
0 v > vco
voltage U and the load current I:

 
where V is the local wind speed. Vci is the cut-in wind speed of
U þ IRs the WTG. Vco is the cut-out wind speed of the WTG. Vr is the
I ¼ IL  ID ¼ IL  I0 exp 1 (15)
a
rated wind speed of the WTG. a ¼ Pr =ðv3r  v3ci Þ, b ¼ v3ci =ðv3r  v3ci Þ.
7840 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 2 ( 2 0 1 7 ) 7 8 3 6 e7 8 4 6

The PV output power is calculated as follows: where Iw ¼ a


Pw ¼ PWTG  AW  hW (19) wAw , Pwhich is the initial cost of the WTG;
N
1þb
SPw ¼ Sw Aw 1þg , which is the salvage value of WTG;
where AW is the swept area of WTG (m2); hw represents the P P  1þn i
OMPw ¼ aOMw  As  N i¼1 1þg , which is the maintenance
Efficiency of WTG.
cost for WTG.
Battery model
(2) Solar energy conversion system cost model
A simplified battery model is employed in this paper, consid-
The total cost of the solar energy conversion system can be
ering the lead-acid batteries have the same charge efficiency
obtained from:
and the discharge efficiency, which value is 0.85 in this study.
To avoid the deep discharge, 25% capacity of the lead-acid Iw  SPw þ OMPw
batteries does not participate in the charging and discharg- COSTw ¼ (24)
NP
ing processes. The remaining electricity of battery in the
where Is ¼ asAs, which is the initial cost of the solar energy
charging state can be calculated by:
conversion system;
 NP
Pbsoc ¼ Pbsoc þ Pb;in  0:85 Pb;in  0:85  Pbcap 1þb
(20) SPs ¼ Ss As 1þg , which is the salvage value of the solar
Pbsoc ¼ Pbcap Pb;in  0:85  Pbcap
energy conversion system;
The battery remaining electricity of the lead-acid battery in P P  1þn i
OMPs ¼ aOMs As N i¼1 1þg , which is the maintenance cost
the discharging state can be calculated by:
for the solar energy conversion system.

Pbsoc ¼ Pbsoc  Pb;in  0:85 Pb;soc  Pb;out 0:85  0:25Pbcap
(21)
Pbsoc ¼ Pbcap Pb;soc  Pb;out 0:85  0:25Pbcap (3) Battery cost model

where Pb,soc is the remaining electricity of the lead-acid bat-


The total cost of the lead-acid battery can be calculated by:
tery. Pbcap is the battery capacitance of the lead-acid battery.
Pb,in is the electric energy charging into the battery. Pb,out is the Ib þ OMPb
COSTb ¼ (25)
electric energy that the battery discharged. NP
PNb  1þn ði1Þ=Nb
where Ib ¼ ab Pbcap i¼1 1þb , which is the initial cost of
lead-acid battery;  
Multi-objective optimization model P P 1þn i
OMPb ¼ aOMb Pbcap N i¼1 1þg , which is the maintenance cost
for the lead-acid battery;
For the hybrid wind/solar/fuel cell DG system, the cost control
and the electricity efficiency should be taken into consider-
(4) SOFC cost model
ation during the design phase. And in order to ensure the
stable power supply, the reliability of energy supply needs to
According to Ref. [29], the total cost of the fuel cell system
be guaranteed. In such a consideration, the capital cost, the
can be calculated by:
electricity efficiency and the reliability of energy supply were
  
selected as three main objectives for the optimization of the pur Nunits 0:362
Csofc þ Cmaint
sofc þ Csofc Tlife
amort
500000
hybrid wind/solar/fuel cell DG system. And the three objective COSTsofc ¼ (26)
NP
functions were described as follows:
pur
where Csofc ¼ 1:1  ½ð0:02687Asofc þ 0:88ÞNsofc þ 248:6 þ 30:8
Objective function of the cost þ0:2465Nsofc , which is the initial cost of the SOFC;
pur
Cmaint
sofc ¼ fmaint Csofc , which is the maintenance cost of the
The objective function of the total cost of the hybrid wind/ SOFC;
pur
solar/fuel cell DG system can be described by: sofc ¼ icap Csofc , which is the amortization expense of the
Camort
SOFC per year.
COST ¼ COSTw þ COSTs þ COSTb þ COSTsofc (22) The parameters used in the cost model of the hybrid DG
system are presented in Table 1.
Refer to Ganesan et al. [19], the costs for the wind energy
conversion system, solar energy conversion system and the Objective function of the electricity efficiency
storage battery can be calculated as follows:
The electricity efficiency of the hybrid DG system within 24 h
(1) Wind energy conversion system cost model can be calculated by:

Pgenerated
The total cost of the wind energy conversion system can be h¼ (27)
Psolar þ Pwind
obtained from:
where Pgenerated represents the electricity energy which
Iw  SPw þ OMPw generated by the solar, wind energy conversion system and
COSTw ¼ (23)
NP the fuel cell stacks. Psolar is the solar energy absorbed by the
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 2 ( 2 0 1 7 ) 7 8 3 6 e7 8 4 6 7841

Table 1 e Input parameters included in the cost model of wind/PV/fuel cell hybrid system.
Parameters Parameters Description Values Parameters Parameters Description Values
aw Initial cost of WTG ($/m2) 100 ab Initial cost of batteries ($/m2) 100
aOMw Yearly operation and maintenance 2.5 aOMb Yearly operation and maintenance 10
cost for wind ($/m2 year) cost for batteries ($/m2 year)
SW Salvage value of WTG ($/m2) 10 As Area of a single PV cell (m2) 1.5
NP Lifespan of project (year) 20 Ss Salvage value of PV panels ($/m2) 45
b Inflation rate 9% Nb Lifespan of batteries (year) 10
g Interest rate 12% Tlife Lifespan of SOFC (year) 10
n Escalation rate 12% Nunits Production volume of SOFC 1482
as Initial cost of PV panels ($/m2) 450 fmaint Maintenance factor of SOFC 0.06
aOMs Yearly operation and maintenance 4.3 icap Capitalization ratio 10%
cost for solar ($/m2 year)

solar energy conversion system. Pwind is the wind energy of power supplied by the distributed generation system and
absorbed by the wind energy conversion system. the user demand power. The reliability of energy supply can
The electric energy Pgenerated which generated by solar, wind be calculated by:
energy conversion system and the fuel cell system can be
Pload  Pshortage
calculated by: reliability ¼ (31)
Pload
 
Pgenerated ¼ Pload  Pshortage þ PH2 surplus
þ Pbsoc  0:25Pbcap  0:85 The design constraints of the multi-objective optimization
(28) are presented in Table 2.

where Pload is the total electric energy that the users demand in
24 h; Pshortage represents the mismatch between the users de-
mand and the electric energy provided by the hybrid system
Methodology
and can be calculated by the hybrid system model presented
Multi-objective optimization method based on Hammersley
in section 2; PH2_surplus is the electric energy that the surplus
Sequence Sampling:
hydrogen can generate after 24 h's running and can be
The three objectives: the capital cost, the electricity effi-
calculated by the solid oxide fuel cell model shown in section
ciency and the reliability of energy supply often constraint
2.2. The formula (Pbcap0.25Pbcap)*0.85 represents the elec-
with each other. Thus, there is not a unique solution for the
tricity energy remaining in the storage battery after 24 h's
multi-optimization problem of the hybrid wind/solar/fuel cell
running of the hybrid system.
DG system. The capital cost, the electricity efficiency and the
The input solar energy of the solar energy conversion
reliability of energy supply were selected to be optimized
system Psolar can be calculated by:
simultaneously under constraints. The multi-objective opti-
Z24h mization method was applied in the design phase to optimize
Psolar ¼ fAs Ns dt (29) design parameters for the hybrid wind/solar/fuel cell DG
0 system under a full day condition to find the Pareto-optimal
solution. The Description for this multi-objective optimiza-
The input wind energy of the wind energy conversion
tion problem statement is shown as below:
system Pwind can be obtained by:
Minimize : capital cost ðCostÞ
Z24h
1 power supply reliability ðReliabilityÞ
Pwind ¼ rAW v3 dt (30) Maximize :
2 electricity efficiency ðhÞ
0 Subject to : xj;L xj  xj;U
where r is the local air density. where xj represents the number j constraint; xj,L is the left
boundary of xj; xj,U is the right boundary of xj.
Objective function of the energy supply reliability In this paper, five key design parameters of the hybrid
wind/solar/fuel cell system were selected as the design vari-
In this work, the reliability of energy supply for the hybrid ables for the multi-objective optimization which are sum-
wind/solar/fuel cell generating system is defined as the ratio marized in Table 2. The performance and the cost of the

Table 2 e Definitions and initial constraints of design parameters.


Design parameters Initial Minimum Maximum
2
Aw Swept area of WTG (m ) 50 50 500
Ns Number of cells in series of a PV module 200 50 200
Nsofc Number of cells in SOFC stack module 100 50 500
Asofc Active area of a single cell of SOFC (cm2) 1000 1000 5000
Pbcap Capacity of storage battery 50 50 500
7842 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 2 ( 2 0 1 7 ) 7 8 3 6 e7 8 4 6

Fig. 3 e Multi-objective optimization framework.

hybrid wind/solar/fuel cell system are influenced by the (2) Level 2: NLP optimizer.
values of these parameters greatly. The Pareto-solution can be
obtained by changing the values of these parameters until the The NLP optimizer is employed in this framework to solve
non-inferior solutions are found. The initial values and the each single-objective optimization problem generated in Level
value ranges of design parameters used in this study were 1. In this paper, successive quadratic programming (SQP)
determined by heuristics and previous modeling experience. method is chosen because of its advantage in solving the
Fig. 3 presents the framework of multi-objective optimiza- nonlinear optimization problems with nonlinear constraints
tion strategy based on advanced ε-constraints method, which [33,34]. Only gradient information is required to solve the
is named MINSOOP [30e32]. Because of the high calculation nonlinear constrained optimization problem for SQP. Hes-
efficiency and the great uniformity properties, Hammersley sians are approximated by low-rank updates defined by the
Sequence Sampling is very effective for the optimization re- step taken at each iteration. The new values of the design
searches. The framework consists of three levels: parameters are calculated by SQP method to search the better
values of objective functions until the optimal values are
(1) Level 1: ε-constraints multi-objective optimization found. The Pareto solution set is formed by the optimal design
method based on HSS method. parameters corresponding to the optimal values of objective
functions. The procedures for the whole multi-objective
In this paper, ε-constraints method was employed in the optimization are described as below. When a multi-objective
multi-objective optimization of the hybrid wind/solar/fuel cell optimization problem is given, ε-constraints method based
system. The primary strategy of ε-constraints is to set one on HSS simplified it to a group of single-objective problems.
objective fk as the main objective which needs to be optimized. Then, SQP method is applied to search the optimal values and
The remaining objectives were transformed into a series of the optimal design parameters of objective functions for each
inequality constraints within right boundaries, εi. The poten- single-objective problem. The Pareto optimal set is formed by
tial ranges of εi are the same as the range of each objective the optimal design parameters obtained by SQP. This process
obtained from solving the single-objective optimization is repeated until obtaining a good approximation of the Pareto
problems for each objective under the initial constraints optimal set of the multi-objective optimization problem.
respectively. Table 3 presents the potential ranges of εi for the
capital cost, the electricity efficiency and the reliability of (3) Level 3: System model.
energy supply obtained by optimizing them respectively
under initial constraints. Then, HSS method is applied in The new values of the design parameters calculated by SQP
generating a group of different values of right boundaries {ε1, method in level 2 are set as the inputs for the system
…,εk1,εkþ1, …,εh} (in which εi,L  εi  εi,U) from the potential modeling. The results obtained from system modeling are
ranges of εi. Thus, the multi-objective optimization problem is delivered back to the SQP method to generate inputs for next
simplified to a certain amount of single-objective optimiza- iteration.
tion problems. The Pareto solution set can be derived by In this study, the energy supply reliability is chosen as the
solving the single-objective optimization problems. main objective which needs to be maximized. The electricity

Table 3 e Potential ranges of εj for all objectives: a payoff table.


Objective Maximum Energy Maximum Energy supply Minimum Capital Minimum Maximum
Efficiency Reliability Cost (εi,L) (εi,U)
Capital Cost ($/year) 2981.12 7413.65 2484.58 2484.58 7413.65
Electricity Efficiency (%) 17.19 2.539 16.71 2.539 17.19
Power Supply Reliability 25.24 91.71 72.67 25.24 91.71
(%)
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 2 ( 2 0 1 7 ) 7 8 3 6 e7 8 4 6 7843

efficiency and the capital cost are converted to inequality


constraints with two boundaries, εreliability and εcost. The HSS
method is used to create a certain amount of the single-
objective optimization problems by changing the values of
εreliability and εcost between their upper and lower bounds
(εreliability: 2.539%e17.19%; εcost: 2484.58e7413.65 $/year in Table
3). In this way, the multi-objective optimization problem of
the hybrid wind/solar/fuel cell DG system is converted to a
certain amount of single-objective problems as follows:

Maximum Energy Supply Reliability


Subject to Electricity Efficiency (h)  εreliability
Capital Cost COST  εCOST Fig. 5 e Hourly solar irradiance.
2.539%  εreliability  17.19%
2484.58  εCOST  7413.65
solutions is formulated by the 300 optimal solutions obtained
by solving 300 single-objective problems generated by ε-con-
Wind and solar energy influenced by environment greatly, straints method based on HSS. The electricity efficiency and
Figs. 4e6 present the hourly wind speed data, the solar irra- the capital cost are plotted on the two axes in Fig. 8, the
diance data and the power demand data respectively (refer- contours represent the values of energy supply reliability at
enced from others' publication [24]) used as inputs of the the corresponding point. These 300 optimal designs can pro-
hybrid wind/solar/fuel cell system model simulation study. vide the appropriate reference information for decision
makers about the design of the hybrid system.
As seen in Fig. 8, the capital cost decreases rapidly with the
Result and discussion decreasing of the energy supply reliability. This is because the
low capital cost decides the low capacity of the wind and solar
In this work, we assumed that the sampling numbers of the energy conversion system. As a result, the electricity gener-
two constraints εreliability and εcost are the same. In order to ated by the hybrid system cannot meet the energy demand of
reduce the computational load and without reducing the ac- users. It is presented in Fig. 8, with the increase of the capital
curacy of the calculation, the appropriate sampling numbers cost and the energy supply reliability, the electricity efficiency
of the two constraints need to be determined. Fig. 7 presents decrease obviously. This is because the high capital cost and
the mean and the variance with different sampling number. the high energy supply reliability means there is more energy
As seen in Fig. 7, the mean and the variance of energy supply participating in the storage and release process in the hybrid
reliability converge with the increase of the sampling number. system and result in the decrease of electricity efficiency.
When the sampling number comes to 300, the mean and the In order to achieve more detailed information of the Pareto
variance of energy supply reliability converge to less than 2%. solution set, the objective values of the 300 optimal designs
Therefore, the sampling number of the two constraints is set were normalized between 0 and 1. The normalized data is
to 300. presented in Fig. 9 and Table 4.
As mentioned in Section 2, the capital cost, the electricity As presented in Fig. 9 and Table 4, three different groups of
efficiency and the energy supply reliability were chosen as the data are illustrated for designs with different purposes, which
three objectives for multi-objective optimization of the hybrid are the maximum electricity efficiency (Group I), the mini-
wind/solar/fuel cell DG system. Approximate Pareto set of mum capital cost (Group II) and the maximum energy supply

Fig. 4 e Hourly wind speed profile. Fig. 6 e Hourly power demand profile.
7844 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 2 ( 2 0 1 7 ) 7 8 3 6 e7 8 4 6

Fig. 7 e Mean and variance of energy supply reliability.

reliability (Group III). The high energy efficiency from 16.77% processes in the hybrid system can be obtained in this paper.
to 17.10% is shown in Group I, with the corresponding cost Therefore, the approximated Pareto surface obtained from the
from 2928 $/year to 3085 $/year and the energy supply reli- multi-objective optimization can show a more in-depth un-
ability from 74.92% to 75.45%. The low capital cost from 2556 $/ derstanding of the characteristics of the hybrid system. De-
year to 3441 $/year is presented in Group II, with the electricity cision makers can backtrack and find the values of the
efficiency ranging from 14.16% to 15.14 and the energy supply decision variables from the Pareto surface to fit their target
reliability from 76.16% to 79.27%. The high energy supply performance, they can easily understand what they will sac-
reliability from 84.82% to 87.74% is summarized in Group III, rifice when they want to improve one of the three objectives
with the electricity efficiency ranging from 8.115% to 11.16% (the cost, energy efficiency and the energy supply reliability).
and the capital cost from 2976 $/year to 3441 $/year. Furthermore, once the design variables have been deter-
When decision makers emphasize the high electricity ef- mined, the dynamic response of each unit in the hybrid sys-
ficiency of the hybrid wind/solar/fuel cell system and do not tem according to the weather and load conditions can be
have a request for the low capital cost and the high energy observed by running the models developed in this paper.
supply reliability, Group I is a better choice to fulfill the need. Then, the design makes can understand how to adjust the unit
When a lower capital cost is required while the energy supply sizing of the hybrid system to reduce the fluctuation caused by
reliability could be sacrificed, Groups II is more suitable for it. the renewable energy according to the Pareto surface.
If the higher energy supply reliability is the main concern The energy supply reliability represents the mismatch be-
comparing with the other two aspects, then Groups III is a tween the demand and the energy produced by the hybrid
proper candidate. system. Thus, for a standalone energy system, this multi-
Because the detailed models of each unit for the wind/
solar/fuel cell hybrid system was established in the paper, the
better understanding of the complex energy conversion

Fig. 8 e Pareto surface of for electricity, energy supply Fig. 9 e Normalized objectives of the 300 designs for wind/
reliability and capital cost. PV/fuel cell hybrid fuel cell system.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 2 ( 2 0 1 7 ) 7 8 3 6 e7 8 4 6 7845

references
Table 4 e Objective function values for different designs
purpose.
Objectives Group I Group II Group III
[1] http://www.bp.com/en/global/corporate/energy-economics/
Electricity Efficiency (%) 16.77e17.10 14.16e15.14 8.115e11.16 statistical-review-of-world.
Energy Supply Reliability (%) 74.92e75.45 76.16e79.27 84.82e87.74 [2] Li F-F, Qiu J. Multi-objective optimization for
Cost ($/year) 2928e3085 2556e3197 2976e3441 integrated hydroephotovoltaic power system. Appl Energy
4/1/2016;167:377e84.
[3] Kellogg W, Nehrir MH, Venkataramanan G, Gerez V. Optimal
objective optimization framework can also serve for the design unit sizing for a hybrid wind photovoltaic generating system.
of a Net Zero Energy Building. When considering integrating the Electr Power Syst Res Oct 1996;39:35e8.
[4] Chong WT, Naghavi MS, Poh SC, Mahlia TMI, Pan KC.
hybrid system into the electricity grids, the detailed models
Techno-economic analysis of a windesolar hybrid renewable
established in this paper can explain the dynamic response of energy system with rainwater collection feature for urban
each power unit (wind energy, solar energy and the fuel cell). high-rise application. Appl Energy 11//2011;88:4067e77.
Thus, the fluctuation of the electricity energy generated by each [5] Yang H, Wei Z, Chengzhi L. Optimal design and techno-
power unit and the whole hybrid system can be estimated economic analysis of a hybrid solarewind power generation
easily and can be taken into consideration of the multi- system. Appl Energy 2//2009;86:163e9.
objective optimization. According to the weather condition of [6] Kellogg WD, Nehrir MH, Venkataramanan G, Gerez V.
Generation unit sizing and cost analysis for stand-alone
a certain area, the multi-objective optimization framework can
wind, photovoltaic, and hybrid wind/PV systems. IEEE Trans
be used to balance supply and demand and secure frequency Energy Convers Mar 1998;13:70e5.
and voltage on the electricity grids when considering inte- [7] Kandil MS, Farghal SA, El-Alfy AE. Optimum operation of an
grating the hybrid system into the electricity grids. autonomous energy system suitable for new communities in
developing countries. Electr Power Syst Res 1991;21:137e46.
[8] Lund H. Large-scale integration of wind power into different
energy systems. Energy Oct 2005;30:2402e12.
Conclusion
[9] Notton G, Cristofari C, Poggi P, Muselli M. Wind hybrid
electrical supply system: behaviour simulation and sizing
In this paper, a multi-objective optimization framework based optimization. Wind Energy 2001;4:43e69.
on the cost, the electricity efficiency and the energy supply [10] Chedid R, Rahman S. Unit sizing and control of hybrid wind-
reliability models of the hybrid wind/solar/fuel cell DG system solar power systems. IEEE Trans Energy Convers Mar
is established. Detailed models of each unit in the hybrid 1997;12:79e85.
wind/solar/fuel cell DG system were built to reveal the com- [11] Lund H, Andersen AN, Østergaard PA, Mathiesen BV,
Connolly D. From electricity smart grids to smart energy
plex energy conversion processes in the hybrid system and
systems e a market operation based approach and
ε-constraints method based on Hammersley Sequence Sam-
understanding. Energy 6//2012;42:96e102.
pling is employed in the multi-objective optimization of the [12] Lund H, Marszal A, Heiselberg P. Zero energy buildings and
hybrid system. As a result, the Pareto surface of the multi- mismatch compensation factors. Energy Build 7//
objective optimization problem for the wind/PV/fuel cell 2011;43:1646e54.
distributed generation system was given in the end of this [13] Huang P-H, Kuo J-K, Wu Z-D. Applying small wind turbines
paper for searching the global optimum solution for decision and a photovoltaic system to facilitate electrolysis hydrogen
production. Int J Hydrogen Energy 6/1/2016;41:8514e24.
makers. In this way, the multi-objective optimization frame-
[14] Sarrias-Mena R, Ferna  ndez-Ramı́rez LM, Garcı́a-Va
 zquez CA,
work of the hybrid DG system was established in this study. Jurado F. Electrolyzer models for hydrogen production from
The complex energy conversion processes in this hybrid DG wind energy systems. Int J Hydrogen Energy 2/23/
system is involved in this optimization and give the decision 2015;40:2927e38.
makers an insight into the feature of the hybrid DG system. [15] Lee TY, Chen CL. Wind-photovoltaic capacity coordination
The mismatch between the demand and the energy produced for a time-of-use rate industrial user. IET Renew Power Gener
Jun 2009;3:152e67.
by the hybrid system can be easily observed. Therefore, the
[16] Fazelpour F, Soltani N, Rosen MA. Economic analysis of
multi-objective optimization framework can also be used to
standalone hybrid energy systems for application in Tehran,
deal with the problem of coordinating fluctuating and inter- Iran. Int J Hydrogen Energy 5/25/2016;41:7732e43.
mittent energy generated by the hybrid DG system and reduce [17] Wang L, Singh C, IEEE. PSO-based multi-criteria optimum
the shock produced by the renewable energy when consid- design of A grid-connected hybrid power system with
ering integrating the hybrid system into the electricity grids. multiple renewable sources of energy. 2007.
With this method, this paper finished the multi-objective [18] Chedid R, Akiki H, Rahman S. A decision support technique
for the design of hybrid solar-wind power systems. IEEE
optimization of the hybrid wind/solar/fuel cell DG system
Trans Energy Convers Mar 1998;13:76e83.
and provided an efficient method for decision makers in the
[19] Ganesan T, Vasant P, Elamvazuthi I. Hybrid neuro-swarm
design of the hybrid wind/solar/fuel cell system. optimization approach for design of distributed generation
power systems. Neural Comput Appl Jul 2013;23:105e17.
[20] Khare V, Nema S, Baredar P. Optimization of hydrogen
based hybrid renewable energy system using HOMER,
Acknowledgements BB-BC and GAMBIT. Int J Hydrogen Energy 10/15/2016;41:
16743e51.
The authors gratefully acknowledge the support of the Natu- [21] Wang J, Yang F. Optimal capacity allocation of standalone
ral Science Foundation of China (No. 51576020). wind/solar/battery hybrid power system based on improved
7846 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 2 ( 2 0 1 7 ) 7 8 3 6 e7 8 4 6

particle swarm optimisation algorithm. IET Renew Power [28] Ming D, Liu-jin Y, Mei-qin M. Modeling and control of fuel
Gener Sep 2013;7:443e8. cells in distributed generation system. Power Syst Technol
[22] Kalagnanam JR, Diwekar UM. An efficient sampling 2009;33. in Chinese.
technique for off-line quality control. Technometrics Aug [29] Tan L, Yang C, Zhou N. Synthesis/design optimization of
1997;39:308e19. SOFCePEM hybrid system under uncertainty. Chin J Chem
[23] Singhee A, Rutenbar RA. Why quasi-Monte Carlo is better Eng 2015;23:128e37.
than Monte Carlo or Latin hypercube sampling for statistical [30] Subramanyan K, Diwekar UM, Goyal A. Multi-objective
circuit analysis. IEEE Trans Computer-Aided Des Integr optimization for hybrid fuel cells power system under
Circuits Syst Nov 2010;29:1763e76. uncertainty. J Power Sources 5/20/2004;132:99e112.
[24] Wang CS. Modeling and control of hybrid wind/photovoltaic/ [31] Subramanyan K, Diwekar UM. Optimizing model complexity
fuel cell distributed generation systems. Montana: Montana with application to fuel cell based power systems. Chem Eng
State University; 2006. Process 2007;40:1116e28.
[25] Parsons R. Kortum, G - Treatise on electrochemistry. J [32] Fu Y, Diwekar UM, Young D, Cabezas H. Process design for
Electroanal Chem 1967;13. 466e&. the environment: a multi-objective framework under
[26] Chen Yang, Yu-ying Zhang, Ze-ya Xu. Numerical simulation uncertainty. Clean Technol Environ Policy 2000;2:92e107.
of heat and mass transfer in SOFC. J Syst Simul [33] Murray W. Sequential quadratic programming methods for
2007;19:3206e9. in Chinese. large-scale problems. Comput Optim Appl 1997;7:127e42.
[27] Fuel cell handbook. 6th ed. EG&G Services, Inc., Science [34] Gill PE, Murray W, Saunders MA. SNOPT: an SQP algorithm
Applications International Corporation, DOE, Office of Fossil for large-scale constrained optimization. Soc Ind Appl Math
Energy, National Energy Technology Lab; 2002. 2005;47:99e131.

Вам также может понравиться