Research on Loss Reduction Method for UHVDC

System Based on Digital Twin Technology and

Improved Eagle Strategy Particle Swarm
2024 9th Asia Conference on Power and Electrical Engineering (ACPEE) | 979-8-3503-0963-8/24/$31.00 ©2024 IEEE | DOI: 10.1109/ACPEE60788.2024.10532545

Optimization Algorithm
1st Yuan Ma 2nd Xiaolin Song 3rd Liang Sun 4th Xin Li
State Grid Shaanxi Electric State Grid Shaanxi Electric State Grid Shaanxi Electric School of Electrical Engineering,
Power Company Marketing Power Company Marketing Power Company Marketing Xi'an Jiaotong University
Service Center (Metering Center) Service Center (Metering Center) Service Center (Metering Center) Xi'an, China
Xi’an, China Xi’an, China Xi’an, China 0009-0003-9450-9288
[email protected] [email protected] [email protected]

5th Wenbo Li 6th Xiaojun Tang*

School of Electrical Engineering, School of Electrical Engineering,
Xi'an Jiaotong University Xi'an Jiaotong University
Xi'an, China Xi'an, China
0000-0001-9266-4354 [email protected]

Abstract—The ultra-high voltage direct current (UHVDC) have experienced rapid development and widespread
transmission system demands highly sophisticated mathematical application. Traditional digital technologies struggle to achieve
models and meticulous data monitoring for energy loss energy consumption situational awareness and ultra-real-time
measurement, and the standard level of digitization often falls virtual deduction analysis in virtual spaces. However, digital
short in supporting advanced loss reduction research. To tackle twin technology enables the feedback from actual physical
this challenge, this paper proposes a digital twin based loss systems to information-physical space digital models,
reduction system architecture that integrates real-time sensing minimizing the deviation between actual data and fitted data
monitoring technology, cloud collaboration technology, and [1]. Numerous scholars have elaborated on the key
typical applications. Subsequently, a detailed mathematical
technologies and system architectures combined with digital
model encompassing converter stations and DC transmission
twins in areas such as the energy internet, high-proportion
lines was developed. Using data collected by measurement
equipment, an energy consumption measurement module was
photovoltaic grids, digitalized distribution systems, and power
implemented in the local computing kernel. This module equipment base systems, and have summarized and forecasted
interfaces with the CloudPSS digital twin platform for batch their development [2-5]. Therefore, constructing a UHVDC
simulation, facilitating the creation of an equipment energy digital twin system based on real physical space and virtual
consumption database. Concurrently, an enhanced Eagle digital space can provide a solid foundation for real-time
Strategy Particle Swarm (ESPSO) Optimization algorithm was energy consumption metering and loss reduction research.
introduced to optimize loss reduction parameters. The
Common research mainly focuses on the AC side energy
optimization results indicated a significant reduction in loss
40.9% in converters and 64.7% in AC filters, with an overall
metering of UHVDC systems, with very limited research on
system loss reduction of 28.27%. Therefore, the effectiveness of the development of devices related to energy metering of
the optimized loss reduction method and the virtual loss converter station components and the DC side, as well as
reduction technology was validated, leading to enhanced safety studies on UHVDC digital twin energy consumption metering
and reliability in the entire UHVDC transmission system. systems [6-10]. Only a few documents have analyzed the
energy consumption mechanisms of individual components of
Keywords—Digital twin; Energy loss measurement; Ultra-high UHVDC systems [11-12]. Therefore, conducting research on
voltage direct current transmission system; ESPSO; Virtual loss energy consumption metering and loss reduction techniques
reduction technology based on digital twins is of significant importance.
I. INTRODUCTION Long-distance transmission and AC-DC conversion in
UHVDC systems result in significant energy losses. Current
Nowadays, the ultra-high voltage direct current (UHVDC) research primarily focuses on the integration and analysis of
transmission systems based on line commutated converters historical energy consumption data of the system, the design of
Project Supported by State Grid Corporation of China Headquarters
overall loss reduction schemes for large AC-DC hybrid grids,
Technology Project (5700- 202122200A-0-0-00). and loss reduction in transmission line type design [13-14].

979-8-3503-0963-8/24/$31.00 ©2024
Authorized licensed use limitedIEEE 1417
to: Xian Jiaotong University. Downloaded on May 27,2024 at 10:06:53 UTC from IEEE Xplore. Restrictions apply.
Liang Xuming et al. [15] have shown that resistance losses are II. LOSS REDUCTION SYSTEM ARCHITECTURE BASED ON
a major factor in system losses, concluding that some losses are DIGITAL TWIN TECHNOLOGY SELECTING A TEMPLATE
proportional to the DC current, and that losses in back-to-back
DC systems are less influenced by factors such as power A. Construction of Digital Twin Loss Reduction System
transmission and ambient temperature. However, their research Digital twin technology, by integrating precise physical
was limited to statistical analysis of historical data on DC models, real-time data from sensors, and historical data sets
transmission system losses, without corresponding research on from measuring equipment, establishes a virtual simulation
loss reduction technologies. Liang Hanqing et al. [16] analyzed model with large steps and high accuracy. It displays the entire
the operational losses of DC transmission lines, deriving a lifecycle process of the physical system on a twin visualization
formula for calculating the economic current density of platform. In collaboration with cloud platform technology, it
conductors and suggesting economical choices for DC facilitates the processing of vast data, elastic computation and
transmission lines to reduce overall system losses. Various optimization of twin models, remote access and management,
intelligent computational methods for power loss optimization security protection, and collaborative sharing, further
have been proposed internationally, such as particle swarm promoting real-time online interaction between 'information-
optimization, differential evolution, and ant colony algorithms. physical-human' [20]. Currently, this technology is
Du Hongwei et al. [17] proposed a distribution network underutilized in ultra-high voltage direct current (UHVDC)
reactive power voltage self-coordination control method based transmission systems. This paper further proposes a
on dynamic cluster partitioning, which minimizes network corresponding digital twin model system architecture for
losses in reactive operation by suppressing voltage fluctuations UHVDC energy consumption metering and loss reduction, as
within the cluster through real-time local voltage control. Deng shown in Figure 1. The architecture consists of four parts: The
Wei et al. [18] presented a power grid loss prediction method physical layer, composed of key energy consumption device
for distribution networks based on grey relational analysis and metering technology and devices, installs sensors at
long short-term memory networks, providing a decision- measurement points in the actual UHVDC system, responsible
making basis for energy saving and loss reduction in for collecting data such as voltage, current, and power of AC
distribution networks. This paper studies energy consumption filters, converters, and DC transmission lines, and transmits it
optimization techniques based on UHVDC, proposing a virtual to the communication layer through the converter station
loss reduction plan that considers both safety and stability, intranet. The communication layer relies on the digital
which can enhance the level of energy efficiency management converter station system, responsible for data invocation and
of the power grid to a certain extent. feedback with the virtual layer. It can achieve synchronous
sampling of physical entity signals and support multi-device
With the rapid development of technologies such as big interoperability, serving as an integrated platform for multiple
data, artificial intelligence, and digital twins [19], technical functions and shared information, providing important
support has been provided for the pilot application of energy technical support for digital twin energy consumption metering
consumption metering and loss reduction systems in ±800kV and loss reduction. The virtual layer includes the UHVDC twin
ultra-high voltage direct current (UHVDC) transmission model layer and visualization system, capable of real-time
systems. To this end, this paper first proposes a digital twin energy consumption metering and loss reduction optimization,
loss reduction system architecture. This framework integrates and visualizes historical and real-time energy consumption,
energy consumption component monitoring technology and parameter assessment monitoring, abnormal energy
devices, remote data transmission technology, and digital twin consumption warnings, loss reduction plans, and effects. The
system software and hardware technologies to achieve application layer is responsible for feeding back the energy
functions like component-wise energy consumption metering, consumption metering and loss reduction optimization results
loss reduction, and data visualization. Additionally, the use of from the virtual layer, assisting operational and maintenance
the CloudPSS twin platform with a larger step size for batch personnel in understanding the system status, ensuring the safe
simulation of the UHVDC system, considering real parameters and stable operation of the overall system.
such as environmental impact and operational conditions,
closely approximates the actual physical system, enhancing the The key technologies of UHVDC loss reduction based on
accuracy and speed of energy consumption metering and digital twins include the following aspects: 1) Defining the
establishing a database for equipment energy consumption. metering point scheme for key energy-consuming components
Subsequently, an energy consumption optimization objective such as converter transformers and converter valve groups, and
function for the UHVDC system, which considers multiple developing ultra-high voltage direct current energy metering
objectives including safety and stability, is proposed, selecting devices that take into account the impact of parameters like
the control angles of converters and rectifiers as parameters to ripple, harmonics, and AC-DC intermixed under atypical
optimize, and establishing optimal loss reduction boundary complex load characteristics. This involves addressing the
conditions. Finally, the ESPSO algorithm is proposed for issue of asynchronous simultaneous clock synchronization of
combined optimization based on loss characteristics, validating multiple input signal sources. 2) Utilizing deep learning
the effectiveness of loss reduction technology and providing combined with high-dimensional coupled interpolation in
guidance for the operational management and control decisions model identification methods to construct a highly accurate
of UHVDC. UHVDC digital twin model. 3) Collaborating with cloud
platform technology to optimize parameters of the twin model,
implementing energy consumption optimization techniques,

1418
Authorized licensed use limited to: Xian Jiaotong University. Downloaded on May 27,2024 at 10:06:53 UTC from IEEE Xplore. Restrictions apply.
and proposing virtual loss reduction plans that consider both of the digital twin’s energy-related results are realized on the
safety and stability. CloudPSS platform. Utilizing cloud platform collaborative
technology and software development interfaces, the digital
The energy consumption component monitoring device is twin is interconnected with local computing cores, expanding
capable of collecting AC and DC real-time operational energy its loss measurement and reduction capabilities. Intelligent
parameters from converter stations and DC transmission lines.
algorithms are used to optimize model parameter
It facilitates edge computing and remote uploading, interfacing configurations, achieving optimized energy consumption
with the communication layer through the communication techniques. This also provides theoretical support for areas
interfaces of the twin system's hardware devices. The data such as intelligent operation and maintenance, clean energy
service platform preprocesses and computes high-precision,
consumption, and fault location [22]. Based on these
multi-channel synchronized sampling signals, including data technologies, this paper establishes a detailed mathematical
latency and accuracy analysis, deployment of Linux-based model of energy consumption metering based on UHVDC,
simulation computing programs, and data storage functions, proposes an improved ESPSO loss reduction algorithm for AC-
seamlessly integrating measured data into the digital twin [21].
DC loss characteristics, optimizes parameters such as converter
The most energy-relevant parameters are selected for transformer and rectifier control angles, and conducts loss
identifiability analysis and correction. Iterative optimization is reduction verification experiments on the twin platform,
employed to make the simulation results of the digital twin as thereby enhancing the system's stability and operational
close as possible to the actual physical system measurement economic efficiency.
results, and the historical inversion and real-time presentation

Equipment State Smart Operations and Clean Energy Product

Fault Localization
Sensing Maintenance Integration Optimization
Application
Layer
CLOUDPSS Energy Digital Twin Platform
Digital Twin Framework

Energy Optimization and Loss Reduction Techniques

Loss Reduction Model CLOUDPSS Simulation Data Analysis

Energy Power Environment Parameter Historical Performance

Parameters Constraints Configuration Optimization Backtracking Evaluation

Intelligent Intelligent Decision

SDK Interface SDK Interface Data Mining
Algorithms Algorithms Support

Virtual
Layer
UHVDC Visualization Digital Twin System
Model High-Dimensional Accurate Historical Energy Tracking
Digital Twin Large-Step Electromagnetic Deep
Identification and Coupled Interpolation
Framework Transient Simulation Model Learning
Optimization Simulation Real-time Energy Status Awareness

Online Parameter Assessment Monitoring

Digital Twin Configuring Monitoring Data Datav Energy Consumption
Visualization Communication Channels Changes Technology User Interface
Timely Abnormal Energy Consumption Warning

Data Feedback Data Retrieval

Digital Twin Data Service Platform

Analog-to-Digital High-Precision Clock Continuous Acquisition of Multi-

Conversion Module A/D Synchronization Channel Analog Signals

High-Precision
Communication Control Forwarding TDC Phase
Multi-Channel FPGA Control Clock Acquisition Mode
Module Correction
Layer Synchronized
Sampling Signals
Upper Computer
Data Processing Module DSP Chip Spectrum Analysis Digital Converter Station
Display

Remote Upload Remote Upload

Energy Consumption Critical Equipment Metering Technology and Devices

Multiple Digital Signals Electronic Voltage

Data Acquisition Ripple Fiber Optics
Transformer
Converter Electronic Current
Physical Converter Transformer
Valve Group
Harmonics
Multiple Analog Signals
Transformer
Integration Unit
Data Monitoring
Layer
Synchronized Clock
Digital Energy Meter Supplement
Synchronized Sampling
Data Detection AC Filter DC Line AC/DC Hybrid
Energy Consumption Critical Equipment Energy Consumption Subcomponent Metering Device

Fig. 1. A framework for the digital twin model of energy consumption measurement and loss reduction for UHVDC transmission

1419
Authorized licensed use limited to: Xian Jiaotong University. Downloaded on May 27,2024 at 10:06:53 UTC from IEEE Xplore. Restrictions apply.
B. Analysis of Key Energy-Consuming Components in Where R is the resistance value of the resistor, determined by
UHVDC factory measurement; I R represents the root mean square
current flowing through the filter resistor.
The topology of the ultra-high voltage direct current
(UHVDC) transmission system is shown in Figure 2. It mainly 2) CONVERTER TRANSFORMER LOSS
consists of components with high energy consumption such as The loss of the converter transformer can be determined by
converter transformers, converter valves, AC filters, and DC measuring the current flowing into the transformer and
transmission lines. The AC and DC filters are used to remove calculating the effective resistance of the transformer. The
AC noise and clutter from the power input, stabilizing the effective resistance can be measured by applying a source
voltage to enhance the performance and reliability of the power voltage of appropriate frequency to the low voltage winding of
system. The transmission of DC electrical energy is realized the transformer under test during the short circuit of the high
through commutation converters at both ends of the system, voltage winding, as shown in equation (3).
and the voltage is raised or lowered through equipment like
transformers to meet varying electrical demands. The station Ptra   1n I n2  R n 
control protection system is utilized to monitor and control the Where Ptra is the loss of the converter transformer; I n and R n
operation of the converter station, ensuring its safe and reliable are the transformer current and effective resistance measured at
th
operation. The computational formulas in this paper all refer to the n harmonic, respectively. The current flowing into the
the IEEE standards for measuring power loss in high-voltage transformer can be obtained by calculating the amplitude of the
direct current converter stations[23]. The specific description six-pulse characteristic harmonic current of the converter
of the relationship between these devices' energy consumption transformer, as shown in equation (4).
metering and the system's operational state is as follows.
 3  F  Eph
In 
1 1
j

n 5 p 5
 π  n  Xt

4
 2 1
 F   k1  k2  2k1k2 cos(2   )  2
7 7 2
2 3 3 2
m
 

  
6 6 sin[(n  1) ]
i
 2
k1 
8 8
AC Power
Supply І
AC Power
Supply ІІ
 n 1
 
 sin[(n  1) ]
2
k2 
Converter Converter
Station ІІ
 n 1
Station І
Where Eph is the effective value of the AC bus phase voltage of
the transformer; k represents the effective resistance
1-AC Filter; 2-Converter Transformer; 3-Rectifier; coefficient; X t represents the reactance of the converter
4-DC Transmission Line; 5-Smoothing Reactor; 6-Capacitor; transformer at the fundamental frequency;  represents the
7-DC Filter; 8-Station Control Protection System
commutation overlap angle of the converter;
Fig. 2. UHVDC transmission system topology F、k1、k2 represents an intermediate variable.
1) AC FILTER LOSS 3) THYRISTOR VALVE LOSS
The loss of the AC filter considers two parts: the losses of the Nt Id 2π  
reactor and the resistor, which can be represented by equations Pval  [V0  R 0 I d ( )] 
3 2π
(1) and (2), respectively. It is necessary to calculate the Where Pval is the conduction loss of the thyristor valve, which is
fundamental and harmonic currents in the filter reactor, and the
impedance of the reactor at the fundamental frequency, as well the main loss caused by the load current flowing through the
as the quality factor at both the fundamental and harmonic valve; Nt represents the number of thyristors connected in
frequencies, should refer to the test values at the time of series in a single valve; I d represents the average current of the
manufacture. DC bridge; V0 represents the average value of the on-state
voltage of the average thyristor; R0 represents the slope
I L2n  X Ln resistance of the on-state characteristics of the average thyristor.
Prea   nn 149 
Qn
4) THYRISTOR DAMPING LOSS
Where n is the harmonic number; I Ln is the calculated current Thyristor damping loss is generated by the 'step' change in the
through the reactor at the nth harmonic; X Ln is the reactance of valve blocking voltage with the energy stored in the valve
the reactor at the nth harmonic; Qn is the average quality factor capacitor, and is represented as follows:
of all reactors in the same category measured at the
n th harmonic. Vn2N fChf (7  6m2 )
Pdam  [sin 2   sin 2 (   )] 
4
Pres  RI R2  In the formula: Pdam represents the thyristor damping loss; VnN
represents the effective value of the transformer valve side

1420
Authorized licensed use limited to: Xian Jiaotong University. Downloaded on May 27,2024 at 10:06:53 UTC from IEEE Xplore. Restrictions apply.
open-circuit line voltage; f is the frequency of the AC system; deviation of the measured voltage VN relative to the nominal
Chf is the sum of the effective end-to-end capacitance inside the voltage , which should be within ±5% under stable system
valve plus the sum of the effective stray capacitance between conditions. The aim is to minimize v under the conditions
the external valve terminals; m is the electromagnetic coupling of parameter optimization changes. dct represents the tap
coefficient;  is the firing angle. position of the converter transformer. The voltage deviation at
different measurement points can be represented by the
5) LOSS OF DC TRANSMISSION LINE following function:
Pline  I 2 dn R L 
In the formula: Pline represents the loss of the DC transmission  N Ve  (1  d ct )
line; I dn is the effective value of the DC transmission line; R L is  f (d ct )   e 1  100
 VN 
the total impedance of the DC transmission line.
 min max
Based on the aforementioned detailed AC and DC energy d ct ＜d ct＜d ct
consumption mathematical models, a loss metering program is
developed in the local kernel. The program, through a software In the formula: N represents the number of system
development interface, calls the virtual operating parameters, measurement points; Ve represents the voltage at the
eth measurement point; dctmin and dctmax represent the upper and
fixed equipment parameters, and other computational data of
each component in the UHVDC twin model to complete the lower limits of the converter transformer tap position,
historical energy consumption deduction of the AC-DC system. respectively.
This UHVDC twin model employs deep learning combined The constraints of the multi-objective energy consumption
with high-dimensional coupled interpolation to identify input optimization model are as follows:
parameters, maximizing the approximation to the actual
UHVDC operating scenarios. 4) AC-DC Power Balance Constraint
To achieve the objective of virtual loss reduction in the  PGi  s pi Pdi  Pi  0, i  S AC
 
entire AC-DC system, it is necessary to further select QGi  Qcri  sQi Qdi  Qi  0, i  S AC
parameters that affect energy consumption, optimize them in
the loss metering module using intelligent algorithms, and feed In the formula: PGi and Pdi respectively represent the active
the optimal solutions back into the twin. This provides power of the generator and converter connected to node i ; Qcri ,
Qdi ,and QGi respectively represent the reactive power of the
decision-making support in a visual format for operational and
scheduling personnel. generator, AC filter, and converter connected to node i ; when
node i is an AC-DC coupling node, s pi  1 , sQi  1 ; when node i
C. Multi-Objective Energy Consumption Optimization Model is a pure AC, s pi  sQi  0 ; Pi and Qi respectively represent the
The optimization technology for UHVDC energy active and reactive power injection equations of node i ;
consumption metering based on digital twins needs to consider S AC represents the set of AC nodes.
multiple objectives such as economic efficiency, safety, and 5) Converter Station Operating Characteristics Equation
stability. It also involves determining the optimal loss
reduction boundaries based on the operating conditions and Constraint
adjustable parameter range of UHVDC. This primarily 3 2 3
Vd  k kV cos   mXs I d  0 
includes the following three objective functions:  b t 
1) Minimum Energy Consumption Where Vt and Vd are the fundamental component of the AC
commutation side line voltage and the DC voltage of the
f1  ( Pfil  Ptra  Pval  Pdam  Pline )min  converter station; k b represents the number of single-stage
six-pulse converters in the converter station; m is the number
Includes the minimization of the total energy consumption of
of stages of the converter station; k is the transformation ratio
each component in the DC transmission system based on
of the converter transformer; Xs represents the commutation
digital twins.
reactance of the converter station.
2) Maximization of Loss Reduction Benefits
6) Converter Station Control Mode Constraint Conditions
f 2  ( PA  PB )  c  T 
 6U jN
Considering economic factors, this includes maximizing the  I dN  [cos   cos(   )]
 4ωLc
loss reduction benefits of the UHVDC transmission system  
based on digital twins. Where PA and PB are the total losses  3 6 cos   cos(   )
before and after parameter optimization, c represents the VdN  U mN
  2
corresponding electricity price; T represents the operating time.
Where Lc is the equivalent reactance of the converter
3) Minimization of Voltage Deviation transformer; IdN and VdN respectively characterize the effective
f3  min(v(dct ))  values of the current and voltage of the converter valve side
Considering stability, this includes the minimization of voltage bus; U jN represents the effective value of the converter grid
deviation at various measurement points in the DC side bus voltage; U mN represents the effective value of the
transmission system. In the formula: v represents the converter transformer bus voltage.

1421
Authorized licensed use limited to: Xian Jiaotong University. Downloaded on May 27,2024 at 10:06:53 UTC from IEEE Xplore. Restrictions apply.
 7) Upper and Lower Limit Constraints for Control the grid, optimizes voltage distribution under different loads,
Variables and State Variables thereby reducing the overall system's energy consumption.
umin  u  umax B. Energy Consumption Optimization and Loss Reduction
  Algorithm Based on ESPSO
 xmin  x  xmax
Building upon the foundation of the high-precision energy
The control variable u represents the transformer tap ratio and consumption measurement method and virtual system
the control angle of the converter station; the constraints of simulation model mentioned above, in order to further reduce
control variable x consider the upper and lower limit system losses and enhance system stability, it is necessary to
constraints of voltage and current at both ends of each integrate intelligent optimization algorithms. The optimal
component in the DC transmission system. solutions obtained should be applied in the field to provide
personnel with virtual loss reduction measures.
For the above-mentioned multi-objective energy
consumption optimization model of the ultra-high voltage The Eagle Strategy (ES), inspired by the hunting behavior
direct current transmission system based on digital twins, it of eagles such as golden eagles, is used to address optimization
can be simplified into the form of equation (16). problems in continuous parameter spaces. It does not rely on
gradient information and optimizes solutions through random
min  f1 ( x), f 2 ( x), f 3 ( x ) searching and evaluation of candidate solutions. This strategy
 has two important parameters: random searching and dense
 s.t r ( x)  0  tracking, and the combination of these two parameters allows
 s  s ( x)  s for more effective exploration of the global space. Initially, ES
 employs the Levy distribution function for random global
Where r ( x) represents the equality constraints, s( x) represents searches. If promising solutions are found, it uses optimization
the inequality constraints. algorithms with local search capabilities for dense tracking.
This process is repeated with new global searches in different
III. OPTIMIZATION OF LOSS REDUCTION DIGITAL TWIN SYSTEM regions. In this study, the ES strategy was applied to improve
USING ESPSO ALGORITHM optimization algorithms and was used in power optimization
problems, resulting in improved outcomes.
A. Virtual Loss Reduction Method Based on OLTC
With the development of UHVDC transmission systems, The energy consumption measurement of the UHVDC
there is a growing demand for increased transmission capacity, system involves the collection, input, and processing of multi-
supply safety, reliability, and energy efficiency. Most scholars dimensional data. Therefore, the quality of the initial particles
have approached this from the perspective of power flow plays a crucial role in the learning speed and search
control, altering the power transmission in the system by performance of the algorithm. Assuming that there are D
controlling voltage amplitude, phase angle, and series particle subgroups, and the feature sets corresponding to
impedance, thereby ensuring the safe and reliable operation of particle M i are represented as ( xi(1) , xi(2) , , xi( N ) ) , and the target
the system[24]. Unified power flow controllers play a values for these subgroups are Q  ( y1 , y2 , , yi ) , the probability of
significant role in this context, capable of independently selecting a sample can be expressed as follows:
controlling active and reactive power flows, but their yi
manufacturing and maintenance costs are relatively high, pi  
making widespread use challenging. Traditional mechanical max QD ( y )
j 1,..., D
on-load tap changers (OLTC) have lower costs and, due to
The numerator and denominator respectively represent the
their mechanical components, offer high durability and
reliability, making them suitable for various load and target value of the particle in question and the maximum target
transformer applications. However, their regulation precision is value among all subgroups. If the probability value is greater
lower, and they are prone to arcing during tap changing, with than a random number within the [0,1] range, then the particle
slower response times [25]. With the application of new is selected, i.e., xi,D  1 ; otherwise, xi,D  0 . Based on the
electronic tap changers and digital control systems, fully probability values of particles in each subgroup, samples are
electronic OLTC can offer higher precision and intelligent selected from the original data to form the initial particle group.
control [26]. Based on real-time operational data of the system, The method of introducing the Levy function is used to
the control system analyzes this data and generates appropriate mutate particle position information, enhancing the particle's
digital control signals to adjust the tap position of the global search capability. The probability density function is as
transformer. In this scenario, the converter transformer can be follows:
considered a continuous variable. This paper uses a fully
electronic OLTC to construct a virtual simulation model,  
simulating rapid voltage regulation control without power  ( ) sin( )
2 1
outages. By optimizing the transformer tap ratio and phase ( s ) ~  s1 

angle, control over the firing angle and commutation angle is 
achieved. Under specific power transmission and system ( s  s0  0), (1    3)
stability requirements, the method reduces voltage deviation in

1422
Authorized licensed use limited to: Xian Jiaotong University. Downloaded on May 27,2024 at 10:06:53 UTC from IEEE Xplore. Restrictions apply.
Where ( ) is the standard gamma function, and s is the step
size. When   3 , it transforms into Brownian motion. When
  2 , it transforms into the Cauchy distribution. In the equation, min and max are pre-defined minimum and
maximum inertia coefficients, typically min set to 0.4 ,and max
d
set to 0.9. f average represents the average fitness value of all
Begin particles at the d th iteration, where fmind  min  f ( x1d ), f ( x2d ),..., f ( xnd ) is
the iteration number."
Using the Q-probability to Initialize the Population, Set the switching parameter p  2 to toggle between global
Initialize the particle's velocity vij(0) and position xij(0) :
and local search phases. If p＜2 , switch to the local search
phase, where the search is focused around the potential optimal
(k )
Calculate vij and xij based on the PSO formula,
(k )
solution. Calculate the new velocity and position of each
calculate the fitness value of the particle,
(k )
update the best position pij of the i t h particle and the overall
particle using Equation (19), compute their fitness functions,
best position pgj( k ) among all particles and update pi(,kj) and pg(k, )j . The flowchart is shown in Figure 3.

Use the Levy function for random global search.

IV. CASE STUDY ANALYSIS
Determine a random number and set the switching
parameter p between global search and local search. A. System Simulation Parameters
k=k+1
The experimental data is sourced from the Shaanbei-
no
p  0.2?
Wuhan ±800kV UHVDC transmission project, which includes
multidimensional data such as the root mean square voltage
yes
and current values at critical nodes of the system, converter
Enter local search and introduce the adaptive
inertia weight factor
transformer parameters, firing angle, commutation angle, and
various fixed equipment parameters. The recorded waveform
Calculate the particle's fitness function value;
data is processed internally and transmitted to the
(k ) (k )
Calculate the new particle's velocity vij and Position xij aforementioned digital twin model, where batch simulations
update pij( k )and pgj( k )
are conducted for energy consumption measurement and
k=k+1 optimization for loss reduction. The adjustable parameter limits
no
k  kmax for the system are shown in Table 1. Training data is taken
from February 2023 to October 2023, with a time step of 30
yes
(k )
minutes for loss analysis.
Output pgj
On the basis of energy consumption measurement, an
improved Particle Swarm Optimization (PSO) algorithm was
End employed for loss reduction. This method is suitable for
continuous variables. To achieve the desired search
Fig. 3. Flow diagram of EPSO optimization algorithm performance, in ESPSO, the actual positions of particles are
represented by vectors consisting of the values of transformer
If a potential solution is found during random global search, tap positions and rectifier control angles. These vectors are
an improved Particle Swarm Optimization (PSO) algorithm is used to calculate the objective function. The parameter settings
employed to perform a local search. Each particle in the for the improved model based on ESPSO are presented in
algorithm possesses three attributes: velocity, position, and Table 2.
fitness. During iterations, particles update their velocities and
The multi-dimensional parameter data collected in real-
positions by tracking individual best (P_best) and global best time by the converter station is input into the energy
(G_best) values, as shown in Equation (19). The fitness consumption measurement twin model, resulting in the energy
function evaluates the quality of each particle, thereby distribution of each component and the total energy
determining the solution to the optimization problem. consumption of the system. An improved optimization
algorithm is then applied to optimize the variable parameters.
vijk 1   vijk  c1r1 ( pbestijk  xijk )  c2 r2 ( gbestijk  xijk ) The fitness function is defined as the weighted sum of the three
 k 1  
k 1 objective functions given in equations (8-10):
 xij  xij  vij
k

F ( x)   3i 1 i f i ( x) 
In the equation, r1 and r2 are random numbers within the
interval [0,1], c1 and c2 are acceleration coefficients used to Where i represents the weighting coefficients for
assess the particle's acceleration. To enhance the algorithm's different optimization objectives, with values of 0.6, 0.2, and
search capability and convergence speed, an adaptive inertia 0.2, respectively.
weight is introduced, as shown in the following equation:
 f ( x d  f mind
)
min  (max  min ) d i , f ( xid )  f average
d

i  
d f average  f mind

 , f ( xi )  f average
d d
 max

1423
Authorized licensed use limited to: Xian Jiaotong University. Downloaded on May 27,2024 at 10:06:53 UTC from IEEE Xplore. Restrictions apply.
TABLE I. CONSTRAINTS ON PARAMETERS FOR ULTRA-HIGH VOLTAGE losses of various components of the system obtained by
DIRECT C URRENT TRANSMISSION LOSS REDUCTION SYSTEM
different optimization algorithms.
Parameters to be Optimized Upper and Lower Limits Based on Table 3, Figure 4 shows a comparison of the loss
Converter Transformer 0.9125≤k≤1 results obtained from the energy consumption measurement
5≤≤30 and loss reduction model. The results show that the total power
Firing Angle
losses of the system obtained by IMA, ASO, PSO, ISA, and
5≤≤20
Commutation Angle the improved ESPSO are 33.354MW, 30.171MW, 32.021MW,
26.198MW, and 25.159MW, respectively. Compared to before
algorithm optimization, their loss reduction rates are 4.9%,
TABLE II. ESPSO OPTIMIZES MODEL PARAMETERS 13.98%, 8.71%, 25.31%, and 28.27%, respectively. The
improved ESPSO algorithm exhibits better robust performance.
Parameters Value/Range Figure 5 shows a comparison of the discrete loss results
Parameters D 5 randomly extracted from various components of the UHVDC
30
system. It can be observed from the figure that under constant
Number of Particles N current control mode, the transmission line losses do not vary
[ 10,10]
Velocity vi significantly. The losses of the AC filter and the converter
Learning Factor c1 , c2 2 transformer directly affect the fluctuations in the output DC
[0.4, 0.9] voltage and the voltage and current of the entire AC-DC
Inertia Weight 
system. Therefore, the AC filter and the converter transformer
Switching Parameter p 0.2 are the main objects of loss reduction. Among them, the
Number of Evolutionary Generations gen 50 improved ESPSO algorithm achieves the most significant loss
reduction, with reduction rates of 64.7% and 40.9%,
respectively.
The loss reduction twin system achieves precise control of
voltage and current by optimizing the combination of 80 Ptotal
Pdam
parameters for converter transformers, firing angles, and 70 Pline
commutation angles. It also improves the quality of AC and Pval
DC waveforms, reduces the system's harmonic levels, and 60 Ptra
Prea
lowers overall system losses, thereby enhancing the system's
50
stability and efficiency while meeting load demands and
Loss ( MW )

network conditions. In this study, the relative error is used as 40

an indicator to evaluate the safety and stability of the system's
operation: 30

yˆi  yi 20
 100% 
yi 10

In the equation, yˆi and yi represent the discrete values and 0

IMA ASO PSO ISA ESPSO
rated values of the AC and DC waveforms, respectively. A Optimization Algorithm
smaller value of the indicator indicates a safer and more stable
operation of the system under these conditions. Fig. 4. Loss reduction effect of different optimization algorithms

B. Validation Test of Energy Consumption Measurement and The improved ESPSO accelerates the algorithm's
Loss Reduction Twin Model optimization speed by optimizing the initial population using
Population-based intelligent optimization algorithms such the proposed Q-probability method and combines it with the p
as IMA, ASO, PSO, and ISA are widely used in engineering parameter for global and local searches until it converges to the
optimization problems. To validate the superiority and optimal solution. The convergence curve of the fitness function
reliability of the continuous voltage regulation strategy based for the energy consumption measurement and loss reduction
on ESPSO, the above-mentioned optimization algorithms and system is shown in Figure 6.
the proposed improved ESPSO were used to solve the energy Furthermore, to clearly present the results of the continuous
consumption measurement and loss reduction model. The voltage regulation strategy based on improved ESPSO, 2500
maximum number of iterations was set to 100, and after data points were collected in a time series, and a comparison
running 50 times, the best optimization solutions are shown in chart of DC current and voltage fluctuations was plotted, as
Table 3. The loss reduction twin model simulates continuous shown in Figures 7 and 8. From the fluctuation effects, it can
voltage regulation mechanisms based on OLTC control be observed that the twin system improves the waveform
strategies, which effectively expand the optimization range of quality of DC voltage and current by optimizing the
multidimensional parameters compared to traditional tap-level combination of control angles on the converter and rectifier
regulation, providing more flexibility. The table displays the sides, reducing some of the losses caused by harmonics.
high-voltage winding coefficients, rectifier control angles, and

1424
Authorized licensed use limited to: Xian Jiaotong University. Downloaded on May 27,2024 at 10:06:53 UTC from IEEE Xplore. Restrictions apply.
 TABLE III. CONTROL VARIABLE VALUES AND POWER LOSSES

Variable Tap Changer Style IMA ASO PSO ISA ESPSO

Winding Factor 0.015 0.048 0.066 0.053 0.091 0.099
Firing Angle/° 23.354 18.629 16.049 17.967 11.443 10.042
Commutation Angle/° 6.961 6.997 6.894 7.017 7.006 6.853
Converter Transformer 1.116 1.157 1.178 1.162 1.208 1.218
AC Filter/MW 6.403 5.394 4.879 5.261 5.067 3.778
Converter Transformer/MW 10.789 9.777 7.347 9.192 4.321 3.808
Thyristor Valve/MW 6.816 7.181 7.198 6.948 6.887 7.261
DC Transmission Line/MW 9.188 9.683 9.704 9.369 9.285 9.788
Thyristor Damping/MW 1.879 1.319 1.042 1.248 0.637 0.524

7.5 20
Initial Initial

Transformer Loss ( MW )
7.0
AC Filter Loss ( MW )

ESPSO ESPSO
6.5 15
6.0
5.5
10
5.0
4.5
4.0 5

3.5
3.0 0
0 50 100 150 200 0 50 100 150 200

Sequence Number Sequence Number

8.0 10.5
Thyristor Valve Loss ( MW )

Initial Initial
7.6 ESPSO 10.0 ESPSO
Line Loss ( MW )

7.2 9.5

6.8 9.0

6.4 8.5

6.0 8.0
0 50 100 150 200 0 50 100 150 200
Sequence Number Sequence Number

2.4 45
Initial Initial
Thyristor Damping Loss ( MW )

Total System Loss ( MW )

ESPSO ESPSO
2.0 40

1.6 35

1.2 30

0.8 25

0.4 20
0 50 100 150 200 0 50 100 150 200
Sequence Number Sequence Number

Fig. 5. Comparison of the loss reduction effect of each sub-component sequence of the EPSO algorithm

On the AC side, 2500 sampling points were also taken for optimization, the THD values improved to 12.6% and 0.88%
harmonic analysis. Before optimization, the distortion of for current and voltage, respectively. This improvement in
current and voltage waveforms, known as Total Harmonic waveform distortion of voltage and current outputs enhances
Distortion (THD), was 19.4% and 1.19%, respectively. After the safety, stability, and efficiency of the system.

1425
Authorized licensed use limited to: Xian Jiaotong University. Downloaded on May 27,2024 at 10:06:53 UTC from IEEE Xplore. Restrictions apply.
 1.19%, respectively. After optimization, the THD values
60
improved to 12.6% and 0.88% for current and voltage,
ESPSO respectively. This indicates that the waveform distortion of
55
voltage and current outputs has been improved to some extent,
making the system operate more safely, stably, and efficiently.
Fitness Function

50

V. CONCLUSION
45
This paper proposes a strategy for energy consumption
measurement and loss reduction based on digital twin
40
technology, exploring new applications of this technology in
UHVDC systems. Initially, the paper presents the digital twin
35
0 10 20 30 40 50 60 70 80 90 100 architecture of the loss reduction system and constructs models
Number of Iterations for real-time energy consumption measurement and loss
reduction. The improved ESPSO algorithm optimizes the
Fig. 6. Convergence curve of fitness function of energy consumption
converter and rectifier control angle parameters, enhancing
metering and loss reduction system
power quality and system stability. The robustness of this
approach is verified by comparing actual loss measurement
300
Initial
data from the Shaanbei-Wuhan converter station with the
ESPSO losses obtained from this optimization strategy, achieving an
overall system loss reduction rate of 28.27%. Moreover, after
DC Voltage Fluctuation

290 parameter optimization, the quality of the DC voltage and

current waveforms is enhanced, and the harmonic content of
the AC current is reduced, resulting in safer and more stable
280
system operation. Finally, through practical case validation,
this method demonstrates its ability to monitor and measure the
energy consumption of DC transmission systems in real time,
thereby improving system stability and reducing overall losses.
270
0 500 1000 1500 2000 2500
ACKNOWLEDGMENT
Time ( min )
Special thanks to the generous support from Professor
Fig. 7. Comparison chart of DC voltage fluctuation Xiaojun Tang's laboratory at the School of Electrical
Engineering, Xi'an Jiaotong University.
0.08
Initial
ESPSO REFERENCES
0.04 [1] S. G. Hao, Q. Yong, and L. Lingen, “Key technologies and
DC Current Fluctuation

application prospects for power equipment operation and

maintenance in new power systems,”. High Voltage Technology, vol.
0.00
47, no. 9, 2021, pp. 3072-3084.
[2] P. T. Jiao, C. Sheng, and Z. Qi, “Design and application prospects of
digital twin system framework for energy internet,”. Chinese Journal
-0.04 of Electrical Engineering, vol. 41, no. 6, 2021, pp. 2012-2029.
[3] D. X. Dong, Z. J. Li, L. and X. Xue, “Overload risk warning method
for photovoltaic high proportion distribution networks based on
-0.08 digital twins,”. Power system protection and control, vol. 50, no. 9,
0 500 1000 1500 2000 2500
2022, pp. 136-144.
Time ( min ) [4] F. Z. Ming, J. Ling, and Z. X. Rui, “Design of digital twin base
system for power equipment in distribution systems,”. Integrated
Fig. 8. Comparison chart of DC current fluctuations Smart Energy, vol. 45, no. 3, 2023, pp. 50-56.
[5] S. Y. Wei, Z. L. Mei, and M. Zhao, “Preliminary exploration of
Furthermore, to clearly present the results of the improved digital active distribution systems”. Chinese Journal of Electrical
ESPSO continuous voltage regulation strategy, 2500 data Engineering, vol. 42, no. 3, 2022, pp. 1760-1773.
points were collected in a time series, and the comparison [6] T. D. Ping, Z. Xin, and L. C. Zhu, “Research and implementation of
Back-to-back DC transmission network loss monitoring system,”
graphs of DC current and voltage fluctuations were plotted, as Electrical Measurement and Instrumentation, vol. 58, no. 3, pp. 74-80.
shown in Figures 7 and 8. From the waveform fluctuations, it 2021.
can be observed that the twin-system, through the optimization [7] B. Tan, J. Liu, W Luo, H. Zhou, and J. Zhao, “A method for
of the combination of the commutation transformer and predicting power loss of HVDC converters based on support vector
rectifier-side control angle parameters, improved the waveform regression,” The 13th International Congress on Image and Signal
quality of DC voltage and current, reducing partial losses Processing, BioMedical Engineering and Informatics (CISP-BMEI),
Chengdu, China, 2020. pp. 1037-1041.
caused by harmonics. Similarly, 2500 sampling points were
taken on the AC side for harmonic analysis. Before [8] A. Tosatto, T. Weckesser, and S. Chatzivasileiadis, “Market
Integration of HVDC lines: internalizing HVDC losses in market
optimization, the distortion of current and voltage waveforms, clearing,”. IEEE Transactions on Power Systems, vol. 35, no.6, pp.
known as Total Harmonic Distortion (THD), was 19.4% and 451-461. January 2020.

1426
Authorized licensed use limited to: Xian Jiaotong University. Downloaded on May 27,2024 at 10:06:53 UTC from IEEE Xplore. Restrictions apply.
[9] M. A. Odhano, M. S. Jokhio, N. A. Unar, N. H. Mirjat, S. A. Khatri, Networks Based on Cluster Dynamic Partition,” Power System
and B. Aslam, “Modelling and analysis of power losses in Automation, 2023, vol. 12, pp. 1-11.
transmission system of pakistan: a case study of matiari-lahore VSC- [18] D. Wei, L. Li, and L. Yong, “Distribution network loss prediction
HVDC Link," The 6th International Multi-Topic ICT Conference based on GCA and LSTM neural networks,” Guangdong Electric
(IMTIC), Jamshoro & Karachi, Pakistan, 2021. pp.1-5. Power, 2019, vol. 32, no. 09, pp. 154-159.
[10] M. Nian, W. G. Da, and C. Zhao, “Analytical calculation of steady [19] Z. Peng, P. T. jiao, and W. X. ying, “Key technologies and prospects
state current stress in key components of modular multilevel of power internet of things for energy internet digital twins,” Chinese
converter submodules,” Power Grid Technology, 2022, vol.46, no.10, Journal of Electrical Engineering, 2022, vol. 42, no. 2, pp. 447 -458.
pp. 3712-3720.
[20] L. Peng, X. Wei, and C. T. tian, “The concept, architecture, and key
[11] F. Z. Qiang, Z. Z. Ming, and S. B. Chen, “Loss analysis of multi port technologies of digital grid,” Chinese Journal of Electrical
power electronic transformers based on DSIM simulation,” Journal of Engineering, 2022, vol. 42, no. 14, pp. 5002-5017.
Tsinghua University (Natural Science Edition), 2023, vol. 63, no. 1,
[21] R. Q. Kun, X. X. Li, and L. J. Ju, “Research on digital twin
pp. 94-103.
technology in cyberspace security,” Journal of System Simulation,
[12] X. J. Bo, Y. C. Xiang, and S. Zhan, “Identification of key lines in 2023, vol. 12, pp. 1-15.
AC/DC hybrid power grid based on static safety domain,” Power
[22] Z. Y. Jun, Y. Y. Qi, and L. L. Qian, “Outlook on New Low Voltage
System Protection and Control, 2020, vol. 48, no.6, pp. 165 -172.
Distribution System Technology Driven by Dual Carbon Goals,”
[13] D. P. Fei, Y. L. Zhi, and L. H. Jun, “Analysis of line loss and Power System Automation, 2022, vol. 46, no. 22, pp. 1 -12.
capacity expansion in AC/DC same line transmission technolog,”
[23] "IEEE Recommended Practice for Determining Safe Distances From
Power Supply Technology, 2017, vol. 41, no.1, pp. 142 -145.
Radio Frequency Transmitting Antennas When Using Electric
[14] L. H. Qing, L. Y. Dong, and W. Lei, “Research on the optimal ratio Blasting Caps During Explosive Operations," in IEEE Std C95.4 -
and networking performance of hybrid half wavelength AC 2002 , vol., no., pp.1-52, 9 Feb. 2003.
transmission lines,” High voltage technology, 2018, vol. 44, no. 1, pp.
[24] T. X. Gao, W. Feng, and Z. Fang, “Review of research on partial
65-74.
power DC converters,” Journal of Electrical Engineering Technology,
[15] L. H. Qing, W. Xiong, and L. X. Ming, “Analysis of operational 2023, vol. 12, pp. 1-17.
losses and economic efficiency of ultra-high voltage AC and high-
[25] S. D. Dong, D. L. Wei, and D. Biao, “Design and research on
voltage DC transmission systems,” High Voltage Technology, 2013,
improvement schemes for mechanical on load tap changer,” High
vol. 39, no. 3, pp.630-635.
Voltage Technology, 2019, vol. 45, no.7, pp. 2141-2149.
[16] L. H. Qing, L. Y. Dong, and S. G. hao, “A method for reproducing
[26] W. He, and L. J. Xiang, “Overview of Power Electronics
single-phase grounding faults in transmission lines based on traveling
wave inversion,” High Voltage Technology, 2019, vol. 45, no. 01, pp. Evolution of Transformers and Its Impact on Voltage Stability,”
209-2017.
Power System Protection and Control, 2020, vol.48, no. 16, pp.
171-187.
[17] D. H. Wei, Wei T. Zheng, and X. Dong, “Autonomous and
Collaborative Control of Reactive Power and Voltage in Distribution

1427
Authorized licensed use limited to: Xian Jiaotong University. Downloaded on May 27,2024 at 10:06:53 UTC from IEEE Xplore. Restrictions apply.