Received 9 January 2023, accepted 28 January 2023, date of publication 31 January 2023, date of current version 3 February 2023.

Digital Object Identifier 10.1109/ACCESS.2023.3241228

Multi-Objective Optimization for Distribution

Network Reconfiguration With Reactive Power
Optimization of New Energy and EVs
RENBO WU AND SHUQIN LIU
School of Electrical Engineering, Shandong University, Jinan 250061, China
Corresponding author: Renbo Wu ([email protected])

ABSTRACT The rise of new energy and the wide application of electric vehicles (EVs) have led to
the substantial expansion of distribution network in recent years. The problems such as the decline of
transmission reliability and the rise of power loss in distribution network are becoming increasingly serious.
Therefore, distribution network can greatly improve its reliability and quality of power supply voltage
through changing topology. This work built a distribution network reconfiguration (DNR) model with new
energy and EVs firstly. Then, the position of bus tie switches and the reactive power regulation range of new
energy and EVs were proposed as decision variables. A multi-objective evolutionary algorithm would be
applied to this DNR model and the optimization results will be obtained when considering the line loss and
voltage deviation as the objective function. In order to get different optimal compromise solutions with the
changes of actual environment, this work employed a new decision-making method named Prevalence Effect
Method (PEM). Finally, a high-quality strategy of DNR and reactive power regulation will be obtained.

INDEX TERMS Distribution network reconfiguration, new energy, electric vehicles, multi-objective
optimization.

I. INTRODUCTION irradiation, cloud cover, photovoltaic panel orientation or dust

A. IMPACT OF NEW ENERGY AND EVS ON DISTRIBUTION diffusion and other factors may interfere with the normalized
NETWORK PV to a great extent [3]. Besides, high penetration of PV
Renewable energy sources (RES) such as photovoltaic system may lead to the issues of voltage rise, reverse power flow and
(PV) play a great important part in reducing the environment increased energy loss [4].
pollution in recent years due to its reduction of greenhouse New energy has been strongly supported by various
effect [1]. As a kind of mature and widely used power countries in recent years because of its cleanness and
generation method, photovoltaic power generation conforms circularity. At the same time, the wide application of new
to the strategy of sustainable development and the concept energy drives the explosive growth of electric vehicles (EVs).
of safe power generation perfectly. With the development Such phenomenon has led to a problem that the dispatching
of distributed generations (DGs), PV can be operated at a operations of distribution networks will be affected [5]. New
smaller scale called distributed energy resources (DER). This energy has great uncertainty due to the weather factors, which
form of PV aims to be closer to the load that needs to consume is the issue fluctuating the power quality of distribution
power, which uses the idea of decentralized investment to network [6]. However, with the development of vehicle to
reduce the loss in the transmission [2]. However, PV has grid (V2G), EVs present their marvelous energy storage
the characteristics of intermittence and instability. Solar capacity [7]. On the one hand, a large number of energy
storage batteries may realize peak cut, so as to balance power
The associate editor coordinating the review of this manuscript and generation and consumption [8]. On the other hand, EVs can
approving it for publication was Salvatore Favuzza . achieve the function that it can provide a certain guarantee

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
10664 VOLUME 11, 2023
R. Wu, S. Liu: Multi-Objective Optimization for DNR With Reactive Power Optimization of New Energy and EVs

for the instability of new energy in the event of extreme system average interruption duration index (SAIDI), and
weather. expected energy not supply (EENS). Study [19] considered
Actually, the new energy can not only provide active the uncertainty of DGs, the safe and economic operation of
power to the distribution network, but also regulate their the loads. Authors in [20] had designed a hybrid market-based
reactive power output to achieve a local reactive power DNR methodology to obtain optimal configuration of the
balance for the distribution network [9]. As a result, it is network and Locational Marginal Prices (LMPs) at DG
beneficial to reduce the active power loss and voltage connected buses. Later in [21], research evaluated the
deviation. post-fault distribution network in different scenarios and fully
formulate the influences of demand variations, uncertainty
B. DEVELOPMENT AND RESEARCH STATUS OF DNR of distributed generations and protection failures on the
Experimental research has shown that most outages will reliability indices. Literature [22] studied two different
occur at distribution network within the entire power system DNR scenarios with using minimum switching operations
because of the instability of new energy and EVs [10]. to resupply the maximum energy of loads. A pioneering
Besides, serious power loss is a major problem in different paper [23] explored some acceleration strategies in distribu-
size of distribution networks [11]. In order to improve tion networks with binary variables indicating line-switching
dependability and reduce power loss, one of the methods statuses and tap positions for enhancing the computational
is to make rational use of the characteristics of open-loop performance. With the under-load tap changer and shunt
operation and closed-loop structure of distribution net- capacitors participating in voltage control, literature [24]
work [12]. Distribution network reconfiguration (DNR) with investigated the coordinated voltage control problem for
the peculiarity mentioned above is to maintain the bus smart distribution grid with the integration of DG. However,
voltage of distribution network and improve its operation they seldom consider that new energy and EVs will play a
performance by changing the close/open status of tie role or have an impact on the DNR.
switches [13]. Thus, with high-penetration of new energy
and EVs connecting to distribution network, it is urgent to C. CONTRIBUTIONS
maintain its normal operation and reduce power loss through The main contributions of this paper are twofold:
DNR. (1)Compare to the former work, this paper presents a DNR
Researchers in different regions have had different model with new energy and EVs. New energy and EVs are
understandings of DNR over time, and performed numer- hot themes in the power system, and they may probably
ous meaningful viewpoints in modeling and optimization, participate in all processes of power transmission [25]. Many
as discussed below. From the perspective of the solution researchers focus on improving the method of optimization
procedure, literature [14] employed an efficient Genetic in DNR, but often ignore the actual situation of distribution
Algorithm (GA) to accelerate the calculation and show network. Especially, the model proposed can not only obtain
higher accuracy in DNR. In [15], researchers proposed the optimal status of tie switches, but also form a reliable
an improved Particle Swarm Optimization (PSO) which reactive power regulation strategy of the distribution network.
involved a new acceleration coefficient. Although the solu- (2)The conventional DNR only considers the status
tions of two algorithms above had indicated both a rapid optimization of the bus tie switches. In contrast, the reactive
convergence and a good robustness of these approaches, power regulation outputs of the new energy and EVs are
they mainly focused on a single optimization problem. also considered as the controllable variables for DNR in
In other words, they could not proposed an optimization this work. The constraints of tie switches in DNR and the
strategy based on solving multi-objective problems (MOPs) reactive power regulation involved in the model will be
in DNR. Reference [16] suggested a recent meta-heuristic introduced in detail below. This work employs a relatively
sine-cosine algorithm to reconfigure the distribution network new multi-objective optimization algorithm (MOA) proposed
with simultaneous allocation of multiple DGs. Inspired by another study to compare with the single-objective
by heuristic algorithm, an efficient meta-heuristic method optimization algorithm in many researches for this nonlinear
for reconfiguration of distribution systems named modified problem. Besides, the influence of climate changes on new
Tabu Search (TS) was presented in [17]. This kind of energy is considered in this paper, then the decision-making
methods provides an idea of rapid convergence to optimal method named Prevalence Effect Method (PEM) [26] will
solutions. be used to this model for the search of different optimal
Different focuses of research lead to different presentation solutions.
of objective in DNR studies. An approach for DNR and the The remainder of this paper is organized as follows:
optimal placement of capacitors based on a mixed-integer Section II presents the model of DNR with high-penetration
linear programming (MILP) model was shown in [18]. This of new energy and EVs. In section III, the MOA and other
work emphasizes the loss of DNR, enhance reliability and single-objective optimization algorithms employed in this
improve the voltage profile. In literature [10], some reliable model will be briefly introduced. Section IV analyzes the
indexes of distribution network were presented including numeral tests result in different systems or situations. The
the system average interruption frequency index (SAIFI), conclusion of this work will be provided in Section V.

VOLUME 11, 2023 10665

 R. Wu, S. Liu: Multi-Objective Optimization for DNR With Reactive Power Optimization of New Energy and EVs

FIGURE 1. Illustrative example of different switches in distribution

network.

II. MATHEMATICAL MODEL OF DISTRIBUTION

NETWORK RECONFIGURATION
A. DESCRIPTION OF DNR AND THE NETWORK
SIMPLIFICATION
Actually, tie switches and sectionalizing switches play
essential roles in distribution network for reducing the power
loss and duration of customer interruptions in the process
of electric energy transmission [27]. Figure 1 presents the
vivid example of branches and multiple candidate tie switches
locations of normal distribution.
The example system shown in Figure 1 is composed of
two substation nodes and four load nodes. Substation nodes
included Bus1 and Bus 5 refers to the nodes where voltage
FIGURE 2. Process of DNR simplifying in IEEE 69-bus system.
and current are transformed, electric energy is received and
distributed in distribution network [28]. The remaining buses TABLE 1. Switches Contained in Different Loops in 69-Bus System.
are all load nodes in this example. The branches with S5,
S6, S13 and S14 can be seem as tie lines, then the whole
distribution network will be connected into a ring. The
advantage of the ring network is that it can greatly improve
the reliability of the distribution network [29]. The status of
tie switches and sectionalizing switches can be changed in
process of DNR or after the fault in distribution network. For
will be formed into five different loops. It is obvious from
example, circuit breaker B3 in service will be tripped by the
the figure that there are partial intersections between the five
fault in Bus 4-5. However, the power that loads need in Bus
different loops. For example, loop 1 and loop 4 have the
4 can be transmitted from Bus 3 when the tie switches S5 and
overlap that includes the area between node 4-9. Thus, this
S6 are closed.
overlap only needs to be collected by any one of the loops as
However, there will be some invalid solutions in opti-
shown in Table 1.
mization of DNR because of the different islands [30].
The invalid solutions that can not meet the constraint of B. OBJECTIVE FUNCTION
distribution network topology will greatly reduce the search DNR with the non-linear characteristics has various objec-
efficiency [15]. Therefore, the topologies of different systems tives in MOP. Equation (1) minimizes the objective function,
need to be simplified via the method presented below. PLoss , assuming that active power loss is equal to the thermal
Particularly, this work takes the 69-bus system as an losses generated during the electrical energy transmission
example for simplification in DNR. The original topology of [31], [32]. To ensure that the voltage deviation of distribution
this system is shown in Figure 2 (a) with several branches. network will not affect the normal power consumption of
Then, the specific branches that have no relationship with users, equation (2) specifies the sum of voltage deviation of
others have been highlighted in green while the rest of them each node [33].
have tie switches with another line in Figure 2 (b). Since  Nl
the green branches cannot be connected to other nodes in X P2 + Q2
rl l 2 l

 f1 = min(PLoss ) =

a ring, they should be deleted from the original topology. Ui (1)
Meanwhile, the switches in a ring are considered as decision  Xl=1 2
 f2 = min(Vd ) = ∗
(Vj − V )

variables in Section IV. In Figure 2 (c), the remaining nodes j∈Ni j

10666 VOLUME 11, 2023

R. Wu, S. Liu: Multi-Objective Optimization for DNR With Reactive Power Optimization of New Energy and EVs

where Nl is the set of all branches of distribution network equation (4)(5).

and l means the lth branch of it; P2l and Q2l represent the
current active power and reactive power flow in lth branch, Qw,min ≤ Qw ≤Qw,max (4)
(
respectively; Ui2 refers to the actual voltage of the ith node; Qw,max = Qs,max + Qc,max
rl is the resistance of the lth branch; Vj and Vj∗ denote the (5)
Qw,min = Qs,min +Qc,min
current voltage and rated voltage of the jth node.
where Qw,max and Qw,min are the upper and lower limits of
C. CONSTRAINT CONDITION reactive power output range, respectively; Qw is the current
The objective function presented in equation (1) is subject to reactive power transmitted from wind turbines; Qs,max and
some constraints mentioned below: Qs,min denote the upper and lower limits of the reactive power
flow through the stator, respectively; Qc,max and Qc,min refer
1) ACTIVE POWER AND REACTIVE POWER BALANCE
to the upper and lower limits of reactive power flow in grid
side inverter, respectively.
The distribution network should ensure that the active power
and reactive power flow of the whole system are in a dynamic
4) REACTIVE POWER LIMIT OF PV
equilibrium state at any time [34]. In other words, the total
active and reactive power outputs generated by the power Photovoltaic (PV) equipment that transmits the solar energy
supply should be continuously equal to the total active and into electric energy can not only inject more active power
reactive power consumptions, respectively, as well as for each into the distribution network, but also regulate the reactive
node. power balance of it [37]. In this work, the adjustment range
of reactive power output for PV system is determined by its
internal inverter.
 N Nth Nbr
 XDG X X
 PSlack +

 PDG,j = PD,j + PL,j


j=1 j=1 j=1
Qpv,min ≤ Qpv ≤Qpv,max (6)
N DG Nth Nbr (2)  q
 Q
pv,max = (S )2 − (Ppv )2
q pv
 X X X
 QSlack +

 QDG,j = QD,j + QL,j (7)
= − (S )2 − (P )2
 Q
j=1 j=1 j=1 pv,min pv pv

where PSlack and QSlack are the active power and reactive where Qpv,max and Qpv,min are the upper and lower limits
power supplied from the slack bus, respectively; NDG means of reactive power for PV, respectively; Qpv is the current
the set of distribution generators (DGs) in distribution reactive power injected into the distribution network via PV;
network; PDG,j and QDG,j represent the active power and Spv represents the capacity of PV inverter; Ppv means the
reactive power output of jth DG, respectively; Ntb is the set active power output of the PV system.
of total buses and Nbr is the set of total branches; PD,j and
QD,j denote the active and reactive power demand of the jth 5) REACTIVE POWER LIMIT OF EVS
node, respectively; PL,j and QL,j mean the active and reactive With the wide spread of the V2G, EVs gradually has
power loss of the jth node, respectively. the ability of mutual transmission with the distribution
network through the charging piles [38]. In fact, the EVs
2) BUS VOLTAGE CONSTRAINTS can be operated at the states of charge or discharge since
Every node in distribution network has a fluctuation range of the active power inputs or ouputs of EVs are not taken
voltage [35]. The distribution network will enter the abnormal as the optimization variables in the proposed DNR. The
state or fault working state when the voltage is out of the participation of EVs is benefit to achieve a local reactive
following fluctuation range. power balance for the distribution network, thus the voltage
deviation and power loss can be dramatically reduced. Since
Vkmin ≤ Vk ≤ Vkmax (3) the inverters of EV charging piles are similar to the PV
system, the principle of reactive power regulation is almost
where Vkmin and Vkmax mean the minimum voltage and the same as that of the PV system when EVs output
maximum voltage at kth node, respectively; Vk is the voltage electric energy as the power supply terminal. Hence, the
magnitude at kth node. reactive power regulation range of the EVs can be written as
Eqs. (8)-(9).
3) REACTIVE POWER LIMIT OF WIND POWER QEV,min ≤ QEV ≤QEV,max (8)
The reactive power of wind turbines fed into the distribution  q
2 2
EV,max = q(SEV ) − (PEV )
Q
network has the relationship with stators and the grid side
(9)
inverters of frequency converter [36]. The capacity limits Q = − (S )2 − (P )2
EV,min EV EV
are obtained by the rated stator and rotor currents in the
steady state. Thus, the upper and lower limit of reactive where QEV,max and QEV,min refer to the upper and lower
power transmitted from wind turbines can be calculated as limits of reactive power for EVs, respectively; QEV means

VOLUME 11, 2023 10667

 R. Wu, S. Liu: Multi-Objective Optimization for DNR With Reactive Power Optimization of New Energy and EVs

the current reactive power transmitted into the distribution 1) An fpfs-matrix [aij ] will be constructed as:
network; SEV denotes the capacity of charging pile inverters;  
a01 · · · a0n
PEV is the current active power output when they are
[aij ] =  ... . . . ...  (12)
 
considered as the power supply.
am1 · · · amn
6) MAXIMUM CURRENT CONSTRAINT
where i ∈ {1, 2, . . . ,m − 1} and j ∈ {1, 2, . . . ,n}, m ≥ 2
The current flow in each branch could not exceed a standard
and n ≥ 1. Each column of this matrix corresponds to an
value that is applied in actual distribution network in the
objective, and the a0j that ranges from 0 to 1 represents the
process of optimization.
essential of the objective in this column. Besides, aij refer to
0 ≤ Im ≤ Im,max , m = 1, 2, . . . Nbr (10) the solutions of different algorithms or different solutions on
the same algorithm Pareto front (PF) such that i ̸= 0.
where Im means the current flow in mth branch; Im,max refers
2) The fpfs-matrix listed in equation (12) will change with
to the maximum standard in different distribution network.
the essential of each objective, and then a score matrix
can be obtained by calculation:
7) NUMBER OF SWITCHES WITH STATUS CHANGE
CONSTRAINT Xn  1 Xm−1   1 Xn 
The topology of distribution network has been separated into si1 = [ akj ait a0j aij ]
j=1 m−1 k=1 n t=1
several loops such as the 69-bus system in Figure 2. Then only (13)
one state of switch should be changed in each loop because of
1 P m−1
the guarantee of the connectivity in whole network. If there where m−1 k=1Pakj represents the prevalence effect value of
are several branches opened in a same loop, there would jth parameter; 1n nt=1 ait is the prevalence effect value of ith
appear islands [15]. In summary, the quantity of action switch algorithm or situation; si1 denotes the score of ith row and i ∈
must be equal to 1. {1, 2, . . . ,m − 1}. The matrix [si1 ] which is usually expressed
as [si1 ] = [s11 , s21 , . . . , si1 ]T means the score set ofeach row.
Numn = 1, n = 1, 2, . . . loopmax (11) s
k1
maxsi1
3) Then, a decision matrix uk |uk ∈ U will be
where Numn is the switch that changes its state in the nth
loop; loopmax means the set of total loop. obtained via the two steps above. where u means the
set of algorithm or different situation of the same
III. OPTIMIZATION DESIGN OF DNR algorithm; uk is one of the algorithm or situation
A. BRIEF DESCRIPTION OF ARSBX of the same algorithm. this matrix sorted according
DNR is a typical constrained multi-objective optimization to the scores will apply to the decision-making that
problems (CMOP), which not only needs to meet the active analyzes and judges the solutions of dnr under different
and reactive power balance of distribution network, but also conditions.
has certain range requirements on key points such as node
voltage amplitude. In order to highlight the advantages of the C. APPLICATION DESIGN OF ALGORITHMS IN DNR
latest MOA in DNR, this work will compare the optimization The optimization process of ARSBX for DNR can be
results of following algorithms presented below. summarized in Figure 3. Noted that the optimization bounds
The algorithm named adaptive rotation-based simulated of controllable variables include the discrete switch numbers
binary crossover (ARSBX) [39] is applied for DNR in of different loops, the reactive power outputs of new energy
this paper. Crossover which is a crucial operation for and EVs. Thus, the initial individuals can be randomly
generating promising offspring solutions in evolutionary generated within the optimization bounds. Besides, the
MOP is greatly emphasized in this algorithm. One of the power balance and bus voltage constraints will be checked
widely used crossover operators, simulates binary crossover based on the power flow calculation for each searching
(SBR), can not solve the rotated Pareto problems. Thus, individual. For each feasible solution, the fitness function
ARSBX improved from SBR is addressed this kind of can be determined according to the line loss and the voltage
problems. deviation. Particularly, the line loss can be obtained through
the difference between input power and output power of all
B. BRIEF DESCRIPTION OF PEM the branches in the distribution network. Besides, the voltage
To deal with the problems in the event of parameters or deviation is also obtained via adding the difference between
objects with uncertainties, some researchers proposed the the voltage of each node and the reference voltage. Finally,
fuzzy parameterized fuzzy soft sets (fpfs-sets) [40]. However, the optimal strategy of DNR can be obtained after multiple
the concept of fpfs-matrices and their basic properties are iterations with the optimization operators of ARSBX.
introduced in [25] recently. The author first describe an
effective algorithm named PEM via the fpfs-matrices in IV. CASE STUDIES
decision-making. In short, the process of PEM can be In this paper, IEEE 33-bus and 69-bus system are applied for
simplified as the following steps. this DNR model to test different algorithms. Two different

10668 VOLUME 11, 2023

R. Wu, S. Liu: Multi-Objective Optimization for DNR With Reactive Power Optimization of New Energy and EVs

FIGURE 4. Process of DNR simplifying in IEEE 33-bus system.

TABLE 2. Switches Contained in Different Loops in 33-Bus System.

Solar irradiation has a great impact on PV systems while the

wind speed plays an essential role in the generation of wind
turbines. Thus, it is also a complex problem to reasonably
point out the specific buses that new energy or EVs connected
to. It should be noted that the EVs have two operating states
(i.e., charge or discharge), which will greatly change the
FIGURE 3. Flow chart of optimization in DNR. active power output of the EV station. However, it can be
seen from Eq. (9) that the sign of active power output has no
systems and their simplified topologies are shown in Figure 2 effect on the regulation range of reactive power output. Thus,
and Figure 4, respectively. In 33-bus system, the topology the connection and detailed parameters of new energy can be
is also formed into 5 loops coincidentally via the simplified obtained from the Appendix.
method demonstrated in Section II. The different switches In this work, the irritation and the population size of
will be evenly distributed into different loops as far as ARSBX are set to be 50 and 50, respectively. The states of the
possible without repeated statistics are shown in Table 2. switches, the reactive power outputs of new energy and EVs
Hence, the changeable switches will be selected in 5 different are taken as the optimization variables, then the optimization
loops as decision variables. The regulation resources include dimensions of the algorithms are equal to be 15 and 25 for
wind turbine, PV system, and EVs. Their reactive power the IEEE 33-bus and 69-bus systems, respectively. Besides,
outputs can be ensured to be the assigned control commands the Newton’s method [44] is applied to load flow calculation
via the reactive power control of grid side and charging pile in the case studies and the presented DNR is done hourly as
inverters [41], [42], [43]. There are 4 PV systems, 3 wind an operating problem.
turbines, and 3 EV power stations in 33-bus system. The
inverters capacity of PV system and EV charging piles are A. DNR ANALYSIS IN DIFFERENT SYSTEM
set as 0.25MW and 0.28 MW, respectively. However, there Table 1 and 2 have shown the included switches in simplified
are 8 PV systems, 6 wind turbines and 6 EV charging and loops for different systems, respectively. The proposed model
replacement power stations in 69-bus system. The inverters allows only one changeable status of switch for each loop.
capacity of PV system are set as 30kW and the capacity of In 33-bus system, the tie switches that consist of switches
EV charging piles are set as 30kW, respectively. 33-37 and sectionalizing switches are all closed before
In general, active power injected in distribution network DNR. Then, the results obtained from different algorithms
from new energy is closely related to nearby climatic factors. have some subtle differences in Table 3. At the same time,

VOLUME 11, 2023 10669

 R. Wu, S. Liu: Multi-Objective Optimization for DNR With Reactive Power Optimization of New Energy and EVs

TABLE 3. DNR solutions of different algorithms in 33-bus system. TABLE 4. DNR solutions of different algorithms in 69-bus system.

the switches 69-73 are closed in initialization in 69-bus

system and three kinds of DNR regulation occur in Table 4.
However, the time cost among three has absolute differences.
ARSBX usually spends 6-7 seconds in optimization while
others spend almost double in 33 and 69-bus system.
ARSBX has extremely high efficiency in DNR for this
model.

B. ANALYSIS OF POWER LOSS

In general, the power loss of a distribution network is
directly influenced by the operation states, such as the load
demand and the network topology. For example, there is
a large difference between the testing IEEE 33-bus system
and the original IEEE 33-bus system. Firstly, The original
IEEE 33-bus system does not include the new energy and
EVs. The testing IEEE 33-bus system is extened based on
the original system, while 4 PV systems, 3 wind turbines,
and 3 EV power stations are connected. Hence, the power
demand or generation of the connected buses will vary
largely, as shown in APPENDIX. Particularly, the active
power demands of the buses with EV power stations (i.e.,
the buses 24, 25, and 32) increase largely. As a result, the
power loss of the testing IEEE 33-bus system easily increases.
Secondly, the original IEEE 33-bus system is a simple radial
network. In contrast, the testing IEEE 33-bus system contains
multiple switches to change the network topology, as shown
in Figure 4. Hence, it also results in a large difference on the
power loss between these two systems with different network
topologies.
One of the objectives, power loss, plays an essential role in
solutions of DNR. It is closely related to the transmission cost FIGURE 5. Active power loss of each branch with ARSBX;(a) IEEE 33-bus
of the distribution network. The result obtained form ARSBX system; (b) IEEE 69-bus system.

shows that the power loss reduces 143.80kW and reactive

power reduces 50.12kVar in 33-bus system in Table 5. The
ratio of power loss reduction reaches more than 70% while the Figure 5 (a) and (b) present the active power loss in
ratio of reactive power loss reduction exceeds 40%. However, each branch, respectively. The red bars in three figures refer
GA, PSO and sequential quadratic programming (SQP) have to the active power loss of different branches before DNR
just a little difference for optimization strategy in line loss. while the green bars represent the branches active power loss
Nevertheless, the reactive power reduction in PSO and SQP with ARSBX after DNR. In 33-bus system, the loss often
seems a few unsatisfactory compared with GA and ARSBX. concentrates in branches 1-9 and 22-30 before DNR and
The reduction did not exceed 10%. the maximum loss reaches the 50kW in branch 2. The loss
In 69-bus system, although the power loss reduction from mainly distributes in branches 1-2 and 18-24 after DNR, yet
different methods reaches more than 50%, ARSBX has great the maximum one just reaches 15kW in branch 2. In 69-bus
advantages compared with other algorithms in Table 6. The system, the total loss will increase because of the increase
reactive power loss reduction in 69-bus is generally lower of lines and loads. Thus, the active power loss of each
than that of 33-bus. For example, the reactive power loss branch will obviously rise before DNR. The loss amasses
reduction of ARSBX is just over 25% while the GA and PSO in branch 72 but the total loss has sharply decline after
are just over 11% and 5%, respectively. DNR.

10670 VOLUME 11, 2023

R. Wu, S. Liu: Multi-Objective Optimization for DNR With Reactive Power Optimization of New Energy and EVs

FIGURE 6. Comparison of voltage profile; (a) IEEE 33- bus system;

(b) IEEE 69-bus system.
FIGURE 7. Voltage deviation ratio of each bus with ARSBX; (a) IEEE 33-
bus system; (b) IEEE 69-bus system.
TABLE 5. Loss generated during transmission in 33-bus system.

TABLE 6. Loss generated during transmission in 69-bus system.

C. IMPOROVEMENT OF VOLTAGE PROFILE TABLE 7. Change of voltage after DNR in 33-bus system.

Table 7 and Figure 6 (a) show that different voltage profile can
be obtained from several methods. In 33-bus system, results
of ARSBX can maintain a relatively stable fluctuation range
except for bus 19-24, yet ARSBX has been terribly improved
compared with other situations. The voltage deviation of
ARSBX has a great decreasing trend, which is 54.8%
less than that before DNR. Besides, Figure 7 (a) also In 69-bus system, the results obtained via three algorithms
reflects the difference between ARSBX and initial situation. are relatively close. Figure 6 (b) shows that ARSBX is

VOLUME 11, 2023 10671

 R. Wu, S. Liu: Multi-Objective Optimization for DNR With Reactive Power Optimization of New Energy and EVs

TABLE 8. Change of voltage after DNR in 69-bus system. TABLE 11. Convergent iteration in each system with different algorithms.

TABLE 12. Cost reduction in each system with different.

TABLE 9. Reactive power output of different algorithms in IEEE 33-bus

system (kVar).

TABLE 10. Reactive power output of different algorithms in IEEE 69-bus

system (kVar).

FIGURE 8. (a) Buses connected with new energy or EV power stations in

33-bus system; (b) Buses connected with new energy or EV power
stations in 69-bus system.

E. ANALYSIS OF COST AND OTHER RESULTS

Financial analysis can be equal to the calculation of active
dominant in bus 10-27 and 61-69. Although the result of power loss in some specific aspects. In this work, the active
initial situation has advantage in bus 40-50, its total voltage power loss calculated above can be seem to be generated
deviation is considered to be the largest. At the same within one hour. The electric energy cost in summer can
time, voltage deviation ratio of initial situation in 69-bus generally be set as 0.1 dollar per kW·h. Therefore, Table 11
system has several steep peak while the ARSBX has the presents the different cost reduction for several algorithms in
relatively smooth broken line. Although the optimization 33 and 69-bus systems. Table 12 shows the cost reduction of
results of ARSBX in the IEEE 69-bus system is somewhat ARSBX is more than the others in same system. In addition,
unsatisfactory compared with other algorithms, it can still ARSBX has the far fewer convergent iteration in optimization
achieve a 49.1% voltage deviation reduction rate in Table 8. among the different algorithms.

D. OPTIMIZATION OF REACTIVE POWER V. CONCLUSION

Tables 9 and 10 present the reactive power optimization In this work, a novel DNR model with new energy and EVs
results in different systems. It can be seen that the optimiza- is proposed for multi-objective optimization, which contains
tion results of reactive power outputs are vastly different the following contributions:
by different methods. It directly reveals that the constructed 1) The constructed DNR can further reduce the power loss
DNR is a complex optimization and the optimization methods and voltage deviation for the distribution network as
easily trap into different local optimums. it can fully exploit the excellent regulation capacities

10672 VOLUME 11, 2023

R. Wu, S. Liu: Multi-Objective Optimization for DNR With Reactive Power Optimization of New Energy and EVs

TABLE 13. Parameters of new energy and EVs in IEEE 33-bus system. APPENDIX
Besides, the connected locations of new energy and EV power
stations are shown in Figure 8. In the IEEE 33-bus system,
three wind turbines, four PV systems, and three EV stations
has been added, while six wind turbines, eight PV systems,
and six EV stations in the IEEE 69-bus system. In the case
studies, the EV power station connected at the 46th bus of the
IEEE 69-bus system is set to be operated at the state of dis-
charge (i.e., V2G), the rest are operated at the state of charge.
At the same time, since the available load or power supply
at each node may be different, the parameter settings of new
energy and EVs that are shown in Tables 13 and 14 also need
to be changed through the model above.
REFERENCES
TABLE 14. Parameters of new energy and EVs in IEEE 69-bus system. [1] J. Wang, W. Wang, H. Wang, and H. Zuo, ‘‘Dynamic reconfiguration of
multiobjective distribution networks considering DG and EVs based on a
novel LDBAS algorithm,’’ IEEE Access, vol. 8, pp. 216873–216893, 2020.
[2] H. Liu, J. Jiang, S. Zhan, X. Yang, D. Yu, and G. Yan, ‘‘Study on energy
management model of integrated new energy-storage-charging system
considering the influence of uncertainties,’’ Fuel, vol. 331, Jan. 2023,
Art. no. 125784.
[3] H. Wei, Y. Zhang, Y. Wang, W. Hua, R. Jing, and Y. Zhou, ‘‘Planning
integrated energy systems coupling V2G as a flexible storage,’’ Energy,
vol. 239, Jan. 2022, Art. no. 122215.
[4] S. S. Trivedi and A. V. Sant, ‘‘Comparative analysis of dual active bridge
DC–DC converter employing Si, siC and GaN MOSFETs for G2V and
V2G operation,’’ Energy Rep., vol. 8, pp. 1011–1019, Nov. 2022.
[5] A. Kavousi-Fard, M. A. Rostami, and T. Niknam, ‘‘Reliability-oriented
reconfiguration of vehicle-to-grid networks,’’ IEEE Trans. Ind. Informat.
vol. 11, no. 3, pp. 682–691, Jun. 2015.
[6] R. Fathi, B. Tousi, and S. Galvani, ‘‘A new approach for optimal allocation
of photovoltaic and wind clean energy resources in distribution networks
with reconfiguration considering uncertainty based on info-gap decision
theory with risk aversion strategy,’’ J. Cleaner Prod., vol. 295, May 2021,
Art. no. 125984.
[7] Q. Hu, S. Bu, and V. Terzija, ‘‘A distributed p and q provision-based
voltage regulation scheme by incentivized EV fleet charging for resistive
distribution networks,’’ IEEE Trans. Transport. Electrific., vol. 7, no. 4,
pp. 2376–2389, Dec. 2021.
of new energy and EVs. As a result, the active power [8] W. Sun, Y. Qiao, and W. Liu, ‘‘Economic scheduling of mobile energy stor-
loss, reactive power loss, and the voltage quality can age in distribution networks based on equivalent reconfiguration method,’’
Sustain. Energy, Grids Netw., vol. 32, Dec. 2022, Art. no. 100879.
be further improved. Particularly, the active power loss [9] B. Xu, G. Zhang, K. Li, B. Li, H. Chi, Y. Yao, and Z. Fan, ‘‘Reactive power
reductions of ARSBX are up to 71.26% and 68.93% optimization of a distribution network with high-penetration of wind and
for the IEEE 33-bus and 69-bus systems, respectively. solar renewable energy and electric vehicles,’’ Protection Control Mod.
Power Syst., vol. 7, no. 1, p. 51, Dec. 2022.
Meanwhile, the reactive power loss reductions are up [10] D. Anteneh, B. Khan, O. P. Mahela, H. H. Alhelou, and J. M. Guerrero,
to 44.88% and 25.54%, respectively. ‘‘Distribution network reliability enhancement and power loss reduction
2) Different from the conventional single-objective opti- by optimal network reconfiguration,’’ Comput. Electr. Eng., vol. 96,
Dec. 2021, Art. no. 107518.
mization algorithms for DNR, the presented ARSBX [11] A. P. Hota, S. Mishra, and D. P. Mishra, ‘‘Active power loss allocation in
can acquired multiple Pareto optimal solutions. Then radial distribution networks with different load models and DGs,’’ Electr.
a proper balance between power loss and voltage Power Syst. Res., vol. 205, Apr. 2022, Art. no. 107764.
deviation can be achieved after selecting a compromise [12] B. Amanulla, S. Chakrabarti, and S. N. Singh, ‘‘Reconfiguration of power
distribution systems considering reliability and power loss,’’ IEEE Trans.
solution from the optimal solutions via PEM. Hence, Power Del., vol. 27, no. 2, pp. 918–926, Apr. 2012.
the operation flexibility of the distribution network can [13] W. Zheng, W. Huang, D. J. Hill, and Y. Hou, ‘‘An adaptive distributionally
be significantly improved. In addition, the convergence robust model for three-phase distribution network reconfiguration,’’ IEEE
Trans. Smart Grid, vol. 12, no. 2, pp. 1224–1237, Mar. 2021.
rate of ARSBX is much smaller than the comparative [14] M. Mahdavi, H. H. Alhelou, A. Bagheri, S. Z. Djokic, and R. A. V. Ramos,
algorithms, which will reduce the optimization time for ‘‘A comprehensive review of metaheuristic methods for the reconfiguration
DNR. of electric power distribution systems and comparison with a novel
approach based on efficient genetic algorithm,’’ IEEE Access, vol. 9,
In the future, the generation randomness of new energy and pp. 122872–122906, 2021.
EVs will be further considered in DNR. To better adapt the [15] H. Hizarci, O. Demirel, and B. E. Turkay, ‘‘Distribution network
reconfiguration using time-varying acceleration coefficient assisted binary
complex random distribution network, our future works will particle swarm optimization,’’ Eng. Sci. Technol., Int. J., vol. 35, Nov. 2022,
focus on the reinforcement learning based DNR. Art. no. 101230.

VOLUME 11, 2023 10673

 R. Wu, S. Liu: Multi-Objective Optimization for DNR With Reactive Power Optimization of New Energy and EVs

[16] U. Raut and S. Mishra, ‘‘An improved sine–cosine algorithm for simul- [36] D. Santos-Martin, S. Arnaltes, and J. L. R. Amenedo, ‘‘Reactive power
taneous network reconfiguration and DG allocation in power distribution capability of doubly fed asynchronous generators,’’ Electr. Power Syst.
systems,’’ Appl. Soft Comput., vol. 92, Jul. 2020, Art. no. 106293. Res., vol. 78, no. 11, pp. 1837–1840, Nov. 2008.
[17] A. Y. Abdelaziz, F. M. Mohamed, S. F. Mekhamer, and M. A. L. [37] M. Habib, A. Gram, A. Harrag, and Q. Wang, ‘‘Optimized management of
Badr, ‘‘Distribution system reconfiguration using a modified Tabu search reactive power reserves of transmission grid-connected photovoltaic plants
algorithm,’’ Electr. Power Syst. Res., vol. 80, no. 8, pp. 943–953, Aug. driven by an IoT solution,’’ Int. J. Electr. Power Energy Syst., vol. 143,
2010. Dec. 2022, Art. no. 108455.
[18] L. A. Gallego, J. M. Lopez-Lezama, and O. G. Carmona, ‘‘A mixed-integer [38] M. J. E. Alam, K. M. Muttaqi, and D. Sutanto, ‘‘Effective utilization of
linear programming model for simultaneous optimal reconfiguration and available PEV battery capacity for mitigation of solar PV impact and
optimal placement of capacitor banks in distribution networks,’’ IEEE grid support with integrated V2G functionality,’’ IEEE Trans. Smart Grid,
Access, vol. 10, pp. 52655–52673, 2022. vol. 7, no. 3, pp. 1562–1571, May 2016.
[19] Q. Chen, W. Wang, H. Wang, J. Wu, X. Li, and J. Lan, ‘‘A social [39] L. Pan, W. Xu, L. Li, C. He, and R. Cheng, ‘‘Adaptive simulated binary
beetle swarm algorithm based on grey target decision-making for a crossover for rotated multi-objective optimization,’’ Swarm Evol. Comput.,
multiobjective distribution network reconfiguration considering partition vol. 60, Feb. 2021, Art. no. 100759.
of time intervals,’’ IEEE Access, vol. 8, pp. 204987–205013, 2020. [40] N. Çağman, F. Çıtak, and S. Enginoğlu, ‘‘Fuzzy parameterized fuzzy soft
[20] E. Azad-Farsani, I. G. Sardou, and S. Abedini, ‘‘Distribution network set theory and its applications,’’ Turkish J. Fuzzy Syst., vol. 1, no. 1,
reconfiguration based on LMP at DG connected busses using game theory pp. 21–35, 2010.
and self-adaptive FWA,’’ Energy, vol. 215, Jan. 2021, Art. no. 119146. [41] M. Rahimi, ‘‘Coordinated control of rotor and grid sides converters in
[21] Z. Li, W. Wu, B. Zhang, and X. Tai, ‘‘Analytical reliability assessment DFIG based wind turbines for providing optimal reactive power support
method for complex distribution networks considering post-fault network and voltage regulation,’’ Sustain. Energy Technol. Assessments, vol. 20,
reconfiguration,’’ IEEE Trans. Power Syst., vol. 35, no. 2, pp. 1457–1467, pp. 47–57, Apr. 2017.
Mar. 2020. [42] W. Liang, Y. Liu, and J. Peng, ‘‘A day and night operational quasi-Z
[22] H. Sekhavatmanesh and R. Cherkaoui, ‘‘A multi-step reconfiguration source multilevel grid-tied PV power system to achieve active and reactive
model for active distribution network restoration integrating DG start-up power control,’’ IEEE Trans. Power Electron., vol. 36, no. 1, pp. 474–492,
sequences,’’ IEEE Trans. Sustain. Energy, vol. 11, no. 4, pp. 2879–2888, Jan. 2021.
Oct. 2020. [43] Y. Bi, C. Wu, T. Zhao, H. Li, J. Xu, G. Shu, and Y. Wang, ‘‘Modified
[23] Y. Liu, J. Li, and L. Wu, ‘‘Coordinated optimal network reconfiguration deadbeat predictive current control method for single-phase AC–DC PFC
and voltage regulator/DER control for unbalanced distribution systems,’’ converter in EV charging system,’’ IEEE Trans. Ind. Electron., vol. 70,
IEEE Trans. Smart Grid, vol. 10, no. 3, pp. 2912–2922, May 2019. no. 1, pp. 286–297, Jan. 2023.
[24] W. Sheng, K.-Y. Liu, S. Cheng, X. Meng, and W. Dai, ‘‘A trust region SQP [44] W. F. Tinney and C. E. Hart, ‘‘Power flow solution by Newton’s method,’’
method for coordinated voltage control in smart distribution grid,’’ IEEE IEEE Trans. Power App. Syst., vol. PAS-86, no. 11, pp. 1449–1460,
Trans. Smart Grid, vol. 7, no. 1, pp. 381–391, Jan. 2016. Nov. 1967.
[25] J. R. Pillai, B. Bak-Jensen, and P. Thogersen, ‘‘Electric vehicles to support
large wind power penetration in future Danish power systems,’’ Presented
at the IEEE Vehicle Power Propuls. Conf., Seoul, South Korea, Oct. 2012.
[26] S. Enginoğlu and N. Çağman, ‘‘Fuzzy parameterized fuzzy soft matrices
and their application in decision-making,’’ TWMS J. Appl. Eng. Math.,
vol. 10, no. 4, pp. 1105–1115, 2020.
[27] M. Jooshaki, S. Karimi-Arpanahi, M. Lehtonen, R. J. Millar, and RENBO WU received the B.Eng. degree in com-
M. Fotuhi-Firuzabad, ‘‘An MILP model for optimal placement of sec-
puter science and technology from the Huazhong
tionalizing switches and tie lines in distribution networks with complex
University of Science and Technology, Wuhan,
topologies,’’ IEEE Trans. Smart Grid, vol. 12, no. 6, pp. 4740–4751,
Nov. 2021. China, in 2006, and the M.Eng. degree in electrical
[28] B. Morsy, A. Hinneck, D. Pozo, and J. Bialek, ‘‘Security constrained OPF engineering from Shandong University, Jinan,
utilizing substation reconfiguration and busbar splitting,’’ Electr. Power China, in 2013, where he is currently pursuing
Syst. Res., vol. 212, Nov. 2022, Art. no. 108507. the Ph.D. degree in advanced manufacturing
[29] K. H. Youssef, ‘‘Optimal routing of ring power distribution systems,’’ with the School of Electrical Engineering. His
Electr. Power Syst. Res., vol. 199, Oct. 2021, Art. no. 107392. major research interests include power system
[30] M. A. Samman, H. Mokhlis, N. N. Mansor, H. Mohamad, H. Suyono, operation and control, operation and management
and N. M. Sapari, ‘‘Fast optimal network reconfiguration with guided of distribution automation systems, and distributed optimization.
initialization based on a simplified network approach,’’ IEEE Access,
vol. 8, pp. 11948–11963, 2020.
[31] R. Pegado, Z. Ñaupari, Y. Molina, and C. Castillo, ‘‘Radial distribution
network reconfiguration for power losses reduction based on improved
selective BPSO,’’ Electr. Power Syst. Res., vol. 169, pp. 206–213,
Apr. 2019.
[32] J.-S. Pan, H.-J. Wang, T.-T. Nguyen, F.-M. Zou, and S.-C. Chu, ‘‘Dynamic SHUQIN LIU received the B.Eng. degree in
reconfiguration of distribution network based on dynamic optimal period radiometry from Shandong University, Jinan,
division and multi-group flight slime mould algorithm,’’ Electr. Power Syst.
China, in 1982, and the M.Eng. degree in industrial
Res., vol. 208, Jul. 2022, Art. no. 107925.
electronic technology and electromagnetic mea-
[33] S. Arulprakasam and S. Muthusamy, ‘‘Reconfiguration of distribution
networks using rain-fall optimization with non-dominated sorting,’’ Appl. surement and the Ph.D. degree in mechanical and
Soft Comput., vol. 115, Jan. 2022, Art. no. 108200. electrical engineering from Xi’an Jiaotong Univer-
[34] T. T. Tran, K. H. Truong, and D. N. Vo, ‘‘Stochastic fractal search algorithm sity, Xi’an, China, in 1990 and 2000, respectively.
for reconfiguration of distribution networks with distributed generations,’’ She is currently a Professor with the School of
Ain Shams Eng. J., vol. 11, no. 2, pp. 389–407, Jun. 2020. Electrical Engineering, Shandong University. Her
[35] S. Lei, Y. Hou, F. Qiu, and J. Yan, ‘‘Identification of critical switches research interests include magnetic suspension
for integrating renewable distributed generation by dynamic network bearing theory and high-speed rotating machinery, vertical axis wind turbine,
reconfiguration,’’ IEEE Trans. Sustain. Energy, vol. 9, no. 1, pp. 420–432, new energy microgrid design, and operating control.
Jan. 2017.

10674 VOLUME 11, 2023