Wu et al.

Journal of Big Data (2024) 11:172 Journal of Big Data

https://doi.org/10.1186/s40537-024-01034-0

RESEARCH Open Access

Dual‑weight decay mechanism

and Nelder‑Mead simplex boosted RIME
algorithm for optimal power flow
Huangying Wu1, Yi Chen1, Zhennao Cai1*, Ali Asghar Heidari2, Huiling Chen1* and Guoxi Liang3*

*Correspondence:
[email protected];
Abstract
[email protected]; The increasing demand for electricity presents substantial challenges in power system
[email protected]
planning, particularly optimizing the Optimal Power Flow (OPF) problem. The OPF
1
Department of Computer problem entails establishing the best settings for control variables in a power system
Science and Artificial
Intelligence, Wenzhou University, to reduce objectives such as generating cost and transmission losses while meet-
Wenzhou 325035, China ing operational restrictions. This research introduces an upgraded RIME optimization
2
School of Surveying algorithm (WDNMRIME) to address these challenges. WDNMRIME integrates a dual-
and Geospatial Engineering,
College of Engineering, weight decay mechanism and the Nelder-Mead simplex (NMs), enhancing popula-
University of Tehran, Tehran, Iran tion diversity and mitigating the risk of local optima. Additionally, NMs expedites
3
Department of Artificial convergence by refining the population’s optimal solution set. Experimental validation
Intelligence, Wenzhou
Polytechnic, Wenzhou 325035, on the IEEE 30-bus test system demonstrates that WDNMRIME achieves a generation
China cost of $806.00298 per hour and reduces total power loss from 1.43 MW to 1.39 MW.
These results surpass the performance of the original RIME algorithm, showcasing
a 15% improvement in convergence speed. The algorithm effectively optimizes multi-
ple concurrent Flexible Alternating Current Transmission Systems (FACTS) devices, even
under the uncertain nature of wind energy resources modeled using the Weibull prob-
ability density function. These findings highlight WDNMRIME’s significant contribution
to improving OPF optimization in dynamic power systems.
Keywords: FACTS devices, RIME, Optimal power flow, Global optimization

Introduction
Society’s emphasis on environmental protection has led scholars to develop a wide
interest in renewable energy sources [1]. However, the current power grid faces chal-
lenges, such as the increasing demand for distributed generation and congestion, mak-
ing it essential to ensure safety, power quality, efficiency, and stability [2]. Modern power
systems are increasingly adopting Flexible AC Transmission System (FACTS) devices
to address these issues. Integrating the FACTS technology combination allows for the
flexible and rapid control of AC power transmission. FACTS transforms initially mini-
mally controllable power networks into fully controllable ones by dynamically regulating
power, voltage, flow, impedance, and phase angle [3].

© The Author(s) 2024. Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0
International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long
as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you
modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of
it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise
in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted
by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy
of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by-​nc-​nd/4.​0/.
Wu et al. Journal of Big Data (2024) 11:172 Page 2 of 33

When it comes to the operation and planning of power systems, the Optimal Power
Flow (OPF) problem emerges as a prominent challenge that is pivotal for ensuring reli-
ability and cost-effectiveness [4, 5]. It finds wide application in areas like secure system
operation, economic dispatch, grid planning, and reliability analysis, combining security
and economic considerations effectively [6, 7]. This optimization is typically proactive,
minimizing generation costs, or reactive, minimizing network losses. Researchers have
investigated a range of techniques, including nonlinear programming, quadratic pro-
gramming, and integer programming. Despite these efforts, these methods encounter
limitations in effectively addressing the problem [8]. Linear programming approaches
struggle to capture the intricacies of OPF, leading to significant errors and rendering
them unsuitable for resolution [9, 10]
To solve these problems, researchers have come up with many efficient algorithms,
such as Particle Swarm Optimization (PSO) [11] and Differential Evolution (DE) [12], to
newer approaches such as RIME optimization algorithm (RIME) [13], Weighted Mean of
Vectors (INFO) [14], Runge Kutta Optimizer (RUN) [15], Hunger Games Search (HGS)
[16], Harris Hawks Optimization (HHO) [17], Slime Mould Algorithm (SMA) [18, 19],
Colony Predation Algorithm (CPA) [20], Liver Cancer Algorithm (LCA) [21], and Parrot
Optimizer (PO) [22]. This category of algorithms has demonstrated outstanding opti-
mization performance in various optimization domains, such as the dispatch problem
[23], economic emission [23, 24], etc. However, these metaheuristic algorithms have
slow convergence speeds and insufficient accuracy, requiring appropriate classification
and configuration for effective optimization [25, 26]. Consequently, researchers have
introduced enhanced and hybrid variants of these algorithms, demonstrating notable
improvements in solving the OPF problem. These innovative approaches have shown
promising performance and effectiveness in tackling the challenges posed by the OPF,
leading to more reliable and efficient solutions. Ehsan et al. [27] proposed and care-
fully studied a novel fuzzy adaptive hybrid configuration algorithm for the self-adap-
tive PSO and DE algorithms to solve the multi-objective OPF problem. Birogu et al.
[28] introduced a hybrid algorithm based on HHO and DE. Taher et al. [29] proposed
an improved multi-flower pollination algorithm to solve the OPF problem effectively.
Birchfield et al. [30] presented a memory-based association learning mechanism and
lévy flight mechanism in the moth flame optimization and demonstrated its effective-
ness. Attia et al. [31] incorporated the levy flight mechanism into the original sine cosine
algorithm, improving its convergence speed and demonstrating its effectiveness. Naderi
et al. [32] proposed a fuzzy-based enhanced comprehensive learning PSO algorithm and
provided evidence of its efficacy. Yuan et al. [33] proposed an improved strength Pareto
evolutionary algorithm for multi-objective OPF. Roy et al. [34] proposed a biogeogra-
phy-based optimization approach for the OPF problem. Niknam et al. [35] introduced
an improved shuffled frog leaping algorithm for multi-objective OPF. Narimani et al.
[36] proposed a hybrid algorithm based on particle swarm optimization and shuffle frog
leaping algorithm for OPF and demonstrated its effectiveness. Naderi et al. [37] intro-
duced a unique fuzzy adaptive heterogeneous comprehensive learning particle swarm
optimization technique for the OPF problem, and showed its efficacy. Naderi et al. [38]
developed a hybrid method designed to the shuffling frog leaping procedure and PSO,
named hybrid wavelet mutation-based SFLA-wavelet mutation-based PSO. Weng et al.
Wu et al. Journal of Big Data (2024) 11:172 Page 3 of 33

[39] proposed a cross-pattern search algorithm-based improved sine cosine algorithm

for optimizing multiple objectives in the OPF problem. Given the reported results in the
literature, further research in this field is promising and encouraging. Naderi et al. [40]
proposed a hybrid fuzzy-based enhanced comprehensive learning PSO-DE, to address
the OPF. The advantages and disadvantages of all algorithms are summarized in Table 1.
Considering all the aforementioned improvements and newly proposed algorithms,
this paper introduces an enhanced version of RIME, termed WDNMRIME. This algo-
rithm integrates advanced enhancement techniques and the latest algorithms to tackle
the complexity and diversity challenges in optimization problems. Firstly, we present a
central weight decay mechanism that modulates the probability and magnitude of indi-
vidual movements by incorporating the individual’s fitness and distance from the central
position, using a precise decay factor. This mechanism improves the algorithm’s global
search capability and enhances population diversity, thereby effectively promoting the
exploration and discovery of solutions. Secondly, we employ the Nelder-Mead simplex
(NMs) mechanism to concentrate the algorithm’s exploitation efforts on the solution
space near the optimal solution. This approach ensures a more efficient and focused
search, further improving the algorithm’s overall performance. This mechanism helps
improve the convergence accuracy of the algorithm, allowing it to approach the opti-
mal solution more effectively and obtain superior solutions within a limited number of
iterations. To assess the effectiveness of WDNMRIME, the study employs it to optimize
three categories of FACTS devices: Thyristor-Controlled Series Compensator (TCSC),
Thyristor-Controlled Parallel Compensator (TCPS), and Static Var Compensator (SVC).
The optimization objective aims to minimize the generation cost and power transmis-
sion losses. The paper employs a suitable wind speed probability function to model

Table 1 References summary

Refs. Basic algorithm Advantage Disadvantages

[27] PSO + DE Strong adaptability and multi-objective High computational complexity

optimization
[28] HHO + DE Enhanced global search capabilities, balanc- Difficulty in parameter selection and
ing exploration and development high computational complexity
[29] MFO Parameter adaptation Local optima trap
[30] MFO Adapting to complex problems Difficult parameter setting
[31] SCA Accelerate convergence speed, simple and Parameter selection dependencies
effective
[32] PSO Fuzzy enhancement, comprehensive learning Increased complexity
[33] SPEA Multi-objective optimization, Pareto optimiza- Increased dependencies
tion
[34] BBO Strong global search capability, multi-modal High computational complexity, local
optimization problems optimality problem
[35] SFLA High optimization efficiency Parameter setting challenges
[36] PSO + SFLA Excellent search efficiency and convergence Hybrid algorithm complexity increases
speed
[37] PSO Fuzzy adaptability, heterogeneous compre- Parameter tuning challenges
hensive learning
[38] PSO + SFLA Multi-format OPF problem-solving skills Parameter tuning challenges
[39] SCA Strong global search capability Computing resource consumption
[40] PSO Fuzzy Adaptability High computational complexity
Wu et al. Journal of Big Data (2024) 11:172 Page 4 of 33

the randomly generated wind speeds in the optimization process. This incorporation
allows for a realistic representation of the uncertainty associated with renewable energy
sources. Additionally, the study considers reserve and penalty costs to simulate sce-
narios of potential overestimation and underestimation of uncertain renewable energy
generation. By factoring in these costs, the optimization model becomes more robust
and capable of handling the variability inherent in renewable energy resources. To tackle
the constraints inherent in the OPF problem, the paper leverages the advantages of the
Superior Feasible (SF) solutions approach. Combining this SF treatment with the pro-
posed WDNMRIME algorithm allows the researchers to discover optimal solutions for
the OPF problem under various objective functions. Through optimizing actual param-
eters within the power system, the paper achieves significant cost reductions, showcas-
ing the practicality and effectiveness of their algorithm. This paper makes the following
contributions:

1. This paper applies the RIME to the OPF problem for the first time, leveraging its
global search capability and diversity in solving the OPF problem.
2. A central weight decay mechanism is introduced in the RIME to enhance population
diversity and facilitate solution discovery. Using an accurate decay factor, this mech-
anism regulates the probability and magnitude of individual movement by consid-
ering the individual’s fitness and its distance from the central position. It effectively
maintains population diversity during search and prevents premature convergence to
local optima.
3. To improve the convergence accuracy of the RIME and its ability to solve the OPF
problem, the NMs mechanism is introduced. By combining the NMs mechanism
with the RIME, this paper improves the effectiveness of solving complex OPF prob-
lems.
4. The enhanced WDNMRIME is employed for devices like SVC, TCSC, and TCPS,
analyzing its capability to address OPF objectives across various objective functions.
Through these application experiments, the paper evaluates the performance of the
algorithm in practical scenarios and demonstrates its effectiveness in reducing gen-
eration costs and power transmission losses.

The paper’s structure unfolds as follows: Initially, it delves into utilizing swarm intel-
ligence algorithms in OPF. Subsequently, comprehensive modeling definitions of FACTS
devices are outlined. The fifth section elaborates on the WDNMRIME. Experimental
results and analysis are then presented in the sixth section, and lastly, the study con-
cludes by summarizing findings and discussing future prospects.

FACTS devices modeling

TCSC
By manipulating the control signal, the TCSC effectively exerts control over the distri-
bution of current within the system. In this paper, the effective reactance of TCSC can
be expressed by using Eq. (1), XC represents the fixed capacitive reactance and XL (γ )
represents the variable inductive reactance, and Fig. 1 shows the basic circuit diagram of
TCSC [41, 42].
Wu et al. Journal of Big Data (2024) 11:172 Page 5 of 33

XL

Line
XC
Fig. 1 TCSC equivalent circuit

Fig. 2 TCSC circuit model

XC XL (γ )
XTCSC (γ ) = (1)
XL (γ ) − XC

In this paper, this paper uses Fig. 2 to represent the schematic diagram of the TCSC
model. Equation (2) is the value of transmission line correction reactance , Xmn denotes the
line inductive reactance. Additionally, Eq. (3) involves the parameter λ, which represents
the degree of series compensation [43, 44].

X = Xmn − XTCSC = (1 − )Xmn (2)

XTCSC
= (3)
Xmn

In this paper, Eqs. (4–7) are employed to define the active and reactive power flow
between buses m and n via TCSC. These equations serve as the basis for expressing TCSC’s
line current equations. Here, Vm and Vn represent the voltage magnitudes at buses m and
n respectively, while δm and δn denote the phase angles at those buses [45]. Additionally,
Eq. (8) defines Gmn as the line conductance, and Eq. (9) expresses Bmn as the susceptance
coefficient, where Rmn signifies the resistance [46].

Pmn = Vm2 Gmn − Vm Vn Gmn cos(δm − δn ) − Vm Vn Bmn sin(δm − δn ) (4)

Qmn = −Vm2 Bmn − Vm Vn Gmn sin(δm − δn ) + Vm Vn Bmn cos(δm − δn ) (5)

Pnm = Vn2 Gmn − Vm Vn Gmn cos(δm − δn ) + Vm Vn Bmn sin(δm − δn ) (6)

Wu et al. Journal of Big Data (2024) 11:172 Page 6 of 33

Qmn = −Vn2 Bmn + Vm Vn Gmn sin(δm − δn ) + Vm Vn Bmn cos(δm − δn ) (7)

Rmn
Gmn = (8)
2
Rmn + [(1 − )Xmn ]2

(1 − )Xmn
Bmn = −
2 + [(1 − )X ]2 (9)
Rmn mn

TCPS
TCPS is an apparatus employed within power systems to govern power transmission,
accomplished by manipulating the phase angle between voltage and current waveforms
[47, 48]. It is commonly employed in high-voltage transmission systems to regulate
power flow, enhance system stability, and control reactive power. Equations (10–11)
delineate the flow equations, while the flow equations are expressed by Eqs. (12–13) [49].

Vm2 Gmn Vm Vn
Pmn = 2
− [Gmn cos(δm − δn + �) + Bmn sin(δm − δn + �)] (10)
cos � cos�

Vm2 Bmn Vm Vn
Qmn = − 2
− [Gmn cos(δm − δn + �) − Bmn cos(δm − δn + �)] (11)
cos � cos�

Vm Vn
Pnm = Vn 2 Gmn − [Gmn cos(δm − δn + �) − Bmn sin(δm − δn + �)] (12)
cos�

Vm Vn
Qnm = −Vn 2 Bmn + [Gmn sin(δm − δn + �) + Bmn cos(δm − δn + �)] (13)
cos�

The reactive and active power of TCPS between bus m and n are expressed by Eqs.
(14–17), with the Φ signifying the phase shift angle of the TCPS. These equations com-
prehensively understand the TCPS device’s impact on the system’s power flow dynamics.
Figure 3 illustrates a schematic diagram of the TCPS model.

Fig. 3 Equivalent structural model of TCPS

Wu et al. Journal of Big Data (2024) 11:172 Page 7 of 33

Pms = −Gmn Vm2 tan2 � − Vm Vn tan�[Gmn sin(δm − δn ) − Bmn cos(δm − δn )] (14)

Qms = Bmn Vm2 tan2 � + Vm Vn tan�[Gmn cos(δm − δn ) + Bmn sin(δm − δn )] (15)

Pns = −Vm Vn tan�[Gmn sin(δm − δn ) + Bmn sin(δm − δn )] (16)

Qns = −Vm Vn tan�[Gmn cos(δm − δn ) − Bmn sin(δm − δn )] (17)

SVC
SVC is a power electronic device used in power systems to control and regulate reac-
tive power. It primarily enhances voltage stability, improves power quality, and supports
power transmission and distribution [50]. Figure 4 illustrates the basic model of SVC, and
Eq. (18) represents its line equivalent reactance, where γ represents the ignition angle of the
thyristor.

BSVC = BC + BL (γ ) (18)

BC = ωC (19)

 
1 2γ sin2γ
BL (γ ) = 1− − (20)
ωL π π

For power flow calculation problems, the representation of reactive power supplied by
SVC is as follows:

QSVC = −Vm2 BSVC (21)

Fig. 4 Equivalent structural model of SVC

Wu et al. Journal of Big Data (2024) 11:172 Page 8 of 33

SVC can respond quickly to changes in system conditions, effectively regulate volt-
age, and control power factors. It can mitigate voltage fluctuations, suppress voltage
oscillations, and enhance transient stability [51]. Additionally, SVC is commonly used
in industrial facilities with load fluctuations, such as steel mills and mining operations,
to maintain a stable power supply and improve power quality. Overall, the SVC plays a
crucial role in balancing the power system and supporting efficient power transmission
and distribution through dynamic control of reactive power. It is essential for maintain-
ing the reliability and stability of the power system [52].

Problem formulation
This research focuses on exploring system optimization within the IEEE 30-bus test
system, widely acknowledged as a representation of power systems in the field. A criti-
cal aspect of the optimization procedure includes the incorporation of wind turbines,
in particular FACTS devices, into this system [53]. Key parameters associated with this
optimization configuration are succinctly summarized in Table 2. Additionally, detailed
explanations are provided in the paper concerning the precise arrangement of generator
units, in particular FACTS devices, within the context of the IEEE 30 system [54].

Thermal generator fuel costs

Equation (22) represents the electricity generation cost ($/h) for the i-th thermal power
generator, where ai,bi and ci are the cost coefficients for the i-th power generator group
PTGi.
2
CTOi (PTGi ) = ai + bi PTGi + ci PTGi (22)

While Eq. (22) has practical applications, it must be acknowledged that it has limitations
in omitting valve point loading. Ignoring such loads subsequently indirectly impacts the
overall input–output cost curve [56]. To alleviate this concern and address the aforemen-
tioned drawbacks, further expansion is proposed in Eq. (23). This improved formulation of
the objective function incorporates valve effects, thus encompassing a more comprehen-
sive representation of system dynamics. Here, di and ei are the calculation coefficients for
the valve point loading PTGi. PTGi
min
represents the minimum power generation for the i-th

Table 2 IEEE30 system

Components Description Quantity

Generation Buses (TG) Buses: 1, 2, 8, and 13 4

Wind Generation Buses (WG) Buses: 5 and 11 2
Branches See Appendix Table 10 for details [55] 41
Tap Changing Transformers Branches with tap changing capability: 11, 12, 15, 36 4
TCPS Devices Branches: 24 and 25 2
Load Demand Total power: 283.4 MW, 126.2 MVAr -
SVC Devices Branches: 16 and 17 2
TCSC Devices Branches: 20 and 21 2
Load Bus Voltage Range Range of voltage for load buses: [0.95–1.05] p.u 24
Wu et al. Journal of Big Data (2024) 11:172 Page 9 of 33

generator. Table 3 illustrates the parameters utilized for cost computation in each generator
cluster.
   
2 min
CTi (PTGi ) = ai + bi PTGi + ci PTGi + di × sin ei × PTGi − PTGi  (23)
 

Model of wind power probabilities

The Weibull Probability Density Function (PDF) is utilized to depict the probability distri-
bution of continuous random variables [57]. It is commonly employed to model the lifespan
or failure characteristics of various systems and phenomena. The Weibull PDF is often uti-
lized in reliability analysis to simulate the failure rate of components or systems over time. It
allows for flexible capturing of different failure modes, ranging from early failures to wear-
out failures. Within this paper, the purpose entails the emulation of unfamiliar wind veloci-
ties [58]. Equation (24) represents this very purpose, wherein the scaling coefficient and
configurational factor find representation as c and k, correspondingly.
   
k v k−1  −( v )k 
fv (v) = × × e c 0<v<∞ (24)
c c

In this study, each turbine is allocated a fixed power output of 3 MW, as denoted by
Eq. (25). Here, vr , vout , and vin respectively signify the turbine’s rated wind speed, cut-out
wind speed, and cut-in wind speed. Pwr represents the rated output of an individual wind
turbine. For this investigation, vin is prescribed as 3 m/s,vout as 25 m/s,vr ​as 16 m/s. Equa-
tion (25) comprehensively illustrates the correlation between wind speed and turbine power
output, delineating three distinct operational segments [59].

0
 � �v < vin andv > vout
v−vin
pw (v) = pwr vr −vin vin ≤ v ≤ vr (25)

 p
wr vr < v ≤ vout

In this study, Eqs. (26–28) are employed to compute the probabilities associated with
wind power reaching either zero or its rated capacity.
      
vin k vout k
fw (pw ){pw = 0} = 1 − exp − + exp − (26)
c c

      
vr k vout k
fw (pw ){pw = pwr } = exp − − exp − (27)
c c

Table 3 Genset cost parameters

Bus Generators a ($/h) b ($/MWh) c ($/MW2h) d ($/h) e (rad/MW)

1 TG1 0 2 0.00375 18 0.037

2 TG2 0 1.75 0.0175 16 0.038
8 TG8 0 3.25 0.00834 12 0.045
13 TG13 0 3 0.025 13.5 0.041
Wu et al. Journal of Big Data (2024) 11:172 Page 10 of 33

� � � � � �k−1
k × (vr − vin ) pw
fw (pw ) = × vin + × (vr − vin )
ck × pwr pwr
k 
(28)
  � �
pw
vin + pwr × (vr − vin )
× exp−
  
c


Direct cost assessment of wind power plants

Equation (29) delineates the incurred expenses linked with wind energy production at
the i-th wind farm. Here, PWS,i denotes the anticipated power output from the wind
farm, while wi represents its respective loss factor.
 
Cw,i PWS,i = wi PWS,i (29)

Assessment of uncertainty costs in wind power

The inherent unpredictability of wind energy poses challenges when the realized power
output of wind turbines diverges from anticipated values. Consequently, independent
system operators (ISOs) are prompted to formulate tailored policies to manage such dis-
crepancies. For instance, in cases where a wind farm exceeds its initial electricity gen-
eration projections, the ISO implements reserve measures to regulate the surplus supply
[60, 61]. In this study, Eq. (30) is used to evaluate the uncertain cost of wind power
output.

   
 Pws,i    
CRw,i PWs,i − Pwav,i = KRw,i PWs,i − Pwav,i = KRw,i PWS,i − pw,i fw pw,i dpw,i
0
(30)
   
 Pwr,i    
CPw,i Pwav,i − PWS,i = KPw,i Pwav,i − PWS,i = KPw,i pw,i − PWS,i fw pw,i dpw,i
pws,i
(31)
KRw,i represents the reserve cost coefficient for the i-th wind farm; its value is set to
3 in this paper [55]. Pwav,i probability density function reflecting the wind power dis-
tribution for the i-th turbine. Equation (31) is used in this study to represent the pen-
alty cost for the excess electricity delivered to the grid when the wind farm’s generation
falls below expectations [62]. KPw,i represents the penalty cost factor concerning the i-th
wind farm, its value is set to 1.5 in this paper.

Formulation
Objectives
Aggregate generation cost analysis
The main objective of this study is to minimize the combined cost of fuel consumption
and wind power generation. This study focuses on reducing the total cost of electricity
generation by integrating the secondary cost of thermal power [as shown in Eq. (23)],
direct costs [given in Eq. (29)], reserve costs [calculated in Eq. (30)], and penalty costs
[described in Eq. (31)]. The comprehensive representation of the total cost of electricity
Wu et al. Journal of Big Data (2024) 11:172 Page 11 of 33

generation is clearly given in Eq. (32). Where NTG represents the number of fuel genera-
tors and NWG represents the number of wind turbines in the system.
NTG NWG      
Cgen = CTi (PTGi ) + Cw,i Pws,i + CRw,i Pws,i − Pwav,i + CPw,i (Pwav,i − Pws,i )
i=1 i=1
(32)

Assessing real active power losses

The main goal of this study is to reduce the incurred power losses, as measured by Eq. (33).
The transmission conductance of the m-th branch route connecting buses i and j is
denoted by Gm(i,j).
nl  
Gm(ij) Vi 2 + Vj2 − 2Vi Vj cos δi − δj (33)

Ploss =
m=1

In pursuit of attaining the optimal resolution for the aforementioned objective within the
OPF problem, this study utilizes an optimization algorithm [63]. It is essential to emphasize
that throughout the entire solving process, multiple intricate equations and inequality con-
straints must be adhered to ensure the accuracy and feasibility of the obtained results.

Equivalence constraint
When accounting for the integration of FACTS devices, the power equilibrium constraints
can be mathematically articulated as follows:
NB NTCPS NB NB     
i=1
(PGi − PLi ) +
i=1
Pis =
i=1 j=1
|Vi |Vj Yij cos θij + δij (34)

NB NTCPS NB NB     

i=1
(QGi − QLi ) +
i=1
Qis = − |Vi |Vj Yij sin θij + δij
i=1 j=1
(35)

The active and reactive power of the generating unit of group I is represented by PGi and
QGi, respectively. PLi and QLi represent the active and reactive load demand at bus i. Pis and
Qis represent the active and reactive power of the TCPS at bus i. The conductance of the
transmission line is symbolized as Yij, while the relative phase angle of this connection is
represented by θij. Moreover, δij signifies the voltage magnitude difference between bus i
and bus j. The total number of buses in the system is denoted by NB.

Inequality constraints
Inequality constraints establish the maximum and minimum boundaries for the active and
reactive power outputs from generator clusters, along with the voltage magnitude of these
designated generator clusters [64]. These limits are mathematically depicted by Eq. (36–38).
min max
PGi ≤ PGi ≤ PGi , ∀i ∈ NG (36)

min max
QGi ≤ QGi ≤ QGi , ∀i ∈ NG (37)

min max
VGi ≤ VGi ≤ VGi , ∀i ∈ NG (38)
Wu et al. Journal of Big Data (2024) 11:172 Page 12 of 33

The inequality constraint for transformers can be represented by Eq. (39):

Ttmin ≤ Tt ≤ Ttmax , ∀i ∈ NT (39)

To ensure the reliability of the wind power generation system, specific safety condi-
tions must be adhered to for each transmission line. These conditions are of utmost
importance as they can reduce the risk of operational or catastrophic failures [65].
The inequality constraint conditions, which encompass these crucial safety considera-
tions, can be formulated as follows:
min max
VLp ≤ VLp ≤ VLp , ∀p ∈ NL (40)

max
Slp ≤ Slq , ∀p ∈ NL (41)

In the inequality constraints mentioned above, VLp

min
and VLpmax
represent the lower
and upper limits, respectively, of the voltage magnitude at load bus i. Moreover, Slq
and Slq
max
correspond to the apparent power flow and the maximum permissible limit
of apparent power flow on line i. It is crucial to emphasize that NG and NT signify the
overall count of generator buses and transformers within the system, respectively. The
acceptable operational ranges of TCSC, TCPS, and SVC are determined by the subse-
quent inequality constraints, where NTCSC, NTCPS, and NSVC denote the respec-
tive quantities of three devices existing in the system [66, 67].

TCSC: min max

TCSCm ≤ TCSCm ≤ TCSCm , ∀m ∈ NTCSC (42)

TCPS: min max

TCPSn ≤ TCPSn ≤ TCPSn , ∀n ∈ NTCPS (43)

SVC: Qmin max

SV Cj ≤ QSVCj ≤ QSVCj , ∀j ∈ NSVC (44)

FACTS location
When adding FACTS devices, the following conditions must be met: no two FACTS
devices are installed at the same location, SVCs, which can exchange reactive power,
are not installed at generator buses, and TCSCs and TCPSs are not installed on
branches with tap-changing transformers [68, 69]. In this paper, we optimized the
locations and ratings of FACTS devices to minimize generation costs. Typically, these
locations are randomly assigned by researchers. As initial conditions, we randomly
set them as follows:

• SVC: bus 14 and bus 23.

• TCSC: branch 5 and branch 26.
• TCPS: branch 28 and branch 32.

These are the randomly assigned locations, not the final system results. Subse-
quently, we analyzed technical and economic factors to achieve cost minimization.
Wu et al. Journal of Big Data (2024) 11:172 Page 13 of 33

Extensive simulations were conducted on the system to verify the optimal proposed
locations for the FACTS installations. Table 4 summarizes the comparison between
the initial and final locations, showing a difference of 1.73 ($/h). Although this differ-
ence may seem minor, the optimized allocation of FACTS device locations proposed
in this paper can yield substantial benefits due to the characteristics of power system
operations.

Integrating feasible constraint handling techniques with WDNMRIME

In this section, this paper will begin with SF mothed [70, 71]. The study will then delve
into an enhanced adaptation of the RIME. This algorithm incorporates dual weight
decay mechanisms and the NMs method, with the objective of proficiently tackling intri-
cate, difficult optimization tasks.

The dominance of feasible solutions

In addressing the OPF problem, feasible solution methodologies have demonstrated
remarkable abilities in managing constraint conditions. As such, it can be mathemati-
cally expressed through the subsequent formulation [72, 73]:
Mathematical expression:

minx f (x) (45)

gi (x) ≤ 0, i = 1,2, ......, m (46)

hj (x) = 0, i = 1,2, ......, n (47)

here, x represents the control parameters to be optimized, f (x) denotes the objective
function to be minimized, gi (x) are the inequality constraint functions, and hj (x) are the
equality constraint functions. The equation constraints are normalized and transformed
into inequality constraints to ensure consistency and uniformity. Thus, the combined
expression for the total set of constraints is as follows, Nc − k represents the number of
equation constraints:

max[g i , nor(x),
 0],  i = 1,2....k
Yi (x) = 
max hi , nor(x) − δt , 0 , i = k + 1, ...Nc (48)

Within the mentioned equation, δt represents the tolerance parameter utilized

when transforming the inequality into a constraint equation. Additionally, ’nor’ is the

Table 4 Comparison of power generation costs between the preset and optimized locations of
FACTS equipment
Location presets Cost ($/h)
Best Worst Average

Initial position 811.1526 811.9721 811.3975

Final Position 809.4179 810.0175 809.8.62
Wu et al. Journal of Big Data (2024) 11:172 Page 14 of 33

normalization process. The aggregate weighted mean of all constraints, representing the
total count of violated constraints, is presented as Eq. (49):
 Nc
i=1 wi (Yi (x))
vio(x) =  Nc (49)
i=1 wi

The weight parameters, denoted as wi , play a crucial role in the SF method. In SF, cer-
tain comparison rules exist to assess the superiority between Solution Xi and Solution Xj ,
and these rules can be summarized as follows:

1. Solution Xi is feasible, but the Solution Xj is not feasible.

2. Both Solution Xi and Solution Xj are feasible solutions, but the objective value of
Solution Xi is better than Solution Xj.
3. Both Solution Xi and Solution Xj are classified as infeasible solutions, yet Solution
Xi exhibits a relatively lower overall constraint violation compared to Solution Xj ,
denoted as vio( Xi) < vio( Xj).

Therefore, in this research, the focus is inclined towards feasible individuals rather
than infeasible ones. If both candidates can be selected, their selection will be based on
comparing their individual fitness. However, if both candidates are infeasible, the selec-
tion process will be influenced by the magnitude of their population constraint viola-
tions [74].

The suggested method

This section provides a comprehensive introduction to the enhanced WDNMRIME
method, which includes the original RIME, dual-weight decay mechanism, and NMs
mechanism.

RIME
The conceptual framework of RIME is primarily drawn from the formation dynamics
of rime. It can be broadly categorized into three components: the exploration strategy

Fig. 5 RIME model

Wu et al. Journal of Big Data (2024) 11:172 Page 15 of 33

for soft rime, the mechanism for penetrating hard rime, and the mechanism for positive
greedy selection. Figure 5 illustrates a schematic depiction of RIME.

Exploration strategy for soft rime The development of soft rime demonstrates notable
stochasticity, with its particles having the ability to spread freely over the majority of sur-
faces on attached objects while advancing slowly in a uniform direction. Specifically, this
research outlines the growth dynamics of rime through the incorporation of dedicated
updating formulas, enhancing the algorithm’s capability to mimic the formation process
of soft rime.

Rijnew = Rbest,j + r1 cosθ ∗ β ∗ h ∗ Ubi,j − Lbi,j + Lbi,j , r2 < E

   
(50)

t
θ =π∗ (51)
10 ∗ T

 
w∗t
β =1− /w (52)
T


E= (t/T ) (53)

The parameters θ, β, and E undergo alterations with the advancement of iterations, where
t represents the current iteration number and T signifies the final iteration count. The con-
stant w maintains a consistent value of 5. Rijnew denotes the updated position of the parti-
cle, where i denotes the i-th individual and j denotes the j-th dimension. Rbest,j indicates
the value of the best individual in the j-th dimension. Moreover, r1 represents a random
value within the interval (-1, 1), and r2 represents another random value. Additionally, h is a
random number ranging from 0 to 1. Ubi,j and Lbi,j respectively denote the maximum and
minimum boundaries of the i-th individual in the j-th dimension.

Mechanism for penetrating hard rime The formation process of hard rime is comparatively
less complex and more uniform than that of soft rime. As rime particles aggregate to form
hard rime, they exhibit consistency in their growth direction, rendering them susceptible
to penetration by various agents. This phenomenon is termed as rime puncture. To emu-
late this occurrence, scholars have introduced the hard-rime puncture mechanism, aimed
at delving deeper into the solution space and augmenting the algorithm’s capacity for local
exploration. The precise updating formulas for the rime particles are outlined as follows.

Rijnew = Rbest,j , r3 < F normr (Si ) (54)

Rijnew signifies the newly updated position of the individual, Rbest,j represent the best indi-
vidual in the j-th dimension. r3 is a randomly generated number within the range (-1, 1).
F normr (Si ) represents the normalized value of the fitness function for the current particle.

Mechanism for positive greedy selection It evaluates the fitness values of particles before
and after updating. Should the updated fitness value prove superior, the particle’s solution
is substituted, and both particles’ solutions are updated concurrently. Moreover, the global
Wu et al. Journal of Big Data (2024) 11:172 Page 16 of 33

best solution, along with its corresponding fitness value, has also been updated. With each
iteration, this mechanism guarantees that the population advances towards the global opti-
mum. Equations (54–55) illustrate the positive greedy selection mechanism.

F (Ri ) = F Rinew
  
, F Rinew < F (Ri )
 
Ri = Rinew (55)

F (Rbest ) = F Rinew , F Rinew < F (Ri )

    

Rbest = Rinew F Rinew < F (Rbest )

, (56)

Dual‑weight decay mechanism

Although RIME has the ability to adapt its search strategy based on environmental influ-
ences, there is still a lack of effective communication between individuals, leading to a sus-
ceptibility to local optima. Therefore, this paper proposes a dual-weight decay mechanism.
Through key mechanisms such as central weight decay, average position guidance, distance
threshold control, and Gaussian perturbation for enhanced diversity, this mechanism can
effectively adjust the individuals’ movement behavior, allowing the population to better
adapt to environmental changes and conduct more comprehensive searches in the solution
space. Simultaneously, introducing Gaussian perturbation increases population richness,
further expanding the exploration range of the solution space. This optimization mecha-
nism is expected to improve the convergence and search efficiency of the algorithm, pro-
viding better results for solving optimization problems. The dual-weight decay mechanism
is depicted in Fig. 6.

a. Global center point radiation strategy

At the beginning of each iteration, this paper first calculates the current global center
point. Compared to the global best individual, this center position has several advantages.
Firstly, the center position combines information from multiple individuals, while the
global best individual can only represent one solution. Therefore, the center position can
better reflect the overall state and characteristics of the population. Secondly, the center

Fig. 6 Dual-weight decay mechanism

Wu et al. Journal of Big Data (2024) 11:172 Page 17 of 33

position has a more stable guiding effect. As it is calculated based on the positions of mul-
tiple individuals, the center position can smooth out the influence of noise and outliers,
reliably guiding the movement of the population. Finally, the center position helps maintain
population diversity. It not only guides the population towards the global optimum but also
considers the distances between individuals, thus preserving the exploratory and diverse
nature of the population, avoiding premature convergence to local optima. Equation (57)
represents this process:

X 1 + X 2 + X3
G= (57)
3

X1. X2, and X3 represent the current top three individuals. In the subsequent move-
ment strategy, this paper first determines the movement probability by calculating the
decay factor and the distance between individuals and the center position. The decay
factor decreases as the number of iterations increases, indicating a gradual decrease
in the likelihood of individuals moving during the iteration process. The decay factor
is calculated using the following formula:
 2
iter
D =1− (58)
Maxiter

The distance between individuals and the center position reflects the relative posi-
tional relationship between individuals and the center. A larger distance indicates a
decreasing influence of the center’s radiation towards the edges, and, thus a higher
likelihood of movement. Formula 58 yields a movement probability R1, within the
range of [0,1]. Here, distance represents the current individual’s distance from the
center position.

R1 = exp(−Ddistance) (59)

b. Local center point radiation strategy

In summary, when individuals do not follow the global center point radiation strat-
egy, they seek help from their neighbors to move to a better position. First, the decayed
distance threshold is calculated, and then the neighbors of the individual are searched
within that distance threshold range. Equation (61) represents this process:

Dis = Initdistance ∗ exp(−D) (60)

where Initdistance represents the initial distance threshold, which is set to 0.9 in this
paper. For each individual, the average position of the neighboring individuals is calcu-
lated as the movement target. The movement direction and distance between the indi-
vidual and the movement target are determined. By updating the individual’s position, it
is moved a certain distance towards the movement target. The above mechanism utilizes
the radiation from the local center point, allowing individuals to explore the solution
space better within a local range and search for better solutions through interaction with
neighboring individuals. The local center point radiation mechanism maintains local
Wu et al. Journal of Big Data (2024) 11:172 Page 18 of 33

exploration while injecting a certain level of randomness and diversity, thereby enhanc-
ing the algorithm’s global search capability.

c. Individual movement strategy

This paper employs the same individual movement strategy in both the global and
local center point radiation strategies. First, the movement direction is calculated as the
vector difference between the individual’s current position and the global center point
for each individual. The movement distance is then calculated as the distance between
the individual and the current center point. The calculation formulas are described as
follows:

MD = MP − Xi (61)

where MP represents the average center position of all individuals within the distance
threshold range, and MD is the movement direction of the current individual. This paper
uses Gaussian perturbation to adjust the movement direction to increase the diversity
and randomness of movement. The specific formula is described as follows:

MDnew = MD × [1 + Gaussian(0,1)] (62)

The perturbation vector is added to the movement direction by introducing Gauss-

ian-distributed random perturbation, providing individuals with a certain level of
randomness and variation in their movement direction.
Algorithm 1 The pseudocode of Dual Center Weight Decay Mechanism
Wu et al. Journal of Big Data (2024) 11:172 Page 19 of 33

Nelder‑Mead simplex method

This study adopts the NMs method to further enhance the performance of RIME in
terms of local exploitation capability. In essence, the NMs method is a polyhedron,
and for a simplex with n dimensions, its convex hull contains n + 1 vertices. The NMs
method primarily improves the worst position by performing operations such as
reflection, expansion, and contraction, thereby generating a new simplex. The specific
steps are as follows, and a schematic process can be referred to in Fig. 7.

1. Check termination condition: Ensure that the termination condition is satisfied,

such as reaching the maximum predefined number of iterations or if the change in
the objective function value falls below a certain threshold.
2. Determine the best and worst points: Find the best solution Xbest and Xbetter and the
worst solution Xworst in the current set of solutions.
3. Determine the centroid: Calculate the centroid using the best and second-best solu-
tions to obtain an intermediate point. The calculation method is shown in Eq. (63).

(Xbest + Xbetter )
Xc = (63)
2

4. Reflection operation: Reflect the intermediate point Xc with respect to the worst
point Xworst to obtain a new reflected point Xr . Here, α is the reflection coefficient.

Xr = Xc + α ∗ (Xc − Xworst ) (64)

5. Contrast the objective function value of the reflection point with that of the optimal
point: If it’s smaller, proceed with expansion. If it falls between the best and second
best points, replace the worst point and end the iteration. If it exceeds that of the
worst point, perform a shrinking operation.
6. Expansion operation: Extend the reflected point further to obtain a new expanded
point Xy . The expansion is done using formula 64, where β is the expansion coef-
ficient.

Xy = Xc + β(Xr − Xc ) (65)

X2

Xe
Xr
X oc
Xc
X ic
XW

X1
Fig. 7 NMs method
Wu et al. Journal of Big Data (2024) 11:172 Page 20 of 33

7. Compare the objective function value f(X) of the expanded point with the best point:
If f (Xy ) is smaller than f (Xbest ), replace Xworst and end the current iteration. If
f (Xy ) is greater than or equal to f (Xbest ), perform a contraction operation.
8. Contraction operation: Shrink all vertices of the current simplex toward the best
point except for the best point itself, resulting in a new contracted point Xic . The
calculation method is shown in Eq. (66), where γ is the contraction coefficient.

Xic = Xc + γ (Xr − Xc ) (66)

9. Compare the objective function value of the contracted point with the worst point:
If f (Xic ) is smaller than f (Xworst ), replace Xworst and end the current iteration. Oth-
erwise, perform a reduction operation.
10. Reduction operation: Shrink all vertices of the current simplex toward the best point
except for the best point itself by a certain distance, resulting in a reduced point Xoc .
The calculation is done using Eq. (67), where δ is the reduction coefficient.

Xoc = Xc − δ(Xworst − Xc ) (67)

11. Return to step 1 and continue the iteration

Through these operations, the NMs method progressively searches the feasible
region of the problem until it finds the optimal solution that satisfies the termination
condition or can no longer be further optimized [46].

The framework of the proposed WDNMRIME algorithm

Considering the attributes of the RIME, this paper introduces two pivotal strategies,
namely the dual-center weight decay strategy and the NMs strategy, to overcome the
limitations of the RIME and elevate its overall performance. The specific process of
the algorithm can be found in Fig. 8.
Metaheuristic methods are often in risk of an event called immature convergence
and stagnation challenges due to several reasons, including imbalance of diversifica-
tion and intensification phases and weakness of their operators [75]. RIME is not an
exception. RIME faces a key limitation in being prone to local optima, hindering its
ability to find the global optimal solution. To overcome this challenge, we propose
the dual-center weight decay strategy. This method aims to balance global explora-
tion and local exploitation by gradually reducing the influence of dual centers over
time. By lessening their impact, the algorithm becomes less likely to converge to sub-
optimal solutions prematurely, thus broadening its search scope. This enhancement
maintains a harmonious balance between exploration and exploitation, significantly
improving solution quality. Furthermore, RIME suffers from issues regarding conver-
gence accuracy and efficiency. To tackle this, we introduce the NMs strategy, which
enhances the algorithm’s convergence rate and precision through a novel neighbor-
hood search mechanism. This mechanism allows for more efficient navigation of the
solution space, guiding the search toward promising regions. By adaptively adjust-
ing the search process based on problem characteristics and iteration progress, the
NMs strategy improves the algorithm’s ability to converge to superior solutions while
Wu et al. Journal of Big Data (2024) 11:172 Page 21 of 33

Fig. 8 Flow chart of WDNMRIME

reducing computational overhead. This innovative addition significantly boosts the

algorithm’s convergence speed and competitiveness with solution quality.
By incorporating these strategies, the algorithm achieves a better balance between
global exploration and local exploitation, while also enhancing its convergence accu-
racy and efficiency. These innovative contributions enhance the algorithm’s competi-
tiveness and make it more effective in solving optimization problems.
Wu et al. Journal of Big Data (2024) 11:172 Page 22 of 33

Algorithm 2 The pseudocode of WDNMRIME

Experimental results
A comprehensive analysis is conducted, and the experimental results obtained through
its implementation are discussed in detail. Throughout the optimization process, distinct
device-level control variables are assigned to each FACTS device, adhering to particu-
lar rounding guidelines. The configuration of FACTS devices aims to keep the tap range
of transformers within the interval of [0.9–1.1] per unit (pu) and the voltage of load buses
within the range of [0.95–1.1] pu. Figure 9 shows the load bus voltage curves for all case stud-
ies in the adapted system under 100% load conditions. It demonstrates that under 100% load
conditions, the voltage stability of the load bus in a power system is the basis for stable sys-
tem operation, which directly affects the quality of power supply and equipment life. Table 5
lists the allowed ranges of control parameters for the proposed method to provide a compre-
hensive understanding of the optimization process. The TCPS angle can vary between -5°
and 5°. The SVC can supply up to 10 MVAr of reactive power. Experiments are performed
using MATLAB 2018b on a system with Windows 10, 16 GB RAM, and a 3.6 GHz processor.
The program runs for a maximum of 1000 iterations.
Wu et al. Journal of Big Data (2024) 11:172 Page 23 of 33

Fig. 9 Voltage profiles of load buses for scenario 1 to 3

Table 5 Configuring control parameter limits in the IEEE 30 System

Control variables Lower limit Upper limit Parameters Lower limit Upper limit

PTG2 (MW) 20 80 PTG1 (MW) 50 200

PWG5 (MW) 0 75 QTG1 (MVAr) − 20 150
PTG8 (MW) 10 35 QTG2 (MVAr) − 20 60
PWG11 (MW) 0 60 QWG5 (MVAr) − 30 35
PTG13 (MW) 12 40 QTG8 (MVAr) − 15 48.7
V1 (p.u.) 0.95 1.10 QWG11 (MVAr) − 25 30
V2 (p.u.) 0.95 1.10 QTG13 (MVAr) − 105 44.7
V5 (p.u.) 0.95 1.10
V8 (p.u.) 0.95 1.10
V11 (p.u.) 0.95 1.10
V13 (p.u.) 0.95 1.10
T11 (p.u.) 0.95 1.10
T12 (p.u.) 0.95 1.10
T15 (p.u.) 0.95 1.10
T36 (p.u.) 0.95 1.10
FACTS rating
τTCSC1 (%) 0 50%
τTCSC2 (%) 0 50%
�TCPS1 (deg.) − 5° 5°
�TCPS2 (deg.) − 5° 5°
QSVC1 (MVAr) − 10 10
QSVC2 (MVAr) − 10 10
Wu et al. Journal of Big Data (2024) 11:172 Page 24 of 33

Scenario 1: total generation cost minimization

Throughout the optimization process, the primary emphasis is on maintaining the
safety constraints of the power system. Research findings indicate that under Scenario 1,
WDNMRIME achieves a total cost of 806.00298 $/h. However, it is essential to recognize
that wind turbines may offer lower costs, but their reliance on wind energy limits them
from reaching maximum rated capacity. Consequently, this may result in higher reserve
costs. To evaluate WDNMRIME, a comparative analysis is conducted with several lead-
ing metaheuristic algorithms, including SCA [76], SSA [77], TLBO [78], FPA [79], CBA
[80], RIME, and CPSCA. The results of this comparison demonstrate that the proposed
WDNMRIME algorithm surpasses all previously mentioned competitors in terms of
total generation cost and convergence speed, as depicted in Fig. 10. WDNMRIME exhib-
its swift convergence in the initial phases and maintains the highest level of convergence
accuracy. Furthermore, it demonstrates significant improvements compared to the origi-
nal RIME approach. Table 6 provides detailed values. The specific parameters have been
given in the previous article, and the units are standard power system units.

Scenario 2: minimization of active power loss

This section delves deeper into utilizing the WDNMRIME for active power loss optimi-
zation. The experimental control parameters are detailed in Table 7, and the proposed
approach attains a noteworthy reduction in active power loss, achieving a minimum of
1.41 MW. Comparing Case 2 with Case 1, it is observed that Case 2 has higher genera-
tion costs but lower power losses compared to Case 1. Furthermore, the WDNMRIME
is compared with CBA, CPSCA, TLBO, FPA, CPA, RIME, and SaDE [81]. A care-
ful examination of the table reveals that the proposed approach excels in active power
optimization, surpassing all rival algorithms concerning numerical values. Moreover,

Fig. 10 Convergence comparison of WDNMRIME and other enhanced algorithms in Scenario 1

Wu et al. Journal of Big Data (2024) 11:172 Page 25 of 33

Table 6 Evaluating the WDNMRIMR against Other Algorithms in Scenario 1

Parameters WDNMRIME SCA SSA TLBO FPA CBA RIME CPSCA

PTG2 (MW) 39.08144 21.58520 37.27378 40.48124 41.58773 47.91525 43.28067 40.85676
PWG5 (MW) 50.52857 48.86661 49.41338 49.59133 49.08935 40.63919 47.52100 48.56250
PTG8 (MW) 10.00002 11.16133 19.93471 10.00000 10.00000 20.25520 10.00000 10.00000
PWG11 (MW) 42.11213 39.86275 34.57652 42.03184 41.45031 31.11732 41.13230 42.49587
PWG13 (MW) 12.00000 12.00000 12.00000 12.00000 12.00000 14.17110 12.00000 12.00000
V1 (p.u.) 1.10000 1.10000 1.10000 1.07395 1.09002 1.10000 1.10000 1.10000
V2 (p.u.) 1.08907 1.10000 1.10000 1.05892 1.07739 1.08951 1.09652 1.10000
V5 (p.u.) 1.07078 1.10000 1.10000 1.03683 1.06323 1.06480 1.08288 1.08438
V8 (p.u.) 1.07280 1.07516 1.10000 1.03664 1.05033 1.07418 1.08661 1.07645
V11 (p.u.) 1.10000 1.02760 1.10000 1.07253 1.07539 1.09996 1.09773 1.10000
V13 (p.u.) 1.10000 1.10000 1.10000 1.07605 1.02419 1.09992 1.10000 1.09063
T11 (p.u.) 1.00894 1.02248 1.10000 0.99275 1.04019 0.96815 1.10000 1.05523
T12 (p.u.) 0.98519 0.91823 1.10000 0.90026 1.06920 1.05034 0.90284 0.93125
T15 (p.u.) 1.02308 1.10000 1.10000 1.01074 1.03092 1.06313 1.02673 1.02500
T36 (p.u.) 0.98794 0.96339 1.10000 0.95657 1.00799 0.98573 0.98026 0.97500
τTCSC1 (%) 10.44483 17.34982 19.66045 7.12930 19.33441 21.23023 15.37298 24.00000
τTCSC2 (%) 23.78919 16.04843 24.71541 24.27170 7.46264 7.93190 20.82919 18.81047
�TCPS1 (deg.) 10.00000 8.52571 4.37859 9.35656 − 0.91966 − 2.80357 9.19581 10.00000
�TCPS2 (deg.) 10.00000 − 3.93093 − 10.0000 9.87099 10.00000 − 5.76017 9.61948 6.01873
QSVC1 (MVAr) 2.06889 26.74210 40.00000 35.67990 3.39469 29.73375 15.27534 11.24184
PTG1 (MW) 128.4725 124.1578 132.1457 129.6424 131.4851 129.8677 130.4147 132.002
QTG1 (MVAr) 40.99955 41.00000 35.01257 15.89355 3.89941 14.35777 13.13833 13.49786
QTG2 (MVAr) 0.25896 0.01149 0.13091 0.04646 0.00000 0.07297 0.29649 0.00000
QWG5 (MVAr) 0.23866 0.00000 0.25174 0.26473 0.28358 0.14364 0.50000 0.49758
QTG8 (MVAr) 13.56916 1.26878 1.00000 15.52910 15.37154 16.27119 14.29367 4.00000
QWG11 (MVAr) 8.83160 4.81178 1.00000 14.30588 31.40380 25.76243 16.91832 33.86313
QTG13 (MVAr) 2.41171 5.00000 − 5.00000 − 3.77580 2.01870 0.49881 3.26449 − 1.02447
QSVC2 (MVAr) 1.21408 − 3.78915 − 5.00000 2.80548 − 0.27456 − 1.45456 0.16190 5.00000
Cgen ($/h) 806.0029 818.5882 814.6227 807.2833 808.0918 817.5626 806.5732 806.5491

the convergence curve shown in Fig. 11 further demonstrates the superiority of the
proposed algorithm. The SMA algorithm comes next, although its limited exploitation
capability results in lower precision and an inability to achieve better solutions. On the
other hand, the CPSCA algorithm suffers from being trapped in local optima. In sum-
mary, as demonstrated in this section, the proposed WDNMRIME method efficiently
solves the optimization problem of active power loss.

Scenario 3: minimization of gross costs

According to the conclusions in Table 8, Case 3 demonstrates the successful appli-
cation of the proposed WDNMRIME, resulting in the lowest total cost of 1089.093
$/h. This outcome represents a compromise between cost optimization and minimal
power loss. To further evaluate the effectiveness of WDNMRIME, this study com-
pares it with HGWOSCA [82], CPSCA, MSCA [31], FPA, GWO [83], RIME, and
SaDE. It is evident that WDNMRIME outperforms all competing algorithms, ensur-
ing the highest total cost achievement for the optimization solution. Additionally,
Wu et al. Journal of Big Data (2024) 11:172 Page 26 of 33

Table 7 Evaluating the WDNMRIMR against Other Algorithms in Scenario 2

Parameters WDNMRIME CBA CPSCA TLBO FPA CPA RIME SaDE

PTG2 (MW) 65.24286 20.01489 65.27974 26.06323 51.27808 79.99337 65.19643 60.69678
PWG5 (MW) 75.00000 74.97026 74.99890 74.99994 74.68206 74.98971 75.00000 74.87705
PTG8 (MW) 35.00000 34.93725 35.00000 34.99977 35.00000 34.99228 35.00000 34.95449
PWG11 (MW) 60.00000 59.64986 60.00000 59.99999 53.00177 59.93996 60.00000 59.80884
PWG13 (MW) 40.00000 39.98222 39.99888 39.25920 32.80063 39.76809 40.00000 37.09450
V1 (p.u.) 1.10000 1.10000 1.10000 1.05295 1.06279 1.09794 1.10000 1.04301
V2 (p.u.) 1.10000 1.09491 1.10000 1.04740 1.05887 1.10000 1.10000 1.03771
V5 (p.u.) 1.09044 1.08719 1.09023 1.03808 1.05676 1.09061 1.09061 1.02626
V8 (p.u.) 1.09494 1.09337 1.09414 1.04245 1.05079 1.09467 1.09555 1.03236
V11 (p.u.) 1.10000 1.10000 1.09812 1.06923 1.06315 1.10000 1.10000 1.07482
V13 (p.u.) 1.10000 1.09989 1.09851 1.09607 1.03724 1.10000 1.10000 1.07031
T11 (p.u.) 1.07525 1.05712 1.10000 0.99270 0.99798 1.01904 1.08639 0.97863
T12 (p.u.) 0.90119 0.90236 0.90469 0.91819 0.93991 0.92869 0.90000 0.97156
T15 (p.u.) 1.00048 1.03138 1.00625 1.06949 0.96679 1.00476 1.02578 0.99337
T36 (p.u.) 0.97755 1.04085 0.97813 0.98784 0.99553 1.02688 1.00998 0.96724
τTCSC1 (%) 21.00099 28.92350 23.91321 23.80785 24.86378 3.00129 24.03168 24.41430
τTCSC2 (%) 24.21563 3.00420 21.00000 7.40435 22.18007 28.61521 21.01913 20.80095
�TCPS1 (deg.) 10.00000 6.37429 9.01529 9.99986 9.84401 9.79281 9.78489 9.45180
�TCPS2 (deg.) 9.02088 − 9.99938 10.00000 9.95763 9.89420 8.96577 10.00000 9.46189
QSVC1 (MVAr) 15.22453 39.92015 13.00000 10.81756 39.87537 39.97772 14.15482 14.04708
PTG1 (MW) 51.85121 52.45765 61.32752 54.36760 54.61852 54.34511 57.54161 59.51761
QTG1 (MVAr) 11.31783 38.28706 6.41620 28.82034 31.55706 1.70769 1.00305 15.62691
QTG2 (MVAr) 0.50000 0.49117 0.49925 0.31507 0.19258 0.02264 0.50000 0.47954
QWG5 (MVAr) 0.50000 0.49637 0.00000 0.00001 0.00135 0.36700 0.00000 0.22832
QTG8 (MVAr) 35.06994 1.25628 4.66554 20.53273 28.41962 1.04318 34.70151 34.87706
QWG11 (MVAr) 14.02225 1.22487 35.00000 13.67758 35.12037 1.06625 33.68446 20.90990
QTG13 (MVAr) 4.25905 − 4.93659 − 0.52191 0.56950 0.12124 − 4.84115 4.01127 3.95920
QSVC2 (MVAr) − 2.29217 − 4.99962 3.47070 1.41341 3.18683 − 4.90828 4.99955 − 0.05435
Ploss (MW) 1.39469 2.07737 1.41687 1.92283 2.01242 1.80917 1.43050 1.60641

the convergence curve of WDNMRIME, as shown in Fig. 12, supports its superiority.
While CPSCA shows characteristics of rapid convergence during the iteration pro-
cess, it often gets trapped in local optima. Although SaDE demonstrates a good rate,
it fails to evolve sufficiently and reach global optima in later stages. In conclusion,
the application of WDNMRIME in Case 3 confirms its robustness as an optimiza-
tion technique, outperforming other algorithms in effectively solving the considered
optimization problem.

Running time comparison of WDNMRIME and other algorithms

Efficiency of an algorithm is significantly gauged by its time complexity. Substantial
improvements should ensure a minimal time complexity while achieving the opti-
mal solution. Table 9 displays the runtime of WDNMRIME for each model, Fig. 13
Wu et al. Journal of Big Data (2024) 11:172 Page 27 of 33

Fig. 11 Convergence comparison of WDNMRIME and other enhanced algorithms in Scenario 2

visualizes the running time of the algorithm. Although WDNMRIME does not have
the shortest runtime for every model, it consistently performs better compared to
other algorithms. Notably, the final fitness of WDNMRIME is less than that of other
state-of-the-art algorithms. This improvement can be attributed to the incorporation
of the double-weighted recession mechanism, which enhances WDNMRIME’s ability
to escape local optima, thereby improving convergence accuracy and reducing overall
computation.

Conclusions
This paper introduces WDNMRIME, an enhanced RIME method specifically
designed to optimize power systems by solving the OPF problem. WDNMRIME
integrates a novel weighted relationship between global and local individuals, sig-
nificantly enhancing diversity and exploration capabilities. This innovation leads to
superior performance in total cost optimization and substantial reductions in power
losses compared to conventional RIME algorithms. Experimental validation on the
IEEE 30-bus test system demonstrates WDNMRIME achieving a remarkable total
cost of 806.0029 $/h and reducing power losses from 1.43 MW to 1.39 MW, achieving
optimal results. However, applying WDNMRIME to larger and more complex power
systems presents challenges due to its scalability limitations arising from increased
Wu et al. Journal of Big Data (2024) 11:172 Page 28 of 33

Table 8 Evaluating the WDNMRIMR against Other Algorithms in Scenario 3

Parameters WDNMRIME HGWOSCA CPSCA MSCA FPA GWO RIME SaDE

PTG2 (MW) 40.88791 41.12547 43.14193 40.61418 38.95331 41.67524 43.42570 42.23425
PWG5 (MW) 75.00000 75.00000 75.00000 75.00000 73.86171 74.99516 75.00000 74.94401
PTG8 (MW) 35.00000 34.97545 35.00000 35.00000 34.57125 35.00000 35.00000 34.98492
PWG11 (MW) 60.00000 59.92615 60.00000 60.00000 60.00000 59.60019 60.00000 59.36863
PWG13 (MW) 24.25802 24.11087 22.12500 19.72746 28.40216 24.09224 21.86078 23.87537
V1 (p.u.) 1.10000 1.10000 1.10000 0.95000 1.04503 1.10000 1.10000 1.04847
V2 (p.u.) 1.09609 1.09660 1.10000 0.95000 1.03706 1.09672 1.09828 1.04346
V5 (p.u.) 1.08639 1.08642 1.10000 0.95000 1.02741 1.08605 1.09202 1.03326
V8 (p.u.) 1.09041 1.09074 1.10000 0.95000 1.02594 1.09050 1.09417 1.03618
V11 (p.u.) 1.10000 1.09990 1.10000 1.10000 1.09104 1.10000 1.10000 1.09277
V13 (p.u.) 1.10000 1.09981 1.10000 1.10000 1.06136 1.09761 1.10000 1.07728
T11 (p.u.) 1.08989 1.05938 1.06875 0.90000 0.95971 1.03622 1.03790 1.02545
T12 (p.u.) 0.90994 0.90866 0.94375 0.90000 0.96181 0.94417 0.90000 0.92622
T15 (p.u.) 1.00882 1.03946 1.02188 0.90000 0.95347 1.04371 1.01610 1.00145
T36 (p.u.) 0.98623 0.98677 1.00625 0.90000 0.96291 1.00451 0.98801 0.96790
τTCSC1 (%) 21.33523 17.92655 23.70400 3.00000 28.95550 23.91516 19.49348 23.83196
τTCSC2 (%) 24.29869 18.70465 20.52597 14.74937 20.54267 21.51980 24.41496 20.70226
�TCPS1 (deg.) 9.99998 − 1.65464 9.82411 − 10.00000 1.65548 3.26937 2.76878 8.11659
�TCPS2 (deg.) 9.12302 7.14736 9.99336 − 10.00000 10.00000 0.97932 10.00000 9.65257
QSVC1 (MVAr) 13.15552 2.07612 25.00000 8.09852 34.18468 1.44652 35.04700 13.13129
PTG1 (MW) 118.1844 123.2572 125.4125 124.4602 135.1428 133.1745 128.1578 124.2675
QTG1 (MVAr) 15.80801 5.10880 2.87579 2.79906 41.00000 20.24157 14.55697 2.36649
QTG2 (MVAr) 0.50000 0.12709 0.29006 0.00000 0.32747 0.04309 0.12012 0.34489
QWG5 (MVAr) 0.50000 0.01950 0.00000 0.00000 0.47108 0.05407 0.22841 0.18444
QTG8 (MVAr) 14.25572 12.62433 12.00000 1.00000 20.25330 1.12056 9.27865 21.51091
QWG11 (MVAr) 34.99472 11.61122 13.49175 1.13203 13.55522 33.33602 15.43979 32.69353
QTG13 (MVAr) 0.91397 0.15108 2.94229 − 0.99058 0.30432 1.21318 0.49342 0.12573
QSVC2 (MVAr) 2.34259 0.63789 − 4.89735 − 0.08396 1.22323 0.88627 − 3.32370 2.56350
Cgross ($/h) 1089.592 1110.714 1100.162 1194.503 1128.395 1108.793 1101.467 1106.423

Fig. 12 Convergence comparison of WDNMRIME and other enhanced algorithms in Scenario 3

Wu et al. Journal of Big Data (2024) 11:172 Page 29 of 33

Table 9 The comparison between WDNMRIME and various algorithms (seconds)

Item Scenario 1 Scenario 2 Scenario 3

WDNMRIME 70.479 WDNMRIME 64.744 WDNMRIME 76.536

CBA 1461.39 SCA 71.55.01 HGWOSCA 2182.10
CPSCA 60.286 CPSCA 57.536 CPSCA 67.536
FPA 60.593 FPA 56.552 FPA 66.447
CPA 66.635 CBA 76.01 GWO 71.755
RIME 74.098 RIME 55.567 RIME 65.752
SaDE 76.2656 TLBO 70.250 SaDE 81.984
SSA 91.1427 CPA 67.719 MSCA 108.432

Fig. 13 Running time comparison of WDNMRIME and other algorithms

computational complexity. Addressing these challenges requires further research to

enhance operational efficiency and scalability for handling extensive power system
optimization tasks. Moreover, investigating WDNMRIME’s adaptability to uncer-
tain load demands and external environmental changes is crucial, as these factors
can significantly influence optimal system operations. Robust modeling and optimi-
zation techniques are pivotal for augmenting WDNMRIME’s applicability and resil-
ience in practical power system applications. In conclusion, WDNMRIME stands
as an advanced and innovative approach in power system optimization, surpassing
traditional methods in terms of cost reduction and operational efficiency. Overcom-
ing scalability hurdles and fortifying its robustness against uncertain conditions are
imperative for broader adoption and enhanced effectiveness in complex power system
environments.

Appendix 1
See Table 10
Wu et al. Journal of Big Data (2024) 11:172 Page 30 of 33

Table 10 Line loading limit (MVA)

Branches Scenario 1 Scenario 2 Scenario3 branches Scenario1 Scenario2 Scenario3

1 130 130 130 21 16 16 16

2 130 130 130 22 32 32 32
3 65 65 65 23 32 32 32
4 130 130 130 24 32 32 32
5 130 130 130 25 32 32 32
6 65 65 65 26 32 32 32
7 90 90 90 27 32 32 32
8 70 70 70 28 16 16 16
9 16 16 16 29 32 32 32
10 16 16 16 30 16 16 16
11 32 32 32 31 16 16 16
12 65 65 65 32 65 65 65
13 32 32 32 33 16 16 16
14 65 65 65 34 16 16 16
15 65 65 65 35 16 16 16
16 65 65 65 36 32 32 32
17 65 65 65 37 16 16 16
18 32 32 32 38 16 16 16
19 16 16 16 39 16 16 16
20 16 16 16 40 32 32 32
41 32 32 32

Acknowledgements
This work was supported by the Graduate Scientific Research Foundation of Wenzhou University (3162024003083).
This work was also supported in part by the Natural Science Foundation of Zhejiang Province (LTGS23E070001,
LZ22F020005), National Natural Science Foundation of China (62076185, 62301367).

Author contributions
Huangying Wu: writing—original draft, writing—review and editing, software, visualization, investigation. Yi Chen: writ-
ing—original draft, writing—review and editing, software, visualization, investigation, conceptualization, methodology,
formal analysis. Zhennao Cai: writing—original draft, writing—review and editing, software, visualization, investigation.
Ali Asghar Heidari: writing—review and editing, software, visualization, investigation. Huiling Chen: conceptualization,
methodology, formal analysis, investigation, writing—review and editing, funding acquisition, supervision, project
administration. Guoxi Liang: conceptualization, methodology, formal analysis, investigation, writing—review and editing,
Funding acquisition, supervision, project administration.

Data availability
No datasets were generated or analysed during the current study.

Declarations
Competing interests
The authors declare no competing interests.

Received: 11 May 2024 Accepted: 2 November 2024

References
1. Liu Z, et al. A review of common-mode voltage suppression methods in wind power generation. Renew Sustain
Energy Rev. 2024;203:114773.
2. Ju Y, et al. Distributed three-phase power flow for AC/DC hybrid networked microgrids considering converter limit-
ing constraints. IEEE Trans Smart Grid. 2022;13(3):1691–708.
3. Phadke AR, Fozdar M, Niazi KR. A new multi-objective formulation for optimal placement of shunt flexible AC trans-
mission systems controller. Electr Power Compon Syst. 2009;37(12):1386–402.
4. Farrag MEA, Putrus GA. Design of an adaptive neurofuzzy inference control system for the unified power-flow
controller. IEEE Trans Power Deliv. 2012;27(1):53–61.
Wu et al. Journal of Big Data (2024) 11:172 Page 31 of 33

5. Li S, et al. A fast and accurate calculation method of line breaking power flow based on Taylor expansion. Front
Energy Res. 2022;10:943946.
6. Ergun H, et al. Optimal power flow for AC-DC grids: formulation, convex relaxation, linear approximation, and imple-
mentation. IEEE Trans Power Syst. 2019;34(4):2980–90.
7. Zamzam A, Baker K. Learning optimal solutions for extremely fast AC optimal power flow. Arxiv. 2019.
8. Guerrero J, et al. Towards a transactive energy system for integration of distributed energy resources: home energy
management, distributed optimal power flow, and peer-to-peer energy trading. Renew Sustain Energy Rev.
2020;132:110000.
9. Momoh JA, Zhu J. Improved interior point method for OPF problems. IEEE Trans Power Syst. 1999;14(3):1114–20.
10. Yan X, Quintana VH. Improving an interior-point-based OPF by dynamic adjustments of step sizes and tolerances.
IEEE Trans Power Syst. 1999;14(2):709–17.
11. Alrashidi MR, El-Hawary ME. A survey of particle swarm optimization applications in power system operations. Electr
Power Compon Syst. 2006;34(12):1349–57.
12. Deng W, et al. An improved differential evolution algorithm and its application in optimization problem. Soft Com-
put. 2021;25:5277–98.
13 Wu H, et al. Feature selection in high-dimensional data: an enhanced RIME optimization with information entropy
pruning and DBSCAN clustering. Int J Mach Learn Cybern. 2024. https://​doi.​org/​10.​1007/​s13042-​024-​02143-1.
14 Ahmadianfar I, et al. INFO: an efficient optimization algorithm based on weighted mean of vectors. Expert Syst Appl.
2022;195:116516.
15 Ahmadianfar I, et al. RUN beyond the metaphor: an efficient optimization algorithm based on Runge Kutta method.
Expert Syst Appl. 2021;181:115079.
16. Yang Y, et al. Hunger games search: visions, conception, implementation, deep analysis, perspectives, and towards
performance shifts. Expert Syst Appl. 2021;177:114864.
17. Alabool HM, et al. Harris hawks optimization: a comprehensive review of recent variants and applications. Neural
Comput Appl. 2021;33:8939–80.
18 Chen H, et al. Slime mould algorithm: a comprehensive review of recent variants and applications. Int J Syst Sci.
2022. https://​doi.​org/​10.​1080/​00207​721.​2022.​21536​35.
19. Li S, et al. Slime mould algorithm: a new method for stochastic optimization. Future Gener Comput Syst.
2020;111:300–23.
20. Tu J, et al. The colony predation algorithm. J Bionic Eng. 2021;18(3):674–710.
21. Houssein EH, et al. Liver cancer algorithm: a novel bio-inspired optimizer. Comput Biol Med. 2023;165:107389.
22 Lian J, et al. Parrot optimizer: algorithm and applications to medical problems. Computers Biol Med.
2024;172:108064.
23. Dong R, et al. Multi-strategy enhanced kernel search optimization and its application in economic emission dis-
patch problems. J Comput Des Eng. 2024;11(1):135–72.
24 Dong R, et al. Boosting kernel search optimizer with slime mould foraging behavior for combined economic emis-
sion dispatch problems. J Bionic Eng. 2023. https://​doi.​org/​10.​1007/​s42235-​023-​00408-z.
25. Sun G, et al. Low-latency and resource-efficient service function chaining orchestration in network function virtual-
ization. IEEE Internet Things J. 2019;7(7):5760–72.
26. Xu X, et al. Multi-objective robust optimisation model for MDVRPLS in refined oil distribution. Int J Prod Res.
2022;60(22):6772–92.
27. Naderi E, et al. A novel hybrid self-adaptive heuristic algorithm to handle single-and multi-objective optimal power
flow problems. Int J Electr Power Energy Syst. 2021;125:106492.
28. Birogul S. Hybrid Harris Hawk optimization based on differential evolution (HHODE) algorithm for optimal power
flow problem. IEEE Access. 2019;7:184468–88.
29. Taher MA, et al. An improved moth-flame optimization algorithm for solving optimal power flow problem. Int Trans
Electr Energy Syst. 2019;29(3): e2743.
30. Babaeinejadsarookolaee S, et al. The power grid library for benchmarking AC optimal power flow algorithms. Arxiv.
2021.
31. Attia A-F, El Sehiemy RA, Hasanien HM. Optimal power flow solution in power systems using a novel Sine-Cosine
algorithm. Int J Electr Power Energy Syst. 2018;99:331–43.
32. Naderi E, Pourakbari-Kasmaei M, Abdi H. An efficient particle swarm optimization algorithm to solve optimal power
flow problem integrated with FACTS devices. Appl Soft Comput. 2019;80:243–62.
33. Yuan X, et al. Multi-objective optimal power flow based on improved strength Pareto evolutionary algorithm.
Energy. 2017;122:70–82.
34. Roy PK, Ghoshal SP, Thakur SS. Biogeography based optimization for multi-constraint optimal power flow with emis-
sion and non-smooth cost function. Expert Syst Appl. 2010;37(12):8221–8.
35. Niknam T, et al. A modified shuffle frog leaping algorithm for multi-objective optimal power flow. Energy.
2011;36(11):6420–32.
36. Narimani MR, et al. A novel approach to multi-objective optimal power flow by a new hybrid optimization algo-
rithm considering generator constraints and multi-fuel type. Energy. 2013;49:119–36.
37. Naderi E, et al. A novel fuzzy adaptive configuration of particle swarm optimization to solve large-scale optimal
reactive power dispatch. Appl Soft Comput. 2017;53:441–56.
38. Naderi E, et al. Multi-objective optimal power flow incorporating flexible alternating current transmission systems:
application of a wavelet-oriented evolutionary algorithm. Electr Power Compon Syst. 2024;52(5):766–95.
39. Weng X, et al. A vertical and horizontal crossover sine cosine algorithm with pattern search for optimal power flow
in power systems. Energy. 2023;271:127000.
40. Naderi E, et al. Optimization of active power dispatch considering unified power flow controller: application of
evolutionary algorithms in a fuzzy framework. Evol Intel. 2024;17(3):1357–87.
41. Lesieutre BC, Hiskens IA. Convexity of the set of feasible injections and revenue adequacy in FTR markets. IEEE Trans
Power Syst. 2005;20(4):1790–8.
Wu et al. Journal of Big Data (2024) 11:172 Page 32 of 33

42. Singh RP, Mukherjee V, Ghoshal S. Particle swarm optimization with an aging leader and challengers algorithm for
the solution of optimal power flow problem. Appl Soft Comput. 2016;40:161–77.
43. Nadeem M, et al. Optimal placement, sizing and coordination of FACTS devices in transmission network using
whale optimization algorithm. Energies. 2020;13:753.
44. Ruiz F, et al. Surveying solid-state transformer structures and controls: providing highly efficient and controllable
power flow in distribution grids. IEEE Ind Electron Mag. 2020;14(1):56–70.
45. Schauder C, Mehta H. Vector analysis and control of advanced static VAR compensators. In: IEE Proceedings C (Gen-
eration, Transmission and Distribution). IET; 1993.
46. Hallinan AJ Jr. A review of the Weibull distribution. J Qual Technol. 1993;25(2):85–93.
47 Cain MB, O’neill RP, Castillo A. History of optimal power flow and formulations. Fed Energy Regul Comm.
2012;1:1–36.
48. Huneault M, Galiana FD. A survey of the optimal power flow literature. IEEE Trans Power Syst. 1991;6(2):762–70.
49. Biswas PP, Suganthan P, Amaratunga GA. Optimal power flow solutions incorporating stochastic wind and solar
power. Energy Convers Manag. 2017;148:1194–207.
50 Chow JH, Sanchez-Gasca JJ. Power system modeling, computation, and control. Hoboken: John Wiley & Sons; 2020.
51. Ferreira RS, Borges CLT, Pereira MV. A flexible mixed-integer linear programming approach to the AC optimal power
flow in distribution systems. IEEE Trans Power Syst. 2014;29(5):2447–59.
52. Tu S, Wächter A, Wei E. A two-stage decomposition approach for AC optimal power flow. IEEE Trans Power Syst.
2020;36(1):303–12.
53. Veeramsetty V, Venkaiah C, Kumar DV. Hybrid genetic dragonfly algorithm based optimal power flow for computing
LMP at DG buses for reliability improvement. Energy Syst. 2018;9:709–57.
54. Castillo A, et al. The unit commitment problem with AC optimal power flow constraints. IEEE Trans Power Syst.
2016;31(6):4853–66.
55. Alsac O, Stott B. Optimal load flow with steady-state security. IEEE Trans Power Appar Syst. 1974;3:745–51.
56. Shao Z, et al. A linear AC unit commitment formulation: an application of data-driven linear power flow model. Int J
Electr Power Energy Syst. 2023;145:108673.
57. Li S, et al. Adaptive constraint differential evolution for optimal power flow. Energy. 2021;235:121362.
58. Chatzos M, et al. High-fidelity machine learning approximations of large-scale optimal power flow. arXiv preprint
arXiv:​2006.​16356. 2020.
59. Islam MZ, et al. A Harris Hawks optimization based single-and multi-objective optimal power flow considering
environmental emission. Sustainability. 2020;12(13):5248.
60. Nusair K, Alasali F. Optimal power flow management system for a power network with stochastic renewable energy
resources using golden ratio optimization method. Energies. 2020;13(14):3671.
61. Kaymaz E, Duman S, Guvenc U. Optimal power flow solution with stochastic wind power using the Lévy coyote
optimization algorithm. Neural Comput Appl. 2021;33:6775–804.
62. Muhammad Y, et al. Design of fractional swarm intelligent computing with entropy evolution for optimal power
flow problems. IEEE Access. 2020;8:111401–19.
63. Mohan TM, Nireekshana T. A genetic algorithm for solving optimal power flow problem. In: 2019 3rd International
conference on Electronics, Communication and Aerospace Technology (ICECA). IEEE; 2019.
64. Velasquez OS, et al. Optimal power flow in direct-current power grids via black hole optimization. Adv Electr Elec-
tron Eng. 2019;17(1):24–32.
65. Deng Z, Rotaru MD, Sykulski JK. Kriging assisted surrogate evolutionary computation to solve optimal power flow
problems. IEEE Trans Power Syst. 2019;35(2):831–9.
66. Zhao T, et al. DeepOPF+: a deep neural network approach for DC optimal power flow for ensuring feasibility. In:
2020 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids
(SmartGridComm). IEEE; 2020.
67. Durairasan M, Balasubramanian D. RETRACTED: an efficient control strategy for optimal power flow management
from a renewable energy source to a generalized three-phase microgrid system: a hybrid squirrel search algorithm
with whale optimization algorithm approach. Trans Inst Meas Control. 2020;42(11):1960–76.
68. Mahdad B, Srairi K, Bouktir T. Optimal power flow for large-scale power system with shunt FACTS using efficient
parallel GA. Int J Electr Power Energy Syst. 2010;32(5):507–17.
69. Mukherjee A, Mukherjee V. Chaotic krill herd algorithm for optimal reactive power dispatch considering FACTS
devices. Appl Soft Comput. 2016;44:163–90.
70. Held L, et al. An optimal power flow algorithm for the simulation of energy storage systems in unbalanced three-
phase distribution grids. Energies. 2021;14(6):1623.
71 Venkatesan K, Govindarajan U. Optimal power flow control of hybrid renewable energy system with energy storage:
a WOANN strategy. J Renew Sustain Energy. 2019;11(1): 015501.
72. Rao RV, Savsani VJ, Balic J. Teaching-learning-based optimization algorithm for unconstrained and constrained real-
parameter optimization problems. Eng Optim. 2012;44(12):1447–62.
73. Mallipeddi R, et al. Efficient constraint handling for optimal reactive power dispatch problems. Swarm Evol Comput.
2012;5:28–36.
74 Galántai A. A stochastic convergence result for the Nelder-Mead simplex method. Mathematics. 2023;11(9):1998.
75 Meng Q, et al. Revolutionizing photovoltaic consumption and electric vehicle charging: a novel approach for resi-
dential distribution systems. IET Gener Transm Distrib. 2024. https://​doi.​org/​10.​1049/​gtd2.​13232.
76. Wang M, Lu GZ. A modified sine cosine algorithm for solving optimization problems. IEEE Access. 2021;9:27434–50.
77. Wu HY, et al. Super-evolutionary mechanism and Nelder-Mead simplex enhanced salp swarm algorithm for photo-
voltaic model parameter estimation. IET Renew Power Gener. 2024. https://​doi.​org/​10.​1049/​rpg2.​12973.
78. Warid W. Optimal power flow using the AMTPG-Jaya algorithm. Appl Soft Comput. 2020;91:106252.
79. Lei MY, Zhou YQ, Luo QF. Enhanced metaheuristic optimization: wind-driven flower pollination algorithm. IEEE
Access. 2019;7:111439–65.
80. Mugemanyi S, et al. Optimal reactive power dispatch using chaotic bat algorithm. IEEE Access. 2020;8:65830–67.
Wu et al. Journal of Big Data (2024) 11:172 Page 33 of 33

81. Acharjee P. Application of efficient self-adaptive differential evolutionary algorithm for voltage stability analysis
under practical security constraints. Appl Math Comput. 2013;219(23):10882–97.
82. Singh N, Singh S. A novel hybrid GWO-SCA approach for optimization problems. Eng Sci Technol Int J.
2017;20(6):1586–601.
83 Premkumar M, et al. Augmented weighted K-means grey wolf optimizer: an enhanced metaheuristic algorithm for
data clustering problems. Sci Rep. 2024;14(1):5434.

Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.