You are on page 1of 13

Simulation Modelling Practice and Theory 94 (2019) 1–13

Contents lists available at ScienceDirect

Simulation Modelling Practice and Theory


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

Nonlinear stochastic modeling for optimal dispatch of distributed


T
energy resources in active distribution grids including reactive
power
Hasan Mehrjerdia, Reza Hemmatib, , Elahe Farrokhib

a
Electrical Engineering Department, Qatar University, Doha, Qatar
b
Department of Electrical Engineering, Kermanshah University of Technology, Kermanshah, Iran

ARTICLE INFO ABSTRACT

Keywords: This paper deals with energy storage system (ESS) in active distribution networks. The purpose is
AC power flow to install ESSs on the grid to minimize network losses. The problem is expressed as an optimi-
Active power loss zation programming to minimize annualized cost of losses and annualized investment cost of
Distributed energy resource ESSs at the same time. The constraints of the programming are given as security constraints of the
Energy storage system
network and ESS operational constraints. The network is also equipped with distributed energy
Reactive power
resource (DER) and its uncertainty is modeled and dealt by means of stochastic programming.
Different DERs including diesel, wind, and solar resources are modeled and studied. The pro-
posed nonlinear mixed integer stochastic programming is solved by particle swarm optimization
(PSO). AC power flow is adopted to consider both active and reactive powers in the model. The
ESSs are modeled including both active and reactive powers. The introduced planning finds
optimal location, capacity, and power for ESSs. Furthermore, the charging-discharging regime for
active power of ESSs and injection-absorption pattern for reactive power of ESSs are determined.
The introduced methodology is successfully simulated on a typical distribution network. The
simulation results confirm that the planned strategy properly installs ESSs on the grid and
minimizes network losses. The results demonstrate that the ESSs decrease network losses about
22%. Finally, considering reactive power for ESSs results in about 24% cost reduction.

1. Introduction

This paper aims to study energy storage systems (ESS) in active distribution networks. In order to provide more details, the
introduction is divided into three subsections including (i) literature review, (ii) motivation and incitement, (iii) contribution and
paper organization.

1.1. Literature review

The ESSs are being rapidly developed in electric power systems to solve technical and economic problems. There are different
types of ESSs with various applications. The main methods of energy storage can be classified into mechanical [1], electrical [2],


Corresponding author at: Department of Electrical Engineering, Kermanshah University of Technology, Kermanshah, P.O.Box: 6715685420,
Iran.
E-mail addresses: Hasan.mehrjerdi@qu.edu.qa (H. Mehrjerdi), r.hemmati@kut.ac.ir (R. Hemmati), elahefarokhi11@gmail.com (E. Farrokhi).

https://doi.org/10.1016/j.simpat.2019.01.005
Received 27 October 2018; Received in revised form 27 January 2019; Accepted 30 January 2019
Available online 31 January 2019
1569-190X/ © 2019 Elsevier B.V. All rights reserved.
H. Mehrjerdi et al. Simulation Modelling Practice and Theory 94 (2019) 1–13

Nomenclature Mqch e, t
charging reactive power of ESS (kW)
Mr rated power of ESS (kW)
APlines, t , l
apparent power of line (p.u.) of objective function of the planning ($/year)
APmax maximum capacity of line (p.u.) s, t
Ploss active power losses (kW)
b index of buses s, t , l
Pline active power flow in line (p.u.)
B set of buses b, s, t
Pgen generated active power on bus (p.u.)
s, t
Closs hourly cost of losses ($) b, s, t
Pload consumed active power on bus (p.u.)
Cr rated capacity of ESS (kWh) s, t , l
Qline reactive power flow in line (p.u.)
DCloss daily cost of energy losses ($) b, s, t
Qgen generated reactive power on bus (p.u.)
e index of ESSs b, s, t
Qload consumed reactive power on bus (p.u.)
E set of ESSs s, t
Qloss reactive power losses (kVar)
Ess c investment cost on ESS capacity ($/kWh) Rs probability of scenario
Essm investment cost on ESS power ($/kW) s index of scenarios
G e, t = 0 initial energy of ESS (kWh) S set of scenarios
G e, t = T final energy of ESS (kWh) T set of time intervals
l index of lines t index of time intervals
L set of lines Vmin minimum level of voltage (p.u.)
Ltess life time of ESS (year) Vmax maximum level of voltage (p.u.)
Mche, t
charging power of ESS (kW) V s, t , b voltage of bus (p.u.)
e, t
Mdch discharging power of ESS (kW) tt duration of time interval (h)
Mqdch e, t
discharging reactive power of ESS (kW)

thermal [3], electrochemical [1], and chemical technologies [1]. The detailed applications of different ESSs are listed in Table 1 [4].
It is clear that the ESSs can successfully deal with various problems in electrical networks including short-term and long-term issues.
The short term problems are often classified into power quality issues such as load interruption and the long term problems are
mainly classified into operational and energy issues such as energy shifting [5].
As it was stated, one of the main applications of ESSs is to deal with renewable energy intermittency [6] and uncertainty [7]. The
renewable energy resources such as wind turbines and solar panels often show two kinds of uncertainty including fast fluctuations
within seconds-minute and slow variations in time scales of hours. The fast fluctuation is related to alteration of wind speed or solar
irradiation and it provides small scale variations in output power produced by renewable resources. On the other hand, the slow
variation is related to the energy profile of renewable resources and it provides large scale alterations in output power of renewable
unit. For instance, solar energy profile becomes zero during the nighttime while it operates on maximum power during noontime.
There is therefore a large-scale slow-speed alteration in output power of solar system. As a result, the ESSs are utilized to tackle such
uncertainties in renewable resources. The fast response ESSs are proper to smooth fast fluctuations with small-scale magnitude and
the slow response ESSs are useful for slow fluctuations with large-scale magnitude. Sometimes the hybrid ESSs including small-
capacity fast-response and large-capacity slow-response is utilized to achieve better results [8].
Another important application of ESSs can be regarded as loss reduction in electrical networks [9]. The network operators often
pursue towards new techniques to minimize power losses. The ESSs have been efficiently proposed to reduce network losses. The
network losses are often minimized through shifting energy from off-peak hours to on-peak hours. Such procedure not only minimizes
network losses but also improves voltage profile, enhances network operation, and reduces energy costs [10].

1.2. Motivation and incitement

The ESSs have diverse applications and the purpose is to utilize all their benefits at the same time. However, proper utilization and
operation of ESSs need an accurate mathematical programming considering all characteristics of ESSs. In order to install and operate

Table 1
Main storage technologies and their application.
Method Main technologies Operation time and duration Applications
of storage

Mechanical Flywheel Seconds to minutes Spinning reserve, renewable power smoothing


Compressed Air Energy Hours to days Energy arbitrage, load leveling, peak cutting, loss reduction, renewable energy
Storage shifting
Pumped Hydro Storage
Electrical Super Capacitor Seconds to minutes Spinning reserve, renewable power smoothing
Electrochemical Battery Storage System Minutes to hours Spinning reserve, renewable power smoothing, Energy arbitrage, load leveling,
peak cutting, loss reduction, renewable energy shifting
Chemical Hydrogen Storage Hours to days Energy arbitrage, renewable energy shifting
Thermal Heat-Cold Storage Minutes to hours Energy arbitrage, load leveling, peak cutting, renewable energy shifting

2
H. Mehrjerdi et al. Simulation Modelling Practice and Theory 94 (2019) 1–13

ESSs, it is required to carry out optimization planning to determine optimal sizing and siting of ESSs [11]. Furthermore, optimal
operation (i.e., charging-discharging regime) must be determined for the ESSs. Most of the studies only consider active power of ESSs
and conduct their research to active power ability [12]. But the ESSs have also reactive power capability and can absorb or inject
reactive power to the grid [13]. Thus, it is more practical to consider reactive power of ESSs in the planning. The impact of reactive
power is more significant when the network losses are studied. Because the reactive power causes a part of losses. In this regards,
current paper aims to minimize losses while it considers both active and reactive powers for ESSs.
Moreover, many studies use DC power flow to model ESS operation [14]. But DC power flow is a little bit far from realistic and
does not model accurate operation of network and ESSs. As a result, AC power flow would be more practical to model the network
and ESSs. This paper models the network and ESSs by AC power flow to present realistic model and results.
Eventually, DERs such as diesel, wind, and solar are the new technologies in the electrical networks that offer many advantages
such as loss reduction, voltage improvement, reliability improvement, and cost reduction [15]. It may be useful to model the DER in
coordination with ESSs. This paper considers different DERs including diesel, wind, and solar. The impacts of these technologies are
studied on the network and ESS planning and operation.

1.3. Contribution and paper organization

In this paper, optimal location, capacity, and power of ESSs are determined to minimize network losses in active distribution
network. A 30-bus active distribution network equipped with DER is modeled and studied. The DER uncertainties are considered and
dealt by means of stochastic programming. The network and ESS constraints are included in the proposed stochastic planning. Results
demonstrate that optimal ESS planning improves network operation and reduces power losses.
The introduced methodology presents both technical and economic analyses on ESSs. The economic items such as investment cost
and energy price are analyzed. As well, the technical aspects such as voltage, power flow, losses, and ESS capacity are addressed.
It is worth mention that some mentioned items may be already studied by the other researchers. However, this paper models all
these items at the same time and evaluates their impacts on each other. The main contributions of the paper could be presented as
follows:

ü All given data are close to realistic and it makes the results practical.
ü AC power flow including both active and reactive powers is utilized. The AC power flow models reactive power of loads, reactive
power of lines, reactive losses, and reactive power of DER.
ü The ESSs are modeled considering both active and reactive power.
ü Optimal operation pattern (i.e., charging-discharging regime) is determined for active power of ESSs.
ü Optimal operation pattern (i.e., injecting-absorbing regime) is determined for reactive power of ESSs.
ü Location, power, and capacity of ESSs are optimized subject to all operational and technical constraint of network and ESSs.
ü The DERs are modeled in the network and their uncertainty is studied.

Apart from this introductory, rest of the paper is organized as follows: the problem modeling is given in Section 2, Section 3
describes the solution algorithm, and Section 4 presents the problem data. The results and discussion are given in Section 5 and
Section 6 is devoted to the conclusions and future work.

2. Mathematical expression

In this paper, the ESSs are optimized in order to reduce annual cost of network and losses. The ESS planning is expressed as a
constrained nonlinear optimization programming. Eq. (1) calculates cost of losses during planning horizon.
S T
s, t s, t
DCloss = (Ploss × Closs × t t × Rs )
s =1 t=1 (1)

The permitted voltage levels are specified by constraint (2). The apparent power including both active and reactive powers is
calculated by (3). The allowed line capacity is given by (4). It is clear that (4) includes both active and reactive powers. This paper
applies AC power flow including reactive power and reactive losses that makes the results real and practical.

Vmin V s, t , b Vmax b B, s S, t T (2)

s, t , l s, t , l s, t , l
APline = (Pline + Qline ) l L, s S, t T (3)

(4)
s, t , l
APline APmax l L, s S, t T

Constraints (5) and (6) confirm the equilibrium of active and reactive powers on all buses. The active and reactive power losses
are also calculated by (7) and (8).
b, s, t e, t e, t b, s, t s, t , l
Pgen Mdch = Mch + Pload + Pline b B, s S, t T
l lb (5)

3
H. Mehrjerdi et al. Simulation Modelling Practice and Theory 94 (2019) 1–13

b, s, t e, t e, t b, s, t s, t , l
Qgen Mqdch = Mqch + Qload + Qline b B, s S, t T
l lb (6)
B B
s, t b, s, t b, s, t
Ploss = Pgen Pload s S, t T
b=1 b=1 (7)
B B
s, t b, s, t b, s, t
Qloss = Qgen Qload s S, t T
b=1 b=1 (8)
The constraints (9) and (10) indicate the criteria on charged-discharged powers of ESS. The constraint (11) specifies that the
energy of ESS at the initial and final periods must be the same. The energy of ESS is specified by (12) and its capacity is limited by
(13).
e, t
Mch Mr e E, t T (9)
e, t
Mdch Mr e E, t T (10)

G e, t = 0 G e, t = T e E (11)

(12)
e, t e, t
G e, t = G e, t 1 + (Mch Mdch ) × tt e E, t T

G e, t Cr e E, t T (13)
Eventually, objective function is defined by (14) and it must be minimized. This objective function shows annualized cost of
network. It should be noted that the proposed objective function can also be presented as a multi-objective algorithm. However, in
this paper, all terms of objective function are presented as ($/year). As a result, all terms can be properly added because they have
similar units. Eventually, a single objective function including three terms is achieved as shown by Eq. (14).
of = (DCloss × 365) + (Cr × Ess c + Mr × Essm ) × (Ltess ) 1
(14)

3. Optimization method

3.1. Monte–Carlo simulation

This paper utilizes combination of Monte–Carlo simulation (MCS) and particle swarm optimization (PSO) to solve the problem.
The MCS simulation generates a set of scenarios based on uncertain parameters. Then PSO solves the problem under every scenario of
performance. The proposed methodology is a well-known approach to solve stochastic optimization programming. This paper uses
backward scenario reduction technique to reduce number of scenarios and simulation time.

3.2. PSO algorithm

In PSO algorithm, each particle includes two parts as current position (Xi) and current velocity (Vi) as shown in (15). These
random parameters are generated in the initial iteration [16].

x i (t ) = {xi1 (t ), x i2 (t ), ...,xin (t )}
vi (t ) = {vi1 (t ), vi2 (t ), ...,vin (t )} (15)
PSO algorithm starts its iterations and calculates the objective function for each particle in the population. At each iteration, two
fitness values are recorded. One as the best fitness obtained so far as global fitness and one as the best fitness achieved at the current
iteration as local fitness. These two fitness are applied to update the population as expressed by (16) and (17).
vi (t + 1) = wvi (t ) + c1 r1 [Pbi (t ) x i (t )] + c2 r2 [gb (t ) x i (t )] (16)

x i (t + 1) = xi (t ) + vi (t + 1) (17)
where, w indicates the inertia coefficient, c1 and c2 are acceleration coefficients, r1 and r2 specify random variables between zero and
one. These random variables maintain the random nature of algorithm. The inertia coefficient w is adopted as follows:
wmax wmin
w = wmax Itr
Itrmax (18)
where, wmin and wmax are minimum and maximum levels, Itrmax shows maximum iterations of PSO, and Itr indicates current iteration.
The steps of the algorithm are as follows:
Step1 (Initialization): The initial population is created with random number generation and one particle in the population is
nominated. The ESSs related to current particle are installed on the network.
Step 2 (Power flow): current topology of the network is modeled and power flow is performed. The network parameters such as
voltage, powers, and losses are calculated.

4
H. Mehrjerdi et al. Simulation Modelling Practice and Theory 94 (2019) 1–13

Step 3 (checking the constraints): the problem constraints are checked. If the constraints are violated, current particle is
removed and the next one is nominated; otherwise, objective function is calculated for current particle. This procedure is carried on
until calculating the objective function for all qualified particles in the population. The best particle including minimum fitness value
is determined under current iteration. The best particle under all iterations is also determined.
Step 4 (Convergence of algorithm): convergence of algorithm is examined. If the convergence criterion is met, the optimal
results are achieved; otherwise, algorithm updates the population and reiterates the process until finding optimal solution.
Step 5 (Stochastic programming): all the above steps are simulated for each scenario of performance.
The flowchart in Fig. 1 shows solution of the model by PSO algorithm. As shown in the flowchart, the PSO minimizes objective
function (1) subject to all constraints and under all scenarios of performance. The PSO population is updated at every iteration until
attaining the optimal solution and meeting the convergence criterion.
In this paper, an advanced PSO algorithm is used to solve the problem. This advanced PSO uses adaptive inertia weighting as
shown in (18). As well, the characteristics of genetic algorithm such as mutation and mating are added to PSO. As a result, this paper
utilizes modified-adaptive PSO as an advanced optimization algorithm. Moreover, in order to confirm the optimal solution found by
the algorithm, the problem is solved by various optimization techniques and their results are compared and confirmed. As well, each
optimization technique is simulated several times and its setting are changed under each simulation to prevent falling into local
optimums.

Fig. 1. Flowchart for solution of model by PSO.

5
H. Mehrjerdi et al. Simulation Modelling Practice and Theory 94 (2019) 1–13

4. Test network

The test system is 30-bus radial active distribution network [12]. The topology of network is shown in Fig. 2. The network is
installed with DER on bus 23. Three different DERs are simulated including diesel, wind, and solar. The generated energy by diesel is
constant in all hours. But the wind and solar systems follow different generation patterns based on wind speed and solar irradiation.
The maximum power of DER is 100 kW [17]. The energy pattern and operation of different DERs is modeled and discussed in
Section 5.3. The daily load profile is presented in Table 2. The daily hours are divided into 12 time sections each one 2-h. Daily energy
price is regarded as 0.08 $/kWh for first 8 h, 0.18 $/kWh for second 8 h and 0.35 $/kWh for third 8 h. The base apparent power and
voltage for per unit system are 10 MVA and 11 kV, respectively. Voltage limit is 0.92–1.07 p.u. [8].
The buses 8, 11, 18, 21 and 24 are considered as candidate places for ESS. Maximum power and capacity of ESS are 20 kW and
160 kWh, respectively [18]. The candidate powers for ESS are 0, 10, and 20 kW. The candidate capacities for ESS are 0, 80, and
160 kWh. The ESS cost is 100 ($/kwh) and its life-cycle is 5500 charging-discharging cycle [17].

5. Results and discussions

The proposed methodology to find location and sizing of ESSs is simulated on the specified test network. Table 3 lists the results
for losses and costs. The proposed methodology minimizes the cost while it determines best sizing and siting for ESSs as listed in
Table 4. The proposed planning finds optimal capacity, active power, reactive power, and location for all ESSs. It is clear that four
ESSs are installed on the network in buses 8, 11, 18, 24 and bus 21 is not equipped with ESS.
The proposed planning utilizes ESSs to minimize operational cost of the network including cost of losses. In this regard, an optimal
charging-discharging pattern is determined for all ESSs as depicted in Fig. 3. It is demonstrated that the ESSs charge energy during
low-cost off-peak time sections and discharge it during on-peak high-cost periods. As well, the ESSs perform on their full capacity to
minimize losses. Such optimal operation reduces costs as much as possible and improves voltage profile during on-peak loading
conditions. The archived optimal operation discharges energy during on-peak time sections (e.g., time section 10) and helps the
network to face heavy loading conditions.
As it was stated, the proposed methodology optimizes both active and reactive power of ESSs. The reactive power pattern of ESSs
is summarized in Table 5. The results confirm that the installed ESSs only inject reactive power to the grid and they do not absorb
reactive power. As well, they operate only at time section 10 (i.e., on-peak loading condition) to improve voltage profile. The ESSs
inject reactive power to the grid and help the network to satisfy voltage constraints. The voltage profile at two operating conditions
including on-peak and off-peak loading is depicted in Fig. 4. It is clear that both voltage profiles satisfy the permitted voltage limits
(i.e., 0.92 p.u.–1.07 p.u.). Under on-peak loading condition, the voltage profile drops compared to the other cases.
Fig. 5 shows power flow in the network lines and the permitted capacity of lines is also presented. The power flow in all lines is

Fig. 2. Radial distribution network installed with distributed energy resource.

6
H. Mehrjerdi et al. Simulation Modelling Practice and Theory 94 (2019) 1–13

Table 2
Profile of load levels.
Time section 1 2 3 4 5 6 7 8 9 10 11 12

Hour 0–2 2–4 4–6 6–8 8–10 10–12 12–14 14–16 16–18 18–20 20–22 22–24
Load (%) 30 20 30 35 40 45 50 60 50 100 85 70

Table 3
Achieved losses and costs after ESS deployment.
Parameter Optimal level

Annual power losses (kWh) 1.3344e+5


Annual cost of losses ($/year) 6013.6
Annual investment cost on ESS ($/year) 3733.3

Table 4
Optimal sizing and siting of ESSs.
ESS number Location Capacity (kWh) Active power (kW) Reactive power (kVar)

ESS 1 Bus 8 160 20 70


ESS 2 Bus 11 160 20 70
ESS 3 Bus 18 80 10 80
ESS 4 Bus 24 160 20 80

Fig. 3. Optimal charging-discharging pattern of ESSs.

Table 5
Reactive power pattern of ESSs.
Time section ESS 1 and 2 ESS 3 and 4
Injected power (kVar) Absorbed power (kVar) Injected power (kVar) Absorbed power (kVar)

1 to 9 0 0 0 0
10 70 0 80 0
11–12 0 0 0 0

less than the limited level and all the constraints are satisfied. Some lines operate near to congestion state such as line 16 and on the
other hand, some lines have free capacity such as line 1. Line 1 is the most important line of the network and all power of the network
is supplied through this line. As a result, high capacity is selected for line.

7
H. Mehrjerdi et al. Simulation Modelling Practice and Theory 94 (2019) 1–13

Fig. 4. Voltage under on-peak and off-peak loading conditions.

Fig. 5. Power flow in lines and permitted capacity of lines.

5.1. Comparison study 1: with and without ESS

In order to demonstrate the effectiveness of the given methodology, the method (network with ESS) is compared against the
network without ESS. Table 6 shows the comparison results for the networks with and without ESS. It is clear that installing ESS
decreases network losses and costs significantly. The results show that the losses are reduced by about 22%.
Fig. 6 shows the received power from the upstream grid with and without ESS. In the network with ESS, the received power is
increased when the ESSs are charged (i.e., time intervals 1 to 4) and then the power is decreased when the ESSs are discharged (i.e.,
time Sections 8–12). Since the energy price at time sections 8 to 12 is greater than the energy price at time intervals 1 to 4, the costs of
energy and losses are significantly reduced. This point is shown by Fig. 7 where the hourly losses of the network is depicted with and

Table 6
Comparison study on network with and without ESS.
With ESS Without ESS

Annual power losses (kWh) 1.33e+5 1.62E+5


Annual cost of losses ($/year) 6013.6 7459.5

8
H. Mehrjerdi et al. Simulation Modelling Practice and Theory 94 (2019) 1–13

Fig. 6. Received power from the upstream grid with and without ESS.

Fig. 7. Hourly losses of network with and without ESS.

without ESS. It is clear that the losses is increased at time intervals 1 to 4 and it is decreased at time intervals 8 to 12. However, the
decrease in losses is greater than the increase in losses and total loss is decreased as a result.
Fig. 8 shows the voltage profile with and without ESS. The voltage limit is also depicted to provide better comparison. It is clear
that the voltage profile with ESS satisfies the voltage limit while the network without ESS cannot tolerate high loading condition and
its voltage drops under the permitted levels at many buses.

5.2. Comparison study 2: with and without reactive power

In current paper, both active and reactive powers are included for ESSs, however many papers only consider active power for
ESSs. As a result, a comparison study is carried out between the methodology given by this paper and the methodology given by paper
[12]. In paper [12], only active power is modeled for ESSs. It is worth mentioning that the proposed methodology given by paper
[12] (i.e., considering only active power) has been adopted by many other papers. As a result, this comparison is between current
paper and all the researches that have only considered active power for ESSs. However, paper [12] has been selected as a typical
paper that has modeled ESSs with active power. In order to provide more details to the readers, two cases are simulated and
compared as follows:
Case 1: ESSs with reactive power (the methodology introduced by this paper including both active and reactive power for ESS).
Case 2: ESSs without reactive power (the methodology introduced by the other papers such as paper [12] including only active

9
H. Mehrjerdi et al. Simulation Modelling Practice and Theory 94 (2019) 1–13

Fig. 8. Voltage profile with and without ESS.

power for ESS).


Table 7 compares losses and costs for planning with and without reactive power and Table 8 lists the planning output with and
without reactive power. The results verify that the ESSs without reactive power need larger capacity and power to satisfy network
constraints. As a result, the investment cost on ESSs is significantly increased. In contrast, the ESSs with reactive power are able to
inject reactive power to the grid and satisfy the voltage limits. Accordingly, smaller capacity and power is required for such ESSs
resulting in less investment cost. The results verify that the proposed methodology including both active and reactive power for ESSs
is better and more efficient. The given strategy reduces the cost by about 24%.

5.3. Comparison study 3: different distributed energy resources

In order to demonstrate the impacts of DER on the planning, a comparison study is carried out with and without DER as listed in
Table 9. It is shown that DER makes positive impacts on the network and reduces losses and costs considerably. However, the DERs
are different and they have diverse energy profiles. In order to demonstrate the impacts of energy profile of DERs on the planning,
three DERs are modeled and simulated including solar, wind, and diesel. The energy profile for these DERs is depicted in Fig. 9. It is
clear that these DERs follow different generation patterns. The planning outputs under these DERs is listed in Table 10. The results
verify that the DERs with small energy profile (i.e., solar) cannot properly improve the network operation and the DERs with large
energy profile (i.e., diesel) can properly improve the network losses and costs. However, the operational cost of diesel generator is
more than the other DERs. Moreover, the diesel generator releases environment pollutions. As a result, a tradeoff between the DERs is
necessary.
It should be noted that the generated power from a diesel power plant is not constant and it is better to present an optimal
generation profile with different levels at different hours to minimize the operational cost (fuel cost) of diesel power plant. However,
in this paper, the main focus is to study the energy storage systems and their impacts on renewable energy resources and the non-
renewable energy resources such as diesel power plants are not the main concentration of the paper. As a result, a constant generation
pattern is regarded for diesel unit. Moreover, the purpose is to compare the wind, solar, and diesel units. In order to present a fair
comparison under similar conditions, it is assumed that all of these resources including wind, solar, and diesel work on their full
capacity at all hours. It is worth mentioning that the proposed assumption (working on full capacity at all hours) may not be optimal
for the diesel power plant.

Table 7
Comparing losses and costs for planning with and without reactive power.
With reactive power Without reactive power

Annual power losses (kWh) 1.33e+5 1.39e+5


Annual cost of losses ($/year) 6013.6 4535.3
Annual investment cost on ESS ($/year) 3733.3 7466.7
Total cost ($/year) 9747 12,002

10
H. Mehrjerdi et al. Simulation Modelling Practice and Theory 94 (2019) 1–13

Table 8
Planning output with and without reactive power.
ESS number Location Capacity (kWh) Active power (kW) Reactive power (kVar)

Without reactive power ESS 1 Bus 8 320 40 0


ESS 2 Bus 11 320 40 0
ESS 3 Bus 18 160 20 0
ESS 4 Bus 24 480 60 0
With reactive power ESS 1 Bus 8 160 20 70
ESS 2 Bus 11 160 20 70
ESS 3 Bus 18 80 10 80
ESS 4 Bus 24 160 20 80

Table 9
Comparing the results with and without DER.
With DER Without DER

Annual power losses (kWh) 1.33e+5 2.03E+5


Annual cost of losses ($/year) 6013.6 8626.8
Annual investment cost on ESS ($/year) 3733.3 3733.3
Total cost ($/year) 9747 12,360

Fig. 9. Generation pattern of solar, wind, and diesel DERs.

Table 10
Comparing the results under different generation patterns for DERs.
Solar Wind Diesel

Annual power losses (kWh) 1.81e+5 1.68e+5 1.33e+5


Annual cost of losses ($/year) 8626.8 6923.7 6013.6

5.4. Sensitivity analysis

In order to confirm the accuracy of simulations, an error analysis is carried out and its results are listed in Table 11. The results
show that the losses and costs are closely associated with network parameters and changing each parameter can make significant
impacts on the outputs. The results indicate that location of DER is very important. Installing DER on first or last buses is not optimal
and increases the costs.

11
H. Mehrjerdi et al. Simulation Modelling Practice and Theory 94 (2019) 1–13

Table 11
Sensitivity analysis on parameters of the planning.
Analyzed case Annual cost of losses ($/year)

Nominal case 6013.6


Changing location of DER to bus 30 7070.4
Changing location of DER to bus 2 8349.8
Decreasing DER power by 10% 6224.8
Increasing DER power by 10% 5813.1
Increasing loading by 5% 6841.2
Decreasing loading by 5% 5247.2
Increasing energy price by 5% 6314.3

6. Discussion of achievements

This paper presents an optimal plan on ESSs to minimize losses and investment cost. The main achievements of the paper can be
presented as follows:

■ Installing ESS on the network makes positive impacts on the grid. It reduces the network losses, improves voltage profile, and
reduces congestion in the network lines. The results indicate that ESSs reduce the network losses about 22%.
■ Considering reactive power ability for ESSs makes the results realistic. This ability of ESSs help the network to improve voltage. It
reduces the capacity and power of required ESSs and decreases the cost about 24%.
■ The results of this paper indicate that finding optimal sizing and siting for ESSs is not enough. In order to utilize all positive
abilities of ESSs at the same time, it is also necessary to find optimal charging-discharging regime for active power of ESSs and
optimal injection-absorption pattern for reactive power of ESSs.
■ The results demonstrate that AC power flow including reactive power of loads and network is more realistic.
■ The DERs make positive impacts on losses and costs. The impact of DERs is associated with their energy profiles. The larger energy
profile for DERs provides more benefits to the grid.

7. Conclusions

This paper presents an optimization programming to minimize losses of the network. The introduced plan determines sizing and
siting of energy ESSs on the distribution network. The planning also determines operation pattern for both active and reactive power
of ESSs. A typical distribution network installed with DER is regarded as case study and the stochastic planning is carried out and
solved by PSO. The results demonstrate that the plan finds optimal capacity, active power, reactive power, and location for all ESSs.
The network is equipped with four ESSs on buses 8, 11, 18, 24 and bus 21 is not installed with ESS. The results show that the ESSs
charge energy during low-cost off-peak hours and discharge energy at on-peak high-cost hours and they reduce losses and costs by
such operation. The ESSs also inject reactive power to the grid under on-peak loading condition to improve voltage profile. The
results verify that installing ESSs decreases the network losses about 22%. The voltage profile in the network without ESS drops under
the permitted levels at many buses while the network with ESSs successfully satisfies the voltage limits. The results also demonstrate
that the ESSs without reactive power need larger capacity and power to satisfy network constraints resulting in more investment cost.
But the ESSs with reactive power need smaller capacity and power resulting in less investment cost. The results verify that con-
sidering reactive power ability for ESSs reduces the cost about 24%. Different DER are also simulated. The results show that the DERs
with less energy profile cannot appropriately improve the network operation while the DERs with large energy profile make more
positive impacts on the grid.
Further to this work, following topics are suggested as future work: considering multi-objective algorithm instead of a single
objective, providing a generalized model for all kinds of renewable DERs and droop-controlled DG units and storage systems, pre-
senting an optimal generation pattern (including different generating powers at different hours) for diesel generator rather than a
constant generation pattern.

References

[1] R. Amirante, E. Cassone, E. Distaso, P. Tamburrano, Overview on recent developments in energy storage: mechanical, electrochemical and hydrogen tech-
nologies, Energy Convers. Manag. 132 (2017) 372–387.
[2] B. Zakeri, S. Syri, Electrical energy storage systems: a comparative life cycle cost analysis, Renew. Sust. Energy Rev. 42 (2015) 569–596.
[3] B. Cui, D.C. Gao, F. Xiao, S. Wang, Model-based optimal design of active cool thermal energy storage for maximal life-cycle cost saving from demand man-
agement in commercial buildings, Appl. Energy. 201 (2017) 382–396.
[4] H. Saboori, R. Hemmati, S.M.S. Ghiasi, Sh. Dehghan, Energy storage planning in electric power distribution networks – A state-of-the-art review, Renew. Sust.
Energy Rev. 79 (2017) 1108–1121.
[5] H. Saboori, R. Hemmati, V. Abbasi, Multistage distribution network expansion planning considering the emerging energy storage systems, Energy Convers.
Manag. 105 (2015) 938–945.
[6] H.R. Baghaee, M. Mirsalim, G.B. Gharehpetian, H.A. Talebi, Application of RBF neural networks and unscented transformation in probabilistic power-flow of
microgrids including correlated wind/PV units and plug-in hybrid electric vehicles, Simul. Model. Pract. Theory. 72 (2017) 51–68.
[7] R. Hemmati, H. Saboori, S. Saboori, Assessing wind uncertainty impact on short term operation scheduling of coordinated energy storage systems and thermal

12
H. Mehrjerdi et al. Simulation Modelling Practice and Theory 94 (2019) 1–13

units, Renew. Energy. 95 (2016) 74–84.


[8] R. Hemmati, H. Saboori, Emergence of hybrid energy storage systems in renewable energy and transport applications–A review, Renew. Sust. Energy Rev. 65
(2016) 11–23.
[9] M. Fantauzzi, D. Lauria, F. Mottola, A. Scalfati, Sizing energy storage systems in DC networks: a general methodology based upon power losses minimization,
Appl. Energy. 187 (2017) 862–872.
[10] H. Zhao, Q. Wu, S. Hu, H. Xu, C.N. Rasmussen, Review of energy storage system for wind power integration support, Appl. Energy. 137 (2015) 545–553.
[11] M. Zidar, P.S. Georgilakis, N.D. Hatziargyriou, T. Capuder, D. Škrlec, Review of energy storage allocation in power distribution networks: applications, methods
and future research, IET Gener. Transm. Dis. (2016) 645–652.
[12] H. Saboori, R. Hemmati, Optimal management and planning of storage systems based on particle swarm optimization technique, J. Renew. Sust. Energy. 8
(2016) 024105.
[13] R. Hemmati, N. Azizi, Nonlinear modeling and simulation of battery energy storage systems incorporating multiband stabilizers tuned by Meta-heuristic al-
gorithm, Simul. Model. Pract. Theory. 77 (2017) 212–227.
[14] R. Hemmati, M. Shafie-khah, J.P. Catalão, Three-level hybrid energy storage planning under uncertainty, IEEE Trans. Ind. Electronic. (2018).
[15] T. Adefarati, R. Bansal, Integration of renewable distributed generators into the distribution system: a review, IET Renew. Power Gener. 10 (2016) 873–884.
[16] H. Saboori, R. Hemmati, M.A. Jirdehi, Reliability improvement in radial electrical distribution network by optimal planning of energy storage systems, Energy.
93 (2015) 2299–2312.
[17] R. Hemmati, Technical and economic analysis of home energy management system incorporating small-scale wind turbine and battery energy storage system, J.
Clean. Prod. 159 (2017) 106–118.
[18] R. Hemmati, Optimal design and operation of energy storage systems and generators in the network installed with wind turbines considering practical char-
acteristics of storage units as design variable, J. Clean. Prod. 185 (2018) 680–693.

13

You might also like