Alexandria Engineering Journal 112 (2025) 569–584

Contents lists available at ScienceDirect

Alexandria Engineering Journal

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

Original article

Coordinated allocation of distributed generation, soft open points and

energy storage systems in unbalanced distribution networks considering
flexibility deficiency risk
Jie Wang a,* , Xiaochao Fan a, Lijun Xu a , Ruijing Shi a , Weiqing Wang b
a
Department of Energy Engineering, Xinjiang Institute of Engineering, Urumqi, Xinjiang 830023, China
b
Engineering Research Center of Education Ministry for Renewable Energy Power Generation and Grid-connected Control, Xinjiang University, Urumqi, Xinjiang 830047,
China

A R T I C L E I N F O A B S T R A C T

Keywords: Optimizing the location and capacity of flexible resources is critical. The increase in the permeability of
Unbalanced distribution networks distributed generation (DG) causes the net load to drastically fluctuate, which sharply increases flexibility de-
Coordinated planning mand for the distribution network. The asymmetric access of distributed generators, asymmetric line parameters,
Distributed generation
and unbalanced loads aggravate the three-phase voltage unbalance of distribution systems. Energy storage
Soft open points (SOPs)
Energy storage systems (ESSs)
systems (ESSs) and soft open points (SOPs) can flexibly adjust power flow direction, which can address the
Flexibility deficiency risk problem. This study analyzed the characteristics of flexibility resources based on the flexibility theory.
Furthermore, the potential loss of flexibility demand is quantitatively assessed by conditional value-at-risk, and
the coordination planning model of DG, SOP, and ESS for the unbalanced distribution network is constructed
considering flexibility and economy. Because the resulting mathematical formulation belongs to a nonconvex
quadratic programming problem, an adaptive linear relaxation programming method is proposed to improve
computational efficiency. The adaptive linear relaxation technique is constructed based on the properties of
quadratic functions and the mean value theorem, and the non-convex quadratic programming problem is con-
verted to a series of parametric linear relaxation programming problems by the technique. Finally, the effec-
tiveness of the proposed model is verified on IEEE 33-bus and IEEE 123-bus systems.

1. Introduction Asymmetric line parameters, unbalanced load, and DG lead to a

three-phase voltage unbalance in the distribution network, which not
Renewable energy sources, such as wind power and photovoltaic, only increases active power loss but also affects the normal operation of
exhibit zero carbon emission and have zero marginal cost [1]. However, power equipment [5,6]. The active and reactive power of each phase of
the spatial and temporal characteristics and uncertainties of renewable SOP can be controlled independently, and this control feature improves
energy, such as wind power and photovoltaic, result in problems such as load unbalance in three-phase distribution networks [7]. ESS can reduce
power abandonment, load reduction, and local operation safety to the network losses, increase voltage levels, increase equipment utilization,
power grid, causing the tight flexibility of the distribution network [2, and delay system upgrade investment. Thus, the coordinated planning of
3]. In the future, flexibility planning, as special planning for the un- DG, SOP, and ESS in the unbalanced active distribution network (UADN)
certain characteristics of the system, is crucial for realizing a renewable should be studied to combine respective phase power control
distributed energy distribution system with high permeability [4]. An capabilities.
effective method for integrating the positive role of flexible resources Currently, extensive studies have been conducted on the flexibility of
and formulating a coordinated and optimal allocation scheme of large-scale interconnected power grids, flexible mechanism of supply
distributed generation (DG), energy storage system (ESS), and soft open and demand balance, flexibility indexes of supply and demand balance,
point (SOP) is urgently needed in the new distribution network planning and their application to the planning and scheduling of power grids
research. [8–13]. In [14], a two-layer model was established for optimizing ESS

* Corresponding author.
E-mail address: [email protected] (J. Wang).

https://doi.org/10.1016/j.aej.2024.11.039
Received 14 February 2024; Received in revised form 28 October 2024; Accepted 9 November 2024
Available online 13 November 2024
1110-0168/© 2024 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria University. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

allocation based on the conditional value-at-risk theory. However, this semi-definite program [32] and second-order cone program (SOCP)
study only considered the flexible role of energy storage equipment in [33,34] relaxations are proposed based on the power flow formula of the
improving the distribution network without considering controllable distribution network, which can reduce complexity but may cause
loads and high flexibility resources such as SOP, which cannot reflect the infeasible solutions to the original nonlinear power flow model.
impact of current flexibility resource synergy on energy storage alloca- Consequently, although convex relaxation can achieve a high accuracy
tion schemes. A distribution network expansion planning model was solution for the specific conditions of the distribution system, it cannot
established by using the flexibility of end-users to minimize investment be ensured for the three-phase unbalanced system [35,36]. Thus,
costs [15]. A flexible investment strategy based on the Markov decision different new and attractive algorithms are required that can make full
process was developed considering the uncertainty of multiple time use of convex optimization techniques to solve each of the scenarios
scales, and a flexible expansion planning of the distribution system concerned above.
based on an approximate dynamic programming model was established. To address this problem, an adaptive linear relaxation programming
However, this model was only from a deterministic design flexibility (ALRP) model is proposed for optimal coordination planning of DG,
strategy of long-term and short-term loads, not the power distribution. SOP, and ESS. First, a flexibility deficiency risk model is constructed
The effective application of energy storage and SOP flexible resources in based on the flexibility theory to evaluate the potential loss caused by
an active distribution network is critical [16]. Furthermore, energy the stochastic fluctuation of flexibility demand. Next, a single-level
storage and power electronic devices was incorporated into the active planning model is established for DG, SOP, and ESS considering flexi-
distribution network, and the flexibility quantification of the active bility, safety, and economy indexes, considering flexibility improvement
distribution network was analyzed from the perspectives of node flexi- methods such as network reconfiguration, voltage control, and ESS
bility quantification, system flexibility matching, and flexibility of output adjustment. Second, the adaptive linear relaxation optimization
network transmission [17]. SOP and ESS exhibit the spatiotemporal equations are proposed to solve the nonconvex quadratic programming
regulation of power flow, which can effectively alleviate power unbal- problem, which can reach optimal and feasible solutions.
ance and voltage deviation, and improve the flexibility indicators in The paper includes six sections: Section 2 analyzes the characteristics
amplitude, frequency, and intensity dimensions. The analysis revealed of flexible resources and constructs flexibility indexes. Section 3 presents
that the research problem of flexible resource planning is focused on the proposed coordination planning model for DG, SOP, and ESS. Sec-
symmetric distribution networks. To effectively use the collective flex- tion 4 demonstrates the solution method. Moreover, Section 5 reported
ibility of large-scale distributed energy resources equipment, the dis- simulation results of the proposed model. Finally, Section 6 summarizes
tribution level power aggregation strategy for transmission and the main research results of this paper.
distribution interaction was studied [18]. Thus, a feasible flexibility
planning model should be designed by integrating DG, SOP, ESS, and 2. Distribution network flexibility analysis and construction
other flexibility resources in UADN to reduce power loss, voltage un-
balance, and investment and operation costs. 2.1. Distribution network flexibility resources
The optimal planning of DG, SOP, and ESS in distribution networks is
generally considered a bi-level scenario programming problem, which The flexibility resources of the distribution network are distributed
should comprehensively consider the coupling between optimal allo- in the power supply side, grid side, and load side of the topology [37,
cation scheme and operation strategy [19–21]. The planning scheme 38]. The power supply side provides flexibility through the superior
with the best performance can be obtained by using the iterative algo- power grid [39]. Flexible devices represented by the SOP and network
rithm or single-level optimization method dual theory conversion for the reconfiguration can be used to realize the flexibility of the grid side. The
allocation scheme and operation strategy [22,23]. Because the lower load side can satisfy the demand and supply balance through flexible
model involves nonlinear constraints such as a three-phase power flow resources such as ESS and controllable loads [40]. Because the network
equation, ensuring the solving efficiency of the optimization problem reconfiguration cannot realize daily operation, the flexibility resource
based on an iterative algorithm is difficult. This problem can be solved represented by the superior power grid, SOP, ESS, and controllable load
by the single-level planning model obtained by the conversion of duality is used to construct the flexibility index of the distribution network.
theory. Thus, the coordinated planning of three-phase distribution
network contains both continuous variables and discrete variables. The (1) Flexibility resources on the power and load sides
main problem is to find a suitable and efficient optimization power flow The superior power grid provides the upward/downward
algorithm. For the optimization of discrete variables, the common flexibility adjustment ability through the power on the connec-
methods include genetic algorithm, particle swarm optimization algo- tion line between the distribution network and the superior
rithm [24,25] and sensitivity analysis. These intelligent optimization power grid. The ESS provides the upward/downward flexibility
algorithms can effectively deal with discrete constraints and have good adjustment ability through the charge/discharge. The control-
global optimization ability, but the optimization speed for three-phase lable load provides the upward/downward flexibility adjustment
unbalanced ADN is slow. For continuous variables, the primary opti- ability through the cut-off/turn-on. The flexibility adjustment
mization strategies include interior point method [26], linear approxi- ability of each flexibility resource is expressed as follows.
mations and convex relaxations. Linear approximation methods have ⎧ up ∑ { ( )}
been investigated to transform and solve nonlinear unbalanced power ⎪ F = min Pϕess,i − Pess,ϕ
i,t , η Ei,t
ess,ϕ
− Edown,ϕ
⎨ ess ,i,t ϕ∈{a,b,c}
⎪ ess,i
flow formulas, which have good applicability to large-scale systems, but { ( )/ } (1)
∑
they are based on the premise that power loss is ignored, and unbal- ⎪
⎩ Fdown
⎪
ess ,i,t = min Pess,ϕ
i,t + Pϕess,i , Eϕess,i − Eess,ϕ
i,t η
anced distribution systems are not valid [27–29]. Furthermore, for the ϕ∈{a,b,c}

original distribution network power flow problem with original ∑

⎧ up
nonlinear programming, the solution of the linearized model is not ⎪ F =− Pcl,ϕ
j,t
feasible. A continuous linear programming (SLP) method is explored to ⎨ cl,j,t
⎪
ϕ∈{a,b,c}
∑ (2)
solve the OPF problem of large distribution systems, and the nonlinear Pcl,ϕ
⎩ Fcl,j,t =
⎪ down
⎪ j,t
programming problem is regarded as multiple iterations of the approx- ϕ∈{a,b,c}
imate linear programming (LP) problem to obtain a feasible optimal
solution [30,31]. However, it is challenging to SLP algorithm applied to
the OPF problem because of the coupling among phases and the
requirement of solving the OPF model based on the full three-phase.

570
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

Fig. 1. Schematic of SOP installation.

⎧ up ∑ { } √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Fg,t = min ΔPϕg, max , Pg, max − Pg,ϕ
⎪
⎪
⎨ t PL,ϕ
SOP,i = ASOP,i PϕSOP,i 2 + QϕSOP,i 2 (5)
ϕ∈{a,b,c}
∑ { } (3)
⎪
⎪ down
⎩ Fg,t = min ΔPϕg, max , Pg,ϕ
t
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
ϕ∈{a,b,c} PL,ϕ
SOP,j = ASOP,j PϕSOP,j 2 + QϕSOP,j 2 (6)

⎧
whereFup
ess
down
i,t andF ess ,i,t are the upward and downward flexible ⎪
⎪ Qmin ,ϕ ≤ QϕSOP,i ≤ Qmax ,ϕ
⎨ SOP,i
⎪ SOP,i
adjustment capabilities, respectively, provided by the ESS. min ,ϕ min ϕ
QSOP,i = ξ SSOP,ij ; (7)
Here,Eess,ϕ down,ϕ
i,t andEess,i are the capacity and lower capacity limits of ⎪
⎪ /
⎩ Qmax
⎪ ,ϕ
SOP,i = ξ SSOP,ij ; SϕSOP,ij = S3ϕ
max ϕ
SOP,ij 3
the ESS, respectively. Furthermore,Pϕess,i andEϕess,i are the rated
power and rated capacity of the ESS at node i, respectively. Here, ⎧
Pess,ϕ up ⎪ Qmin ,ϕ ≤ QϕSOP,j ≤ Qmax ,ϕ
i,t is the charge/discharge power of the ESS. Furthermore,F cl,j,t
⎪
⎨ SOP,j
⎪ SOP,j
min ,ϕ
andFdown
min ϕ
QSOP,j = ξ SSOP,ij ; (8)
cl,j,t are the upward and downward flexible adjustment ca- ⎪
⎪ /
⎩ Qmax ,ϕ = ξmax Sϕ ; Sϕ
⎪
pabilities provided by the controllable load aggregator at node j, SOP,j SOP,ij SOP,ij = S3ϕ
SOP,ij 3
respectively. Here,Pcl,ϕ
j,t is the outage load of the controllable load
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
aggregator at node j. Furthermore, Fup down
g,t andF g,t are the upward PϕSOP,i 2
+ QϕSOP,i 2
≤ SϕSOP,ij (9)
and downward flexibility adjustment capabilities, respectively,
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
provided by the superior power grid. Furthermore, ΔPϕg, max is the 2 2
PϕSOP,j + QϕSOP,j ≤ SϕSOP,ij (10)
upper limit of the climbing power of the superior power grid.
Furthermore,Pg, max is the limit of the power transmitted from the Constraint (4) is the active power balance of SOP. Constraints (5) and
superior power grid to the distribution network. Here,Pt is the
g,ϕ (6) define the power loss equation of SOP. Constraints (7) and (8) are the
active power injected by the superior power grid. t is the time reactive power constraints of SOP. Constraints (9) and (10) are SOP
interval for a period, i.e. 1 h. capacity constraints. ξmax and ξmin represent the upper limit coefficient
(2) Flexibility resources on the grid side and lower limit coefficient of reactive power respectively.

The flexibility resources on the network side consider SOP, which 2.2. Distribution network flexibility index
can control active and reactive power flexibly and safety in real time,
effectively solving voltage violation and improving the flexibility of the The flexibility demand of the distribution network can be attributed
topology structure. The back-to-back voltage source converter (B2B to the random fluctuation of net load, which refers to the difference
VSC), an implementation of SOP, is considered as an example [41]. between the total load of all nodes in the distribution network and the
Fig. 1 displays the control principle of the SOP, with separate phase total power output of the DG. If the net load uncertain power is greater
power control. One SOP has two VSCs: one is the PQ control mode, and than the downward flexibility adjustment ability of the system, then the
the other is the VdcQ control mode. The VSC with PQ control is used to system is limited by the lack of down-adjustment flexibility and cannot
control the transmitted active power of the SOP and the reactive power absorb too much DG output. If the net load uncertain power is less than
injected by the converter, whereas the VSC with VdcQ control is used to the upward flexibility adjustment ability of the system, then the system
keep the DC bus voltage constant and inject reactive power [42− 43]. cannot guarantee reliable load demand–supply. The distribution
The active and reactive power of each phase can be independently network flexibility deficiency indexISD can be calculated as follows:
controlled by the VSC controller in the UADN. The individual phase
power control model of SOP [42,43] is as follows:

PϕSOP,i + PϕSOP,j + PL,ϕ Lϕ

SOP,i + PSOP,j = 0 (4)

571
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

⎧ ( )
⎪
⎪ ISD = Iup down
SD + I SD 1 ∑∑⃒⃒ a / / / ⃒2
⃒
⎪
⎪ JU = ⃒V 3 + ej4π/3 V bi,t 3 + ej2π/3 V ci,t 3⃒ (17)
⎪
⎪
⎪ T T t∈T i∈Ωn i,t
⎪ up ∑
⎪ ( ( ) )
⎨ ISD = max ut Fup nl
t − Pt , 0
t=1 (11)
⎪
⎪ wheredand LT are discount rate and asset life cycle, respectively.
⎪
⎪
⎪
⎪ ∑ T
(( )( ) ) CEAI is the investment cost, COM is operation maintenance cost,
⎪
⎪
⎪ Idown = max 1 − ut Pnl down
t − Ft ,0 CFRSC,s is the flexibility resource scheduling cost, CNLC,s is the
⎩ SD
t=1
network loss cost, CPRC,s is the power purchase cost, CPC,s is carbon
⎧ up ∑ up ∑ up revenue,Mis the number of days in a year, andsis the number of
⎪ F = Fess,i,t + Fcl,j,t + Fup
⎨ t
⎪
i∈Sess j∈Scl
g,t
scenarios. Furthermore, Si,3ϕ,DG andSij,3ϕ,SOP are the rated capacity
∑ ∑ (12) of DG and SOP, respectively; κ is the cost coefficient of SOP for
⎩ Ft
down
⎪
⎪ = Fdown
ess,i,t + Fdown down
cl,j,t + F g,t operation and maintenance.Eess,3ϕ,i andPess,3ϕ,i are the rated ca-
i∈Sess j∈Sel
pacities and rated power of ESS respectively, cI,DG andcI,SOP are
whereISD is the difference between flexibility adjustment capacity and unit capacity investment cost respectively of DG and SOP,
flexibility demand at each time period. Iup down respectively; ce andcp are unit capacity and unit power investment
SD andI SD are the sum of the
difference between flexibility supply and flexibility demand, respec- cost, respectively.Ploss l
ij,ϕ,t is operation loss, Pi,ϕ,DG is the active power
tively. ut is the state variable of the change of net load power during the of the DG; cO,DG andcO,ess are unit power operation cost of DG and
period T, and 1 indicates the increase of net load, and 0 indicates the ESS, respectively; cel is unit power interruption load compensa-
decrease of net load. Fup down tion price;closs is the loss cost of unit power;cg is the purchase price
t andF t are the total upward and downward
flexibility adjustment capacity of distribution network, respectively. T is to the superior power; andJU is three-phase voltage unbalance
the total time in a scheduling period, which is 24 in this paper. degree.
(2) Constraints
3. Flexible DG, SOP, and ESS coordinated planning model
a) Constraints on the rated capacity and rated power of the DG, SOP,
3.1. Single-layer flexible coordination planning model and ESS are as follows:

xDG
i SDG, min ≤ SDG,i ≤ xi SDG, max , ϕ ∈ {a, b, c}
ϕ DG
(18)
The objectives of the single-layer model include an economic index,
flexibility index, and three-phase unbalanced voltage index. The eco-
xSOP
ij SSOP, min ≤ SSOP,ij ≤ xij SSOP, max , ϕ ∈ {a, b, c}
ϕ SOP
(19)
nomic cost considers the equal annual investment cost of DG, SOP, ESS,
and the total annual operation and maintenance cost of the distribution ⎧
network, as follows: ⎨ xess Pess, min ≤ Pϕ ≤ xess Pess, max
i ess,i i
, ϕ ∈ {a, b, c} (20)
⎩ xess Eess, min ≤ Eϕess,i ≤ xess Eess, max
i i
(1) Objective function
The objectives of the single-layer model mainly include an
where Pess, max and Pess, min are the upper and lower limits of the rated
economic index, flexibility index and three-phase unbalance
power of the ESS, respectively. Furthermore,Eess, max andEess, min are the
voltage index. The economic cost mainly considers the equal
rated capacity upper and lower limits of the ESS, respectively. Here, xDG
i ,
annual investment cost of DG、SOP、ESS and the total annual
operation and maintenance cost of the distribution network, as xSOP
ij and xess
i are 0–1 variables, which represent whether DG, SOP, and
shown in (13). ESS are installed at nodei.
b) The simplified Distflow model is adopted to simulate the power
min FTotal = ω1 (CEAI + COM ) + ω2 ISD + ω3 JU (13) flow of the distribution network, and the branch power flow constraints
( are as follows [44− 45]:
nb
d(1 + d)y ∑ ∑
CEAI = cI,DG Si,3ϕ,DG + cI,SOP Sij,3ϕ,SOP ⎧ ( )
(1 + d)y − 1 i=1 ij∈nSOP
⎪
⎪ H
⎪ vi = vj + Sij zij + zij Sij − zij lij zij
H H
⎪
) (14) ⎪
⎪
⎪ ( ) ∑ ( )
∑( )
⎪
⎪
⎪
⎪ diag Sij − zij lij − sL,j + sDG,i = diag Sjk
+ ce Eess,3ϕ,i + cp Pess,3ϕ,i ⎪
⎨ k:j→k
i∈Sess ⎡ ⎤ [ ][ ]H (21)
⎪
⎪
⎪ v i Sij Vi Vi
∑ ( ) ⎪
⎪ ⎣ ⎦ =
COM = M ps CNLC,s + CPRC,s + CPC,s + CFRSC,s (15) ⎪
⎪
⎪
⎪ SHij lij Iij Iij
s∈Φ
⎪
⎪
⎪
⎩
V i − V j = Zij Iij i∼j
⎧
⎪ ∑ ∑ ∑ T
⎪
⎪ CNLC = closs
⎪
⎪ Ploss
ij,ϕ,t where vi and vj are square of voltage at nodes i and j respectively; vi =
⎪
⎪
⎪ ij∈Sb ϕ=a,b,c t=1
⎪
⎪
⎪
⎪ |Vi |2 , Vi is complex voltage; i ∼ j represents the branch of the network; zij
⎪ ∑ ∑ T
is the impedance of the branch; Iij is complex line current; lij = Iij (Iij )H ; Sij
⎪
Pgϕ,t
⎪
⎪ CPRC = cg
⎪
⎪
⎪
⎪
⎪
⎪ ϕ=a,b,c t=1 is complex power flow in branch i ∼ j; Sij = Pij + jQij ; sL,j is complex
⎪
⎪
⎪
⎨ ∑∑ T power demand at bus j; sDG,i is per-phase complex power for DG con-
CPC = cap Pap
k,ϕ,t (16) nected at bus i.
⎪ k∈SDG t=1
⎪
⎪
⎪ Because of the positive semidefinite (PSD) matrix rank being 1 in
⎪
⎪ nb
⎪
⎪ ∑ ∑ ∑ T ∑ (21), the SDP model cannot be well extended, and the voltage and
⎪
⎪ CFRSC = cO,DG Pli,ϕ,DG + κ cI,SOP Sij,3ϕ,SOP
⎪
⎪
⎪ current information of each phase cannot be obtained. Therefore, the
⎪
⎪ i=1 ϕ=a,b,c t=1 ij∈nSOP
⎪
⎪
⎪ three-phase unbalance of the system cannot be evaluated.
⎪
⎪ ∑ ∑ ∑ T ⃒ ⃒ ∑∑ ∑ T
⎪ ⃒ ess ⃒
⎪
⎪
⎪
⎪ + cO,ess ⃒Pi,ϕ,t ⃒ + cel Pclj,ϕ,t
⎩ i∈S ϕ=a,b,c t=1 j∈Sjl ϕ=a,b,c t=1

572
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

Table 3
Planned locations of flexible devices for IEEE 33-bus unbalanced distribution
system.
Case Flexible (Location, Capacity/kVA)/(Location, Capacity/kVA)/(
devices Location, Capacity /MWh; Power/MVA)

1 Wind power (9 A,200); (13B,100); (14 C,100); (25 A,100); (28B,100);

(30 A,100); (30B,100); (30 C,200); (31 A,100);
(32 A,100); (32B,100)
Photovoltaic (8 A,200); (8B,100); (8 C,200); (11 C,100); (12B,100);
(14 A,200); (14B,200); (15 C,200); (18 A,100);
(25 A,200); (25B,300); (25 C,200); (29 C,100);
(30 A,100); (30B,100); (30 C,100); (31B,100); (32 A,200)
Fig. 2. Structure of the modified IEEE 33-bus system.
2 Wind power (17 C,200); (25 C,200); (29B,200); (30 A,500)
Photovoltaic (9B,200); (10 A,300); (14B,200); (17 A,200); (25 A,200);
(25 C,200); (30 C,400); (33B,200)
Table 1 SSOP (8− 21,500); (18− 33,500)
DG, SOP, and ESS parameters. 3 Wind power (8B,200); (24 C,200); (25 A,200); (25B,200)
Photovoltaic (16 C,300); (17B,400); (22 C,400); (24B,500); (25 A,200);
Parameters Wind Photovoltaic SOP ESS
(26B,400);
power
(28 A,200); (32 A,500);
Discount rate 0.08 - SSOP (8− 21,300); (9− 15,100); (18− 33,300)
Economic service life 20 - ESS location 7; 12; 15; 23; 25; 32; 33
(Year) Eess ,Pess (0.5515,2); (0.5479,2); (0.5479,2);(0.5525,2); (0.5516,2);
Unit rated capacity 100/kW 100/kVA - (0.5519,2);(0.5515,2)
Investment cost(RMB) 7800/kW 8200/kW 1000/ - 4 Wind power (9B,100); (25 A,200); (25 C,100); (31B,300)
kVA Photovoltaic (17 C,500); (18B,300); (20 C,200); (24B,700); (25 A,100);
Operation cost (RMB/ 0.3 0.3 - 0.0195 (32B,500);
kWh) SSOP (8− 21,300); (9− 15,100); (18− 33,200)
Capacity cost (RMB/kWh) - - 0.5 1000 ESS location 7; 11; 24; 32; 33
Power cost (RMB/kW) - - - 1500 Eess ,Pess (0.4379,1.9); (0.5562,2.7); (0.3041,1.4); (0.5008,3.1);
Operation cost coefficient - - 0.01 - (0.3326,1.7);
Loss coefficient - - 0.02 -

Table 2 ⎧
⎪ ∑ ϕϕ ∑ ( ( ) )
DG, SOP, and ESS price.
⎪
⎪
⎪
⎪ Pjk = Pϕϕ
ij − rϕφ
ij lϕr
ij − lij
ϕm
− 2xϕφ
ij lij
ϕrm
− pϕj
⎪
⎪ k:j→k
Parameters Value (RMB/kWh) ⎪
⎪ ϕ={a,b,c}
⎪
⎨ φ={a,b,c}
Loss cost Pick period:0.9, Normal period:0.6, Valley period:0.3 ∑ ( ϕφ ϕrm ( )) (23)
⎪ ∑ ϕϕ
Electricity purchase Pick period:0.58, Normal period:0.42, Valley period: ⎪
⎪
⎪ Qjk = Qϕϕij − 2rij lij + 2xϕφ
ij lϕr
ij − lij
ϕm
− qϕj
price 0.32 ⎪
⎪ k:j→k
⎪
⎪ ϕ={a,b,c}
DG power price 0.35 ⎪
⎪
⎩ φ={a,b,c}
Emission 0.25
Pick period: 10:00–12:00, 7:00–20:00. Normal period: 6:00–9:00, 13:00–16:00.
Valley period: 1:00–5:00. ⎧
⎨ pϕ = pϕ − pϕ − pϕ − pϕ , ϕ = {a, b, c}
j L,j DG,j SOP,j ESS,j
(24)
⎩ qϕj = qϕL,j − qϕDG,j − qϕSOP,j , ϕ = {a, b, c}
To address these issues, effective approximations are introduced for
the variables in the three-phase unbalanced power flow model, where ( )2 ( )2
the voltage is converted into real and imaginary parts to represent the Pϕϕ
ij + Qϕϕ
ij = vϕϕr
i lij + vi
ϕϕr
lij + vϕϕr
ϕϕm ϕϕm
i lij
ϕϕm
+ vϕϕm
i lϕϕr
ij (25)
unbalanced indicator.

⎧
⎪ ∑ ( ϕφ ϕφ ) ∑ ((( ϕφ )2 ( ϕφ )2 )( φφr ))
⎪ vj + vj = vi + vi − 2 rij Pij + xϕφij Qij rij + xij lij + lφφm
⎪
⎪ ϕϕr ϕϕm ϕϕr ϕϕm ϕφ
+
⎪ ij
⎪
⎪
⎪
⎪ ϕ={a,b,c} ϕ={a,b,c}
⎪
⎪
⎪
⎪
⎪
⎪ φ={a,b,c} φ={a,b,c}
⎪
⎪ ∑ (( ϕφ aφ )( ) ( )( ))
⎪
⎪ rr mm φ φ rm φ φ mr
⎪ rij 1 rij 2 + xij 1 xij 2 lij 1 2 + lij 1 2 + rij 1 rij 2 − xij 1 xij 2 − lij 1 2 + lij 1 2
ϕφ ϕφ φ φ φ φ ϕφ ϕφ ϕφ ϕφ
⎪
⎪ +2
⎪
⎪
⎪
⎪ ϕ={a,b,c}
⎪
⎪
⎪
⎨ φ ,φ ={a,b,c}
(22)
1 2

⎪
⎪ φ1 ∕
=φ2
⎪
⎪
⎪
⎪ ( )H ( )2 ϕϕm ( ϕm )2
⎪
⎪
⎪
⎪
⎪ vϕϕr
i + vϕϕm
i j = V ϕi V ϕi , vϕϕr
i = V ϕri , vi = V i
⎪
⎪ ( )H ( )2 ( )2
⎪
⎪
⎪
⎪
⎪ lϕϕr
ij + lϕϕm
ij j = Iϕij Iϕij , lϕϕr
ij = Iϕr
ij , lϕϕm
ij = Iϕmij , lϕϕrm
ij = Iϕr
ij I ij ,
ϕm
⎪
⎪
⎪
⎪ φ φ2 rm φ r φ r φ φ2 mm φ m φ m
⎪
⎪
⎪
⎪ lϕφr
ij = Iϕr
ij I ij , lij
φr ϕφm
= Iϕm
ij I ij , lij
φm 1
= Iij 1 Iij 2 , lij 1 = Iij 1 Iij 2 ,
⎪
⎪
⎪ φ φ2 rm φ r φ m φ φ2 mr φ m φ r
⎩ lij 1 = Iij 1 Iij 2 , lij 1 = Iij 1 Iij 2

573
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

Table 4
Planned cost of flexibility equipment for the IEEE 33-bus unbalanced system.
Case Total investment cost Operation cost Purchase cost Active power/ Flexibility deficiency risk Three-phase voltage
/104RMB /104RMB /104RMB 104RMB /104RMB unbalance

Original - - 1653.7235 94.1618 - 379.66

system
1 550.8356 439.6271 631.9403 59.6723 441.23 451.11
2 464.4086 369.9999 528.1723 55.8394 403.71 405.65
3 497.7238 382.7986 472.6716 47.2672 359.92 381.43
4 392.1105 284.3219 503.2250 51.3223 412.34 390.44

limits of current real part branchij, respectively; andIm m

ij ,Iij are the upper

and lower of current imaginary part branchij, respectively.

c) DG output constraint
⎧ ϕ
⎪
⎪ PDG,i,t = Pϕ,re DG,i , ϕ = {a, b, c}
⎪
⎪
⎪
⎪
⎪
⎪ QDG,i,t = PϕDG,i,t tan θϕDG,i , ϕ = {a, b, c}
ϕ
⎪
⎨ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( )
2 ( 2 ) (29)
⎪
⎪ PϕDG,i,t + QϕDG,i,t ≤ SϕDG,i , ϕ = {a, b, c}
⎪
⎪
⎪
⎪ ∑ ϕ
⎪ 3ϕ
⎩ SDG,i = SDG,i
⎪
⎪
ϕ∈{a,b,c}

where Pϕ,re
DG,i is the rated power of DG connected to phase ϕ; QDG,i,t is the
ϕ

reactive power of DG connected to phase ϕ; and SϕDG,i is the capacity of

DG of phase ϕ.
Fig. 3. Proportion of different objectives in four cases for IEEE 33-bus system.
d) ESS output constraint
( )2
lϕφ
ij ≤ lϕϕ
ij lij
φφ
(26)

⎧
⎪
⎪
⎪
⎪
⎨ ( )2 ( ) ( )2 ( )2 ( ) ( )2 ( ) ( )
V ri ≤ diag vϕϕr
i,t ≤ V ri Vm
i ≤ diag vϕϕm
i,t ≤ Vm
i diag vϕϕm
i,t ≤ diag vϕϕr
i,t (tan(θ))2 (27)
⎪
⎪
⎪
⎪
⎩

⎧
⎪
⎪
⎪
⎪
⎨ ( )2 ( ) ( )2 ( )2 ( ) ( )2 ( ) ( )
Irij ≤ diag lϕϕr
ij,t ≤ Irij Im
ij ≤ diag lϕϕm
ij,t ≤ Im
ij diag lϕϕm
i ≤ diag lϕϕr
i (tan(θ))2 (28)
⎪
⎪
⎪
⎪
⎩

⎧
whereV ϕr
i andV i are real and imaginary parts of voltage with phaseϕ,
ϕm
⎪ (
bess,ϕ
)
⎪ SOCi,t = SOCi,t− 1 + Pi,t × Δt, ϕ = {a, b, c}
⎪
⎪ ϕ ϕ

respectively; Iϕr
ij andI ij are the real and imaginary parts of current with
ϕm ⎪
⎪
⎪
⎪ down,ϕ ess,ϕ
⎪
⎪ E ≤ Ei,t ≤ Eess,i , ϕ = {a, b, c}
ϕ
phaseϕ, respectively; rϕφ
ij andxij are resistanceϕ and reactance of
ϕφ
⎨ ess,i
⎪
− Pess,i ≤ Pbess,ϕ
ϕ
≤ Pϕess,i , ϕ = {a, b, c} (30)
branchij, respectively; andPϕϕ
ij ,Qij are active and reactive power of
ϕϕ
⎪
⎪
i,t
⎪ ch,ϕ
⎪P × P dch,ϕ
⎪ = 0, ϕ = {a, b, c}
branchij, respectively. pϕj andqϕj are active and reactive loads of the nodej, ⎪
⎪
⎪
i,t i,t
/
⎪ bess,ϕ
respectively;V ri andV ri are the upper and lower limits of voltage real part ⎩ Pi,t
⎪
⎪ = η Pi,t − Pdch,ϕ
ch ch,ϕ
i,t ηdch , ϕ = {a, b, c}
nodei, respectively. V m m
i andV i are the upper and lower limits of voltage

imaginary part nodei, respectively; Im r whereSOCϕi,t is the charged state of ESS access to phase ϕ; Pch,ϕ
i,t and
ij andIij are the upper and lower

574
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

Fig. 4. Voltage distribution of Cases 1–4 for IEEE 33 system at 16:00.

Fig. 5. SOP 18–33 power distribution of Case 2 for IEEE 33 system.

575
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

Fig. 6. SOP 18–33 power distribution of Case 3 for IEEE 33 system.

Fig. 7. ESS active power distribution of Case 3 for IEEE 33.

Pdch,ϕ
i,t are charging and discharging power of ESS respectively;ηch and andvcl,ϕ
t represents the binary variable.

ηdch are the charging and discharging efficiency of ESS, respectively. g) Constraint of active power injected by the superior main network
e) Operation constraints of SOP are presented in (4)–(10). − ΔPgmax ≤ ΔPg,ϕ ≤ ΔPgmax (33)
t
f) Controllable load constraint
The proposed model is essentially a nonconvex quadratic program-
cl,ϕ
vcl,ϕ cl cl,ϕ cl
t Pmin ,i ≤ Pi,t ≤ vt Pmax ,i , ϕ = {a, b, c} (31) ming problem, which should be solved accurately and efficiently.

T
∑
vcl,ϕ ≤ Tcl (32)
t=1
t max
3.2. Risk control in proposed coordinated planning

wherePcl cl The uncertainty of the DG output power and load demand is the
max ,i ,Pmin ,i are the upper and lower limits of the controllable load
primary reason for the flexibility deficiency of the active distribution
interruption capacity at the nodei, respectively. Here, T cl
max is the
system. Therefore, to prevent the risk of potential flexibility deficiency
maximum period of controllable load interruption at the nodei,
and investment cost, the conditional value-at-risk (CVaR) theory is

576
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

Fig. 8. Structure of the modified IEEE 123-bus system.

introduced into the coordination planning scheme of DG, SOP, and ESS.
Table 5 The CVaR is a coherent risk measure [46− 47]. If Uis a continuous
Planned positions of flexible devices in the IEEE 123-bus unbalanced system. random variable bounded with probability 1, then holds for α ∈ (0, 1).
Case Flexible (Location, Capacity /kVA)/(Location, Capacity /kVA)/( CVaRα (U )represents the conditional expectation of Ua given that it is
devices Location, Capacity /MWh; Power/MVA) greater than the (1 − α)left quantile VaRα (U ) =
1 Wind power (65B,300); (66 C,100); (103 C,400); (113 A,100) inf{u ∈ ℝ : ℙ(U ≤ u) ≥ 1 − α}for a given α ∈ (0, 1). To reduce
Photovoltaic (49B,200); (64B,300); (74 C,200); (76B,100); (76 C,300); computational complexity, the CVaR can be written as the following
{
(88 A,200); (110B,100); (117 C,300) optimization problem, CVaRα (U ) = inf t + α− 1 E[ max {U − t, 0}]}.
2 Wind power (65B,200); (69 A,300); (73 C,200); (76B,100); (96B,300); t∈ℝ
(98 A,400);
Photovoltaic (48 A,100); (48 C,300); (49B,200); (66 C,200);
(71 A,100); (73 C,200); (76 A,300); (103 C,200)
SSOP (8− 21,500); (18− 33,500)
3 Wind power (26 A,200); (51 A,200); (66 C,300); (67 A,100);
(103B,500)
Photovoltaic (16 C,600); (17B,300); (22 C,200); (24B,500); (25 A,200);
(26B,400);
(28 A,300); (32 A,500);
SSOP (117− 112,600); (54− 95,700)
ESS location 4; 41; 45; 75; 88; 102
Eess ,Pess (0.1413,1.3485); (0.5,1.8117); (0.0913,0.62);
(0.5201,1.8117); (0.4596,1.6331); (0.4973,1.0645)
4 Wind power (9B,100); (25 A,200); (25 C,100); (31B,300)
Photovoltaic (17 C,500); (18B,300); (20 C,200); (24B,700); (25 A,100);
(32B,500);
SSOP (117− 112,400); (54− 95,400)
ESS location 7; 11; 24; 32; 33
Eess ,Pess (0.4379,1.9); (0.5562,2.7); (0.3041,1.4); (0.5008,3.1);
(0.3326,1.7);

Fig. 9. Proportion of various objectives in four cases of the IEEE 123 system.

Table 6
Planned cost of flexibility devices for the IEEE 123-bus unbalanced system.
Case Total investment cost Operation and maintenance cost Purchase cost Active power/ Flexibility deficiency risk Three-phase voltage
/104RMB /104RMB /104RMB 104RMB /104RMB unbalance

Original - - 1379.0920 84.4486 - 501.27

system
1 259.1120 207.1945 752.6469 55.8633 133.89 487.43
2 281.7232 234.1407 716.8949 29.1332 102.43 293.88
3 321.9047 273.6545 538.0721 27.3849 69.27 262.61
4 304.1915 242.9187 633.7402 31.2152 101.78 312.42

577
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

Fig. 10. Voltage distribution of Case 1 for the IEEE 123 system.

Fig. 11. Voltage distribution of Case 2 for the IEEE 123 system.

Fig. 12. Voltage distribution of Case 3 for the IEEE 123 system.

Fig. 13. SOP 54–94 active and reactive power distribution of Case 2 for the IEEE 123 system.

578
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

Fig. 14. SOP 117–122 active and reactive power distribution of Case 2 for the IEEE 123 system.

Fig. 15. SOP 54–94 active and reactive power distribution of Case 3 for the IEEE 123 system.

Fig. 16. SOP 117–122 active and reactive power distribution of Case 3 for the IEEE 123 system.

Fig. 17. ESS active power distribution of Case 3 for IEEE 123-bus.

579
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

Table 7
Comparison of various algorithms for IEEE 33 and IEEE 123 bus system.
Total investment cost/104 Flexibility deficiency risk /104 Three-phase voltage Time (s) Number of iterations
(RMB) (RMB) unbalance

IEEE 33 PSO 539.4004 391.08 411.32 2641.70 5560

Proposed 497.7238 359.92 381.43 1096.05 1935
algorithm
IEEE 123 PSO 392.1910 70.1921 294.88 2913.53 3987
Proposed 321.9047 69.27 262.61 1574.33 2033
algorithm

( )2 ( )2 ( )2
Therefore, CVaRα represents a very natural and flexible risk measure and ( )2 ( )2 ϕϕr
vϕϕr
i = V ri , vϕϕm
i = Vm
i , vϕϕm
i = Vm
i , lij = Irij , lϕϕr
ij = Irij ,
can be used to explore the impact of various risk tolerance levels and ( )2
model conservativeness through sensitivity analysis on the α parameter. lϕϕm = Im .
ij ij
The CVaR can be incorporated into the primary objective function as
⎧
follows: ⎪ ϕ(xd ) = x2d
⎪
⎪
⎪
⎨
Min β1 FTotal + β2 CVaR(FTotal ) (34) (ld + ud )2
ϕ2 (xd ) = (ld + ud )xd − (36)
⎪
⎪
⎪ 4
where β1 and β2 are the weighting coefficients. The risk value of various ⎪
⎩
ϕ2 (xd ) = (ld + ud )xd − ld ud
degrees can be calculated by adjusting weighting coefficients.
{
Xd (xd ) = ϕ(xd ) − ϕ2 (xd )
3.3. Adaptive linear relaxation solution methodology (37)
Xd (xd ) = ϕ2 (xd ) − ϕ(xd )

To reduce the problem’s complexity caused by the nonconvex

quadratic equality constraints in the proposed optimization model (26),
an adaptive linear relaxation method is proposed based on the quadratic
equality constraints, which can achieve superior solving efficiency.

⎧ ( )
⎪ ϕ xd , xq = xd xq
⎪
⎪
⎪
⎪ [ ( ( ))2 ]
⎪
⎪
⎪ ( ) 1 ( ) (ld + ud ) + δjq lq + uq ld ud + δ2dq lq uq
⎨ ϕ2 xd , xq = lq + uq xd + (ld + ud )xq − +
2 4δdq δdq (38)
⎪
⎪ [
⎪
⎪ ( ( )) 2 ( )( )]
⎪
⎪ ( ) 1 ( ) (ld + ud )2 + δdq lq + uq ld + δdq lq ud + δdq uq
⎩ ϕ2 xd , xq = lq + uq xd + (ld + ud )xq +
⎪
⎪ −
2 4δdq δdq

3.4. Adaptive delimitation technology

{ ( ) ( ) ( )
Xdq xd , xq = ϕ xd , xq − ϕ2 xd , xq
Formula (25) is a nonconvex quadratic function, which can be ( ) ( ) ( ) (39)
Xdq xd , xq = ϕ2 xd , xq − ϕ xd , xq
expressed as follows:
( )2 ( )2 ( ) Thus,
F(x) = Pϕϕ
ij + Qϕϕij − vϕϕr
i lij + vi
ϕϕr
lij + vϕϕr
ϕϕm ϕϕm
i lij
ϕϕm
+ vϕϕm
i lϕϕr
ij 2 2 6 ∑
6
∑ ∑ ∑ ( )
(35) fl = ϕ(xd ) + ϕ(xd ) + ϕ xd , xq , d ∕
=q (40)
d=1 d=1 d=3 q=3
To solve the nonconvex quadratic problem, an adaptive delimitation
technique is proposed to achieve a new lower bound function. Let F = The coordination planning model of DG, SOP, and ESS for the un-
balanced distribution network is as follows:
{x ∈ Rn |l ≤ x ≤ u} ⊆ F0 , where l = (l1 , …, ln )T , u = (u1 , …, un )T . For any
⎧
x ∈ F,d ∈ {1, 2, …, n},q ∈ {1, 2, …, n},d ∕
= q, thusδdq > 0. ⎨ Minmize : (35)
Subject to : (4) − (10), (18) − (20), (22) − (33) (41)
⎩
4. The variables are defined as follows fl ≤ 0

{ } ( )2
X = x = [Pϕϕ ,lij ,lij ],ϕ = {(a,b,c)} . vϕϕr
ϕϕ ϕϕr ϕϕm ϕϕr ϕϕm
= V ri , 4.1. Algorithm solving
ij ,Qij ,vi ,vi i

The proposed ALRP algorithm is implemented based on three basic

580
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

operations, namely branch operation, bound operation, and pruning active distribution system.
operation using MATLAB and solved using CPLEX.
The flexibility deficiency risk is considered for Cases 1–3.
Algorithm. Three-phase operation optimization using adaptive linear
Case 4. DG, SOP, and ESS are planned for a three-phase unbalanced
relaxation programming (ALRP) algorithm

5. Results and discussion active distribution system without considering flexibility deficiency risk.
Table 3 details the planned locations and capacities of DG, SOP, and
To verify the effectiveness of the coordinated planning model of DG,
ESS. The installed capacity of DG for Cases 1, 2, and 3 are 4100, 3000,
SOP, and ESS considering the flexibility deficiency risk, simulation tests
and 3400 kVA, respectively. Compared with Case 3, the total SOP
are performed on the three-phase unbalanced distribution system of
installed capacity for Case 4 is reduced by 300 kVA. This phenomenon
IEEE 33 [48] and IEEE 123-bus [49].
could be attributed to the necessity to install more SOP units to regulate
reactive power and improve voltage curves without ESS access. For
5.1. Modified IEEE 33-bus three-phase unbalanced distribution system example, although SOP 9–15 in Cases 3 and 4 installed only a capacity of
100 kVA, voltage distribution can be improved through single-phase
Fig. 2 displays the modified IEEE 33-bus distribution system. The reactive power regulation. Furthermore, the capacity of DG in Case 4
maximum installed unit capacity of the DG and SOP are 1MVA and is less than that in Case 3 because the risk index of flexibility deficiency
1MVA. The installation location of the SOP is the tie branches. The risk is not introduced, which severely affects the permeability of DG in
maximum installed capacity and power of ESS are 1 MWh and 1 MVA, the distribution system and the reduction of flexibility deficiency risk.
respectively. The parameters of DG, SOP, ESS, and electricity price are Table 4 details the planning costs of DG, SOP, and ESS. As presented
presented in Tables 1 and 2. Four cases are given to analyze the per- in the table, the installation cost of Case 1 is high, and the upward
formance of the coordination planning model for DG, SOP, and ESS in regulation flexibility is deficient without the adjustment of SOP and ESS.
three-phase UADN. The investment and operation costs increase in Case 3, but the purchase
cost of the superior grid is the lowest. The coordination planning of DG,
Case 1. Only DG is planned for a three-phase unbalanced active dis- SOP, and ESS can improve upward and downward flexibility. Among all
tribution system. cases, Case 3 has the lowest cost of flexibility deficiency risk; in detail,
Case 2. DG and SOP are planned for a three-phase unbalanced active the cost of Case 3 is 813,100 RMB, 437,900 RMB, and 524,200 RMB
distribution system. lower than those of Cases 1, 2, and 4, respectively.
From the perspective of three-phase voltage unbalance, compared
Case 3. DG, SOP, and ESS are planned for a three-phase unbalanced

581
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

with Case 1, Cases 2–4 three-phase voltage unbalance decreased by phase voltage unbalance degree of Case 4 clearly cannot satisfy opera-
45.46, 69.68, and 60.67. Especially in Cases 3 and 4, SOP can regulate tional requirements. Thus, the single-phase power regulation capability
the active power of each phase independently, which is conducive to of the SOP for Case 3 enables optimal performance for each index. Fig. 9
balancing the load between feeders in the same phase and reducing displays the proportion of carbon emission, total economic cost, flexi-
power loss as much as possible. Thus, the simultaneous planning of DG, bility deficiency risk cost, and three-phase voltage unbalance to the
SOP, and ESS can improve flexibility and three-phase unbalance, maximum values in the four cases of Cases 1–4, which directly proves
thereby promoting DG absorption capacity. the superiority of the Case 3 planning scheme and its applicability in
Fig. 3 displays the proportion of carbon emission, total economic larger three-phase unbalance system.
cost, flexibility deficiency risk cost, and three-phase voltage unbalance Figs. 10–12 display the voltage distribution of Cases 1–3, respec-
in the maximum value of Cases 1–4. The planning model of Case 3 has tively, for the IEEE 123-bus unbalanced system at 12:00 and 18:00.
more advantages. Comparing the three cases revealed that reasonable planning of the
Fig. 4 gives the voltage profiles of Cases 1–4 and the original model at location and capacity of DG, SOP, and ESS in each phase can effectively
12:00 for the IEEE 33- bus three-phase unbalanced system. It can be seen alleviate the three-phase unbalance of the system and improve voltage
from Case 1 and the original model that access to DG will cause a more distribution.
serious three-phase unbalance. The access of SOP and ESS in Case 2 and Figs. 13 and 14 display the active power and reactive power distri-
Case 3 effectively alleviates the three-phase voltage unbalance. The bution of the SOP for Case 2. During peak load hours, the SOP satisfies
minimum B-phase voltage of the unbalanced distribution system the load requirements of each phase by active power regulation, while
without SOP and DG in Case 1 is 0.9771 p.u. However, the minimum alleviating the three-phase voltage unbalance by reactive power regu-
voltage value of Cases 2–3 are increased to 0.985 p.u. The voltage dis- lation. Figs. 15 and 16 show the active and reactive power distribution of
tributions of phase A, phase B, and phase C in Case 3 are more similar, the SOPs for Case 3. Similar to Case 2, the SOP can alleviate the three-
and the voltage unbalances degree is significantly lower than that in the phase voltage unbalance by regulating the active and reactive power
other cases. and improving the flexibility of the system operation. However, the
To describe the dispatching distribution of DG, SOP, and ESS for each installation capacity of the SOP for Case 3 is less than that for Case 2
phase, Figs. 5–6 show the active power and reactive power distribution because the ESS can discharge during peak load and charge during off-
of SOP 18–33 for Cases 2–3. Fig. 7 shows the active power distribution of peak load, mitigating the increase in SOP transmission capacity. Thus, a
ESS for Case 3. As can be seen from the power distribution of the SOP in reasonable planning scheme for various flexible resources can alleviate
Cases 2–3, the SOP improves voltage distribution and three-phase the three-phase voltage unbalance, improve the operation economy, and
voltage unbalance by transferring active power and reactive power enhance system flexibility, which can increase DG permeability and
compensation during peak load periods. During 0:00–6:00 and promote DG consumption. Fig. 17 displays the active power distribution
22:00–24:00, due to the small load, reactive power compensation and of ESS in Case 3.
three-phase voltage unbalance only rely on DG, and the power trans-
mitted by SOP is close to 0. ESS of Case 4 is discharged during the peak 5.3. Algorithm validation
load period and charged during the off-peak load period. However, due
to the excess power provided by DG and SOP, A-phase ESS is charged at To further verify the accuracy and the computational efficiency of
19:00, which helps to alleviate the voltage unbalance of the distribution the proposed algorithm, the particle swarm optimization (PSO) algo-
system. rithm is used to solve the problem. PSO is a widely recognized intelligent
algorithm [35,36]. A parameter matrix is used, and performance com-
5.2. Modified IEEE 123-bus three-phase unbalanced distribution system parison is presented in Table 7.
Table 7 reveals that compared with the PSO algorithm, the proposed
The IEEE 123-bus unbalanced test system is displayed in Fig. 8, and PLRP algorithm promotes the computing speed for the studied cases.
its system voltage is 4.16 kV. Assume that the peak load is 1.3 times that Because of the linear relaxation to the original problem, the adopted
of the original system. The candidate nodes of DG are all nodes in the PLRP model shows improved computational efficiency and solution
system. Candidate nodes for SOP installation are tie switches 54–94 and quality by ensuring convexity and linearization.
117–122. The maximum capacity of each candidate node of DG and SOP
is 1 MVA. The validity of the proposed model is verified on a large IEEE 6. Conclusions
123-bus distribution system with four cases.
The IEEE 123-bus unbalanced test system is displayed in Fig. 8, and The coordinated planning method of multiple flexible resources has
its system voltage is 4.16 kV. Assume that the peak load is 1.3 times that studied considering the risk of flexibility deficiency in the three-phase
of the original system. The candidate nodes of DG are all nodes in the unbalanced system. First, an ALRP method has proposed to solve the
system. Candidate nodes for SOP installation are tie switches 54–94 and problem of voltage and current phase angle of a three-phase power flow
117–122. The maximum capacity of each candidate node of DG and SOP model based on the semi-positive definite. Second, a coordinated plan-
is 1 MVA. The validity of the proposed model is verified on a large IEEE ning model of DG, SOP, and ESS in the three-phase UADN has estab-
123-bus distribution system with four cases. lished considering the risk of flexibility deficiency.
Table 5 presents the DG and SOP planning results for the IEEE 123 The proposed coordination of DG, SOP, and ESS could increase the
system. The DG capacities in Cases 1–4 are 2500, 3100, 4300, and installed capacity of DG through active and reactive power regulation,
3500 kVA, respectively. The SOPs installed in Cases 3 and 4 are 1300 reduce the total economic cost, and considerably improve the three-
and 1000 kVA, respectively. The results revealed that Case 3 improved phase unbalanced operation of distribution systems. The integrated
the acceptance capacity of the distribution network to DG. In Case 2, the economic cost of effectively coordinated DG, SOP, and ESS are reduced
three-phase unbalanced grid requires the installation of more SOP ca- by 31.06 % and 12.49 %; flexibility deficiency is reduced by 18.43 %
pacity to improve voltage distribution because reactive power cannot be and 48.26 %; and the three-phase voltage unbalance was reduced by
generated without ESS access. 15.45 % and 46.12 % in the IEEE 33-bus system and IEEE 123-bus
Table 6 presents the total cost, active power loss, and three-phase system compared with only planning DG.
voltage unbalance results for the IEEE 123 system. Because of the The flexibility deficiency risk, the three-phase unbalance, and the
installation of DG, SOP, and ESS units, the indexes of Case 3 are comprehensive economic cost of Case 3 are reduced by 31.94 %,
considerably better than those of Cases 1 and 2. Without considering the 15.94 %, and 6.22 %, respectively, compared with those of Case 4 in the
flexibility deficiency risk, the flexibility regulation ability and the three- IEEE 123-bus system. Thus, the optimal allocation of flexible resources

582
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

in the unbalanced active distribution network considering the risk of [8] K. Guerra, P. Haro, R.E. Gutiérrez, A. Gómez-Barea, Facing the high share of
variable renewable energy in the power system: flexibility and stability
flexibility deficiency is helpful to improve operation economy, security,
requirements, Appl. Energy 3 (10) (2022) 11–16.
and flexibility of the system. [9] N. Paul, The effects of power system flexibility on the efficient transition to
The comprehensive consideration of node and grid side flexibility renewable generation, Appl. Energy 28 (3) (2021) 116–128.
resources of the distribution network can flexibly optimize the three- [10] Jiang T., Wu C., Zhang R. Flexibility clearing in joint energy and flexibility markets
considering TSO-DSO coordination. IEEE Transactions on Smart Grid. doi:
phase power flow of the distribution network, which is helpful to 10.1109/TSG. 2022. 3153634.
develop a more reasonable planning scheme of DG, SOP, and ESS in the [11] Z. Lu, H. Li, Y. Qiao, Probabilistic flexibility evaluation for power system planning
UADN, enhance the economy and flexibility of the distribution network considering its association with renewable power curtailment, IEEE Trans. Power
Syst. 33 (3) (2018) 3285–3295.
operation, improve the voltage quality, and promote the consumption of [12] H.K. Wang, S.X. Wang, et al., Optimized dispatching method for flexibility
DG. improvement of distribution network with high-penetration distributed generation,
Moreover, the uncertainty of DG is not considered in the process of Autom. Electr. Power Syst. 42 (15) (2018) 86–93.
[13] C. Gianni, P. Fabrizio, P. Giuditta, Risk-oriented planning for flexibility-based
model establishment in this paper, and it is more real and reliable to use distribution system development, Sustain. Energy, Grids Netw. (30) (2022)
stochastic programming or robust technology to explore the impact of 100–114.
extreme uncertainty of DG on the operation security of distribution [14] X. Ye, et al., Planning for energy storage system considering risk of inflexibility in
active distribution network, 2019 IEEE 3rd Int. Electr. Energy Conf. (CIEEC) (2019)
network. Meanwhile, for the three-phase regulation of discrete variable 259–263.
devices, an UADN operation strategy and optimization planning based [15] R. Faia, B. Canizes, P. Faria, Distribution network expansion planning considering
on parameter linear relaxation algorithm should be also worthy of the flexibility value for distribution system operator, 2019 Int. Conf. Smart Energy
Syst. Technol. (SEST) (2019) 1–6.
further studied, and the effectiveness of the model and method should be
[16] Q. Sun, Z. Wu, W. Gu, Flexible expansion planning of distribution system
verified through various distribution network topology cases to improve integrating multiple renewable energy sources: an approximate dynamic
the flexibility of scheduling, the economy and safety of system programming approach, Energy 22 (6) (2021) 120–217.
operation. [17] H.R. Ji, C.S. Wang, P. Li, Quantified analysis method for operational flexibility of
active distribution networks with high penetration of distributed generators, Appl.
Energy 239 (2019) 706–714.
CRediT authorship contribution statement [18] X. Chen, E. Dall’Anese, C. Zhao, Aggregate power flexibility in unbalanced
distribution systems, IEEE Trans. Smart Grid 11 (1) (2020) 258–269.
[19] L. Zhang, C. Shen, Y. Chen, Coordinated allocation of distributed generation,
Jie Wang: Writing – review & editing, Writing – original draft, capacitor banks and soft open points in active distribution networks considering
Validation, Software, Methodology, Formal analysis, Conceptualization. dispatching results, Appl. Energy 231 (2018) 1122–1131.
[20] J. Wang, N. Zhou, C.Y. Chung, Coordinated planning of converter-based DG units
Xiaochao Fan: Validation, Resources, Methodology, Conceptualization. and Soft open points incorporating active management in unbalanced distribution
Lijun Xu: Writing – review & editing, Validation, Resources, Concep- networks, IEEE Trans. Sustain. Energy 11 (3) (2020) 2015–2027.
tualization. Ruijing Shi: Writing – review & editing, Validation, [21] V.B. Pamshetti, S.P. Singh, Coordinated allocation of BESS and SOP in high PV
penetrated distribution network incorporating DR and CVR schemes, IEEE Syst. J.
Methodology, Data curation. Weiqing Wang: Writing – review & edit- 16 (1) (2022) 420–430.
ing, Validation, Software, Resources. [22] J.M. Arroyo, Bilevel programming applied to power system vulnerability analysis
under multiple contingencies, IET Gener. Transm. Distrib. 4 (2) (2010) 178–190.
[23] M. Asensio, G. Muñoz-Delgado, J. Contreras, Bi-Level approach to distribution
network and renewable energy expansion planning considering demand response,
Declaration of Competing Interest IEEE Trans. Power Syst. 32 (6) (2017) 4298–4309.
[24] Q. Qi, J. Wu, C. Long, Multi-objective operation optimization of an electrical
distribution network with soft open point, Appl. Energy 208 (15):) (2017)
The authors declare that they have no known competing financial
734–744.
interests or personal relationships that could have appeared to influence [25] Avirup Maulik, Probabilistic power management of a grid-connected microgrid
the work reported in this paper. considering electric vehicles, demand response, smart transformers, and soft open
points, Sustain. Energy Grids Netw. 30 (2022) 100–636.
[26] H.T. Zheng, T.J. Shi, Bi-level optimization of distribution network based on soft
Acknowledgments open point and reactive power compensation device is presented, Autom. Electr.
Power Syst. 43 (2019) 117–123.
[27] A. Bernstein, E. Dall’Anese, Linear power-flow models in multiphase distribution
This work is supported by Tianchi talent introduction plan project of networks, 2017 IEEE PES Innov. Smart Grid Technol. Conf. Eur. (ISGT-Eur.) (2017)
Xinjiang Autonomous region (2024XGYTCYC12), in part by the Na- 1–6.
tional Natural Science Foundation of China (52266018), in part by the [28] Z. Yang, H. Zhong, A. Bose, et al., A linearized OPF model with reactive power and
voltage magnitude: a pathway to improve the MW-Only DC OPF, IEEE Trans.
doctoral project of Xinjiang Institute of Engineering under Grant Power Syst. 33 (02) (2018) 1734–1745.
2024XGYBQJ01, in part by the Key research and development project of [29] S. Bolognani, S. Zampieri, On the existence and linear approximation of the power
Xinjiang Uygur Autonomous Region under Grant 2022B01018-1. flow solution in power distribution networks, IEEE Trans. Power Syst. 31 (01)
(2016) 163–172.
[30] Lipka P. Castill Ao, J. Watson, et al., A successive linear programming approach to
References solving the IV-ACOPF, IEEE Trans. Power Syst. 31 (04) (2016) 2752–2763.
[31] M. Zhao, M. Barati, Low-order moment relaxation of ACOPF via algorithmic
successive linear programming, 2021 IEEE Tex. Power Energy Conf. (TPEC) (2021)
[1] Z. Bingxu, P.F. Duan, M.D. Fen, Optimal operation of distribution networks and
1–6.
multiple community energy prosumers based on mixed game theory, Energy 278
[32] X. Bai, H. Wei, Semi-definite programming-based method for security-constrained
(2023) (Part B).
unit commitment with operational and optimal power flow constraints, IET Gener.
[2] H.R. Ji, C.S. Wang, P. Li, et al., Quantified analysis method for operational
Transm. Distrib. 3 (2) (2009) 182–197.
flexibility of active distribution networks with high penetration of distributed
[33] R.A. Jabr, Radial distribution load flow using conic programming, EEE Trans.
generators, Appl. Energy 23 (9) (2019) 706–714.
Power Syst. 21 (03) (2006) 1458–1459.
[3] S. Dalhues, et al., Research and practice of flexibility in distribution systems: a
[34] Zhao J., Li Y., Li P., et al. Sequential voltage regulation of soft normally open point
review, CSEE J. Power Energy Syst. 5 (3) (2021) 285–294.
in active distribution network based on second-order cone programming. High
[4] Li H.B., Lu Z.X., Qiao Y., et al. Bi-level optimal planning of generation-load-storage
Voltage Engineering, 2016, 42(07):2134-2141．.
integrated generalized flexibility resource. Automation of Electric Power Systems
[35] L. Gan, S.H. Low, Convex relaxations and linear approximation for optimal power
2017, 41(21): 46-54(in Chinese).
flow in multiphase radial networks, 2014 Power Syst. Comput. Conf. (2014) 1–9.
[5] Prieto-Araujo E., Junyent-Ferré A, G. Clariana-Colet. Control of modular multilevel
[36] Wang Z., Kirschen D.S., Zhang B. Accurate semidefinite programming models for
converters under singular unbalanced voltage conditions with equal positive and
optimal power flow in distribution systems. arXiv e-prints, 2017.
negative sequence components. 2020 IEEE Power & Energy Society General
[37] Liu J., Cheng H.Z., Zeng P.L., et al. Decentralized stochastic optimization based
Meeting (PESGM) 2020:1-1.
planning of integrated transmission and distribution networks with distributed
[6] A. Zhou, H. Zhai, M. Yang, Three-phase unbalanced distribution network dynamic
generation penetration. Applied Energy 2018, 220:800-813, https: //doi.org/
reconfiguration: a distributionally robust approach, IEEE Trans. Smart Grid 13 (3)
10.1016/j. apenergy. 2018.03.016.
(2022) 2063–2074.
[7] S.M. Fazeli, H.W. Ping, A.R.N. Bin, et al., Individual-phase control of 3-phase 4-
wire voltage-source converter, Power Electron. IET 7 (9) (2014) 2354–2364.

583
J. Wang et al. Alexandria Engineering Journal 112 (2025) 569–584

[38] A. Hamidreza, S.S. Mohammad, V. Vahid, An aggregated model for coordinated [44] R.R. Jha, A. Dubey, Network-Level optimization for unbalanced power distribution
planning and reconfiguration of electric distribution networks, Energy 94 (2016) system: approximation and Relaxation, IEEE Trans. Power Syst. 36 (05) (2021)
786–798. 4126–4139.
[39] He H.J., Ershun D., Zhang Nik, et al, 2021. Enhancing the power grid flexibility [45] M. Pirnia, C.A. Canizares, K. Bhattacharya, et al., A novel affine arithmetic method
with battery energy storage transportation and transmission switching. Applied to solve optimal power flow problems with uncertainties, IEEE Trans. Power Syst.
Energy 290. 29 (6) (2014) 2775–2783.
[40] Y. Gao, Q. Ai, Y. Muhammad, et al., Source-load-storage consistency collaborative [46] W. Fan, Z.F. Tan, F.Q. Li, et al., A two-stage optimal scheduling model of integrated
optimization control of flexible DC distribution network considering multi-energy energy system based on CVaR theory implementing integrated demand response,
complementarity, Int. J. Electr. Power Energy Syst. 107 (2019) 273–281. Energy 263 (2023) 125–133, https://doi.org/10.1016/j.energy. 2022. 125783.
[41] H. Bastami, M. Shakarami, M. Doostizadeh, A decentralized cooperative [47] X. Ran, J. Zhang, K. Liu, An interval–probabilistic CVaR (IP-CVaR) and modelling
framework for multi-area active distribution network in presence of inter-area soft for unknown probability distribution of some random variables, IEEE Trans. Power
open points, Appl. Energy 300 (2021). Syst. 38 (03) (2023) 2035–2045.
[42] W.Y. Cao, J.Z. Wu, N. Jenkins, Operating principle of soft open points for electrical [48] K.C. Bevin, A. Verma, Decentralized local electricity market model using
distribution network operation, Appl. Energy 164 (2016) 245–257. automated market maker, Appl. Energy 334 (2023) 120–131, https://doi.org/
[43] Deakin M., Sarantakos I., David G.W., et al. Comparative analysis of services from 10.1016/j.apenergy. 2023. 120689.
soft open points using cost–benefit analysis. Applied Energy 2023, 333:120-128, [49] F.Z. Mehdi, S.N. Mehrdad, S. Miadreza, Integrated framework for modeling the
https://doi.org/10. 1016/j. apenergy. 2022.120618. interactions of plug-in hybrid electric vehicles aggregators, parking lots and
distributed generation facilities in electricity markets, Appl. Energy 334 (2023)
120703, https://doi.org/10.1016/j.apenergy, 2023.120703.

584