Received: 20 October 2022 Revised: 23 November 2022 Accepted: 10 December 2022 IET Generation, Transmission & Distribution

DOI: 10.1049/gtd2.12721

ORIGINAL RESEARCH

A zero-sample state evaluation model for valve-side bushing of

UHV converter transformer oriented to digital twin under
attribute analysis

Zheng Li1 Kai Liu1 Mu Lin1 Dongli Xin1 Hao Tang2

Guangning Wu1

1
School of Electrical Engineering, Southwest Abstract
Jiaotong University, Chengdu, China
The valve-side bushing of the UHV converter transformer is the key equipment in the
2
China Electric Power Research Institute, Beijing, DC transmission project, and its running state directly affects the security and stability
China
of the power system. This paper analyses the attributes of the physical entities, uses the
digital twins to establish the state feature set and proposes a zero-sample state evaluation
Correspondence
Kai Liu, School of Electrical Engineering, Southwest algorithm for the valve-side bushing. First, the geometric, materials and electrical character-
Jiaotong University, Chengdu, China. istics are analyzed, and the detailed components of carrier current are obtained by empirical
Email: liukai@swjtu.edu.cn
mode decomposition. Then, COMSOL is used to establish digital twins, verify the validity
of twins with axial heat distribution of bushing, and establish a state feature set with the
Funding information
Science and Technology Project of Headquarter of extreme temperature inside and outside bushing. Finally, the fuzzy clustering algorithm is
SGCC, Grant/Award Number: SGTYHT/19- used to classify the state feature set, and the similarity is used as the index to realize the
JS-215. zero-sample state evaluation of the valve-side bushing. Through the demonstration and
analysis of examples, the evaluation model solves the problems of difficulty in extract-
ing the internal features, fewer fault samples, and training difficulty, which is conducive
to improving the operation and maintenance management level of power transmission
equipment.

1 INTRODUCTION At this stage, most of the research on the valve-side bushing of

UHV converter transformers remains in the fault occurrence
Under the background of low carbon and energy structure mechanism, fault analysis, and physical structure optimization
adjustment, higher requirements are put forward for the opti- [8–10]. However, due to the variety of fault forms, difficulties
mal allocation of power resources [1]. The key to realizing the in data collection, and poor correlation, there is less research on
safe transmission of large capacity and long-distance power is the evaluation of the running state.
to improve the operation and maintenance management level In the state research of valve-side bushing of UHV converter
of power equipment. The valve-side bushing of the ultra-high transformers, the acquisition of its state data is the first pre-
voltage (UHV) converter transformer is the “bridge” and weak requisite. With the development of digital twin technology in
link to realize power conversion [2], and its running state sta- aerospace technology [11], medical equipment [12], rail tran-
bility directly affects the reliability of DC transmission projects. sit [13], water conservancy engineering [14], and other fields,
According to statistics, in recent years, power system failures the integration technology between digital twin technology and
caused by converter valve-side bushing of converter trans- electric power field is becoming closer. Tao et al. [15] used digital
former account for 16.7% of all equipment failures [3], and the twin technology to establish a health status management sys-
forms of failures also show diversity, such as SF6 gas leakage, tem for wind turbines, and real-time updates and fusion of twin
aging of capacitor core, oil leakage, discharge, and so on [4–7]. data can be carried out through this system. Jia et al. [16, 17]

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is
properly cited.
© 2022 The Authors. IET Generation, Transmission & Distribution published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology.

IET Gener. Transm. Distrib. 2023;17:1123–1134. wileyonlinelibrary.com/iet-gtd 1123

1124 LI ET AL.

established the CloudEPS platform by using digital twin tech- Based on the above analysis, this paper proposes to obtain
nology to plan and process physical entities at the virtual level, the twin data of the valve-side bushing of the UHV converter
thus providing a feasible scheme for the optimal control of transformer under the attribute analysis based on the digital
the energy internet. To improve the operation and maintenance twin technology and realize the zero-sample status evaluation
level of high-voltage cables, He et al. [18] proposed the use of combined with the fuzzy clustering algorithm. First, the phys-
the digital twin framework to carry out full-cycle state manage- ical entity is analyzed, including attribute parameters, electrical
ment of high-voltage cables. As the heavy equipment in power features, ambient temperature, and current-carrying. To extract
systems, the development of digital twin technology also brings more details of the carried current of the bushing, empiri-
new opportunities and challenges to the valve-side bushing of cal mode decomposition (EMD) is used to decompose the
UHV converter transformers. fundamental and harmonic components. Then, the COMSOL
To ensure the reliability of the state evaluation of the valve- software is used to complete the mapping from physical entity
side bushing of the UHV converter, the analysis of its effective to digital twin, which is verified and analyzed according to the
state feature information is a key step. Elsisi et al. [19] con- axial heat distribution of the bushing, and use different carrier
ducted fault diagnosis for transformers based on an effective current components to obtain the temperature extremum inside
deep-learning platform. Cong et al. [20] tested the infrared spec- and outside the bushing to establish the twin data. Finally, the
trum of the transformer in running and established the electric initial state feature sets of valve-side bushing under different
heating fault diagnosis model of the transformer oil. Leong et al. defect degrees are established, the fuzzy clustering algorithm
[21] studied the transformer health state index and proposed is used to process them, and the state evaluation of valve-side
the method of transformer state feature extraction using the bushing is completed with the similarity matching degree in the
ultraviolet-visible spectrum for the first time. Du et al. [22] used dynamic clustering diagram. Through the verification and analy-
currents of different frequencies as testing tools to analyse the sis of examples, the state evaluation method can ensure the safe
influence of water in oil on bushing aging. Monga et al. [23] running of equipment and further improve the operation and
proposed the calculation method for bushing electric fields and maintenance level of the valve-side bushing.
studied the influence of key components on electric field distri- The main contributions of this paper are as follows: (a)
bution in combination with the bushing structure model. The Analyse the basic physical attributes of the bushing, and then
research on bushing state characteristics also includes insulation establish the physical entity in the digital twin technology,
features [24], dielectric features [25], partial discharge [26], and including not only the fundamental and harmonic components
other objects, but the above feature objects have problems such of the input current after EMD processing, but also the analy-
as fewer data and lower sensitivity to training parameters in the sis of the bushing structural materials, electrical characteristics,
process of bushing state evaluation. and geometric dimensions. (b) The overheating mechanism of
In the state evaluation model of the valve-side bushing of the the bushing is simulated by COMSOL to obtain the overheat-
UHV converter transformer, Záliš et al. [27] proposed to estab- ing characteristics of internal defects, and then the bushing state
lish an expert mapping system using historical fault data, but due feature set under different defects is established. (c) Aimed at
to the differences between valve-side bushing models and run- the characteristic of a few bushing fault samples, a zero-sample
ning environments, the stability of its weight parameters is poor. state evaluation model based on a fuzzy clustering algorithm
Liao et al. [28] used the dielectric model to analyse the water is proposed. The bushing state evaluation is completed by
content of oil-penetrating bushing and took the grey correlation similarity matching degree, which does not require advanced
degree as a parameter index to complete the state evaluation. training.
Elsisi et al. [29–34] applied machine learning and optimization The structure of this paper is as follows: (1) Analysis of basic
algorithms to automatic driving, hybrid battery control, robot algorithms, mainly including digital twin framework, attribute
arm cooperative control, etc., while there are relatively few state analysis algorithm, and fuzzy clustering algorithm; (2) state eval-
assessment optimization algorithms in ultra-high voltage power uation of valve-side bushing of UHV converter transformer,
equipment. including physical entity analysis, virtual twin, state evaluation
The valve-side bushing of the UHV converter transformer process; (3) example demonstration, mainly including the simu-
is different from other power insulation equipment. It has a lation of virtual space, the establishment of state feature set of
large internal current and a complex external running envi- data twins, and the evaluation of running state; 4) conclusion.
ronment, which will cause serious accidents when it fails [35].
Due to the rapid occurrence of valve-side bushing accidents
and the low accident rate, it is difficult to collect status data. 2 ALGORITHM ANALYSIS
The combination of digital twin technology and status feature
analysis can solve the problem of insufficient status representa- 2.1 Digital twin framework for state
tion. When using the fuzzy clustering algorithm to establish the assessment system
state evaluation model, it has the characteristics of zero-samples
[36], and the state evaluation results can be intuitively expressed The concept of digital twins was first proposed by Grieves
through the similarity matching degree in the dynamic clustering [37, 38] and used to establish the mapping relationship
graph. between physical entities and abstract virtual entities. With the
LI ET AL. 1125

2.2 Attribute analysis algorithm

The attribute analysis algorithm is the mathematical basis for

the dynamic analysis of the research object. The attribute
analysis of the valve-side bushing of the UHV converter trans-
former includes both the description of physical entities and
the dynamic display principle of digital twins. The input cur-
rent decomposition algorithm and heat distribution theory are
introduced in detail below.

1) Output current decomposition algorithm.

EMD is the frequency domain analysis of the original signal

according to the time scale. Assuming that the original signal
FIGURE 1 System frame of digital twin technology
sequence is x(t) and the natural mode function component is
imfi (t), the EMD can be formulated as Equation (1).

∑
n
development of information technology, the internet of things, x (t ) = im fi (t ) + r (t ), (1)
cloud computing, communications, and other technologies, i=1
higher requirements are put forward for digital twin technol-
where i is the variable of decomposition layer n, r(t) is the
ogy in the operation and maintenance of specific equipment.
residual component, and different intrinsic mode function
The application of digital twin technology in the field of power
components imfi (t) meet the Dirichley conditions:
equipment mainly consists of five parts: physical entity (PE),
The termination condition of the whole decomposition pro-
digital twin (DW), operations management (OM), digital link
cess is that r(t) is a monotone signal. The corresponding EMD
(DL), and Data twin (DT). The system framework is shown in
is shown in Algorithm 1.
Figure 1.
The specific meanings of digital twin technology are as
follows:

1. PE: When the research object of the physical entity is the ALGORITHM 1: EMD
valve-side bushing, the characteristic description of the input Input: Original signal x(t); Decomposition layers i; Intermediate
object includes not only the geometric, material, and envi- variable k.
ronment, but also the output current characteristics, defect Output: Intrinsic mode function component imfi (t), residual
distribution, sensing equipment, and other elements. component r(t).
2. DW: Digital twin is to complete the mapping from physical 1 // initialization
entity to virtual space, including product simulation, dynamic
2: i←1, k←0
display, and operation parameters.
3: set r(t) = x(t)
3. OM: Operation and maintenance management is the
premise to ensure the safe and stable running of equipment, 4: r(t) local maximum point emax (t), r(t) local minimum
point emin (t)
mainly including fault diagnosis, state evaluation, and state
prediction. 5: // evaluate
4. DT: Data twinning runs through the whole physical entity, 6: m(t ) = (emax (t ) + emin (t ))∕2, update k = k+1
digital twinning, operation and maintenance management, 7: pk (t ) = r (t ) − m(t ), r (t ) = pk (t )
and other links. 8: If pk (t )satisfies two conditions then
5. DL: As a bridge, the digital link realizes the data interac-
9 im fi (t ) = pk (t )
tion between physical entities, digital twins, operation, and
10 r (t ) = r (t ) − im fi (t )
maintenance management.
11 and if r(t) is the monotone signal then
∑n
The digital twin framework based on the state evaluation 12 x(t ) = i=1 im fi (t ) + r (t )
system fully maps the running characteristics of physical enti- 13 else i = i+1
ties into the virtual space. While establishing the digital twin, it 14 Return r(t) local maximum point emax (t), r(t) local minimum
obtains the internal characteristics to realize the expansion of point emin (t)
the running status features and solves the problems of online 15 else
monitoring difficulty and a small sample size of fault defects.
16 Return r(t) local maximum point emax (t), r(t) local minimum point
Thus, the whole life cycle of physical entities and digital twins emin (t)
can be synchronously evolved and evaluated.
1126 LI ET AL.

2) Calculation principle of heat distribution. fuzzy clustering algorithm based on the initial feature matrix Xn
The heat source on the valve-side bushing of the UHV * m is shown in Algorithm 2.
converter transformer is generated by the Joule effect of the
conductive tube; the heat conduction between different solid
materials, the convection heat transfer between solid and
non-solid materials, and the thermal radiation of the solid itself ALGORITHM 2: Fuzzy Clustering Algorithm
need to be comprehensively considered in the heat distribution Input: Original signal x(t); Decomposition layers i; Intermediate variable
calculation. k.
Heat conduction exists between two solids with a cer- Output: Intrinsic mode function component imfi (t), residual component
tain temperature difference. The solid materials of the r(t).
valve-side bushing include a catheter, capacitor core, flange, 1: // initialization
watchband contact, and other devices, and the heat con- 2: i←(1,2, …n), j←(1,2, …m)
duction satisfies the mathematical definition shown in
3: k←0
Equation (2) [39]:
4: // evaluate
( ) 6: Standardization (X→X′′ )
𝜆t 𝜕 𝜕T 𝜕2 T
⋅ r + 𝜆t ⋅ + Qtc = 0, (2)
r 𝜕r 𝜕r 𝜕z 2 xi j − x̄ j
′
7 ∶ xik = (i = 1, 2, … , n; j = 1, 2, … , m) (5)
sj
where r is the geometric radius; T the temperature; λt the ther-
mal conductivity; Qtc the amount of heat conduction between
solids in contact. x ′ i j − min {x ′ i j }
′′ 1≤i≤n
The convective heat transfer of the valve-side bushing 8∶ xik = ( j = 1, 2, … , m) (6)
max {x ′ i j } − min {x ′ i j }
1≤i≤n 1≤i≤n
includes two contact modes: the external natural environment
and the transformer insulating oil; the convective heat transfer 9: Similitude matrix (X′′ →R)
satisfies the mathematical model of Equation (3) [40]:
[ ]
1 ∑
2
3 (xik − x jk )
m

2𝜋kNu (Ti − To ) 10 ∶ ri j = exp − · (7)

Qi = , (3) m k=1 4 s2 k
ln(Do∕Di )
11: Equivalent Boolean Matrix (R→R∗ )
where Qi is the convective heat conduction; k the convective
thermal conductivity; Nu is the Nussel constant; Ti , To the tem- 12 ∶ R2 = R◦R , R4 = R2 ◦R2 , … , (8)
perature inside and outside the bushing surface; Di , Do the
diameter inside and outside the bushing.
13: If there is a natural K, which makes R2K = R2(K +1) then
The thermal radiation of the valve-side bushing is also a
part of the thermal distribution. Compared with heat con- {
1, ri j ≥ 𝜆
duction and convective heat transfer, thermal radiation does 14 ∶ ri j = (9)
0, ri j < 𝜆
not need to contact the medium to transfer. The solid
thermal radiation based on electromagnetic energy can be
calculated by the Stefan–Boltzmann equation as shown in
Equation (4) [41]: Equation (5) of Algorithm 2 solves the dimensional problem
so that the elements of the fuzzy characteristic matrix are in the
Qr = 𝜀𝜎A1 F12 (T1 4 − T2 4 ) (4) interval [0, 1]; x̄ j and sk , respectively, represents the mean value
and variance, which are calculation as follows:
where Qr is the thermal radiation; 𝜀 the emissivity (boldface);
1∑
n
𝜎 the Stefan–Boltzmann factor; A1 the surface area of the
x̄ j = x ( j = 1, 2, … , m), (10)
radiator; F12 the shape coefficient between surfaces of differ- n i−1 i j
ent radiating radiators; T1 , T2 the absolute temperature of the
1∑
n
surface of different radiators. sk = (x − x̄ j )2 . (11)
n i=1 i j

2.3 State evaluation algorithm Finally, when the confidence factor 𝜆 changes from 1 to 0 (the
computer automatically reduces by 0.001 steps), the columns
Suppose there are n objects to be inspected xt (t = 1, 2, …, n), with the same equivalent Boolean matrix are classified into one
and there are m characteristic indicators under each object to category. When 𝜆 = min {ri j } completes the classification, the
1≤i, j ≤n
be inspected. According to the theory of fuzzy mathematics, a dynamic clustering graph is established accordingly.
LI ET AL. 1127

TABLE 1 Attribute parameters of the valve-side bushing of the UHV

converter transformer

Component Attribute data

Aluminium Length/outside diameter/inside 1 400/176/140

tube diameter/mm
Copper tube Length/outside diameter/inside 12 600/140/92
diameter/mm
Insulator skirt Outside diameter/mm 708
Bushing Length/mm 14850
Rated voltage Transformer(AC+DC)/kV 844

TABLE 2 Electrical characteristics of the valve-side bushing of UHV

converter transformer

Relative Conductivity
Material inductivity (S/m)

Silicon rubber 2.8 5 × 10−16

Aluminium 2000 3.5 × 107
Copper 2000 5.7 × 107
Epoxy resin 4.5 1 × 10−13
Converter transformer 2.2 5 × 10−14
insulating oil
SF6 1.002 1 × 10−20

In Figure 2, the connecting terminal is used to connect exter-

FIGURE 2 Schematic structure of valve side bushing of converter
nal input signals. The conservator is used to adjust the volume
transformer change of insulating oil. The upper and lower porcelain bush-
ings are used to resist external discharge. The capacitor core is
used to improve the electric field distribution of the bushing.
3 STATE EVALUATION OF The flange is used to install the bushing, and a measuring termi-
VALVE-SIDE BUSHING OF UHV nal is installed on one side to achieve effective grounding. The
CONVERTER TRANSFORMER grading ring is used to eliminate the electric field imbalance.
To realize the mapping between physical entities and twins,
The valve-side bushing of the UHV converter transformer this paper takes the ±800kV valve-side bushing of an electric
belongs to the current-carrying connection component. As a power company as the experimental object to conduct phys-
weak link in power transmission, its working state is not only ical analysis. The physical parameters are mainly composed
affected by its high voltage, high temperature, mechanical vibra- of attribute parameters, electrical characteristics, and external
tion, material properties, and other factors but also affected by inputs.
the external environment, insulation oil corrosion, and other 1) Attribute parameters
factors. To realize the state evaluation of the valve-side bush- Basic attribute parameters of valve-side bushing of UHV
ing, this paper first analyses the basic attributes of the physical converter transformer are shown in Table 1
entity. Then, use COMSOL to establish digital twins. Finally, the 2) Electrical characteristics
zero-sample status evaluation process is analyzed. The capacitor core is made of epoxy resin, and the electrical
characteristics of other materials are shown in Table 2.
3) External inputs
3.1 Physical entity of valve-side bushing of The insulation structure of the bushing can be divided into
UHV converter transformer two parts: the side of the converter transformer oil and the side
of the valve hall. Therefore, the external environment includes
The main insulation of the valve-side bushing is oil impregnated both the ambient temperature of the valve hall and the insulating
paper capacitor core, the air side is a silicone rubber composite oil characteristics of the converter transformer. According to
insulator, and the transformer insulation oil side is a porcelain the provisions of IEC 60137 and GB/T 4109, in the experiment
insulator or composite insulator. Its overall structure is shown of thermal characteristics analysis, the temperature of the valve
in Figure 2. hall is 50◦ C±2◦ C, and the temperature of the oil is 90◦ C±2◦ C,
1128 LI ET AL.

FIGURE 3 Current waveform at both ends of the valve-side bushing of

the UHV converter transformer

and the theoretical difference between the two is not more than
60K±2K. The bushing excitation is also an important part of
external input. Figure 3 shows the current waveforms at both
ends of the bushing when a UHV power plant is running at FIGURE 4 Construction of digital twin for valve-side bushing of UHV
±800kV full load. converter transformer

3.2 Establishment of a digital twin for the

valve-side bushing UHV converter transformer where Rg (t ′ )is the harmonic resistance, whose calculation
method is shown in Equation (13):
To realize the establishment of a physical entity into a digital
twin, COMSOL software is used as the experimental simulation ⎧R (t ′ ) = 𝜌t ′ L
platform here. The model constraint problem is described as ⎪ g 𝜋 ▵ dg (D− ▵ dg )
follows: ⎪
⎨ √ (13)
First, the physical entity of valve-side bushing is analyzed, 𝜌t ′
⎪
including physical attribute parameters: geometric dimension, ⎪ ▵ dg = 𝜋 fg 𝜇
material attribute, electrical characteristics, etc., as well as ⎩
environmental parameters: insulation oil temperature, valve
chamber temperature, etc. where D is the bushing diameter; 𝜌t ′ the resistivity; L the
Then, there are many reasons for bushing failure, such as length of the conduit; 𝜇 the magnetic resistance; fg the current
contact finger corrosion of the strap, SF6 gas leakage, and so frequency.
on. Through the simulation tests, the temperature change of Based on the analysis of the above steps and combined
the failed bushing is the most obvious electrical connection with the physical experiment platform, the digital twin built by
device. Therefore, this paper defines the contact resistance of COMSOL is shown in Figure 4.
the electrical connector as the bushing degradation feature. The
boundary conditions are set according to the basic principles of
heat conduction, convective heat transfer, and thermal radiation 3.3 Zero-sample state evaluation process
in Section 2.2. based on attribute analysis
Finally, the rated current of the valve-side bushing is
AC+DC, when the spectrum analysis of the measured current The factors that affect the running state of the valve-side
waveform at both ends of the bushing in Figure 3 is carried out; bushing are diverse, which inevitably leads to problems such
the current includes other harmonic components in addition as diverse failure forms and weak correlation. Typical failures
to the 50 Hz basic component, so the skin-effect of harmonic include SF6 gas leakage, partial discharge, aging of the con-
current must be considered in the heat loss calculation [40]. nection parts, and contact corrosion of the watchband. In the
Assuming that the g-th harmonic current component is Ig , its face of different fault types, their fault mechanisms are also very
heat loss is calculated as shown in Equation (12): different, so it is difficult to extract fault feature data.
To solve the above problems, this paper proposes a zero-
sample state evaluation method for valve-side bushing based on
∑
G
P (t ) = Ig2 Rg (t ′ ) (12) digital twin technology and attributes analysis. The specific state
g=1 evaluation flow diagram is shown in Figure 5.
LI ET AL. 1129

FIGURE 6 EMD decomposition of bushing current at valve-side of

UHV converter transformer

The initial feature matrix is established by using the bushings

with different defect degrees, and the running state evaluation
of the valve-side bushing is completed by combining the fuzzy
clustering algorithm flow in Section 2.3.

4 EXAMPLE VERIFICATION
AND ANALYSIS
FIGURE 5 Evaluation process of valve-side bushing state of UHV 4.1 Construction of digital twins for the
converter transformer
valve-side bushing
It can be seen from Figure 5 that the zero-sample state evalu- 4.1.1 Current decomposition of the valve-side
ation of valve-side bushing based on attribute analysis is mainly bushing
composed of four parts: physical entity analysis, simulation anal-
ysis of virtual space, the establishment of state feature set based The carrier voltage and current of the valve-side bushing are
on data twins, and running state evaluation. related to the working mode of the converter transformer. To
1) Physical entity analysis establish the corresponding digital twins of the bushing and
Based on the analysis of physical characteristics (attribute extract more detailed features, it is necessary to decompose
parameters and electrical characteristics of ±800kV valve-side other detailed components except for the 50 Hz fundamental
bushing in Section 3.1), the physical entity analysis of valve-side component.
bushing is completed in combination with running environment Taking the dry-type SF6 gas insulated bushing of phase A
temperature (insulation oil temperature and hall temperature). at Y-Y side of ±800kV UHV as an example, its input signal
2) Simulation of virtual space waveform is shown in Figure 3. The initial current waveform is
First, COMSOL is used to set geometric, material prop- decomposed at different centre frequencies using EMD algo-
erties and physical fields; then, the measured carrier current rithm. The decomposed intrinsic mode function component
is analyzed in the frequency domain, and it is decomposed imf(t) and residual component r(t) are shown in Figure 6.
into decomposition components with different frequency cen- As shown in Figure 6, after the bushing current is pro-
tres by EMD; Finally, combined with the heat loss calculation cessed by the EMD algorithm, there are four intrinsic mode
model in Section 3.2, the digital twins of valve side bushing are componentsim f1 (t )∼im f4 (t )and one residual component r(t),
established. and r(t) meets the termination condition of EMD.
3) Establishment of state feature set based on data twin
Through the analysis of the digital twin, the bushing heat
distribution is taken as the research object, and the charac- 4.1.2 Simulation analysis of virtual space based
teristics data of the bushing external maximum temperature, on sample attributes
external minimum temperature, internal maximum tempera-
ture, and internal minimum temperature under different carrier When the valve-side bushing has fault defects, the most direct
current decomposition components are obtained. features are the changes in DC impedance and carrier cur-
4) Running status evaluation rent. Although the external thermal feature distribution of the
1130 LI ET AL.

FIGURE 7 Simulation results of axial heat distribution of valve-side

FIGURE 8 Simulation results of radial heat distribution of valve-side
bushing
bushing

bushing can be obtained through infrared testing, it is difficult to

obtain the internal features. Combined with the basic attributes,
DC impedance, and other features of the physical entity of the
valve-side bushing in Section 3.1, the EMD results of the car-
ried current are input into the virtual model, and the digital twin
technology is used to obtain the internal characteristics under
the simulation analysis.
The simulation results of UHV ±800kV dry-type SF6 gas-
insulated bushing in COMSOL are shown in Figure 7.
As shown in Figure 7, the temperature of the simulated
bushing at the insulating oil side of the transformer is about
73◦ C, and the highest temperature in the axial temperature
FIGURE 9 Comparison of axial temperature distribution between the
distribution of the bushing occurs at the connection of the
physical entity and digital twin of the valve-side bushing
aluminium–copper conduit. When the bushing fails, the most
obvious change is the temperature at the connection. Therefore,
to obtain the state features of the bushing, the section at the
connection is simulated and intercepted as shown in Figure 8.

4.1.3 Effectiveness analysis of thermal

distribution of twins

To verify the effectiveness of the digital twin simulation analysis

in the above section, the axial thermal distribution of physical
entities and virtual entities of UHV ±800kV valve-side dry-
type SF6 gas-insulated bushing is compared and analyzed. The
maximum temperature results are shown in Figure 9.
As shown in Figure 9, the maximum temperature range of the
bushing is 69.37–100.15◦ C, while the normal reference value is FIGURE 10 The capacitor core of the valve-side bushing is faulty
120◦ C, so the bushing is under normal operation. By comparing
LI ET AL. 1131

TABLE 3 State features of valve-side bushing under normal conditions

Decomposition of
component imf1 (t) imf2 (t) imf3 (t) imf4 (t) r(t)

Maximum temperature 75.8 75.8 75.8 75.8 75.8

of conduit (◦ C)
Minimum temperature 48.9 48.9 48.9 48.9 48.9
of conduit (◦ C))
Maximum external 35.1 35.1 35.1 35.1 35.1
temperature (◦ C)
Minimum external 20.8 20.8 20.8 20.8 20.8
temperature (◦ C)

the axial heat distribution of physical entities and digital twins,

the average error is not more than 5%, which meets the one-to- FIGURE 11 Distribution of grey correlation coefficient
one mapping relationship. Then, the internal characteristics of
the bushing can be obtained through the digital twin, which lays
the foundation for state data twinning. 4.3 Case study on evaluation of bushing
state at converter valve-side

4.2 Defect analysis of bushing and To realize the state evaluation of the valve-side bushing of the
establishment of state feature set UHV converter transformer, the state feature vectors of the
bushing under different defects are established according to the
Taking a ±500 kV power plant as an example, the picture of the method of establishing the feature vectors under normal con-
capacitor core when the valve-side bushing is faulty is shown in ditions in the above section. In the example demonstration,
Figure 10. x1 ∼x5 are used to represent the state feature vectors of the dif-
According to field detection, the blackening substance of ferent bushing. x1 is the bushing with the impedance of 52.2 μΩ
the capacitor core is the substance produced after the local under normal condition, and x2 ∼x5 are the bushing with defect
temperature rises due to the rise of the impedance at the impedance of 120.4, 180.6, 200.3, and 260.8 μΩ.
copper–aluminium connection, and then local flashover occurs.
Analyse other faults such as SF6 gas leakage, and the change
of impedance of the bushing conductive tube is the first. 4.3.1 Grey relation analysis
Therefore, in the face of the diversity of fault defect forms
of the bushing and the features of fewer fault data in oper- To further verify the effectiveness of the fuzzy clustering algo-
ation and maintenance management, the use of digital twins rithm in the state evaluation, this paper analyses the initial
to obtain twin data under different impedance changes is con- matrix X by citing the grey correlation coefficient, in which
ducive to achieving zero-sample state evaluation under online the calculation method of the grey correlation coefficient is as
monitoring. follows:
In the process of establishing the state feature set, first, the min min |x1 (k) − xi (k)| + 𝜌 max max |x1 (k) − xi (k)|
digital twin is established according to the decomposition result i k i k
𝜉i (k) = ,
of the carrier current and the impedance of the catheter in the |x1 (k) − xi (k)| + 𝜌 max max |x1 (k) − xi (k)|
i k
physical entity, and then the maximum temperature and mini-
mum temperature of the catheter are obtained. Then, the initial where 𝜉i (k)is the correlation coefficient of x0 to xi at point k,
state feature set is established. Finally, to reduce the dimension min min |x0 (k) − xi (k)|is the absolute value of the second level
of the feature set, principal component analysis is used to map i k
the initial state features into one-dimensional space. Based on minimum difference of x1 to xi at point k, ρ Is the grey reso-
the analysis of ±800kV valve-side dry-type SF6 gas-insulated lution coefficient, generally 0.5, and the final grey correlation is
bushing under normal conditions in Section 4.1, the state fea- ri :
tures under different decomposition components are shown in
1∑
n
Table 3. ri = 𝜉 (k).
To complete the dimension reduction analysis of the fea- n k=1 i
ture set, the principal component analysis of the feature set in
Table 3 under normal conditions is carried out, and the results Finally, the grey correlation coefficient of x1 in the initial
are as follows: characteristic matrix X is shown in Figure 11.
As shown in the Figure 11, the grey correlation coefficients of
xt = [184.37 6.19 0.51 0.51] x1 are very similar; the grey correlation degrees are 0.776, 0.795,
1132 LI ET AL.

As shown in Figure 12, when λ is 0.664, the bushing x1 and

other bushings are classified, so the similarity λ = 0.664 can
be used as a criterion for bushing evaluation. The higher the
similarity with normal conditions, the better the bushing status.
The lower the similarity with normal conditions, the lower the
bushing status.
In conclusion, using the attribute parameters of physical enti-
ties to establish a digital twin, the internal features of the data
twin can be obtained, and then the zero-sample state evaluation
of the valve-side bushing can be realized.

5 CONCLUSION

The valve-side bushing of the UHV converter transformer is

the key equipment to realize the safe transmission of electric
power. This paper conducts attribute analysis and research on
FIGURE 12 Cluster diagram of valve-side bushing statue evaluation the physical entity and uses the digital twin technology to obtain
the internal state feature set under the defect state. Through case
0.785, and 0.795, respectively. Moreover, the correlation degrees analysis and research, when the similarity of the fuzzy clustering
of x3 and x4 are the same, so effective bushing state evaluation algorithm is taken as the index, its zero-sample state evaluation
cannot be conducted. can be achieved, and the following conclusions are drawn:

1. In the physical entity analysis of the valve-side bushing, the

4.3.2 Fuzzy clustering analysis research includes both the basic physical characteristics and
the electrical characteristics. To decompose the different cen-
According to the fuzzy clustering algorithm shown in Algo- tre frequency components of the carrier current, the EMD
rithm 2 in Section 2.3, the zero-sample state evaluation process algorithm is used to decompose the 50 Hz fundamental
of the valve-side bushing is as follows. component and other harmonic components, which further
improves the accuracy of establishing the digital twin.
1) Initial feature matrix X. 2. In the process of establishing the digital twin of the valve-
side bushing, the mechanism of the bushing with different
defects is analyzed, its state characteristics are analyzed
X = [x1 ; x2 ; x3 ; x4 ; x5 ]
according to the thermal distribution phenomenon, and
the validity of the data twin is verified by using the axial
temperature distribution of the digital twin.
2) Standardize with Equations (5) and (6).
3. In the zero-sample state evaluation of valve-side bushing, the
initial state feature set is established by using the temperature
⎡ 0 0 0.214 0.505⎤ distribution at the copper aluminium joint and the exter-
⎢0.336 0.533 0 0 ⎥ nal temperature distribution under different decomposition
′′ ⎢ ⎥
X = ⎢0.556 0.721 0.622 0.774⎥ components of bushing current carrying, and the fuzzy clus-
⎢0.735 0.874 0.654 0.777⎥ tering algorithm is used to calculate the valve-side bushing
⎢ ⎥ under different defect degrees. Finally, the zero-sample state
⎣ 1 1 1 1 ⎦
evaluation is realized by similarity matching, which provides
3) Use Equation (7) to establish matrix R a guarantee for equipment maintenance and safe operation.

Herein, the preliminary assessment and analysis of the valve-

⎡ 1 0.663 0.444 0.267 0.004⎤
side bushing state have been realized. The following problems
⎢0.663 1 0.780 0.603 0.340⎥
⎢ ⎥ need to be solved in future research:
R = ⎢0.444 0.780 1 0.822 0.559⎥
⎢0.267 0.603 0.822 1 0.736⎥ 1. More problems such as stability and applicability of valve-
⎢ ⎥
⎣0.004 0.340 0.559 0.736 1 ⎦ side bushing state assessment under different application
environments need to be solved urgently.
4) As shown in Equations (8) and (9), the matrix is established 2. How to realize the residual life prediction under complex
after making matrix R transitive, and the cluster diagram formed running conditions is conducive to the establishment of a
when 𝜆 = min {ri j }is shown in Figure 12. modern health management system.
1≤i, j ≤n
LI ET AL. 1133

AUTHOR CONTRIBUTIONS 13. Jim, G.: Software specialist, rFpro develops digital modelling tool of
Zheng Li: writing - original draft. Kai Liu: data curation; for- Applus+ IDIADA proving grounds for real world trials of autonomous
vehicles. Truck & Bus Builder: The International Newsletter of Commer-
mal analysis; writing - review and editing. Mu Lin: investigation;
cial Vehicle Manufacturing, Developments. 40(7), 10–20 (2018)
validation. Dongli Xin: data curation. Hao Tang: resources. 14. Shi, Y., Xu, J., Du, W.: Discussion on the new operation management
Guangning Wu: methodology. mode of hydraulic engineering based on the digital twin technique. J. Phys.:
Confer. Seri., 1168, 022044 (2019)
ACKNOWLEDGEMENT 15. Tao, F., Zhang, M., Liu, Y., et al.: Digital twin driven prognostics and health
management for complex equipment. CIRP Anna. 67(1), 169–172 (2018)
This research was supported in part by the Science and
16. Jia, M., Shen, C., Chen, Y., et al.: Digital twin of the energy internet and its
Technology Project of Headquarter of SGCC: SGTYHT/19- application. Global Ener. Inter. 3(1), 1–13 (2020)
JS-215. 17. Zhou, M., Yan, J., Feng, D.: Digital twin framework and its application to
power grid online analysis. CSEE J. Power Ener. Syst. 5(3), 391–398 (2019)
CONFLICT OF INTEREST 18. He, W., Wu, Z., Ren, X., et al.: Research on the application of digital twin
technique in high voltage cable. In 2020 4th International Conference on
There is no conflict of interest.
Power and Energy Engineering (ICPEE), pp. 90–93. IEEE, Piscataway, NJ
(2022)
DATA AVAILABILITY STATEMENT 19. Elsisi M., Tran M., Mahmoud K., et al.: Effective IoT-based deep learn-
The data that support the findings of this study are available ing platform for online fault diagnosis of power transformers against
from the corresponding author upon reasonable request. cyberattacks and data uncertainties. Measurement 190, 110686 (2022)
20. Li, S., Cao, B., Cui, Y., et al.: Terahertz-based insulation delamina-
tion defect inspection of vehicle cable terminals. IEEE Transactions on
ORCID Transportation Electrification, 1–9 (2022)
Zheng Li https://orcid.org/0000-0001-7484-2794 21. Leong, Y., Ker, P., Jamaludin, M. M.: UV-vis spectroscopy: A new
Kai Liu https://orcid.org/0000-0002-1804-3418 approach for assessing the color index of transformer insulating oil.
Mu Lin https://orcid.org/0000-0001-7795-6906 Sensors. 18(7), 2175 (2018)
22. Du, Z., Nie, D., Zhang, L., et al.: Insulation state evaluation of high volt-
Guangning Wu https://orcid.org/0000-0002-5845-0392
age bushing based on frequency domain dielectric spectrum analysis. High
Voltage Appliances. 49(11), 6–11 (2013)
REFERENCES 23. Monga, S., Gorur, R. S., Hansen, P., et al.: Design optimization of high
1. Wang, L., Tao, Z., Zhu, L., et al.: Optimal dispatch of integrated energy sys- voltage bushing using electric field computations. IEEE Trans. on Diel. &
tem considering integrated demand response resource trading. IET Gener. Elect.l Insul. 13(6), 1217–1224 (2006)
Transm. Distrib. 16, 1727–1742 (2022) 24. Sarkar, S., Sharma, T., Baral, A., et al.: An expert system approach for
2. Chen, M., Liu, X., Shao, Y., et al.: Cracking risk analysis and control for transformer insulation diagnosis combining conventional diagnostic tests
high-voltage dry-type valve-side bushings. IET Gener. Transm. Distrib. 14, and PDC, RVM data. IEEE Transactions on Dielectrics and Electrical
6555–6561 (2020) Insulation. 21(2), 882–891 (2014)
3. Mikhak-Beyranvand, M., Faiz, J. Rezaeealam, B.: Thermal analysis and der- 25. Nie, D., Zhang, H., Chen, Z., et al.: Optimization design of grad-
ating of a power transformer with harmonic loads. IET Gener. Transm. ing ring and electrical field analysis of 800 kV UHVDC wall wall
Distrib. 14, 1233–1241 (2020) bushing. IEEE Trans. on Dielec. & Elec. Insulation. 20(4), 1361–1368
4. Maximov, S., Olivares-Galvan, J. C., Magdaleno-Adame, S., et al.: New (2013)
analytical formulas for electromagnetic field and eddy current losses in 26. Radakovic, Z., Cardillo, E., Schaefer, M., et al.: Design of the winding-
bushing regions of transformers. IEEE Trans. on Magn. 51(4), 1–10 (2015) bushing interconnections in large power transformers. Elect. Engin. 88(3),
5. Su, C., Branch, D.: On-site Testing requirements of ultra-high voltage 183–190 (2006)
converter transformer insulating Oil and SF_6 Gas. Elect. Engin. (07): 27. Záliš, K.: Using expert systems in evaluation of the state of high voltage
146–149 +160 (2016) machine insulation systems. Acta Polytechnica. 40(5), 185–190 (2000)
6. Zhang, S., Peng, Z., Peng, L., et al.: Design and dielectric characteristics 28. Liao, R., Du, Y., Yang, L., et al.: Quantitative diagnosis of moisture con-
of the ±1100 kV UHVDC wall bushing in china. IEEE Trans. on Diel. & tent in oil-paper condenser bushing insulation based on frequency domain
Elect. Insul. 22(1), 409–419 (2015) spectroscopy and polarisation and depolarisation current. IET Gener.
7. Chen, M., Liu, X., Liang, C., et al.: Study on surface charge Accumulation Trans. Distr. 11(6), 1420–1426 (2017)
characteristics of resin impregnated paper wall bushing core under positive 29. Elsisi M.: Improved grey wolf optimizer based on opposition and quasi
DC voltage. Energies. 12, 4420 (2019) learning approaches for optimization: Case study autonomous vehicle
8. Zhang, L., Shuai, X., Han, C., et al.: Valve side lead exit insulation design including vision system. Arti. Intel. Revi. 55, 5597–5620 (2022)
used in ±800 UHV converter transformer. In Proceedings of the Inter- 30. Ismail M. M., Bendary A. F., Elsisi M.: Optimal design of battery charge
national Conference on Condition Monitoring and Diagnosis (CMD), pp. management controller for hybrid system PV/wind cell with storage
452–455. IEEE, Piscataway, NJ 2016. battery. Int. J. Power Ener. Conv. 11(4), 412 (2020)
9. Liu, K., Yang, Z., Wei, W., et al.: Novel detection approach for thermal 31. Elsisi M., Tran M. Q.: Development of an iot architecture based on a
defects: study on its feasibility and application to vehicle cables. High deep neural network against cyber attacks for automated guided vehicles.
Voltage. 1–10 (2022) Sensors 21(24), 8467 (2021)
10. Shi, Y. U., Nie, D., Weimin, M. A., et al.: Overvoltage and insulation coor- 32. Tran M. Q., Elsisi M., Liu M. K., et al.: Reliable deep learning and iot-
dination of Jinping-Sunan ±800 kV UHVDC project converter stations. based monitoring system for secure computer numerical control machines
High Volt. Engin. 38(12), 3219–3223 (2012) against cyber-attacks with experimental verification. IEEE Access 10,
11. Liu, Z., Chen, C., Zhang, C., et al.: Data super-network fault prediction 23186–23197 (2022)
model and maintenance strategy for mechanical product based on digital 33. Elsisi M., Soliman M., Aboelela M., et al.: ABC based design of PID con-
twin. IEEE Access 7, 177284–177296 (2019) troller for two area load frequency control with nonlinearities. Telk. Indo.
12. Alarcón, Serrano, ngel, Madrid, et al.: The role of digital twins in J. Elec. Eng. 16(1), 58–64 (2015)
personalized sleep medicine. In German-Italian Workshop, Social Inno- 34. Elsisi M., Zaini H. G., Mahmoud K., et al.: Improvement of trajectory
vation in Long-Term Care through Digitalization. Springer, Cham tracking by robot manipulator based on a new co-operative optimization
(2022) algorithm. Mathematics 9(24), 3231 (2021)
1134 LI ET AL.

35. Li, S., Cao, B., Cui, Y., et al.: Nonintrusive inspection of moisture damp 41. Du, B., Tian, M., Su, J., et al.: Electrical tree growth characteristics in epoxy
in composited insulation structure based on terahertz technology. IEEE resin with harmonic superimposed DC voltage. IEEE Access 7, 47273–
Transactions on Instrumentation and Measurement. 70, 1–10 (2021) 47281 (2019)
36. Yu, Z., Chen, H., You, J., et al.: Adaptive fuzzy consensus clustering frame- 42. Li, C., Cerrada, M., Cabrera, D., et al.: A comparison of fuzzy clustering
work for clustering analysis of cancer data. IEEE/ACM Trans. Comp. Biol. algorithms for bearing fault diagnosis. J. Intell. Fuzzy Syst. 34(6), 3565–
Bioin. 12(4), 887–901 (2015) 3580 (2018)
37. Grieves, M. W., Virtually Perfect: Driving Innovative and Lean Products
through Product Lifecycle Management. Space Coast Press, Cocoa Beach
(2011)
38. Grieves, M. W., Vickers, M.: Digital twin :mitigatingunpredictable, undesir-
How to cite this article: Li, Z., Liu, K., Lin, M., Xin,
able emergent behavior in complexsystems. Transdisciplinary Perspectives
on Complex Systems, pp. 85–113. Cham, Springer (2016) D., Tang, H., Wu, G.: A zero-sample state evaluation
39. Maximov, S., Escarela-Perez, R., Magdaleno-Adame, S., et al.: Calculation model for valve-side bushing of UHV converter
of nonlinear electromagnetic fields in the steel wall vicinity of transformer transformer oriented to digital twin under attribute
bushings. IEEE Trans. on Magn. 51(6), 1–11 (2015) analysis. IET Gener. Transm. Distrib. 17, 1123–1134
40. Wen, M., Lin, X., Zhong, S.: Numerical simulation and experimental study
(2023). https://doi.org/10.1049/gtd2.12721
of heat transfer characteristics in transformer bushing, High Volt. Eng.
42(9), 2956–2962 (2016)