Received: 13 September 2022

DOI: 10.1049/hve2.12333

ORIGINAL RESEARCH
- -
Revised: 11 December 2022 Accepted: 13 January 2023

- High Voltage

Dynamic thermal rating assessment of oil‐immersed power

transformers for multiple operating conditions

Chen Zhang | Xuzhu Dong | Jiangjun Ruan | Yongqing Deng

School of Electrical Engineering and Automation, Abstract

Wuhan University, Wuhan, Hubei, China
In this article, the dynamic thermal rating assessment method of oil‐immersed power
transformer with multiple operating conditions is proposed, considering the constraints
Correspondence
Xuzhu Dong and Jiangjun Ruan, School of Electrical
of hot spot temperature (HST), top oil temperature, losses of life and the maximum
Engineering and Automation, Wuhan University, allowable current of on‐load tap changer (OLTC) or bushing, which can determine the
Wuhan, Hubei, China. dynamic load curves under different operating conditions and give the most sensitive
Email: dongxz@[Link] and ruan308@[Link]
constraints to limit the dynamic load capacity. To improve the accuracy of HST esti-
mation, the temperature estimation model is also improved and the thermal parameters
Associate Editor: Behzad Kordi
are optimised using the HST measured by optical fibre. Finally, several application ex-
Funding information amples are studied for transformers in different scenarios. The results show that the
National Natural Science Foundation of China ‐ normal cyclic dynamic transformer rating (DTR) is mainly limited by the losses of life
State Grid Corporation Joint Fund for Smart Grid, when the ambient temperature is high, and the average load factor can be increased to
Grant/Award Number: U2066217; Key Research
0.82 with a maximum load capacity of 1.23. The main limiting factor of short‐term DTR
and Development Program of Hubei Province,
Grant/Award Number: 2021BAA182 is the OLTC or bushing current constraint, and the average load factor can be increased
to 1.00. The maximum load capacity of the transformer under both operating conditions
is 23% and 50% higher than its rated load capacity, indicating that the transformer still
has a large load potential available.

1 | INTRODUCTION systems and a real and urgent major need for asset manage-
ment of new power equipment [2]. Oil‐immersed power
In recent years, the ‘double peak’ characteristic of power sys- transformers are crucial equipment in the power grid, and their
tem operation has become more and more apparent [1]. From stable operation is essential to maintain the safety and reliability
2016 to 2019, the maximum duration of a single 5% peak load of the power system. The dynamic transformer rating (DTR)
in the five southern provinces of China is 3–6 h, with a fre- assessment can improve the utilisation rate and provide a basis
quency of 10–40 times a year. A single 3% peak load duration for scientific operation and maintenance strategies for power
is up to 2–6 h and the annual frequency is 6–25 times. A typical supply companies to deal with the relationship between safety,
scenario is to meet the peak load demand, which will only insulation life, and cycle cost. Meanwhile, the whole process
occur tens of hours in the future within a year. It is also and lean management of transformer assets may be realised [3].
required to invest in the corresponding generation, trans- DTR is defined as the maximum load that a transformer
mission, substation, and set a proportional reserve, resulting in may acceptably sustain under varying loads and ambient tem-
wasted investments. peratures. In recent years, most of the research on DTR has
Dynamic thermal rating technology of power equipment is focused on the overload capacity assessment of the trans-
an effective way to improve equipment utilisation. In the case former. Hot spot temperature (HST) is a vital parameter in
of mitigating system investment and ensuring safety, carrying characterising the internal thermal state of the transformer.
out a dynamic thermal rating assessment of power equipment Some literature carried out studies on the dynamic thermal
can enhance the flexibility of system operation, which becomes ratings of transformers from an improved thermal model.
a major issue of concern for the operation of new power Callender et al. took oil‐immersed power transformers as the

-
This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial‐NoDerivs License, which permits use and distribution in any medium, provided the
original work is properly cited, the use is non‐commercial and no modifications or adaptations are made.
© 2023 The Authors. High Voltage published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology and China Electric Power Research Institute.

High Voltage. 2024;9:195–205. [Link]/journal/hve2 195

 23977264, 2024, 1, Downloaded from [Link] by South Korea National Provision, Wiley Online Library on [23/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
196
- ZHANG ET AL.

research object and improved the thermal model by consid- the TOT is 105°C. The two different temperature limits may
ering solar radiation. The results show that a temperature rise appear the TOT produces a smaller margin and becomes a
of 6 K may occur in transformers with high solar radiation limiting factor. The results of the statistical research on the
[4, 5]. Wang et al. corrected H in the transformer HST overload ratio of 132 500 kV transformers show that 9.09%
calculation model using a degree of polymerisation data from of the medium‐voltage side bushings fail to satisfy the
35 scrapped transformer samples and further evaluated 106 requirement of 1.5 times overload [22]. The limiting factor
transformer load capacities. The average load factor could be may be the bushing and on‐load tap changer (OLTC)
increased by 40% with no significant life reduction [6]. In allowable overload current, and these characteristics must be
addition, HST can be combined with real‐time acquisition of fully considered.
transformer state quantities, such as fibre optic real‐time HST In contrast to peak load‐based designs, this paper presents
measurement [7, 8], intelligent algorithm prediction [9], and a comprehensive approach to ensure that the operation is
non‐invasive temperature inversion [10]. Therefore, based on reliable at maximum utilisation without affecting the expected
real‐time data, an online DTR assessment can be carried out. life of the transformer. To fulfil the real‐time thermal capacity
The studies above focused on the estimation of HST and used estimation for power transformers, methods are proposed and
the HST estimates to determine whether the HST exceeded the validated for the refinement of thermal parameters for indi-
limits, ignoring the effects of insulation changes, operating vidual transformers, and HST calculated with refined thermal
conditions, and other factors on load capacity. parameters are compared with measured values under the load
While considering the dynamic thermal rating assessment profile of the heat run test. By accurately considering the
strategy, some literature has evaluated the DTR from different thermal parameters of the transformer, it is possible to allow
ambient temperatures, HST and top oil temperature (TOT) the transient operation of overloads beyond the rating speci-
and ambient temperature data analysis to quantify the DTR's fied in the nameplate. Then, the objective function for DTR
limiting factor [11]. Dong M. proposed a data‐driven DTR assessment is established on this basis, with parameters such as
assessment method based on the ambient temperature and the HST, TOT, loss of life (LOL), bushing and OLTC
load data of power transformers for the past several years, with allowable current overload as constraints. The most ‘realistic’
a daily relative ageing rate of 1 as a constraint, to construct a load data, that is, the typical daily load curve, is used as a basis
future load shape profiles of power transformers, which were for global optimisation. When the preload is the latest
validated by a set of historical load data. The study results show recording in the algorithm, the DTR analyses can be per-
that the maximum error in daily dynamic rating assessment is formed in real‐time or used as an offline planning tool. The
6.6% in winter and 5.7% in summer [12, 13]. It has been output information produced by the thermal algorithm in-
suggested that the transformer load and the ambient temper- cludes the curves of HST, TOT, LOL, load factor, and the
ature were uncertain, and methods were proposed to assess the ultimate load limiting factors by considering the influence of
DTR from the perspective of probabilistic prediction, where operation conditions. DTR is used only as an offline planning
the probability distributions of load and ambient temperature tool when the preload is the 24‐h profile of load and ambient
are first predicted. On this basis, the load and ambient tem- temperature.
perature curves were predicted. By setting the corresponding
constraints, the future load is solved directly or a Monte Carlo
simulation method is used to obtain the future dynamic ther- 2 | DTR ASSESSMENT METHOD
mal rating [14, 15]. Some studies have also quantified trans-
former constraints as the operational risk, which assessed the To evaluate the adaptability of the DTR under different
dynamic thermal rating from the perspectives of risk proba- operating conditions, both long‐term and short‐term effects
bility [16], failure probability [17, 18], and demand side should be considered separately. Increasing HST will reduce
response [19, 20] to grasp the transformer overload risk from a the expected life of power transformer. When the HST rises, it
macroscopic perspective. may generate bubbles and cause insulation failures [23].
From the above study, it can be seen that the HST is a Therefore, the HST, TOT, auxiliary equipment current, insu-
vital parameter in characterising the internal thermal state of lation lifetime and other limitations should be fully considered
the transformer. The accurate acquisition of the HST is in different scenarios.
crucial and the reference values are considered conservative This paper proposes different assessment methods for
regardless of individual differences [21]. Meanwhile, the different DTR application scenarios. Based on the dynamic
research on the DTR is based on temperature rise charac- thermal modelling of the power transformer, a DTR assess-
teristics, taking no account of the effect of insulation ment tool is developed for various load types and scenarios.
changes on load capacity. These are not the only factors that According to the time scale, the typical scenarios were divided
limit the DTR. By accurately considering the thermal pa- into three types, cyclic DTR assessment, short‐term DTR
rameters of the transformer, it is possible to allow the assessment, and short‐time emergency DTR assessment. This
transient operation of overloads beyond the rating specified paper adopts 105 and 140°C as the TOT limit and HST limit
in the nameplate. Considering the temperature difference of transformers [24] respectively. Using the insulation ageing
between the HST and the transformer oil temperature, the rate at the transformer HST of 110°C as the reference value,
limit value of the TOT is used. Typically, the limit value of the insulation relative ageing rate V and LOL are calculated as:
 23977264, 2024, 1, Downloaded from [Link] by South Korea National Provision, Wiley Online Library on [23/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ZHANG ET AL.
- 197

� �
15000 15000
−
V ¼ e 383 θhst þ273 ð1Þ
Z t2 N
X
Ltot ¼ V dt ≈ Vn � tn ð2Þ
t1 n¼1

where θhst is the HST of the oil‐immersed transformer, Ltot is

the insulation life loss in a period from t1 to t2, N is the
number of time intervals in that period, and Vn and tn are the
relative ageing rate and time of the insulation in the time in-
terval respectively.
The literature [2] gives the structure of the DTR assess-
ment. The HST, TOT, and LOL are functions of load factor K,
ambient temperature θamb, and operating time t. The optimi-
sation model is shown in Formula (3). The input of load factor
K is divided into two cases: the first assumes a constant load
factor K and the second assumes a dynamic change. This paper
uses the most ‘realistic’ load data, a typical daily load curve, as a
basis for global optimisation.

max K F I G U R E 1 DTR assessment calculation process: (a) Multiple of load

8 scaling, (b) TOT iteration process, (c) HST iteration process, (d) LOL
>
< θhst ðK; t; θamb Þ ≤ θhlim ; K ∈ ½0; K lim �; t ∈ ½0; T �
>
ð3Þ iteration process. DTR, dynamic transformer rating; HST, hot spot
s:t: θtop ðK; t; θamb Þ ≤ θtlim temperature; LOL, loss of life; TOT, top oil temperature.
>
: Ltot ðK; t; θamb Þ ≤ Llim
>

� 24‐h load profile and ambient temperature profile.

where θtop is the TOT of the oil‐immersed transformer, K is � Transformer thermal parameter data.
load factor, θhlim, θtlim, Llim, Klim is the limit of HST, TOT, � Limit values of each constraint.
LOL, bushing, and OLTC current.
In the iterative process, the preload is a 24 h profile, and the The specific steps are as follows:
ultimate goal is to determine the maximum value. A propor-
tional increase rather than a linear increase is assumed. If a Step 1: Input primary data mentioned above;
linear factor is used, the off‐peak load will increase by the same Step 2: Determine the time interval for dynamic assess-
proportion as the peak load. This would result in a higher base ment according to the transformer operating conditions;
load. Therefore, a multiple of load scaling is used as shown in Step 3: Set the cyclic factor C;
Figure 1a. At the same time, calculate the corresponding Step 4: The set of load factors multiplied by the cyclic
constraint values and determine the maximum value as shown factor C;
in Figure 1b–d; the load as a group needs to be scaled up or Step 5: Use the corresponding estimation formula to
scaled down. When the given profile does not exceed any determine the initial TOT rise and HST rise;
specified limit or exceeds an arbitrary limit. The load must be Step 6: Use the corresponding estimation formula to es-
increased or decreased proportionally until the arbitrary limit is timate the HST, TOT, LOL;
reached. Step 7: Evaluate the maximum value of HST, LOL, TOT,
The proposed method has the following functions: K;

� Determine the initial TOT rise ∆θoi and initial HST to TOT (1) If any of the values exceed the constraint, decrease C by
rise ∆θhi. 0.01 and then move to step 4 to continue the process;
� Determine the maximum dynamic load profile. (2) If none of the constraints is exceeded, increase C by 0.01
� Determine the factor limiting the load. and go to step 4 to continue the process;
(3) If any constraints are approximately equal to the corre-
Figure 2 shows the flow chart of the DTR assessment sponding limit, the load factor data are multiplied by the
method. The evaluation method focuses on 1 day, the time step current cyclic loading coefficient C. The process is com-
in cyclic loading, and the short‐term conditions are 0.5 h. In a plete and the outputs include curves of θhst, θtop, Ltot, K,
short‐time operation, the time step is 1 min. To achieve the and each maximum value, as well as the ultimate limiting
above function, the following data needs to be prepared: factors.
 23977264, 2024, 1, Downloaded from [Link] by South Korea National Provision, Wiley Online Library on [23/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
198
- ZHANG ET AL.

Step 2: Calculate the HST and TOT rise for each time step
as stated above, and repeat the iteration in steps after the
last time of each day;
Step 3: Compare the given initial value to the temperature
value obtained in step 2. If they are equal, the iteration
ends, and if they are not equal, the value obtained in step 2
is used as the initial value and proceeds to step 1.

3 | DETERMINATION OF HST

3.1 | Thermal parameter classification and

determination method

The dynamic thermal model in GB/T1094.7‐2008 [25] China,

IEEE C57.91 [26], and IEC 60076‐7: 2005 [27] requires the
input of several thermal parameters, and once the thermal
circuit topology is determined, many thermal parameters need
to be determined subsequently. Transformer thermal parame-
ters can be classified into two categories: first, dimensionless
parameters related to computational fluid dynamics and flow
heat transfer analysis, which are mainly used for steady‐state
thermal modelling; second, thermal characterisation parame-
ters related to temperature rise tests and nameplate informa-
tion, which are primarily used for dynamic thermal modelling
[28]. These parameters can be classified into three categories:

(1) Ambient temperature θamb, load losses at rated current

Pload, rated, no‐load losses at rated voltage Pno‐load, average
winding to average oil temperature gradient at rated cur-
rent gr, and top‐oil temperature rise in steady‐state at rated
losses ∆θor.
(2) Winding time constant τw and oil time constant τoil.
(3) Oil exponent x, winding exponent y, and hot spot factor
H.

The parameters in (1) can be obtained directly from the

factory's regular heat run test. The parameters in (2) can be
selected or estimated according to the guidelines above if the
HST and the TOT can be continuously measured during the
heat run test, so these two time constants can be obtained
more accurately. Since the parameters in (3) reflect the pa-
rameters affected by the transformer structure and cooling
FIGURE 2 Flow chart of dynamic thermal rating assessment method. capacity, the guidelines and standards give recommended
values of the parameters for different cooling methods.
Due to variations in thermal design, the thermal perfor-
For different operating conditions, the corresponding load mance of different transformers may vary greatly. If the pa-
and ambient temperature data should be converted to a reso- rameters in (3) are to be obtained accurately, an extended heat
lution corresponding to the time step. The initial TOT and run test is required. Therefore, when calculating the HST of a
HST rise can use the actual measured data. Meanwhile, there transformer, the measured data should always be used to
will be a situation that the above initial value is unknown. This obtain the thermal model parameters. The method for deter-
case can be obtained using the HST estimation formula for mining the thermal parameters for different amounts of data is
iterative iterations. The specific steps are as follows: given in Figure 3. When the data is insufficient, the calculation
can be considered using the factory test data. Suppose the data
Step 1: Randomly give an initial value of TOT rise and and the HST cannot be measured. In that case, the recom-
HST rise; mended values given in the guidelines can be directly selected.
 23977264, 2024, 1, Downloaded from [Link] by South Korea National Provision, Wiley Online Library on [23/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ZHANG ET AL.
- 199

TABLE 1 Basic parameters of the transformer.

Parameters Value
Model SZ11‐20000/66

Rated capacity (kVA) 20,000

Rated voltage (kV) (66 � 8 � 1.25%)/10.5

Link group YNd11

Cooling method ONAN

Frequency (Hz) 50

No‐load loss (kW) 14.3

Load loss (kW) 98.9

Number of high voltage (HV) winding layers 68

Number of low voltage (LV) winding layers 78

Number of phases 3

F I G U R E 3 Method for estimating dynamic thermal modelling

parameters of transformers.

3.2 | Transformer thermal parameter

analysis and correction
A more accurate dynamic thermal model is aimed at the mul-
tiple overload period. The improved dynamic thermal model-
ling procedure with optimised thermal parameters is expected
to promise HST accurate estimates. In this paper, the trans-
former is equipped with a fibre optic temperature measurement
system. The specific design parameters are shown in Table 1.
The location of the HST was determined in advance
through multi‐physics field simulation, and the HST is in the
B‐phase low‐voltage winding. The HST measurement method
is quasi‐distributed fibre optic measurement. In the practical
design, fibre optic temperature measurement sensors are
installed in the low‐voltage winding of A, B, and C phase. The
installation instructions are shown in Figure 4. The TOT is
FIGURE 4 Fibre optic temperature measurement probe installation
obtained by thermocouple measurement, and the ambient
schematic.
temperature θamb of the transformer operation is measured by
the temperature sensors at 0.5 m around the transformer. Both
the TOT and the ambient temperature are taken from the The measured data is shown in Figure 5, and the methods
average value. R can be determined by the ratio of load loss to proposed above are applied to estimate the thermal model
no‐load loss. The load was divided into four stages. The HST parameters of the transformer. The comparison between the
tends to be stable for the first three stages, and the load factor measured values and calculated results with IEC 60076‐7
K for the tests stage is known. recommended parameters is shown in Figure 5. It can be seen
 23977264, 2024, 1, Downloaded from [Link] by South Korea National Provision, Wiley Online Library on [23/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
200
- ZHANG ET AL.

Step 4: Calculate the Jacobi matrix J(αk) with the following

expressions:
� �T
∂Eðαk Þ ∂Eðαk Þ ∂Eðαk Þ ∂Eðαk Þ ∂Eðαk Þ
Jðαk Þ ¼ ;
∂k11 ∂k21 ∂k22 ∂τo ∂τw

Sep 5: If E(αk) < ε, then αk can satisfy the identification

requirement and end of the iteration; otherwise, αk+1 is
used as the initial value to calculate the error function
value;
Step 6: Calculate the parameter vector αk+1 at time k +1
using the iterative formula of the L‐M algorithm with
the following expression: αk+1 = αk–(r2E(α)+λI)−1J(αk),
F I G U R E 5 Comparison of measured values of multi‐stage
where: r2E(α) is the Hessian matrix, λ is the damping
temperature rise test and empirical formula results. factor, I is the unit matrix, k is the number of iterations,
and J(αk) is the Jacobi matrix of E(α) to the parameter
vector αk;
Step 7: If E(αk+1)< E(αk), then set k = k+1, λ = λ/β and
that the calculated results of the recommended parameters return to (3); otherwise do without updating the control
have a large error from the measured results. The maximum volume and set αk+1 = αk and λ = λ�β;
error is more than 10°C. The main reason is that the recom- Step 8: At the end of the iteration, the values of the
mended parameters in the guidelines are reference values given optimal thermal parameter vector (τw, τoil, y, k11, k22, H,
by integrating a large number of tests. gr) are obtained;
In this case, the thermal parameters of individual trans- Step 9: Back to step 1, optimal thermal parameter vector
formers should be further determined to refine the dynamic (k11, x).
thermal performance under various loads. A Levenberg‐
Marquardt (L‐M) method is used here to refine thermal pa- Re‐estimation of HST using the optimised thermal param-
rameters by measuring HST and TOT curves during the eters. The optimised thermal parameters are shown in Table 2.
extended heat run test. The L‐M method is one of the most The final results are shown in Figure 6a. As can be seen from
effective methods for estimating the parameters of non‐linear Figure 6a, in the transformer cooling process, if the load change
models. It iterates by searching for the minimum of multiple is small (corresponding to 13 h), the HST calculated error is
variable functions. The method effectively combines the ad- minimal. However, when the load fluctuation is large, the HST
vantages of the steepest descent method and the Gaussian calculation has a large error because the recommended formula
Newton method. only considers the HST to the TOT under the steady‐state,
The process is divided into two steps. The first is to use the ignoring the dynamic changes of the initial HST rise.
HST to TOT gradient curve to estimate τw, τoil, y, k11, k22, H
and gr. The second step is to evaluate the remaining parameters � � �x �
x and k11 with the measurement data of the TOT. For the L‐M 1 þ RK 2
θhst ¼ θamb þ Δθoi − Δθor � � f 3 ðtÞ þ
iteration process is as follows: 1þR
� �x
1 þ RK 2 � �
Step 1: Suppose the expression of the HST, time and Δθor � þ Hgr K y þ Δθhi − Hgr K y � f 4 ðtÞ
characteristic parameters during the test is θh (t) = f (t, α), 1þR
parameter vector α = (τw, τoil, y, k11, k22, H, gr), Use the ð4Þ
measured transformer HST data as input (ti, θi), i = 1, 2 ,
…, n; where f 4 ðtÞ ¼ e−t=τw .
Step 2: Set the initial vector value α0, the coefficient β, the Based on this, the Formula (5) in IEC 60076‐7 [27]is
damping factor λ and the permissible minimal error value improved to Formula (4) in this paper. Using Formula (4) and
ε, according to the recommended values of transformer the refined parameters for the fourth stage of the cooling
characteristics parameters; transient process for HST calculation, the comparison results
Step 3: Calculate the residual sum of squares E(αk) with are shown in Figure 6b. It can be found that the calculated
Pn
the specific expression: Eðαk Þ ¼ ½θi − f ðti ; αÞ� ;
2 results of the cooling process correspond to the measured
i¼1 results, and the maximum error is within 3 K.
 23977264, 2024, 1, Downloaded from [Link] by South Korea National Provision, Wiley Online Library on [23/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ZHANG ET AL.
- 201

It can be seen that the error values of the HST calculated parameters and the thermal parameters calculated from the
using the standard recommended thermal parameters are large. conventional thermal operation test results.
In contrast, the accuracy of the refined thermal parameters
used in this paper is better than that of the IEC recommended
4 | ANALYSIS OF WORK CONDITIONS

TABLE 2 Comparison of the results of different methods.

Condition 1 Cyclic DTR assessment.
Recommended Refined Calculated
Parameters parameters parameters parameters
Principle 1 Based on the time scale, Cyclic DTR assessment is
k11 0.5 0.5 ‐ usually divided into long‐term, and long time. The cycle
k22 2 1.1 ‐ duration is usually 24 h, and the long‐term general assist with
reliable asset investment planning. To avoid overload failure,
k21 2 2 ‐
LOL value of a cycle of the transformer should not be greater
x 0.8 0.75 ‐ than 24 h.
y 1.3 1.15 ‐
Condition 2 Short‐term DTR assessment.
τoil 210 248 196

τw 10 10 6 Principle 2 The TOT, HST, and current limitations of auxil-

H 1.3 1.3 ‐ iary equipment satisfy the constraints and allows the trans-
former relative ageing rate to be greater than 1. Which will
sacrificing part of the transformer insulation life to improve
the transformer's load capacity in the short term.

Condition 3 Short‐time emergency DTR assessment.

Principle 3 Such condition is rare and short‐lived, with tran-

sient changes in HST and TOT. To ensure the power supply,
the assessment can ignore the relative LOL limit and auxiliary
equipment current constraints.

The transformer studied in the previous section was used as

the object. The 24 h load curve and the ambient temperature in
Figure 7 are inputs. The ambient temperature of the day most
similar to the typical load curve is selected as the corresponding
ambient temperature respectively. The maximum daily load
factor is 1 p.u.. The minimum load factor is 0.4 p.u.. The average

F I G U R E 6 Result comparison: (a) Comparison of measured values of

multi‐stage temperature rise test with the method proposed and
(b) Comparison of the results of different calculation methods. FIGURE 7 Typical daily load and ambient temperature profiles.
 23977264, 2024, 1, Downloaded from [Link] by South Korea National Provision, Wiley Online Library on [23/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
202
- ZHANG ET AL.

load factor is 0.67 p.u.. The ambient temperature fluctuates limit of LOL. This advantage is most apparent when the
from 28 to 40°C in summer and from −3 to 9°C in winter. auxiliary equipment allowable overload current satisfies the
conditions. The maximum load factor in winter and summer
can be increased by 53% and 31%. Observation indicates that
4.1 | Cyclic DTR assessment HST and LOL are part of the factors limiting the dynamic
thermal ratings of the transformer. For the case study in
Transformer cyclic operation requires attention to the trans- Figure 9a for a maximum loading rate of 1.28 p.u. in summer,
former's internal temperature rise and insulation life loss. The it can be seen that the TOT reaches the limit while the HST
constraints of the DTR assessment are not obvious. To further still has a margin. In the comparative analysis, Figure 9a,b
analyse the transformer load capacity constraints, Figure 8 shows that when satisfying the LOL, TOT, and HST, the
gives the comparison of each parameter when the θtlim is allowed current increase rate is significantly slower than the
105°C, Llim is 24 h, Kl is 1.5 p.u. and θhlim is limited to 100, 110, increase rate of the three. For summer, the HST reaches
120, 130, 135, and 140°C, respectively, Under different seasons, the limit when the maximum load is 1.23 p.u. For winter, it
it can be seen that when the limit of HST does not exceed still has not reached its limit. When the bushing or tap
130°C, θtop and Ltot have not yet reached the constraint, θh changer does not meet the 1.2 times overload multiplier and
always the constraint limit. Although the constraints are the only the HST constraint is considered, an excessive load
same, the maximum load factor is 1.22 p.u. in summer and 1.47
p.u. in winter. After this, the maximum load limiting factor in
summer becomes LOL and the maximum load limiting factor
in winter becomes auxiliary equipment currently. It can be seen
from the LOL curves that the insulation life loss value shows
exponential form growth with the increase of HST, so the
transformer should be prevented from accelerated ageing than
expected when the transformer is operated under cyclic load.
A comparison of the maximum load allowed and the
limitation of the load of the transformer in winter and summer
is given in Figure 9. When the θtlim is 105°C, Llim is 24 h, Kl is
1.5 p.u. and θhlim is limited to 140°C, The maximum load factor
is 1.23 p.u. in summer and 1.50 p.u. in winter, which is 23%
and 50% higher than its rated capacity, respectively, indicating
that the transformer still has more significant load potential.
The LOL limit utilises the fact that the transformer operates at
a low load factor during regular operation. Thus, excessive
ageing during peak periods is compensated to some extent by
less LOL during off‐peak periods.
It is normally a safer option to use a LOL as a constraint.
However, in peak load scenarios, it is certain to exceed the

F I G U R E 8 Comparison of temperatures and LOL under different F I G U R E 9 Comparison of assessment results with various constraints:
HST limits. HST, hot spot temperature; LOL, loss of life. (a) Summer and (b) Winter.
 23977264, 2024, 1, Downloaded from [Link] by South Korea National Provision, Wiley Online Library on [23/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ZHANG ET AL.
- 203

assessment result is obtained, particularly in winter with a the load capacity. The main limiting factor of load capacity is
margin of 19.46%. the auxiliary equipment.
It can also be seen from Figure 8 that the lower the The above analysis can be summarised that the auxiliary
ambient temperature, the higher the maximum load capacity of equipment current is most likely to be the main factor in the
the transformer when θtop, Ltot, and θh are not exceeding the dynamic load capacity of the transformer. For further
limits. For instance, the maximum loading rate in winter is research, Figure 12 gives the results of the DTR evaluation
14.67% and 14.38% higher than in summer when the TOT or for six different periods (in the same set of curves, the upper
HST exceeds the limit. Therefore, another important obser- curve is always the winter load curve). From the curve, it
vation indicates that the allowable overload current of the can be seen that the maximum load factor of transformer
bushing and OLTC is more likely to be the major factor operation is always 1.5 p.u.. The limiting factors are all
limiting the dynamic load capacity of the transformer, which is auxiliary equipment current, which is similar to the results of
particularly apparent at lower ambient temperature. the cyclic DTR assessment above.

4.2 | Short‐term DTR assessment

When a transformer carries short‐term loads, the duration is

usually a few hours. Suppose that there are two transformers
T1 and T2 in the region where T1 needs to be taken out of
service for maintenance due to abnormal online monitoring
data, scheduled to last 6 h from 7:00–13:00. The dynamic load
capacity of T2 is evaluated during this period so that the system
can ensure the power supply as much as possible under the
premise of safe and stable operation.
Figures 10 and 11 show the results of the short‐term DTR
assessment. Carry out the analysis with summer conditions.
From 7:00 h to 13:00 h, the HST rises rapidly at the beginning,
and the total insulation life loss is 39.28 h, which is 36.97 h
higher than the normal cyclic load operation of 2.31 h. The
maximum load factor was 1.60 times higher than before,
reaching 1.23 p.u., 23% higher than the rated load capacity.
When the TOT is 105°C and reaches the limit, while the HST is
133.32°C, the main limiting load capacity factor is the TOT.
F I G U R E 1 1 HST and TOT changes in short‐term DTR assessment.
When the ambient temperature is lower in the winter, the DTR, dynamic transformer rating; HST, hot spot temperature; TOT, top oil
maximum load factor reaches 1.5 p.u.. The total insulation life temperature.
loss is 52.43 h, sacrificing more insulation life to increase

FIGURE 10 Short‐term dynamic transformer rating assessment

results. F I G U R E 1 2 Comparison of dynamic thermal rating assessment
results under different time periods.
 23977264, 2024, 1, Downloaded from [Link] by South Korea National Provision, Wiley Online Library on [23/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
204
- ZHANG ET AL.

Comparing the load curves in winter and summer, it can be Figure 14. It can be seen that the transformer's short‐time
seen that the lower the ambient temperature, the higher load emergency sustainable time decreases with the increase of
capacity the transformer can carry, the lower the average load the initial load factor and TOT. The TOT of 70°C usually
rating, and the higher the load capacity when the load factor is increases by 30 min compared with the TOT of 85°C. When
closer to the rated value, its maximum load capacity will be the load factor rises from 1.3 p.u. to 1.4 p.u., the recom-
reduced, corresponding to 2–4 h, 10–13 h; If the peak‐to‐valley mended operating time of the transformer short‐term emer-
difference in the load curve becomes more extensive, the lower gency load plummets from 150 min to 75 min. At an initial
load capacity the transformer can carry, corresponding to 2– TOT of 85°C and a starting load factor of less than 1.6 p.u.,
4 h, 2–7 h, 2–9 h or 10–13 h, 10–17 h. The specific influence the transformer can be operated continuously for more than
law will be given in the following paper. 15 min.
The above analysis shows that if the transformer needs to
be maintained or short‐term dynamic thermal rating, priority
can be given to periods with low load factor or large peak‐to‐ 5 | CONCLUSION
valley differences.
This paper proposes an iterative method for DTR assessment
of oil‐immersed power transformers. The strategy is imple-
4.3 | Short‐time emergency DTR mented by a novel approach that does not rely on a single
assessment factor to determine the dynamic load curves of the trans-
former. It matches the loading guidelines and limits for the
Short‐time emergency load is typically aimed at working con- safe operation of transformers under different operating
ditions when the transformer encounters a sudden load surge conditions, including HST, TOT, LOL, auxiliary equipment
with a short duration. The transformer can sustain the peak current limits etc. This is novel in the concept of ‘maximum
load at the minute level. This process is described in Figure 13. utilisation’ of transformers and comprehensive assessment
In the initial stage, the transformer is in thermal balance, and methods.
HST and TOT are in a steady‐state. When the load suddenly The proposed method is validated in three scenarios and
changes to 1.5.p.u., TOT and HST are still transient; this the results show that the iterative method can generate a target
feature can be fully utilised for short‐time emergency DTR control scheme based on the operating conditions of the
evaluation. The analysis above shows that the temperature rise transformer. The strategy can fully use the transformer load
is related to the transformer's load factor and ambient tem- capacity while ensuring the safety of the equipment operation.
perature and is also influenced by the initial temperature of the At the same time, the method is used for accurate modelling.
top oil and winding hot spots. Due to the short duration, The optimisation method of the individual transformer ther-
Short‐time emergency DTR assessment transfer to a maximum mal parameters is used and validated to obtain the transformer
time of operation for short‐time overloads. HST accurately. The thermal parameters are fitted by the data
The short‐term emergency DTR assessment under measured by fibre optics in multiple load stages, while the
different initial TOT and load factor conditions was evaluated HST estimation formula is improved and the maximum HST
at an ambient temperature of 35°C. The results are shown in estimation error does not exceed 3 K. The results of the

FIGURE 13 Interrelations between transformer loading and F I G U R E 1 4 Results of short‐time emergency dynamic transformer
temperatures. rating assessment under different starting conditions.
 23977264, 2024, 1, Downloaded from [Link] by South Korea National Provision, Wiley Online Library on [23/07/2024]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ZHANG ET AL.
- 205

dynamic load capacity assessment of the transformer, in this 9. Alvarez, D.L., Rivera, S.R., Mombello, E.E.: Transformer thermal ca-
case, show that the auxiliary equipment current could be a pacity estimation and prediction using dynamic rating monitoring. IEEE
limiting factor for the maximum load capacity of the trans- Trans. Power Deliv. 34(4), 1695–1705 (2019)
10. Deng, Y., et al.: A method for hot spot temperature prediction of a 10 kV
former. The maximum load factor is 1.23 in summer and 1.50 oil‐immersed transformer. IEEE Access 7, 107380–107388 (2019)
in winter for the typical load curve and ambient temperature, 11. Daminov, I., et al.: Assessment of dynamic transformer rating, consid-
which are 23% and 50% higher than the rated load capacity, ering current and temperature limitations. Int. J. Electr. Power Energy
respectively, indicating that the transformer still has a large Syst. 129, 106886 (2021)
load potential. 12. Dong, M.: A data‐driven long‐term dynamic rating estimating method
for power transformers. IEEE Trans. Power Deliv. 36(2), 686–697 (2021)
It should be noted that the transformer health state also 13. Molina Gómez, A., et al.: Optimal sizing of the wind farm and wind farm
theoretically constrains its load capacity, which is still subject to transformer using MILP and dynamic transformer rating. Int. J. Electr.
further research. Power Energy Syst. 136, 107645 (2022)
14. Bracale, A., Carpinelli, G., De Falco, P.: Probabilistic risk‐based man-
agement of distribution transformers by dynamic transformer rating. Int.
ACKN OW L ED G EME N T S
J. Electr. Power Energy Syst. 113, 229–243 (2019)
This work was supported by the National Natural Science 15. Bracale, A., et al.: SmarTrafo: a probabilistic predictive tool for dynamic
Foundation of China ‐ State Grid Corporation Joint Fund for transformer rating. IEEE Trans. Power Deliv. 36(3), 1619–1630 (2021)
Smart Grid under Grant U2066217, Key Research and 16. Zarei, T., et al.: Reliability considerations and economic benefits of dy-
Development Program of Hubei Province under Grant namic transformer rating for wind energy integration. Int. J. Electr. Po-
2021BAA182. wer Energy Syst. 106, 598–606 (2019)
17. Ariza Rocha, O.D., et al.: Dynamic rating assists cost‐effective expansion
of wind farms by utilizing the hidden capacity of transformers. Int. J.
CON F LIC T OF I N T ER E ST STAT E M EN T Electr. Power Energy Syst. 123, 106188 (2020)
The authors declare no conflicts of interest. 18. Zhang, Y., et al.: Short‐term capacity increasing decision method for oil‐
immersed transformer considering comprehensive risk. Auto Electr.
Power Syst. 41(13), 86–91 (2017) (in Chinese)
DATA AVA IL AB I LI T Y S TAT E M EN T 19. Humayun, M., et al.: Utilization improvement of transformers using
The data that support the findings of this study are available on demand response. IEEE Trans. Power Deliv. 30(1), 202–210 (2015)
request from the corresponding author. 20. Wang, J., et al.: Cooperative overload control strategy of power grid‐
transformer considering dynamic security margin of transformer in
OR CI D emergencies. Int. J. Electr. Power Energy Syst. 140, 108098 (2022)
21. Gao, Y., et al.: Methodology to assess distribution transformer thermal
Chen Zhang [Link] capacity for uptake of low carbon technologies. IET Gener. Transm.
Yongqing Deng [Link] Distrib. 11(7), 1645–1651 (2017)
22. Chen, Q.: Research on Improving the Overload Capacity of Large
R EF ER E NCE S Transformers. Zhejiang University (2010) (in Chinese)
23. Hill, J., et al.: Analysing the power transformer temperature limitation for
1. Zhang, C., et al.: Frequency‐constrained co‐planning of generation and
energy storage with high‐penetration renewable energy. J. Mod. Power avoidance of bubble formation. High Volt. 4(3), 210–216 (2019)
Syst. Clean Energ. 9(4), 760–775 (2021) 24. Sarfi, V., Mohajeryami, S., Majzoobi, A.: Estimation of water content in a
power transformer using moisture dynamic measurement of its oil. High
2. Dong, X., et al.: Research and practices of dynamic thermal rating for oil‐
immersed power transformer. High Volt. Eng. 47(6), 1959–1968 (2021) Volt. 2(1), 11–16 (2017)
(in Chinese) 25. GB/T 1094.7‐2008 Power Transformers‐Part 7: Loading Guide for Oil‐
3. Viafora, N., et al.: Day‐ahead dispatch optimization with a dynamic Immersed Power transformer[S]. Beijing: China Standard Press (2008)
26. C57.91‐2011‐IEEE Guide for Loading Mineral‐Oil‐Immersed Trans-
thermal rating of transformers and overhead lines. Electr. Power Syst.
Res. 171, 194–208 (2019) formers. IEEE (2011)
4. Doolgindachbaporn, A., et al.: A top‐oil thermal model for power 27. Power transformers ‐ part 7: loading guide for oil‐immersed power
transformers that considers weather factors. IEEE Trans. Power Deliv. transformers, in BS IEC 60076‐7:2005 (2005)
28. Wang, L., et al.: Improved dynamic thermal model with pre‐physical
37(3), 2163–2171 (2021)
5. Taheri, A.A., Abdali, A., Rabiee, A.: A novel model for thermal behavior modeling for transformers in ONAN cooling mode. IEEE Trans. Po-
prediction of oil‐immersed distribution transformers with consideration wer Deliv. 34(4), 1442–1450 (2019)
of solar radiation. IEEE Trans. Power Deliv. 34(4), 1634–1646 (2019)
6. Feng, D., Wang, Z., Jarman, P.: Evaluation of power transformers’
effective hot‐spot factors by thermal modeling of scrapped units. IEEE How to cite this article: Zhang, C., et al.: Dynamic
Trans. Power Deliv. 29(5), 2077–2085 (2014) thermal rating assessment of oil‐immersed power
7. Liu, Y., et al.: Spatially continuous transformer online temperature transformers for multiple operating conditions. High
monitoring based on distributed optical fibre sensing technology. High
Voltage. 9(1), 195–205 (2024). [Link]
Volt. 7(2), 336–344 (2022)
8. Ma, G., et al.: Optical sensors for power transformer monitoring: a re- hve2.12333
view. High Volt. 6(3), 367–386 (2021)