Professional Documents
Culture Documents
PII: S2210-6707(18)31517-8
DOI: https://doi.org/10.1016/j.scs.2018.10.021
Reference: SCS 1297
To appear in:
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
Techno-economic analysis and size optimization of an off-grid
T
Department of Energy Management and Optimization, Institute of Science and
IP
High Technology and Environmental Sciences, Graduate University of Advanced
R
Technology, Kerman, Iran
SC
U
Multi-objective optimization of PV/diesel generator/fuel cell
N
system is studied.
A
optimization.
ED
problem.
PT
an efficient system.
E
CC
Abstract
A
This paper presents multi-objective design of a hybrid system composed of photovoltaic (PV), fuel cell
(FC) and diesel generator (DG) to supply electric power of an off-grid community in Kerman, south of
Iran in the presence of operating reserve (OR) and uncertainties of load and solar power. Total net
Corresponding Author. Tel./fax: +98 342 6233176.
E-mail address: a.askarzadeh@kgut.ac.ir, askarzadeh_a@yahoo.com.
1
present cost (TNPC) and loss of power supply probability (LPSP) are considered as the objectives and
the decision variables are the number of PV panels, DGs, FCs, electrolysers and hydrogen tanks. A
recently developed meta-heuristic optimizer, multi-objective crow search algorithm (MOCSA), has
been utilized to find Pareto front. The impact of different parameters (fuel price, FC system equipments
cost, emission cost, etc) has been investigated on the sizing problem. Based on the simulation results, it
is observed that integration of hydrogen energy technology will reduce the total cost of the hybrid
energy systems. Moreover, the impact of OR on the Pareto front is more than that of load and solar
T
IP
power uncertainties.
Keywords: Hybrid energy system, fuel cell, multi-objective optimization, operating reserve, sizing
R
SC
U
N
A
M
ED
1. Introduction
PT
One of the major problems in the world is air pollution and global warming. This is due to the excessive
E
use of fossil fuels. Nowadays, to conquer this problem, renewable energy resources (such as solar,
CC
wind, tidal, water, biomass, etc) are utilized. Using renewable energy resources significantly reduces
CO2 emission.
A
In remote or rural areas where the connection to the grid is impossible, or in areas where power loss
probability is high or power supply is not appropriate, the use of isolated power generation systems
such as an off-grid solar system, a small wind turbine, or a combination of these resources could be
suggested to meet the electricity demand. Due to the uncertainties in renewable resources, it is not
2
possible to use these resources at some hours. Hence, a backup system should be considered to satisfy
the deficit power. In many off-grid applications, diesel generators (DGs) are used as backup systems.
To have a clean, reliable and cost-effective energy system, optimal combination of renewable energy
resources and DGs should be considered to simultaneously reduce CO2 emission, increase reliability
and minimize energy generation cost. Simultaneous use of PV and DG with an energy storage could be
an ideal combination to electrify remote areas. In hybrid power generation systems (HPGSs), energy
storage plays an important role when generated power is more than the load demand. Most often, two
T
IP
types of storages are used in HPGSs: battery andfuel cell (FC). Using battery banks as storage system
has some limitations. Due to short lifetime, the replacement of battery is needed during the project
R
lifetime. In addition, energy losses occur during battery loading and the efficiency of battery decreases
SC
because of ageing or bad operating conditions. In recent years, proton exchange membrane fuel cell
U
(PEMFC) has received a considerable attention as a storage system in HPGSs. In comparison with the
N
other FCs, PEMFC has a number of merits such as low operational temperature, fast start-up and zero
A
pollution if it is run with pure hydrogen. This paper makes use of PEMFC as the storage system. In the
M
investigated hybrid PV/DG/FC system, the excess energy of PV panels is injected to electrolyser to
produce hydrogen. The produced hydrogen is then stored in hydrogen tank to use in deficit conditions.
ED
One of the most challenging problems of HPGSs is optimal sizing. Optimal sizing answers to this
question which combination of the system components is the best one. Optimal sizing can be conducted
PT
by different objectives such as cost, emission or reliability. Study of the literature shows that optimal
E
sizing of HPGSs has been solved with respect to single-objective and multi-objective optimization.
CC
Single-objective optimization leads to a single solution as the optimum configuration of the system
while multi-objective optimization results in a Pareto front which includes a set of equally good
A
solutions. In multi-objective optimization, the designer selects the solution by his/ her preferences.
They are many studies which focus on size optimization of hybrid energy systems. In [1], size
optimization of a hybrid system including PV, wind, tidal and battery has been performed by CSA. In
this research, net present cost (NPC) and equivalent loss factor (ELF) have been selected as objective
3
function and reliability index, respectively. In [2], a hybrid big bang–big crunch (HBB–BC)algorithm
has been applied to optimize a stand-alone hybrid PV/wind/battery system by considering the total
present cost (TPC) as objective function. In [3], optimum design of a grid-connected solar system with
battery storage has been investigated to supply power of residential buildings. The decision variables
are the number of PV panels, batteries and inverters. The objective function is defined to minimize the
cost of energy (COE). In [4], the optimization results of a PV/ FC/electrolyser/inverter system indicate
that the HPGS will be able to meet the daily electrical demand of 150 homes in residential applications.
T
IP
In [5], the size of a PV/DG /battery system has been optimized by grey wolf optimizer (GWO) to find
the minimum value of the total annual cost. In [6], energy management of a hybrid
R
PV/wind/diesel/battery system has been investigated by fuzzy logic control (FLC). It has been shown
SC
that the simple and easy fuzzy logic controller can determine various operating processes of the hybrid
U
system. In [7], cuckoo search (CS) algorithm has been applied to design a hybrid PV/wind/battery
N
generation system with the aim of minimizing total system cost with respect to expected energy not
A
supplied (EENS) and loss of load expected (LOLE) as the reliability indices. In [8], multi-objective
M
optimization of a stand-alone PV/DG hybrid system has been solved by a multi-objective crow search
algorithm (MO-CSA). Objective functions are TNPC, LPSP and emission. In [9], a feasibility analysis
ED
for a PV/DG/battery/biomass hybrid system has been studied in order to supply the electricity of a
village in Morocco. Optimization of the objective functions (TNPC, COE and wood consumption) has
PT
been done by HOMER software. In [10], multi-objective optimization of a PV/DG hybrid system in the
E
isolated sites have been evaluated with the goal of minimizing LCOE and CO2emission. In [11],
CC
optimal sizing of a hybrid system including PV, wind and battery has been solved by considering
uncertainty of wind speed and solar irradiation using a non-dominated sorting genetic algorithm II
A
(NSGA-II). The objective functions are the total cost and LPSP. In [12], multi-objective optimization
of a hybrid PV/wind/FC system has been solved to minimize cost and maximize reliability by multi-
objective particle swarm optimization (MO-PSO). EENS, ELF and LPSP have been considered as the
reliability indices. In [13], optimal sizing of a PV/wind/diesel HPGS with battery storage has been done
4
using multi-objective self-adaptive differential evolution (MOSADE) algorithm. In [14], a new
approach has been applied to optimize a grid-connected PV/FC/battery hybrid energy system. Cost and
CO2emission have been considered as the objective functions. In [15], optimal sizing of a hybrid energy
storage system (HESS) including battery and ultracapacitor has been studied with respect to battery
health. In [16], the optimal size of a PV/biomass hybrid energy system has been determined with
respect to reliability and cost. It has been shown that the hybrid system performance is more promising
than the performance of a single PV system or a single biomass system. In [17], the authors have
T
IP
modelled and controlled a grid-connected PV/DG/FC system. The impact of FC has been studied on
frequency stability of the grid. It has been concluded that FC system can contribute to frequency
R
stability when load increases or decreases. In [18], optimum size of a PV/battery/diesel system has been
SC
carried out for a stand-alone ship power system. Investment cost, fuel cost and CO2 emission are the
U
problem objectives. A methodology has been proposed which takes into account the seasonal and
N
geographical variation of solar radiations and temperatures for PV output correction.
A
Literature survey reveals that the investigation on multi-objective optimization of PV/DG/FC system is
M
rarely found. The present study focuses on multi-objective optimization of a hybrid PV/DG/FC system
(CSA). Authors in [8] have solved multi-objective optimization of a PV/DG hybrid energy system by
MO-CSA and multi-objective particle swarm optimization (MO-PSO). PSO is a well-known and
PT
popular metaheuristic optimization technique which has attracted significant attention for solving
E
complex engineering optimization problems. Simulated results reveal that MO-CSA finds a well-
CC
distributed and widely spread Pareto front and produces more promising results than MO-PSO [8].
Furthermore, in [1], the results of sizing a PV/wind/tidal/battery hybrid system indicate that the solution
A
obtained by CSA is more accurate than the solutions obtained by PSO and genetic algorithm (GA).
Owing to the efficiency of CSA, this algorithm has been selected to solve the size optimization
problem. The advantages of CSA are easy implementation, fast convergence rate and global search
5
ability. This optimizer is utilized to effectively find Pareto front (Cost vs LPSP) of the multi-objective
sizing problem. The main contributions of the present study can be summarized as follows:
paper, a comprehensive modeling is provided, two important objectives are included and the impact of
Optimum number of all the system components (PV, diesel generator, electrolyzer, hydrogen
tank and fuel cell) are determined by the optimization technique to have a cost-effective and reliable
T
IP
energy system.
In real conditions, operating reserve (OR) and uncertainties affect the system performance. In
R
SC
this paper, the multi-objective sizing problem has been solved by consideration of OR and solar and
load uncertainties. OR helps the system to electrify the load when there is sudden change in load
The rest of this paper is organized as follows: Section 2 models the system components and formulates
the optimization problem. In section 3, the proposed algorithm is explained. Results and discussions are
ED
2. Problem definition
E
The hybrid system considered to supply the load demand includes PV, DG, inverter, FC, electrolyser
CC
and hydrogen tank. In this section, equations related to each component are introduced, objective
functions of the problem are defined, system uncertainties and OR are modelled, and energy
A
2.1. PV equations
Photovoltaic is the conversion of light into electricity by use of semiconducting materials. PV panels
produce direct current electricity from sunlight which can be used to electrify different equipments. The
6
power generation of a PV system is directly proportional to the solar radiation. At hour t, the output
𝑅𝑎 2 (𝑡)
𝑁𝑝𝑣 × 𝑃𝑟𝑠 × ( ) 0 ≤ 𝑅𝑎 (𝑡) < 𝑅𝑐𝑟
𝑅𝑠𝑟𝑠 × 𝑅𝑐𝑟
𝑃𝑝𝑣 (𝑡) = 𝑅𝑎 (𝑡) (1)
𝑁𝑝𝑣 × 𝑃𝑟𝑠 × ( ) 𝑅𝑐𝑟 ≤ 𝑅𝑎 (𝑡) < 𝑅𝑠𝑟𝑠
𝑅𝑠𝑟𝑠
{𝑁𝑝𝑣 × 𝑃𝑟𝑠 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
where Ppv(t) is the output power of the PV system at hour t, Npv is the number of PV panels, Prs is the
T
PV panel rated power (kW), Ra(t) is the solar radiation at hour t, Rsrs is the standard solar radiation
IP
(1000 W/m2) and Rcr is the certain radiation (150 W/m2).
R
Since the lifespan of the PV panels is equal to the lifespan of the project, the replacement cost (RC) for
SC
the PV system has not been considered. Net present cost (NPC) of the PV system is calculated by PV
investment cost (ICpv) and operation and maintenance cost (OMCpv) as follows:
𝑵 (4)
𝟏+𝜷 𝒛
𝑶𝑴𝑪𝒑𝒗,𝒏𝒑𝒗 = 𝑶𝑴𝑪𝒑𝒗 × 𝑵𝒑𝒗 × ∑( )
𝟏+𝒓
ED
𝒛=𝟏
where αpv denotes the PV price, N is the project lifetime and subscript npv denotes the net present value.
PT
OMCpv is the yearly operation and maintenance cost of a PV panel. In order to obtain accurate
calculations, escalation (β), inflation (E) and interest (r) rates have been included in the economic
E
calculations.
CC
The inverter capacity is selected so that it can pass the power produced by PV and FC systems. For
A
inverter, NPC (NPCinv) is obtained using its investment cost (ICinv) and replacement cost (RCinv,npv) as
follows:
where
7
𝐼𝐶𝑖𝑛𝑣 = 𝛼𝑖𝑛𝑣 × 𝑃𝑖𝑛𝑣 (6)
1 + 𝛽 𝑛_𝑖𝑛𝑣
𝑅𝐶𝑖𝑛𝑣,𝑛𝑝𝑣 = 𝛼𝑖𝑛𝑣 × 𝑃𝑖𝑛𝑣 × ( )
1+𝑟 (7)
where αinv and n_inv denotes the price and lifespan of the inverter (here 10 years), respectively. Pinv is
2.3. DG equations
Diesel generator is an alternative and reliable source which is used to power different applications such
T
as homes and businesses. It produces electricity by using an alternator and a diesel engine. The engine
IP
utilizes diesel fuel to operate.
R
Fuel consumption of a DG (fc) is modeled by Eq. (8) as a function of the output power (Pdg) and
SC
nominal power (Pdgn) [20]. The fuel cost of the DG (Cf) is the multiplication of the fuel price (Pf) and
fuel consumption. The yearly fuel cost of the DG (FCdg) is obtained by Eq. (10).
𝟖𝟕𝟔𝟎 (10)
M
𝑭𝑪𝒅𝒈 = ∑ 𝑪𝒇 (𝒕)
𝒕=𝟏
ED
where Adg and Bdg are fuel intercept and fuel slope coefficients. These coefficients have been set to
NPC of the DG system (NPCdg) is calculated by use of the investment cost (ICdg), operation and
maintenance cost (OMCdg,npv), replacement cost (RCdg,npv) and fuel cost as follows:
E
(11)
where
A
𝑁
1+𝛽 𝑧
𝑂𝑀𝐶𝑑𝑔,𝑛𝑝𝑣 = 𝑂𝑀𝐶𝑑𝑔 × 𝑁𝑟𝑢𝑛 × ∑( )
1+𝑟 (13)
𝑧=1
8
𝑵
𝟏+𝑬 𝒛
𝑭𝑪𝒅𝒈,𝒏𝒑𝒗 = 𝑭𝑪𝒅𝒈 × ∑( ) (14)
𝟏+𝒓
𝒛=𝟏
where αdg is the DG price, Ndg is the number of DGs and Nrun denotes the number of hours which DG is
running. RC of the DG is determined based on running hours (here 15000 h). OMCdg and FCdg are the
2.4. FC equations
Fuel cell is an electrochemical device which converts the chemical energy of hydrogen and oxygen into
T
IP
electricity. Among different FC types, proton exchange membrane fuel cell (PEMFC) has attracted
significant attention. PEMFC has high current density and operate in low operating temperature.
R
NPC of the FC (NPCfc) is obtained by IC, OMC and RC of the FC as follows:
SC
𝑁𝑃𝐶𝑓𝑐 = 𝐼𝐶𝑓𝑐 + 𝑂𝑀𝐶𝑓𝑐,𝑛𝑝𝑣 + 𝑅𝐶𝑓𝑐,𝑛𝑝𝑣 (15)
where
U
N
𝐼𝐶𝑓𝑐 = 𝛼𝑓𝑐 × 𝑁𝑓𝑐 (16)
A
𝑁
1+𝛽 𝑧
𝑂𝑀𝐶𝑓𝑐,𝑛𝑝𝑣 = 𝑂𝑀𝐶𝑓𝑐 × 𝑁𝑓𝑐 × ∑( )
M
1+𝑟 (17)
𝑧=1
𝟏+𝜷 𝒛 (18)
𝑹𝑪𝒇𝒄,𝒏𝒑𝒗 = 𝑹𝑪𝒇𝒄 × 𝑵𝒇𝒄 × ∑ ( )
ED
𝟏+𝒓
𝒛=𝟓,𝟏𝟎,𝟏𝟓
where αfc is the FC price and Nfc is the number of FCs. OMCfc and RCfc are the operation and
PT
Electrolysis is the process of using electricity to split water into hydrogen and oxygen. This reaction
takes place in a unit called an electrolyzer. NPC of the electrolyser (NPCel) is obtained by IC, OMC and
A
where
9
𝑁
1+𝛽 𝑧
𝑂𝑀𝐶𝑒𝑙,𝑛𝑝𝑣 = 𝑂𝑀𝐶𝑒𝑙 × 𝑁𝑒𝑙 × ∑( )
1+𝑟 (21)
𝑧=1
𝟏 + 𝜷 𝒏_𝒆𝒍 (22)
𝑹𝑪𝒆𝒍,𝒏𝒑𝒗 = 𝑹𝑪𝒆𝒍 × 𝑵𝒆𝒍 × ( )
𝟏+𝒓
where αel is the electrolyser price, Nel is the number of Electrolysers. n_el denotes the lifespan of
theelectrolyser (here 10 years). OMCel and RCel are the operation and maintenance cost and
T
2.6. Hydrogen tank equations
IP
NPC of the hydrogen tank (NPCht) is calculated by using IC and OMC of the hydrogen tank (ICht and
R
OMCht,npv) as follows:
SC
𝑁𝑃𝐶ℎ𝑡 = 𝐼𝐶ℎ𝑡 + 𝑂𝑀𝐶ℎ𝑡,𝑛𝑝𝑣 (23)
where
where αht is the hydrogen tank price, Nht is the number of hydrogen tank and OMCht is the operation
ED
Among the hybrid system components, the DG produces a significant amount of CO2 gas. Hence, a
function for CO2 emission is considered and included in the cost function. This function can be defined
E
as follows [22]:
CC
where
𝐶𝐶 × 𝐸𝐷𝐺 (27)
𝐶𝑂2𝑤 =
1016.04
𝑃𝑇𝑅𝐶
𝐶𝑂2𝑇 = × 1016.04
𝐶𝐶 (28)
10
where PC is the emission cost, EDG is the generation energy by DG, CC is the carbon content (0.6078
kg/kWh), CO2w is the carbon weight in ton, CO2T is the carbon tax ($/ton) and PTRC is the price of the
2.8. Uncertainty
Although load demand and solar radiation are forecasted, there are deviations from the forecasted
values (due to uncertainty). To simulate these deviations close to actual values, the method introduced
T
IP
in [24] has been utilized. Based on this method, to model the uncertainty of the load, the following
R
𝑃𝐿_𝑢𝑛 = 𝑑𝑃𝐿 × 𝑛1 + 𝑃𝐿 (29)
SC
𝑑𝑃𝐿 = 0.6 × √𝑃𝐿 (30)
U
where dPL is the deviation of the load, PL_un is the load demandconsidering the uncertainty and n1 is
N
defined by a standard normal distribution function.
A
2.8.2. Solar power uncertainty
M
To consider the uncertainty of the PV power, the following expressions are taken into account [24]:
where dPpv is the deviation of the PV power, Ppv_un is the PV power considering the uncertainty and n2
In the hybrid system, OR is defined by a safety margin that helps to electrify the demand in hours
where there is sudden change in load demand or solar radiation. Indeed, OR is taken into account to
A
increase the system reliability. In this paper, to consider OR, it is assumed that at each hour the load
demand increases 10% and solar radiation decreases 25%. For instance, if the load is 80 kW and the PV
power is 20 kW, this means that the OR should be 8 kW+5 kW=13 kW. As a results, the hybrid system
The objective functions of the sizing problem are the minimization of the TNPC and LPSP. In this
optimization problem, the number of PVs, DGs, FCs, electrolysers and hydrogen tanks (Npv, Ndg, Nfc,
Nel and Nht) are the decision variables. The objective functions are formulated by Eqs. (33) and (34),
respectively.
𝑀𝑖𝑛. 𝑇𝑁𝑃𝐶(𝑁𝑝𝑣 , 𝑁𝑑𝑔 , 𝑁𝑓𝑐 , 𝑁𝑒𝑙 , 𝑁ℎ𝑡 ) = 𝑁𝑃𝐶𝑝𝑣 + 𝑁𝑃𝐶𝑑𝑔 + 𝑁𝑃𝐶𝑖𝑛𝑣 + 𝑁𝑃𝐶𝑓𝑐 + 𝑁𝑃𝐶𝑒𝑙 + 𝑁𝑃𝐶ℎ𝑡 + 𝑃𝐶
T
IP
(33)
∑8760
𝑡=1 𝐿𝑃𝑆(𝑡)
R
𝑀𝑖𝑛. 𝐿𝑃𝑆𝑃(𝑁𝑝𝑣 , 𝑁𝑑𝑔 , 𝑁𝑓𝑐 , 𝑁𝑒𝑙 , 𝑁ℎ𝑡 ) =
∑8760
𝑡=1 𝑃𝐿 (𝑡) (34)
SC
where LPS is the loss of power supply.
2.10.2. Constraints
U
The constraints related to the system components are defined as follows:
N
0 ≤ 𝑁𝑝𝑣 ≤ 𝑁𝑝𝑣,𝑚𝑎𝑥 (35)
A
0 ≤ 𝑁𝑑𝑔 ≤ 𝑁𝑑𝑔,𝑚𝑎𝑥
M
(36)
At each hour, the capacity of the hybrid system is the summation of PV power obtained by Eq. (1),
CC
rated power of DG system and rated power of FC system with respect to the amount of stored
hydrogen. To supply the demand, the output power of PV system is used and if there is deficit power,
A
FC system is run to satisfy the remaining load. If the load demand is not completely met, DG system is
turned on. After this step, if the load is not entirely met, there will be unmet load. According to the
amount of the produced power and load demand, one of the following states (A or B) may occur. The
12
A. If 𝜼𝒊𝒏𝒗 × 𝑷𝒑𝒗 (𝒕) ≥𝑷𝑳 (𝒕), then electrical load is entirely supplied and there will be excess power.
The excess power produced by PV system (PEx) is injected to the electrolyser for hydrogen production.
where ηinv is the inverter efficiency and PL denotes the load demand.
In this case, if PEx is more than the rated power of the electrolyser (Peln), then 𝑷𝑬𝒙 (𝒕) =𝑷𝒆𝒍𝒏 and the
T
excess power is wasted in a dump load. Since the power injected to the electrolyser cannot exceed the
IP
electrolyser rated power, a dump load is considered in the hybrid system.
R
The hydrogen produced by the electrolyser is stored in hydrogen tanks. The stored H2 (kg) can be
SC
obtained as follows [25]:
U
where Wht(t) and Wht(t - 1) are the amount of hydrogen stored in the tanks at hours t and t - 1,
N
respectively. It must be mentioned that the amount of the stored hydrogen cannot exceed the capacity of
A
B. If 𝜼𝒊𝒏𝒗 × 𝑷𝒑𝒗 (𝒕)<𝑷𝑳 (𝒕), then two states (B.1. and B.2.) may happen.
ED
At first, the deficit power (Pd) is supplied by use of FC system. In this case, the deficit power is
The required H2 (RH2) for meeting all the deficit power is calculated as follows [25]:
E
𝑃𝑑 (𝑡)
CC
𝜂𝑖𝑛𝑣
R𝐻2 = 𝜂𝑓𝑐 ×37.8
(43)
A
B.1. If 𝑹𝑯𝟐 ≤ 𝑾𝒉𝒕 (t), the required hydrogen is less than the stored hydrogen. As a result, the power
supplied by FC is obtained by Eq. (44) and the remained H2 in hydrogen tanks is calculated by Eq. (45).
𝑃𝑑 (𝑡) (44)
𝑃𝑓𝑐 (𝑡) = 𝜂𝑖𝑛𝑣
13
𝑊ℎ𝑡 (t) = 𝑊ℎ𝑡 (𝑡 − 1) −R𝐻2 (45)
B.1.1. Since FC system cannot produce electricity more than its rated power, if 𝑷𝒇𝒄 (𝒕) > 𝑷𝒇𝒄𝒏 , then
𝑷𝒇𝒄 (𝒕) = 𝑷𝒇𝒄𝒏 . The required H2 is obtained by Eq. (46) and the amount of hydrogen in the tanks is
𝑃𝑓𝑐𝑛 (46)
R𝐻2 = 𝜂
𝑓𝑐 × 37.8
To entirely supply the load, DG system should be run. The required power met by the DG system is
T
calculated by Eq. (47).
IP
𝑃𝑑𝑔 (𝑡) = 𝑃𝑑 (𝑡) − 𝜂𝑖𝑛𝑣 × 𝑃𝑓𝑐 (𝑡) (47)
R
Since DG system cannot generate electricity more than its rated power, if 𝑷𝒅𝒈 (𝒕) > 𝑷𝒅𝒈𝒏 , then 𝑷𝒅𝒈 (𝒕)
SC
= 𝑷𝒅𝒈𝒏 and the loss of power supply is calculated as follows:
As before, if 𝑷𝒇𝒄 (𝒕) > 𝑷𝒇𝒄𝒏 , then 𝑷𝒇𝒄 (𝒕) = 𝑷𝒇𝒄𝒏 . To calculate the required H2, Eq. (46) is used and
To entirely meet the load, DG system should be run. The required power met by the DG system is
E
calculated by Eq. (47). As before, if 𝑷𝒅𝒈 (𝒕) > 𝑷𝒅𝒈𝒏 , then 𝑷𝒅𝒈 (𝒕) = 𝑷𝒅𝒈𝒏 and the value of LPS is
CC
calculated by Eq. (48). A brief operation of the energy management process has been shown in Fig. 1.
A
In [26], in order to solve engineering optimization problems a meta-heuristic algorithm was developed
that is based on the intelligent behaviour of crows. In recent years, the efficiency of CSA has increased
14
the attention to utilize this optimizer for solving complex optimization problems. In this study, the
following steps are considered for implementation of MO-CSA for solving the sizing problem.
The adjustable parameters of MO-CSA (flock size (Nc)), flight length (fl), awareness probability (AP)
An initial population of crows in the search space is randomly valued. Position of each crow is
T
IP
indicated by a vector, x = [Npv, Ndg, Nfc, Nel, Nht] where Npv, Ndg, Nfc, Nel and Nht are the number of PVs,
DGs, FCs, electrolysers and hydrogen tanks, respectively. At the first iteration, the initial position of the
R
crows is assumed to be their initial memory.
SC
U
N
A
M
ED
E PT
CC
A
15
T
R IP
SC
U
N
A
M
ED
E PT
The values of objective functions (cost and LPSP) are obtained for each crow.
A
At the first iteration, non-dominated solutions among memory solutions are saved in an EA.
16
At iteration k, crow i ( i =1,2,…,Nc) randomly chooses one of the other crows (crow j) and follows it to
discover its hiding place. Hence, crow i goes to a new position as follows:
where fli,k, mj,k and APj,k are flight length of crow i, the best experience of crow j and awareness
probability of corw j at iteration k, respectively. r is a random number with uniform distribution from [0
1].
T
IP
Step 6: Checking feasibility of new positions
At this step, if the new position of a crow is feasible, the crow updates its position.
R
Step 7: Computation of the objective functions for new positions
SC
The values of cost and LPSP are computed for the new position of each crow.
Step 9: Updating EA
Non-dominated solutions among the solutions of current EA and memory solutions are found and
ED
stored in EA.
Stopping criterion of MO-CSA is to reach to itermax. When the stopping criterion is met, solutions
E
stored in EA are reported as optimal solutions. Otherwise, steps 5–9 are repeated.
CC
Fig. 2 shows the schematic of the PV/DG/FC system. This hybrid system is optimally sized to electrify
an off-grid application in the south of Iran, Kerman. The load profile and solar radiation, during the
year, (8760 h), have been represented in Fig. 3. The parameters used in this study are listed in Tables 1
and 2. The optimization problem has been coded in MATLAB software. For optimal sizing of the
17
hybrid system, MO-CSA is run 10 times and the non-dominated solutions are demonstrated as Pareto
front. The parameter setting of MO-CSA is Nc = 20, itermax = 2000, fl =1 and AP = 0.2. These
parameters have been tuned by trial and error and no attempt has made to optimize them.
DC bus
Load
AC bus
PV
system
× 432
300 - kiden.ir × 432
Inverter 300 - kiden.ir
T
FC system × 432
× 432
IP
300 - kiden.ir
300 - kiden.ir
Diesel
generator
R
SC
Hydrogen Electrolyser
tanks × 432
300 - kiden.ir
U
N
A
M
ED
E PT
CC
A
Fig. 3. (a) Load demand (kW) and (b) solar radiation (W/m2 )
18
Table 1
Parameters used in this study
Economical parameters
Project lifetime 20 years
β 4%
E 9%
r 10%
PV parameters
PV lifespan 20 years
Prs 0.260 kW
Rsrs 1000 W/m2
Rcr 150 W/m2
αpv 468 $/panel
OMCpv 0.02×αpv
T
Inverter parameters
Inverter lifespan 10 years
IP
ηinv 90 %
αinv 100 $/kW
DG parameters
R
DG lifespan 15,000 h
Rated power 1.8 kW
SC
αdg 550 $
OMCdg 0.08 $/kWh
RCdg 550 $
Pf 1.18 $/l
Table 2 U
N
Parameters of the storage system used in this study
FC parameters
A
FC lifespan 5 years
M
ηfc 50%
Rated power 2 kW
ED
OMCfc 0.02×αfc
Electrolyser parameters
Electrolyser lifespan 15 years
E
Rated power 2 kW
CC
19
4.1. Multi-objective optimization of PV/DG and PV/DG/FC systems
In the first scenario, the problem has been solved with and without considering energy storage system.
In this case, the impact of OR, uncertainties and CO2 emission are not included in the sizing problem.
Fig. 4 depicts Pareto front of the sizing problem with and without considering FC. This figure shows
the TNPC of the system at various LPSP levels. In multi-objective optimization, the designer should
select one of the solutions according to his/ her preferences. It is obvious that by decrease of the LPSP
value, the TNPC of the system increases. This is due to the conflict between reliability and cost.
T
IP
Comparison of the Pareto fronts reveals that by using storage system, the cost of the hybrid system
decreases. Indeed, at a same LPSP value, the cost of PV/DG/FC system is less than the cost of PV/DG
R
system. Figs. 5 and 6 indicate the Pareto set obtained by MO-CSA for PV/DG and PV/DG/FC systems,
SC
respectively. In the Pareto set of PV/DG/FC system, only the number of PV panels, DGs and FCs has
U
been shown. Tables 3 and 4 represent the details of non-dominated solutions at LPSP 0% and 10%. It is
N
observed that at LPSP=0, TNPC of the optimized PV/DG system is 1.241 × 106 $. This value has been
A
obtained with respect to 239 PV panels and 31 diesel generators. At LPSP = 0, the rated power of PV
M
and diesel generator systems is 62.14 kW and 55.8 kW, respectively. For the optimized PV/DG/FC
system, at LPSP=0, the TNPC is obtained 1.201 × 106 $ and the optimal number of PV panels, diesel
ED
generators, fuel cells, electrolysers and hydrogen tanks is 443, 30, 7, 26 and 4, respectively. It can be
drawn that if the hybrid system is optimally sized, using FC system leads to decreasing total cost.
E PT
CC
A
20
T
R IP
Fig. 5. Pareto set of the sizing problem without considering storage system
SC
U
N
A
M
ED
E PT
Fig. 6. Pareto set of the sizing problem with considering storage system
CC
Table 3
A
Details of two non-dominated solutions of PV/DG system without considering operating reserve, uncertainty and
emission
LPSP (%) TNPC ($) Npv Ndg
0 1.241 × 106 239 31
10 1.033 × 106 369 13
21
Table 4
Details of two non-dominated solutions of PV/DG/FC system without considering operating reserve, uncertainty
and emission
LPSP (%) TNPC ($) Npv Ndg Nfc Nel Nht
0 1.201 × 106 443 30 7 26 4
10 9.931 × 105 607 5 11 37 8
4.2. Multi-objective optimization of PV/DG/FC system with respect to cost of storage system
T
To investigate the impact of the cost of FC equipments (FC, electrolyser and hydrogen tank) on the
IP
sizing problem, the cost of the FC equipments is decreased 30, 40 and 50 percent and the Pareto fronts
R
are plotted. Fig. 7 shows the impact of the FC system equipments cost on the Pareto front. According to
SC
the results, it can be concluded that the cost of FC equipments has a significant impact on the TNPC
and Pareto fronts. If the cost of FC equipments at LPSP=0 decreases 30, 40 and 50%, consequently, the
U
system TNPC decreases 9.9, 14.5 and 19%, respectively. Table 5 represents details of non-dominated
N
solutions at LPSP 0% and 10%.
A
M
ED
E PT
CC
A
Fig. 7. The impact of the FC system equipments cost on the Pareto front
22
Table 5
Details of two non-dominated solutions of PV/DG/FC system with respect to the cost of storage system
30% reduction
LPSP (%) TNPC ($) Npv Ndg Nfc Nel Nht
0 1.082 × 106 624 30 15 45 12
10 8.855 × 105 605 4 13 41 21
40% reduction
LPSP (%) TNPC ($) Npv Ndg Nfc Nel Nht
1.027 × 106
T
0 632 30 15 45 10
10 8.487 × 105 543 5 12 45 23
IP
50% reduction
LPSP (%) TNPC ($) Npv Ndg Nfc Nel Nht
0 9.731 × 105 622 30 15 46 15
R
10 7.788 × 105 594 5 14 43 5
SC
4.3. Multi-objective optimization of PV/DG/FC system with respect to fuel price
U
In this scenario, the impact of increasing 20, 30 and 40% of DG fuel price is investigated on the Pareto
N
front. Based on Fig. 8, it is observed that at LPSP=0, increase of 20, 30 and 40% of fuel price leads to
A
increasing 6.7, 9.3 and 11.5% in TNPC value, respectively. Table 6 represents details of non-dominated
M
23
Table 6
Details of two non-dominated solutions of PV/DG/FC system with respect to increase of fuel price
20% increase in fuel price
LPSP (%) TNPC ($) Npv Ndg Nfc Nel Nht
0 1.282 × 106 619 30 13 41 7
10 1.076 × 106 474 8 11 27 2
30% increase in fuel price
LPSP (%) TNPC ($) Npv Ndg Nfc Nel Nht
0 1.313 × 106 563 30 15 44 8
10 1.083 × 106 736 3 13 48 5
T
40% increase in fuel price
LPSP (%) TNPC ($) Npv Ndg Nfc Nel Nht
IP
0 1.339 × 106 670 30 15 45 10
10 1.098 × 106 722 3 16 44 9
R
4.4. Multi-objective optimization of PV/DG/FC system with respect to the impact of operating reserve
SC
and uncertainties
4.4.1. PV/DG/FC system with uncertainties of load demand and solar radiation
U
Since during the year, the load demand and solar radiation are uncertain, their uncertainties are taken
N
into account in the sizing problem. Based on Fig. 9, it is observed that at a same LPSP value when
A
uncertainties are included, the cost of PV/DG/FC system increases. Table 7 represents details of two
M
non-dominated solutions at LPSP 0% and 10%. Comparison of Tables 4 and 7 reveals that when solar
ED
24
Table 7
Details of some non-dominated solutions of PV/DG/FC system with respect to uncertainty
LPSP (%) TNPC ($) Npv Ndg Nfc Nel Nht
0 1.218 × 106 427 33 6 24 5
10 9.997 × 105 300 11 3 18 4
Fig. 10 shows Pareto fronts obtained by MO-CSA when OR is included in the sizing problem. It is seen
that when OR is considered, the cost of the hybrid system significantly increases. It originates from this
T
IP
fact that OR helps the hybrid system to compensate the deficit power which may happen due to solar
radiation decrease or load demand increase. In this case, the optimizer attempts to increase the system
R
size to supply the demanded power at deficit conditions. Table 8 represents details of some non-
SC
dominated solutions at LPSP 0% and 10%. It can be drawn that the impact of OR on Pareto front is
Fig. 10. Comparison of Pareto fronts the sizing problem with respect to operating reserve and
uncertainty
A
Table 8
Details of some non-dominated solutions of PV/DG/FC system with respect to uncertainty and operating reserve
LPSP (%) TNPC ($) Npv Ndg Nfc Nel Nht
0 1.462 × 106 277 43 0 0 0
10 1.219 × 106 531 11 5 27 13
25
4.5. Multi-objective optimization of PV/DG/FC system with respect to the impact of CO2 emission
In order to study the impact of emission on the sizing problem, this scenario has been conducted. Fig.
11 shows the Pareto fronts obtained by MO-CSA in four states: (1) in the presence of emission when
there are no OR and uncertainty, (2) in the presence of emission when OR and uncertainties are
considered, (3) in the absence of emission when there are no OR and uncertainty and (4) in the absence
of emission when OR and uncertainties are considered. It is observed that when emission is taken into
account, the cost of the system will increase. In this case, the optimizer will provide a balance between
T
IP
emission cost (namely, using diesel generator) and the cost of using other energy sources. In Fig. 11,
the Pareto front which is at the top of the other ones, is close to the Pareto front of the hybrid system in
R
real conditions.
SC
U
N
A
M
ED
PT
5. Conclusion
Optimal sizing of hybrid energy systems is one of the most challenging problems of power system. In
A
this paper, multi-objective optimization of a hybrid system composed of photovoltaic, diesel generator
and fuel cell by crow search algorithm is the main issue. The impact of operating reserve, emission,
solar power uncertainty, load demand uncertainty and system parameters cost has been investigated on
the Pareto front (TNPC vs LPSP). On an off-grid application, the following results can be concluded:
26
If the hybrid system is optimally sized, combination of PV panels, diesel generator and fuel cell
leads to having a cost-effective and reliable energy system. Indeed, by considering fuel
cell/electrolyser/hydrogen tank in the hybrid energy system, the total cost will decrease.
Operating reserve considerably affects the system cost. Indeed, inclusion of the operating
reserve will increase the system size. Investigations show that the impact of operating reserve is
more than the impact of solar and load uncertainties on the Pareto front.
T
Fuel cost of diesel generator is one of the most important parameters on the Pareto front.
IP
References
R
“Electrical power generation by an optimized autonomous
SC
[1] Askarzadeh, A., 2016:
U
[2] Ahmadi, S., Abdi, S., 2016: “Application of the Hybrid Big Bang–Big Crunch algorithm for optimal
N
sizing of a stand-alone hybrid PV/wind/battery system”, Solar Energy,134, 366–374.
A
[3] Abushnaf, J., Rassau, A., 2018: “Impact of energy management system on the sizing of a grid-
M
[4] Ghenai, ch., Salameh, T., et al., 2018: “Technico-economic analysis of off grid solar PV/Fuel cell
ED
energy system for residential community in desert region”, International Journal of Hydrogen Energy.
[5] Yahiaoui, A., Fedil, F., Benmansour, k., et al., 2017: “Grey wolf optimizer for optimal design of
PT
hybrid renewable energy system PV-Diesel Generator-Battery: Application to the case of Djanet city of
E
[6] Roumila Z., Rekioua D., Rekioua T., Energy management based fuzzy logic controller of hybrid
system wind/photovoltaic/diesel with storage battery, International Journal of Hydrogen Energy, 2017,
A
[7] Sanajaoba, S., Fernandez, E., 2016: “Maiden application of cuckoo search algorithm for optimal
27
[8] Movahediyan, Z., Askarzadeh, A., 2018: “Multi-objective optimization framework of a
photovoltaic-diesel generator hybrid energy system considering operating reserve”, Sustainable Cities
[9] Benchraa, H., Redouane, A., et al., 2018: “Techno-economic feasibility study of a hybrid
Biomass/PV/Diesel/Battery system for powering the village of Imlil in High Atlas of Morocco”, the
[10] Fonseca, M. N., Pamplona, E. D. O., et al., 2018: “Multi-objective optimization applied for
T
IP
designing hybrid power generation systems in isolated networks”, Solar Energy, 161, 207–219.
[11] Kamjoo, A., Maheri, A., et al., 2016: “Multi-objective design under uncertainties of hybrid
R
renewable energy system using NSGA-II and chance constrained programming”, Electrical Power and
SC
Energy Systems, 74, 187–194.
U
[12] Baghaee, H. R., Mirsalim, M., et al., 2017: “Multi-objective optimal power management and sizing
N
of a reliable wind/PV microgrid with hydrogen energy storage using MOPSO”, Journal of Intelligent &
A
Fuzzy Systems, 32, 1753–1773.
M
[13] Ramli, M. A. M., Bouchekara, H. R. E. H., Alghamdi, A. S., 2018: “Optimal Sizing of
[14] Majidi, M., Nojavan, S., et al., 2017: “A multi-objective model for optimal operation of a
PT
battery/PV/fuel cell/grid hybrid energy system using weighted sum technique and fuzzy satisfying
E
[15] Zhang, L., Hu, X., Wang, Z., et al., 2018: “ Multi-Objective Optimal Sizing of Hybrid Energy
Storage System for Electric Vehicles”, IEEE Transactions on Vehicular Technology, 67(2), 1027–1035.
A
[16] A Heydari, A Askarzadeh, Optimization of a biomass-based photovoltaic power plant for an off-
grid application subject to loss of power supply probability concept, Applied Energy, 2016, 165, 601-
611.
28
[17] S. Singh, A.K. Singh, S. Chanana, Operation and control of a hybrid photovoltaic-diesel-fuel cell
system connected to micro-grid, 2012 IEEE Fifth Power India Conference, 19-22 Dec. 2012.
[18] Hai Lan, Shuli Wen, Ying-Yi Hong, David C. Yu, Lijun Zhang, Optimal sizing of a hybrid
PV/diesel/battery in ship power system, Applied Energy, Volume 158, 15 November 2015, Pages 26-
34.
[19] Sawle, Y., Gupta, S.C., et al., 2018: “Socio-techno-economic design of hybrid renewable energy
T
[20] Dufo-López, R., Bernal-Agustín, J. L., 2008: “Multi-objective design of PV– wind– diesel–
IP
hydrogen– battery systems”, Renewable Energy,33, 2559– 2572.
R
[21] Skarstein, O., Ulhen, K, 1989: “Design considerations with respect to long-term diesel saving in
SC
wind/diesel plants”, Wind Energy,13(2), 72–87.
U
[22] Sawle, Y., Gupta, S.C., et al., 2018: “A novel methodology for scrutiny of autonomous hybrid
N
renewable energy system”, International Journal of Energy Research, 42, 570–586.
A
[23] Gupta, S.C., Kumar, Y., Agnihotri, G. R., 2011: “renewable energy analysis and sizing tool”,
M
[24] X. Li, Y.J. Song, S.B. Han, "Frequency control in micro-grid power system combined with
ED
electrolyzer system and fuzzy PI controller", Journal of Power Sources, Vol. 180, pp. 468-475, 2008.
[25] Hakimi, S. M., Moghaddas-Tafreshi, S. M., 2009: “Optimal sizing of a stand-alone hybrid power
PT
system via particle swarm optimization for Kahnouj area in south-east of Iran”, Renewable Energy,34,
E
1855–1862.
CC
[26] Askarzadeh, A., 2016: “A novel metaheuristic method for solving constrained engineering
optimization problems: Crow search algorithm”, Computers and Structures, 169, 1–12.
A
29