energies

Article
Multi-Objective Optimization and Reconstruction of
Distribution Networks with Distributed Power Sources Based on
an Improved BPSO Algorithm
Dan Lu 1 , Wenfeng Li 1, *, Linjuan Zhang 1 , Qiang Fu 2 , Qingtao Jiao 2 and Kai Wang 2

1 State Grid Henan Economic Research Institute, Zhengzhou 450000, China; [email protected] (D.L.);
[email protected] (L.Z.)
2 NARI-TECH Nanjing Control Systems Co., Ltd., Nanjing 210000, China; [email protected] (Q.F.);
[email protected] (Q.J.); [email protected] (K.W.)
* Correspondence: [email protected]; Tel.: +86-199-0388-9591

Abstract: The continuous integration of distributed power into the distribution network has increased
the complexity of the distribution network and created challenges in distribution-network reconfigu-
ration. In order to make the distribution network operate in the optimal mode, this paper establishes
a multi-objective reconfiguration-optimization model that takes into account active network loss,
voltage offset, number of switching actions and distributed power output. For a distribution network
with a distributed power supply, it is easy for the traditional binary particle swarm optimization
algorithm to fall into a local optimum. In order to improve the convergence speed of the algorithm
and avoid premature convergence, this paper adopts an improved binary particle swarm optimiza-
tion algorithm to solve the problem. The IEEE33 node system is used as an example for simulation
verification. The experimental results show that the algorithm improves the convergence speed and
global search ability, effectively reduces the system network loss, and greatly improves the voltage
level of each node. It improves the stability and economy of distribution-network operation and can
effectively solve the problem of multi-objective reconfiguration.

Citation: Lu, D.; Li, W.; Zhang, L.; Fu,

Keywords: binary particle swarm optimization algorithm; multi-objective reconstruction; distributed
Q.; Jiao, Q.; Wang, K. Multi-Objective
power supply; genetic algorithm
Optimization and Reconstruction of
Distribution Networks with
Distributed Power Sources Based on
an Improved BPSO Algorithm.
Energies 2024, 17, 4877. https://
1. Introduction
doi.org/10.3390/en17194877 The power system is a complex system. In the process of power transmission, the
distribution network plays a vital role as a network for carrying and distributing power [1].
Academic Editors: Isabel Jesus, Yuqing
Therefore, its quality and efficiency are crucial to the entire power system. Through re-
Bao, Zhenya Ji and Zhenyu Lv
configuration of the distribution network, the potential of the distribution system can be
Received: 30 May 2024 brought into play without increasing investment in equipment and the performance index
Revised: 29 June 2024 of the system can be improved, with improvements including reduced active power loss in
Accepted: 26 July 2024 the system, reduced voltage difference, and increased output of distributed power. There-
Published: 28 September 2024 fore, distribution-network reconfiguration has good economic benefits for the operation
and management of modern power systems, and it is also an important means by which to
realize the intelligent and sustainable development of power systems [2].
In recent years, with the development of distributed power sources, more and more
Copyright: © 2024 by the authors.
distributed power sources have been connected to the distribution network, leading to
Licensee MDPI, Basel, Switzerland.
significant changes in the magnitude and direction of power flow in the distribution
This article is an open access article
distributed under the terms and
network [3]. In addition, node voltage and network loss will also be greatly affected.
conditions of the Creative Commons
Therefore, the question of how to reconstruct a distribution network containing distributed
Attribution (CC BY) license (https:// power sources has become a hot topic in current smart-grid research [4].
creativecommons.org/licenses/by/ To date, scholars both domestically and internationally have conducted extensive
4.0/). research on reconstruction methods for the integration of distributed power sources into

Energies 2024, 17, 4877. https://doi.org/10.3390/en17194877 https://www.mdpi.com/journal/energies

Energies 2024, 17, 4877 2 of 12

distribution networks. The main methods used include traditional algorithms and artificial
intelligence algorithms [5]. Due to the unsatisfactory nature of the results from the use of
traditional algorithms in large-scale distribution networks, intelligent algorithms including
the ant colony algorithm [6], genetic algorithm [7], and particle swarm algorithm [8], as well
as various combinations of and improvements on these algorithms are currently commonly
used. In reference [9], minimizing network loss is taken as the objective function and the dis-
tribution network is reconstructed by improving the quantum particle swarm optimization
algorithm. Reference [10] takes active power loss as the objective function and incorporates
fuzzy theory into the group search algorithm to solve the optimization and reconstruction
model of photovoltaic distribution networks. Reference [11] addresses the problem of
slow convergence speed in the cuckoo algorithm by introducing a simulated annealing
operation. Simulation results show that this method can effectively improve convergence
speed and reduce network loss. In Reference [12], a multi-objective mathematical model of
distribution-network reconfiguration that considers the network loss, average safety factor,
and minimum safety factor was established. Through improvements to the traditional
mouse swarm optimization algorithm, the global search efficiency of the algorithm was
improved. In Reference [13], an optimal method of distribution-network reconfiguration
based on hybrid simulated annealing and the cuckoo search algorithm is proposed; this
method effectively reduced the active power loss of a distribution network with distributed
generation and improves the voltage quality. The stability and universality of the algorithm
were verified, so the algorithm provides a reference for the safe and economic operation of
smart grid. In Reference [14], an optimal reconfiguration model of a distribution network
with distributed energy based on an improved wolf pack algorithm is proposed. Through
the introduction of Levy flight and adaptive step-size strategy, the convergence perfor-
mance of the algorithm was improved. This paper analyzes and verifies the effectiveness
and superiority of this method for reducing network loss and improving node voltage. In
Reference [15], through the establishment of a distribution-network reconfiguration model
considering demand response and ‘source-storage-load’ structure, the second-order cone
optimization method was used to deal with uncertain factors, which effectively improved
the voltage quality and operational efficiency of the distribution system over different time
periods. These studies provide a theoretical basis for the reconstruction of distributed
power generation integration into distribution networks.
However, most of the above studies in the literature aim at minimizing the network
loss, and there is a problem in that the objective function is relatively simple and the
above heuristic algorithms risk falling into a local optimal solution. In this context, for
distribution-network reconfiguration, this paper proposes an improved algorithm that
combines the dual advantages of the binary particle swarm optimization algorithm and
the genetic algorithm. By introducing the strategy of the genetic algorithm, the improved
binary particle swarm optimization algorithm can better deal with the constraints in
distribution-network reconfiguration. This approach effectively avoids the generation
of invalid solutions and improves search efficiency. This paper not only considers the
traditional minimization of active network loss, but also includes various factors such as
reducing network loss, reducing voltage offset, reducing the number of switching actions,
and increasing the output of distributed power in the objective function. This approach not
only considers the economy of the system but also considers improvements in the stability
of the distribution network’s operation. In this paper, the effectiveness of the method in
different scenarios is verified by simulation.

2. Mathematical Model for Distribution-Network Reconstruction

2.1. Objective Function
Distribution-network reconstruction is a typical nonlinear programming problem in
power systems. Due to the constant changes in load, the distribution network needs to find
the optimal structure to achieve the best solution for active power loss, node voltage offset,
distributed power-generation active output, and switch-operation-frequency coordination.
Energies 2024, 17, 4877 3 of 12

The main goal of reconstruction is to find the optimal network structure, which is of great
significance for reducing distribution network losses and improving voltage quality and
power-supply reliability.
(1) Considering the economic considerations of the system, this article will minimize
the network loss of the system as one of the objectives, and its objective function expression
is as follows:
m P2 + Q2
Ploss = min ∑ k i Ri i 2 i (1)
i =1 Ui
Among these variables, Ploss represents the total network loss of the system; m repre-
sents the total number of switch branches in the network; k i represents the switch status
of the ith branch, with 0 and 1 representing on and off, respectively; Ri represents the
resistance value of the i-th branch; Pi represents the active power of the i-th branch; Qi
represents the reactive power of the ith branch; and Ui represents the voltage value of the
i-th node.
(2) Voltage deviation is an important indicator of power and voltage quality and is
considered as one of the objective functions in this article. The objective function takes the
deviation between the actual node voltage amplitude and the given voltage amplitude
as the measurement standard, aiming to maintain the node voltage within the allowable
range of the system and minimize voltage deviation. Its expression is as follows:
m
Ui − UiN
∆U = min ∑ (2)
i =1
UiN

Among these variables, ∆U represents the voltage deviation of node i, m is the total
number of nodes in the network, and UiN represents the specified voltage amplitude at
node i.
(3) The objective function for optimizing the total number of switch operations in the
distribution network can be expressed as follows:
m
S = min ∑ Si − Si′ (3)
i =1

In the formula, Si and Si′ represent the switch states of branch i before and after
reconstruction, with 1 indicating that the switch is closed and 0 indicating that the switch
is open.
(4) When restructuring the distribution network, it is necessary to optimize the use
of distributed power sources in the system to maximize the utilization of these resources,
thereby reducing the load and energy costs of the entire system and improving the reliability
and efficiency of the system. Specifically, the objective function can be expressed as follows:

Pmin = −min ∑ PiDG (4)

i =n DG

Among these variables, PiDG represents the active output of distributed power source i.
In order to reduce the subjectivity of artificially fixed weights, we use a random
weight-allocation method to normalize the objective function. Through different weight
ratios, different normalization formulas can be obtained; as a result, the final approximate
solution set is also different. Through this method, the multi-objective function can be
normalized simply and quickly, and for the convex Pareto optimal front-end problem, the
optimal solution can be obtained so as to evaluate the performance of the algorithm more
objectively and fairly. Therefore, the total objective function is expressed as follows:
Energies 2024, 17, 4877 4 of 12



 f ( x ) = min[ω1 f 1 + ω2 f 2 + ω3 f 3 + ω4 f 4 ]

ωi = 4λi

∑ λi
(5)

 i =1

 λi = randomi

Because the dimensions of the four constraint conditions are different, the linear scaling
method is used to normalize each objective function and obtain the normalized objective
function f 1 , f 2 , f 3 , f 4 . The random weight normalization algorithm is used to normalize the
objective function by using the randomly generated weights and the weighted summation
method. On this basis, the multi-objective model is solved. wi is the random weight of
the i-th variable, and λi is the ith random number generated, resulting in a total of four
random weights.

2.2. Constraint Condition

When optimizing and reconstructing a distributed power-distribution-network system,
in addition to considering the optimization objectives, it is also necessary to consider
whether the reconstructed system meets the specified operating conditions, which are
listed below.
(1) Node-voltage constraint:

Ui,min ⩽ Ui ⩽ Ui,max (6)

Here, Ui,min and Ui,max are the minimum and maximum effective voltage values
corresponding to the ith node, respectively. The reconstructed distribution network needs
to meet the requirements of stability and reasonable distribution of voltage to prevent
equipment damage or power-quality problems caused by high or low voltage.
(2) Power-flow-equation constraints:

m
Pi + PDGi = Ui ∑ Uj Gij cos δij + Bij sin δij + PLi


j =1
m (7)
Qi + Q DGi = Ui ∑ Uj Gij sin δij − Bij cos δij + Q Li


j =1

Pi and Qi respectively represent the active and reactive power injected by the system
at node i; PDi and Q Di respectively, represent the active and reactive power injected at
node i of the distributed power-generation DG access; PLi and Q Li represent the active and
reactive power of the load at node i, respectively; Ui and Uj represent the effective voltage
values of nodes i and j, respectively; Gij and Bij are the real and imaginary parts of the node
admittance matrix, respectively; δij represents the voltage phase difference between node i
and node j.
(3) Power constraints on branch lines:

Si ≤ Simax (8)
In the formula, Si represents the actual power flowing through branch i and Simax
represents the maximum power allowed to flow through branch i.
(4) DG-capacity constraints

SDGi ≤ SDGimax (9)

In the formula, SDGi represents the actual power injected by DG into branch i and
SDGimax represents the maximum capacity of DG.
(5) Network-topology constraints
Energies 2024, 17, 4877 5 of 12

gn ∈ Gn (10)
Here, gn is the reconstructed network structure and Gn is the set of all allowed radial
network structures in the distribution-network system. After the reconstruction of the
distribution network is completed, its structure should comply with the operational rules
of the power grid; that is, the topology structure should be radial, acyclic, and islanded.

2.3. Method for Judging the Structure of Distribution Networks

One of the most important constraints in the reconstruction of distribution networks
is that the power grid structure must not have islands or loops; that is, its radial structure
must operate, in a manner similar to the tree structure in graph theory. A tree is a special
structure in which any two points are connected and there is no loop between them.
When all switches in the power grid are closed, the system structure forms a connected
graph, denoted as G, and the radial distribution-network structure is a spanning tree of
the connected graph G. A connected graph contains n nodes in the graph, but only (n − 1)
edges are needed to form a tree. If the number of branches in the graph is more than (n − 1),
it indicates the existence of a ring network. If it is less than (n − 1), there must be solitons
in the graph. To carry out distribution-network reconstruction, it is necessary to ensure
that the distribution-network structure is radial.
In the reconstruction of distribution networks, the structure of the distribution net-
work is often complex, with a large number of switches present [12]. In the process of
algorithm optimization, many solutions that do not meet the structural requirements will be
generated. Judging the structure of these solutions is beneficial for improving the efficiency
of algorithm processing.
To avoid the occurrence of a ring-network structure, it is assumed that all switches in
the distribution network are closed, forming some circuits, each with only one switch open.
In order to avoid islanding, at most one switch is disconnected in the common branch of
the loop and at most n-1 switches are disconnected around nodes where n loops converge.
This ensures that the reconstructed distribution network maintains a radial topology.

3. Algorithm Research
3.1. Algorithm Selection
In the reconfiguration of distribution networks in modern power systems, multi-
objective optimization plays a vital role. In a difference from traditional single-objective
optimization, multi-objective optimization needs to consider multiple performance indi-
cators, such as reducing network loss, reducing voltage offset, reducing the number of
switching actions, and increasing the output of distributed power supply, at the same
time. There are often conflicts among these goals. For example, reducing network loss may
increase the number of switching actions, and increasing the output of distributed power
may affect the voltage quality. Therefore, finding a balance point to achieve the coordinated
optimization of these goals is a major challenge in distribution-network reconfiguration.
In order to solve the above multi-objective optimization problem, this paper uses an
improved binary particle swarm optimization algorithm (GA-BPSO). Binary particle swarm
optimization (BPSO) is an optimization algorithm based on swarm intelligence. It searches
for the optimal solution by simulating the social behavior of birds. The BPSO algorithm
shows good performance with optimization problems in discrete space, but it may fall
into local optimal solutions when dealing with complex constraints and multi-objective
problems. The reasons for choosing GA-BPSO as a multi-objective optimization method in
this paper follow.
By combining the crossover and mutation operations of the genetic algorithm (GA),
GA-BPSO enhances the global search ability of the algorithm and helps it jump out of
the local optimal solution, which is particularly important for multi-objective problems.
There is a large number of constraints in the distribution-network reconfiguration prob-
lem. GA-BPSO can effectively deal with these constraints and avoid the generation of
Energies 2024, 17, 4877 6 of 12

invalid solutions. In this paper, the objective function is normalized by random weight
distribution such that the contribution weights of different objective functions to the opti-
mization results are evenly distributed, so as to evaluate the performance of the algorithm
more objectively and fairly. The GA-BPSO algorithm has a fast convergence speed while
maintaining search accuracy, which is particularly important for large-scale power-system
optimization problems.

3.2. Binary Particle Swarm Optimization Algorithm

Traditional particle swarm optimization (PSO) achieves iterative updates by updating
the velocities and positions of particles [16]. The inspiration for this algorithm comes from
bird swarm behavior, which combines the strategies of individual birds with those of other
birds to ultimately search for food in a collective manner. The PSO algorithm exhibits highly
efficient performance in optimization problems in continuous spatial domains. When it is
necessary to optimize a problem in a discrete spatial domain, the standard PSO algorithm
is no longer applicable [17]. To address this issue, Kennedy et al. proposed the binary
particle swarm optimization (BPSO) algorithm in 1997 [18]. The BPSO algorithm uses
binary encoding to represent the solution space in the search space, and its speed and
position updates are done in the same way as in standard PSO algorithms. Due to the
fact that the variable switch-state variables in the optimization of distribution-network
reconstruction are a set of discrete variables, it is more suitable to use the BPSO algorithm.
The formula for updating the velocities and positions of its particles is as follows:

vi,j (k + 1) = w · v
i,j (k) + c1 · rand1 ·

pi,j (k ) − xi,j (k ) + c2 ∗ rand2 ∗ g j (k ) − xi,j (k )
 (11)
1, rand < sigmoid(vi (k + 1))
xi,j (k + 1) =
0, rand ≥ sigmoid(vi (k + 1))

In the formula, the j-th dimensional velocity of the i-th particle at the k-th iteration
is vi,j , where xi,j (k) represents the j-th dimensional position of the i-th particle at the k-th
iteration. The individual’s historical optimal position is pi,j (k); the global optimal position
is g j (k); ω is the inertia weight; and c1 and c2 are the adaptive acceleration constants. The
values of particle position and velocity are represented by converting real numbers to
binary encoding, so particles need to first perform binary encoding and decoding when
updating their position and velocity. That is, each particle uses a binary string to represent
its position and velocity, and some of the operations in the update equation are performed
in binary space.

3.3. Improved Binary Particle Swarm Optimization Algorithm

In dealing with the problem of reconstruction of a power-distribution network, the
BPSO algorithm has better search and convergence capabilities compared to the PSO algo-
rithm. However, there are still many problems in practical applications, one of which is that
the network structure needs to meet a large number of constraint conditions, which results
in a large number of invalid solutions arising during the search process and seriously
affects search efficiency. Another is that the algorithm easily falls into local optima in
distribution-network-reconstruction problems. To address these issues, this study intro-
duces the concepts of “inheritance” and “mutation” from the genetic algorithm (GA) into
the algorithm. The position of attractor p(k) will inherit the optimal individual solution
P(k) and the global optimal solution G(k) with a certain probability [19], and non-optimal
genes will be mutated with a certain probability. Therefore, the iterative process can be
described in the following form:
Energies 2024, 17, 4877 7 of 12

vi,j (k + 1) = ω · vi,j (k) + rand1 · [ pi,j (k) − xi,j (k)]+

rand2 · [ g j (k ) − xi,j (k )] + F1 · [ Pi,j (k) − xi,j (k)]+
F2 · [Vi,j (k) − xi,j (k)] (12)

1, Pi,j (k) + Qi,j (k) < rand
xi,j (k + 1) =
0, Pi,j (k) + Qi,j (k) ≥ rand

Here, F1 and F2 are the “genetic” and “mutation” factors of the genetic algorithm;
Pi,j (k) is the best position of the ith particle in the individual historical optimum at the kth
iteration; Vi,j(k) is the velocity corresponding to the global historical optimum at the kth
iteration; and Qi,j (k) is the mutation value of the ith particle at the kth iteration.
The specific process of GA-BPSO algorithm is shown in the figure.
(1) To initialize the algorithm model, this article considers the topology constraints of the
distribution network when initializing and updating particles, which can narrow the
particle search range and enhance the algorithm’s convergence ability.
(2) To initialize the population, note that in GA-BPSO, the population is composed
of particles from BPSO. First, randomly select some seed points to initialize the
population, with each particle representing a potential solution.
(3) When defining a fitness function, the objective function of distribution-network recon-
struction should be used as the main reference indicator to evaluate the power system.
(4) In the selection operation, use a fitness function to evaluate the fitness of each particle,
select the optimal particle, and make a selection based on the fitness.
(5) In the genetic and mutation operations, to increase randomness and assist in the global
search, mutation operations are performed on certain particles in a new population.
(6) In the particle swarm update, based on the characteristics of the BPSO algorithm
in the population, update the positions and velocities of particles in the population
according to the strategy of the BPSO algorithm.
(7) If the constraint is satisfied, then determine whether the termination condition is
satisfied; if not, repeat steps 3–6 until the stopping condition is reached.

4. Simulation Analysis
To verify the performance of the GA-BPSO algorithm applied in the optimization
and reconstruction of distribution networks, an IEEE33 node distribution-network system
was used as an example, and simulation analysis based on the GA-BPSO algorithm in
Figure 1 was conducted using MATLAB. The original topology structure of the IEEE33 node
distribution network system is shown in Figure 2, which includes 33 nodes, 32 segmented
switches (all closed), and 5 tie switches (all open). The system reference voltage is 12.66
kV, the active load is 3.72 MW, the reactive load is 2.32 Mvar, and the reference capacity
is set to 10 MVA. The initial population size of the GA-BPSO algorithm is set to 50, and
the maximum number of iterations is 40. The convergence accuracy is 10−6 . In order to
verify the performance of the proposed method in distribution-network reconstruction,
this paper conducts simulations under two different application scenarios: one considering
distributed power sources and one not considering distributed power sources.

4.1. Improved Binary Particle Swarm Optimization Algorithm

This scenario involves an IEEE33 node distribution network without distributed
power sources. We applied this method to the reconstruction problem of this type of
distribution network, and the reconstructed system-network diagram is shown in Figure 3.
The disconnect switches for the pre-refactoring system were 33, 34, 35, 36, and 37, while the
disconnect switches for the post-refactoring system were 7, 9, 14, 32, and 37.
 kV, the active load is 3.72 MW, the reactive load is 2.32 Mvar, and the reference capacity
kV, thetoactive
is set loadThe
10 MVA. is 3.72 MW,
initial the reactive
population load
size is 2.32
of the Mvar, and
GA-BPSO the reference
algorithm is set tocapacity
50, and
isthe
setmaximum
to 10 MVA. number of iterations is 40. The convergence accuracy is 10set
The initial population size of the GA-BPSO algorithm is to order
−6. In 50, andto
the maximum
verify number ofofiterations
the performance is 40.method
the proposed The convergence accuracy is 10reconstruction,
in distribution-network −6. In order to

verify the performance

this paper of the proposed
conducts simulations method
under two in distribution-network
different reconstruction,
application scenarios: one consider-
Energies 2024, 17, 4877 8 of 12
this paper conducts simulations under two different application scenarios:
ing distributed power sources and one not considering distributed power sources. one consider-
ing distributed power sources and one not considering distributed power sources.
Start
Start
Input distribution
Input distribution
network information
network information
Initialize algorithm Distribution
Initialize algorithm
model Distribution
network constraints
model network constraints
Generate initial
Generate initial
population
population
Calculate and evaluate
Calculate andof
the fitness evaluate
each
the fitness
particleeach
of
particle
Update individual optimal
Update
positionindividual
and globaloptimal
optimal
position
positionand global
based on optimal
fitness
position based on fitness
Genetics, variation, updating
Genetics, variation,
particles updating
particles
Topology
Topology Power
discrimination
discrimination Power
flow calculation
flow calculation
N
Satisfy constraint N
Satisfy constraint
conditions
conditions
Y
Y
N Meet termination
N Meetconditions
termination
conditions
Y
Y
Output results
Output results
Figure 1. Flowchart of distribution-network-reconfiguration optimization based on the GA-BPSO
Figure
Figure1.1.Flowchart
algorithm. Flowchart of distribution-network-reconfiguration
of distribution-network-reconfiguration optimization
optimization based based on the GA-BPSO
on the GA-BPSO algorithm.
algorithm.
Energies 2024, 17, x FOR PEER REVIEW 9 of 13
23 24 25 37
23 24 25 37
26 27 28 29 30 31 32 33 36
Loop 26 27 28 29 30 31 32 33 36
Loop 34
2 Loop
4.1. Improved Binary Particle
Loop 4
Loop
Swarm Optimization
5
Algorithm
2 Loop 34
1 2 3 4 5 6 7 8 9 10 115 12 13 14 15 16 17 18
4
This scenario involves an IEEE33 node distribution network without distributed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Loop Loop
power sources.
1
We 33
applied this 3
method to the reconstruction problem of this type of dis-
Loop Loop 35
tribution 1network,33and the reconstructed
3 system-network diagram is shown in Figure 3.
19 20 21 22 35
The19disconnect
20 21 22
switches for the pre-refactoring system were 33, 34, 35, 36, and 37, while
the disconnect
Figure 2. switches
Topology for
diagram the
of the post-refactoring
IEEE33 node system
distribution were 7,
system
Figure 2. Topology diagram of the IEEE33 node distribution system 9, 14,reconfiguration.
before
before 32, and 37.
reconfiguration.
Figure 2. Topology diagram of the IEEE33 node distribution system before reconfiguration.
23 24 25
26 27 28 29 30 31 32 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

19 20 21 22

ReconstructedIEEE33
Figure3.3.Reconstructed
Figure IEEE33node
nodedistribution
distributionsystem
systemtopology
topologydiagram.
diagram.

In order to better verify the role of the algorithm in the distribution-network recon-
figuration, the GA-BPSO algorithm is compared with the ant colony algorithm, the PSO
algorithm, and the BPSO algorithm. It can be seen from Figure 4 that the performance of
the GA-BPSO algorithm is better than those of the ant colony algorithm, the particle
swarm algorithm and the binary particle swarm algorithm. The ant colony algorithm, the
 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

19 20 21 22
Energies 2024, 17, 4877 9 of 12
Figure 3. Reconstructed IEEE33 node distribution system topology diagram.

In
In order
order to
to better
better verify
verify the role of thethe algorithm
algorithm in the distribution-network
distribution-network recon-
figuration,
figuration, the GA-BPSO
GA-BPSO algorithm is compared with the the ant
ant colony
colony algorithm,
algorithm, thethe PSO
PSO
algorithm,
algorithm, and the BPSO BPSO algorithm.
algorithm.ItItcan canbebeseen
seenfrom
fromFigure
Figure4 that
4 that
thethe performance
performance of
of
thethe GA-BPSO
GA-BPSO algorithm
algorithm is isbetter
betterthan
thanthose
thoseofofthe
the ant
ant colony
colony algorithm, the particle
particle
swarm
swarm algorithm
algorithm andand the
the binary
binary particle
particle swarm
swarm algorithm.
algorithm. The ant colony algorithm,
algorithm, the
the
particle
particle swarm algorithm and the binary particle swarm algorithm all converge to
swarm algorithm and the binary particle swarm algorithm all converge to the
the local
local
optimal
optimal solution
solution and
and do
do not
not find
find the
the global
global optimal
optimal solution.
solution. The
The GA-BPSO
GA-BPSO algorithm
algorithm hashas
lower
lower fitness
fitness and
and faster
faster convergence
convergence speed.speed. The
The particle
particle swarm
swarm optimization
optimization algorithm
algorithm
converges
converges after
after 28
28 iterations,
iterations, the
the binary
binary particle
particle swarm
swarm optimization
optimization algorithm
algorithm converges
converges
after 19 iterations, and the improved binary algorithm converges after
after 19 iterations, and the improved binary algorithm converges after 5 iterations. 5 iterations.

147
ACO
PSO
146 BPSO
GA- BPSO

145
Active network loss/kW

144

143

142

141

140

139
0 5 10 15 20 25 30 35 40 45 50
Iterations

Figure 4.
Figure 4. Voltage curves
Voltage before
curves and and
before afterafter
reconstruction without
reconstruction considering
without distributed
considering power
distributed
sources.
power sources.

Table
Table 11 lists
lists the
the reconstruction
reconstruction results
results of
of the
the distribution
distribution network
network before
before and
and after
after
reconstruction,
reconstruction, where
where thethe reconstructed
reconstructed data
data were obtained by using GA-BPSO for recon-
struction. According to
struction. According to the
the table,
table, the
the network
network lossloss after reconstruction
reconstruction is 139.4731
139.4731 kW,
which
which isis 31.4%
31.4% lower
lower than
than the
the 202.6471
202.6471 kW
kW from
from before
before the
the reconstruction.
reconstruction. The minimum
minimum
node
node voltage
voltage increased
increased from
from 0.9131
0.9131 p.u.
p.u. to
to 0.9378
0.9378 p.u.,
p.u., an
an increase
increase of
of 2.63%.
2.63%. The
The voltage
voltage
offset
offset decreased
decreased from
from 0.13438
0.13438 p.u.
p.u. to 0.08503
0.08503 p.u.,
p.u., aa decrease
decrease ofof 36.72%.
36.72%. Figure
Figure 55 shows
shows
the
the voltage
voltage curves
curvesofofeacheachnode
nodebefore and
before after
and reconstruction,
after reconstruction,indicating a certain
indicating im-
a certain
provement in node voltage. This indicates that using the method proposed in this article
for distribution-network reconstruction can effectively reduce network losses and provide
some support for node voltages. At the same time, the GA-BPSO algorithm used in this
article has better search ability and convergence speed.

Table 1. Reconstruction results of a distribution network without distributed power sources.

Disconnect Branch Active Network Minimum Node Voltage Switching

Combination Loss/kW Voltage/(p.u.) Offset/(p.u.) Frequency
Pre-reconfiguration 33, 34, 35, 36, 37 202.6471 0.9131 0.13438 \
After reconfiguration 7, 9, 14, 32, 37 139.4731 0.9378 0.08503 8
 Table 1. Reconstruction results of a distribution network without distributed power sources.
Table 1. Reconstruction results of a distribution network without distributed power sources.
Disconnect Branch Active Net- Minimum Node Voltage Switching
Disconnect
Combination Branch work
Active Net-
Loss/kW Minimum Node
Voltage/(p.u.) Voltage
Offset/(p.u.) Switching
Frequency
Pre-reconfiguration 33,Combination
34, 35, 36, 37 work Loss/kW
202.6471 Voltage/(p.u.)
0.9131 Offset/(p.u.)
0.13438 Frequency
\
Energies 2024, 17, 4877
Pre-reconfiguration 33, 34, 35, 36, 37 202.6471 0.9131 0.13438
After reconfiguration 7, 9, 14, 32, 37 139.4731 0.9378 0.08503 8\ 10 of 12
After reconfiguration 7, 9, 14, 32, 37 139.4731 0.9378 0.08503 8
1
1 After
Before
After
0.99
Before
0.99
0.98
0.98
0.97

Voltage amplitude/p.u.
0.97

Voltage amplitude/p.u.
0.96
0.96
0.95
0.95
0.94
0.94
0.93
0.93
0.92
0.92
0.91
0
0.91 5 10 15 20 25 30 35
0 5 10 15Node 20 25 30 35
Node
Figure 5. Voltage profiles before and after reconfiguration without considering distributed power
Figure 5.
Figure
sources. 5. Voltage
Voltageprofiles before
profiles andand
before afterafter
reconfiguration without
reconfiguration considering
without distributed
considering power
distributed
sources.
power sources.
4.2. Reconstruction with Distributed Power Sources
4.2.
4.2. Reconstruction
Reconstruction withwith Distributed
Distributed Power
Power Sources
Sources
To verify whether the algorithm proposed
To verify whether the algorithm proposed in this in article
this article is applicable
is applicable to the
to the optimiza-
To verify
optimization and whether the algorithm
reconstruction of proposed
distribution in this
networks article isdistributed
containing applicablepower
to the
tion and reconstruction of distribution networks containing distributed power sources, five
optimization
sources, fivepowerand reconstruction
distributed power of distribution networksthe containing distributed power
distributed sources weresources
connectedwere connected
to the system,to system,
with access with
nodes access nodes
including node
sources,
including five
node distributed
6, node power
13, node sources
18, node were connected to the system, with access nodes
6, node 13, node 18, node 22, and node22, 25.and node
The 25. Thecapacity
installed installedandcapacity
activeand active
output of
including
output of 5node 6, node 13,
distributed nodesources
power 18, nodewere
22, and node
400300
kVA 25.
andThe installed
300 kW. capacity
During and active
simulation
5 distributed power sources were 400 kVA and kW. During simulation analysis, the
outputnode
analysis,
power of
the5power
distributed
was treated power
node was
as PQ sources
a treated
node a were
asfor PQ 400for
node
processing.kVAWeandused300
processing. kW.
theWe During
used
algorithm simulation
theproposed
algorithm in
analysis,
proposed the
in power
this node
article to was treated
optimize and as a PQ
reconstructnode for
the processing.
distribution We used
system
this article to optimize and reconstruct the distribution system and obtain the reconstructed the
and algorithm
obtain the
proposedstructure,
reconstructed
topology in this article
topology to optimize
structure,
as shown as and 6.reconstruct
shown
in Figure in Figurethe6. distribution system and obtain the
reconstructed topology structure, as shown in Figure 6.
23 24 25
23 24 25
26 27 28 29 30 31 32 33
26 27 28 29 30 31 32 33

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

19 20 21 22
19 20 21 22
Topologydiagram
Figure6.6.Topology
Figure diagramof
ofthe
thereconfigured
reconfiguredIEEE33
IEEE33node
nodedistribution
distributionsystem.
system.
Figure 6. Topology diagram of the reconfigured IEEE33 node distribution system.
Therelevant
The relevantparameters
parametersofofthe
thedistribution
distributionnetwork
networkbefore
beforeand
andafter
afterreconstruction
reconstruction
with The relevant
distributed parameters
power of
sources the
are distribution
shown in network
Table 2 andbefore and
Figure 7.
with distributed power sources are shown in Table 2 and Figure 7. The output after
The reconstruction
output ofoffive
five
distributed powerpower
with distributed sources are asare
sources follows:
showntheinDG output
Table 2 andatFigure
node 67.isThe
91.0output
kvar, the
of DG
five
output at node 13 is 104.6 kvar, the DG output at node 18 is 62.7 kvar, the DG output at
node 22 is 102.9 kvar, and the DG output at node 25 is 103.6 kvar. Based on Tables 1 and 2,
it can be seen that after the IEEE33 node is connected to the distributed power source, the
active power loss decreases from 202.6471 kW to 114.3275 kW, and the minimum node
voltage increases from 0.9131 p.u. to 0.9364 p.u. After the method described in this article
was used to reconstruct and optimize the distribution network system, the active power
loss decreased to 51.5273 kW, a decrease of 74.57%, the lowest node voltage increased to
0.9461 p.u., an increase of 3.61%, and the voltage offset decreased to 0.07034, a decrease of
47.66%. From our simulation results, it can be seen that although the voltage levels of each
node have been significantly improved after the introduction of distributed power sources,
the improvement in network losses is not significant. After using the method described
in this article for reconstruction, the node voltage level has been significantly improved,
 voltage increases from 0.9131 p.u. to 0.9364 p.u. After the method described in this article
was used to reconstruct and optimize the distribution network system, the active power
loss decreased to 51.5273 kW, a decrease of 74.57%, the lowest node voltage increased to
0.9461 p.u., an increase of 3.61%, and the voltage offset decreased to 0.07034, a decrease of
Energies 2024, 17, 4877 47.66%. From our simulation results, it can be seen that although the voltage levels of 11 each
of 12
node have been significantly improved after the introduction of distributed power
sources, the improvement in network losses is not significant. After using the method
described in this
and the active article
power lossfor
hasreconstruction, thereduced.
also been further node voltage level has that
This indicates beenour
significantly
proposed
improved, and the active power loss has also been further reduced. This indicates
algorithm can effectively optimize and reconstruct the distribution network, improving that our
its
proposed algorithm can
efficiency and performance. effectively optimize and reconstruct the distribution network,
improving its efficiency and performance.
Table 2. Distribution-grid-reconfiguration results with distributed power sources.
Table 2. Distribution-grid-reconfiguration results with distributed power sources.
Disconnect Branch Active Network Minimum Node Voltage Switching
Disconnect Branch Active Network Minimum Node Voltage Switching
Combination Loss/kW Voltage/(p.u.) Offset/(p.u.) Frequency
Combination Loss/kW Voltage/(p.u.) Offset/(p.u.) Frequency
Pre-reconfiguration 34, 35, 36, 37, 38 114.3275 0.9321 0.11762 \
Pre-reconfiguration
After reconfiguration 34, 35,8,36,
14,37,
33, 38
34, 35 114.3275
51.5273 0.9321
0.9461 0.11762
0.07034 6\
After reconfiguration 8, 14, 33, 34, 35 51.5273 0.9461 0.07034 6

1
Before(not include DG)
Before(include DG)
0.99 After(include DG)

0.98

0.97
Voltage amplitude/p.u.

0.96

0.95

0.94

0.93

0.92

0.91
0 5 10 15 20 25 30 35
Node

Figure 7.
Figure 7. Voltage
Voltageprofiles
profilesbefore
beforeand
andafter
afterreconfiguration,
reconfiguration,considering
consideringdistributed
distributedpower
powersources.
sources.

5.
5. Conclusions
Conclusions
In
In this
this paper,
paper, an
an improved
improved binary
binary particle
particle swarm
swarm optimization
optimization algorithm
algorithm (GA-BPSO)
(GA-BPSO)
is proposed to study the multi-objective optimization reconfiguration
is proposed to study the multi-objective optimization reconfiguration problem of a distri-
problem of a
bution network with distributed generation. It was found that the algorithm effectively
distribution network with distributed generation. It was found that the algorithm
improves the global search ability and convergence speed by including the strategy of the
effectively improves the global search ability and convergence speed by including the
genetic algorithm and reduces the risk of falling into a local optimum. Through the simula-
strategy of the genetic algorithm and reduces the risk of falling into a local optimum.
tion of the IEEE33 node system, it was proved that GA-BPSO algorithm has a significant
Through the simulation of the IEEE33 node system, it was proved that GA-BPSO
effect in reducing network loss, improving voltage level and optimizing distributed power
algorithm has a significant effect in reducing network loss, improving voltage level and
output. Based on these findings, we can infer that the algorithm is universal and may
optimizing distributed power output. Based on these findings, we can infer that the
be applicable to larger scales or different types of distribution-network reconfiguration
problems. However, in order to further promote the findings of this study, future research
should focus on dynamic reconfiguration, consider the volatility of load and distributed
power output, evaluate the robustness of the algorithm, and apply it to the actual power
grid. In addition, research should also consider environmental impact and sustainability,
as well as integrated applications with other smart-grid technologies to achieve more
comprehensive grid optimization.

Author Contributions: Conceptualization, D.L. and W.L.; methodology, L.Z. and W.L.; validation,
Q.F., Q.J. and K.W.; formal analysis, D.L.; investigation, W.L.; resources, W.L.; data curation, D.L.;
writing—original draft preparation, D.L.; writing—review and editing, W.L.; project administra-
tion, D.L.; funding acquisition, W.L. All authors have read and agreed to the published version of
the manuscript.
Energies 2024, 17, 4877 12 of 12

Funding: This research was funded by the State Grid Henan Economic Research Institute Technology
Project (Nos. SGHAYJ00GHJS2310084).
Data Availability Statement: The raw data supporting the conclusions of this article will be made
available by the authors on request.
Acknowledgments: This research is gratefully acknowledged with technical support and assistance
from State Grid Henan Economic Research Institute.
Conflicts of Interest: Author Dan Lu was employed by the company State Grid Henan Economic
Research Institute. The remaining authors declare that the research was conducted in the absence of
any commercial or financial relationships that could be construed as a potential conflict of interest.

References
1. Wang, Y.; Sun, M. Research on Comprehensive Optimization Planning of Distribution Networks with Microgrids. China Electr.
Power 2018, 51, 118–124. [CrossRef]
2. Liu, D.; Li, Q. Research on multi-objective optimization and reconstruction of distribution networks with distributed power
sources. Electr. Meas. Instrum. 2020, 57, 92–98.
3. Gallego Pareja, L.A.; López-Lezama, J.M.; Gómez Carmona, O. A Mixed-Integer Linear Programming Model for the Simultaneous
Optimal Distribution Network Reconfiguration and Optimal Placement of Distributed Generation. Energies 2022, 15, 3063.
[CrossRef]
4. Xu, Y. Application of Improved Particle Swarm Optimization Algorithm in Optimization and Reconstruction of Distribution
Networks with Distributed Power Sources. Electr. Meas. Instrum. 2021, 58, 98–104.
5. Liu, G.; Ma, N.; Deng, H.; Cao, Y.; Liu, Z. Assessment of New Energy Supportability of Distribution Network Considering
Demand Side Management and Network Reconfiguration. Electr. Power 2023, 56, 186–192.
6. Sekhavatmanesh, H.; Cherkaoui, R. Distribution network restoration in a multi-agent framework using a convex OPF model.
IEEE Trans. Smart Grid 2019, 10, 2618–2628. [CrossRef]
7. Li, Z.; Wu, L.; Xu, Y.; Moazeni, S.; Tang, Z. Multi-stage real-time operation of a multi-energy microgrid with electrical and thermal
energy storage assets: A data-driven MPC-ADP approach. IEEE Trans. Smart Grid 2021, 13, 213–226. [CrossRef]
8. Liu, H.J.; Cheng, L.Y.; Huang, J.G.; Wu, W.; Guan, X.; Huang, T.X. Islanding of multi period active distribution network
considering intermittent DG and load response. Electr. Power Constr. 2018, 39, 50–57.
9. Pan, H.; Yang, L.; Hu, G. Optimization and Reconstruction of Distribution Network Based on Improved Quantum Particle Swarm
Optimization Algorithm. Electr. Meas. Instrum. 2018, 55, 31–36+49.
10. Tang, Y.; Wu, Z.; Gu, W.; Yu, P.; Du, J.; Luo, X. Research on active distribution network fault recovery strategy based on unified
model considering reconstruction and island partition. Power Syst. Technol. 2020, 44, 2731–2740.
11. Wu, Y.; Cheng, X.; Liu, J. Optimization and Reconstruction of Distribution Networks with Distributed Power Sources Based on
SA-CS Algorithm. Sci. Technol. Eng. 2023, 23, 626–632.
12. Kai, W.; Jing, C. Distribution network reconfiguration with distributed generation based on improved rat swarm optimization.
Sci. Technol. Eng. 2024, 24, 1075–1082.
13. Wu, Y.; Cheng, X.; Liu, J. Distribution network reconfiguration based on hybrid simulated annealing cuckoo algorithm. Sci.
Technol. Eng. 2023, 23, 626–632.
14. Hu, S.; Yang, S.; Xu, K.; Qiu, R.; Liu, C.; Liu, H. Optimal reconfiguration of distribution network with nistributed energy based on
improved wolf pack algorithm. Heilongjiang Electr. Power 2024, 46, 45–49.
15. Pan, K.; Zhao, R.; Liu, S.; Lu, J.; Liu, H.; Chen, Y. Distribution network reconfiguration considering power load demand and
distributed generation uncertainty. Renew. Energy 2024, 4, 546–552.
16. Zhang, X.; Cheng, X.; Zeng, J.; Li, J.; Cao, J. Micro grid islanding reconstruction based on dynamic programming and optimal
flow patterns. Power Grid Technol. 2020, 44, 4608–4616.
17. Song, Z.; Zhang, Q.; Liu, X.; Liu, X. Multi-objective reconstruction of distribution network based on improved particle swarm
optimization algorithm. Intell. Comput. Appl. 2020, 10, 77–79.
18. Liu, J.C.; Liu, S.L. Optimal distributed generation allocation in distribution network based on second order conic relaxation and
Big-M method. Power Syst. Technol. 2018, 42, 2604–2611.
19. Chen, K.; Pan, L. Optimization and Reconstruction Design of Distribution Network with Distributed Power Sources Based on
GA-QPSO Algorithm. Lab. Res. Explor. 2022, 41, 111–115+120.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.