JP7801955B2 - Transformer diagnostic device and transformer diagnostic method - Google Patents

Description

This disclosure relates to a transformer diagnostic device and a transformer diagnostic method.

Transformers installed at electric power company substations are important equipment for power supply, and advanced diagnostics are required. The lifespan of an oil-filled transformer is heavily dependent on the deterioration of the coil insulating paper. Operating at high temperatures for long periods of time causes thermal degradation of the coil insulating paper, significantly affecting its lifespan. Deterioration of coil insulating paper is commonly measured by measuring specific components in the insulating oil. Examples of methods for measuring these specific components include the furfural method and the CO + CO2 method. Meanwhile, the degree of deterioration of insulating paper can also be determined by its average degree of polymerization. However, because the average degree of polymerization cannot be measured directly, a method has been adopted in which insulating oil inside the transformer is sampled, specific components in the insulating oil are measured, and the degree of polymerization is estimated based on the correlation between the amount of these specific components and the average degree of polymerization.

Japanese Patent Application Laid-Open No. 2006-250872

However, this correlation is a static relationship that does not depend on time, and has the problem that it does not take into account changes in state over time.

Meanwhile, the temperature rise and resulting lifespan reduction caused by overloading of oil-immersed transformers are described in IEC 60076-7 and IEEE C57.91. These standards specify the loss of life due to the temperature of the hotspot part of the oil-immersed transformer coil, which is the hottest part.

The above correlation is a binary relationship between the amount of specific components and the average degree of polymerization, and there is a problem in that it is difficult to show a three-way correlation, including temperature, even if other measured quantities are used. In addition, it is difficult to calculate the time-series state changes mentioned above, and there is a problem in that it is not possible to estimate changes in lifespan loss over time and transformer abnormalities caused by this lifespan loss.

Therefore, the object of the present disclosure is to provide a transformer diagnostic device and a transformer diagnostic method that can estimate transformer abnormalities with high accuracy.

The transformer diagnostic device disclosed herein includes an analysis device that measures the average degree of polymerization of the insulating paper in the transformer from the amount of dissolved gas of a specific component dissolved in the insulating oil in the transformer, and an abnormality determination device that calculates a probability distribution of the remaining life and a probability distribution of the average degree of polymerization based on a state equation that represents the relationship between the time change of a state variable, which is the remaining life of the transformer based on the life loss of the transformer, and an observation equation that represents the relationship between an observation variable, which is the average degree of polymerization, and the state variable , calculates the likelihood of the average degree of polymerization using the measurement value of the average degree of polymerization measured by the analysis device and the probability distribution of the average degree of polymerization, and estimates the remaining life based on the likelihood of the average degree of polymerization and the probability distribution of the remaining life, thereby determining an abnormality in the transformer.

The transformer diagnosis method disclosed herein includes the steps of measuring the average degree of polymerization of the insulating paper in the transformer from the amount of dissolved gas of a specific component dissolved in the insulating oil in the transformer; calculating a probability distribution of the remaining life and a probability distribution of the average degree of polymerization based on a state equation that represents the relationship between the time change of a state variable, which is the remaining life of the transformer based on the life loss of the transformer, and an observation equation that represents the relationship between the observation variable, which is the average degree of polymerization , and the state variable ; calculating the likelihood of the average degree of polymerization using the measured value of the average degree of polymerization and the probability distribution of the average degree of polymerization; and determining an abnormality in the transformer by estimating the remaining life based on the likelihood of the average degree of polymerization and the probability distribution of the remaining life .

This disclosure makes it possible to estimate transformer abnormalities with high accuracy.

1 is a diagram showing the configuration of a transformer diagnostic device 1 according to a first embodiment; 3 is a flowchart showing the procedure of a diagnostic method for a transformer 2 according to the first embodiment. 10A and 10B are diagrams showing examples of displays on a display device 4. 10A and 10B are diagrams showing examples of the probability distribution of the average degree of polymerization and the abnormal threshold distribution. 10 is a diagram showing the configuration of an abnormality determination device 3 when the functions of the abnormality determination device 3 are realized using software. FIG.

The following describes the embodiments with reference to the drawings.

Embodiment 1.
FIG. 1 is a diagram showing the configuration of a transformer diagnostic device 1 according to a first embodiment.

The transformer diagnostic device 1 includes an oil-borne gas analyzer 5, a temperature sensor 6, a transmission device 8, a measurement database 7, an abnormality determination device 3, and a display device 4.

The temperature sensor 6 is installed near the windings of the transformer 2. The temperature sensor 6 measures the hot spot temperature of the transformer 2 at a frequency of 30 minutes to 1 hour. Instead of the hot spot temperature, the oil temperature, which is easier to measure, may be measured, and the hot spot temperature may be estimated by estimating the difference between the oil temperature and the hot spot temperature. The hot spot temperature can be estimated, for example, using a differential equation represented by a thermal equivalent model with transient characteristics that takes transient phenomena into account, as described in IEC 60076-7.

The gas-in-oil analyzer 5 is installed near the transformer 2. The gas-in-oil analyzer 5 obtains the amount of dissolved gas of specific components dissolved in insulating oil sampled, such as once a year. The gas-in-oil analyzer 5 calculates the average degree of polymerization of the insulating paper based on the amount of dissolved gas.

The measurement database 7 stores the obtained measurement values.

Transmission device 8 transmits the measurement results.

The abnormality determination device 3 determines whether there is an abnormality in the transformer 2.

The abnormality determination device 3 estimates the remaining life, which cannot be directly observed, by subtracting the life loss from the initial life when the transformer 2 began operation. The remaining life is expressed as a state equation, and is represented as a state space model in which the average degree of polymerization, which can be observed semi-directly, is expressed as a function of the remaining life using an observation equation.

In other words, the abnormality determination device 3 estimates the state variable based on a state equation that represents the relationship between the time change of the state variable, which is the remaining life of the transformer 2 based on the life loss of the transformer 2, an observation equation that represents the relationship between the observation variable, which is the average degree of polymerization, and the state variable, and the measured value of the average degree of polymerization measured by the gas-in-oil analyzer 5, and determines whether there is an abnormality in the transformer 2 based on the estimated value of the state variable.

The display device 4 displays abnormality determination results, etc.

The abnormality determination device 3 includes a prediction unit 31, a likelihood estimation unit 32, and a deterioration determination unit 33.

The prediction unit 31 calculates the lifetime loss Vn for the hot spot temperature _θH measured the nth time using equation (1). When _θH = 110°C, Vn = 1, when _θH is greater than 110°C, Vn > 1, i.e., deterioration is accelerated, and when _θH is less than 110°C, Vn < 1, i.e., deterioration is decelerated.

When temperature measurements are taken M times over a one-year period at measurement intervals Δt, the prediction unit 31 calculates the one-year life loss V using equation (2). If Vn is greater than 1, V will be longer than the actual time, and if Vn is less than 1, V will be shorter than the actual time.

The prediction unit 31 calculates the probability distribution of the remaining life at time t using the discrete time t=0, 1, 2, ..., N of the hot spot temperature and the life loss Vt , using the state equation of equation (3) which shows the relationship between time t-1 and time t for the residual life R. The prediction unit 31 calculates the probability distribution of the average degree of polymerization D at time t using the observation equation of equation (4) which shows the relationship between the remaining life R at time t and the average degree of polymerization D.

where Rt is the remaining life at time t, Dt is the average degree of polymerization at time t, Vt is the life loss at time t, ut is the system noise at time t, and vt is the observation noise at time t.

The prediction unit 31 generates system noise using random numbers L times, calculates the life loss at time t using equation (2) with the hotspot temperature measurement, and calculates the life loss at time t using equation (3) with the system noise generated by random numbers from the remaining life at time t-1. L remaining lifespans are obtained. This is called the probability distribution of the remaining lifespan Rt.

For each life loss, the likelihood estimation unit 32 uses the probability distribution of the observation noise and the remaining life Rt at time t to obtain the probability distribution of the average degree of polymerization at time t using the observation equation in equation (4). By substituting the measured value of the average degree of polymerization into equation (4), the probability of measuring the average degree of polymerization is obtained. Since there are L remaining lives at time t, L probabilities of obtaining a measured value of the average degree of polymerization for each remaining life are also obtained. The likelihood of the average degree of polymerization is calculated by Bayesian estimation for the probability distribution of the average degree of polymerization calculated using the equation of state using the measured value of the average degree of polymerization at time t.

That is, the likelihood estimation unit 32 estimates the probability distribution of the remaining life Rt of the transformer 2 based on the measurement data. In Bayesian estimation, X is considered to be a random variable in the probability density function P. P(X) is the prior probability density distribution of the random variable X, P(X|Y) is the posterior probability density distribution of the random variable X when the measurement value Y is obtained, and P(Y|X) is the posterior probability density distribution that the measurement value Y is obtained for the random variable X, and these are called likelihoods.

In this embodiment, a particle filter is used as a method for estimating the posterior probability density distribution in the likelihood estimation process. A particle filter is a method for approximating the posterior probability density distribution P(X|Y) of a random variable X with the distribution of a particle group consisting of a large number of particles, and updates the posterior probability density distribution of the random variable X in a time series manner by repeating the prediction process and likelihood estimation. If the number of particles Xt is L, then Xt can be expressed as Xt = {Xt(1), Xt(2), ..., Xt(L)}. The remaining lifetime Rt corresponds to a particle, and L particles of each type are generated.

Using the observation equation, a probability distribution of the predicted value of the average degree of polymerization Dt at time t is generated for each particle, taking into account the probability distribution of the observation noise. Using the probability distribution of this predicted value and the measured value of the average degree of polymerization YDt, a weight w(i) is calculated as the probability of obtaining this measured value for the i-th particle. Resampling is performed according to the weight as a ratio to the sum of the weights w(i) of each particle Xt. In resampling, each particle is regenerated based on the weight value, and a new particle group is generated according to the weight. In other words, each particle is generated based on the probability of the value obtained by dividing the weight w(i) of particle Xt by the sum of the weights of all particles, resulting in a new particle group.

The deterioration determination unit 33 determines whether there is an abnormality in the transformer 2 based on the estimated remaining lifespan.

Figure 2 is a flowchart showing the steps of the diagnostic method for transformer 2 in embodiment 1.

In step S101, the prediction unit 31 performs initialization processing as follows:

Let D0 = 1100 and R0 = 18000.

In step S102, the prediction unit 31 executes the prediction process.

Specifically, the prediction unit 31 calculates the one-year life loss V based on equations (1) and (2) and sets this as Vt.

The prediction unit 31 calculates the remaining lifetime Rt(i) and average degree of polymerization Dt(i) at time t for the number of particles L based on equations (5) to (8). ut(i) is the system noise at time t, vt(i) is the observation noise at time t, and rand is a uniform random number.

In step S103, the likelihood estimation unit 32 executes a likelihood estimation process. Specifically, the likelihood estimation unit 32 acquires the observed value YDt of the average degree of polymerization at time t. The likelihood estimation unit 32 calculates the likelihood (weight) wt(i) of the average degree of polymerization for each particle based on equation (9).

Next, the degradation determination unit 33 calculates the normalized weight nwt(i) from the weight wt(i) according to equation (10).

In step S104, the deterioration determination unit 33 executes the deterioration determination process. The deterioration determination unit 33 obtains the remaining life MRt by weighting the remaining life Rt(i) of particle i at time t according to equation (11).

The deterioration determination unit 33 obtains the average degree of polymerization MDt according to equation (12).

The deterioration determination unit 33 determines that the transformer 2 is normal if the average degree of polymerization MDt is equal to or greater than the threshold value MD, and determines that the transformer 2 is abnormal if the average degree of polymerization MDt is less than the threshold value MD.

If it is determined that transformer 2 is abnormal, processing proceeds to step S105. If it is determined that transformer 2 is normal, processing proceeds to step S106.

In step S105, the deterioration determination unit 33 displays the abnormality determination result, remaining life MRt, and average polymerization degree MDt on the display device 4.

In step S106, if t is the end time, processing proceeds to step S107. If t is not the end time, processing proceeds to step S108.

In step S107, the deterioration determination unit 33 displays the remaining life MRt and average polymerization degree MDt on the display device 4.

In step S108, the prediction unit 31 samples particle Rt again L times with a probability according to the normalized weight nwt(i). That is, the prediction unit 31 selects particles Rt = {Rt(1), Rt(2), ..., Rt(L)} with a probability of nwt(i) rather than a uniform 1/L probability. The resulting L particles form the probability distribution at time t. If nwt(i) is uniformly 1/L, one Rt(i) is obtained for each particle. If nwt(i) is large, two or more Rt(i) are obtained, and if nwt(i) is small, zero Rt(i) is obtained.

The prediction unit 31 similarly samples particle Dt again L times with a probability according to the normalized weight nwt(i).

In step S109, the prediction unit 31 increments t.

Figure 3 shows an example of the display on the display device 4.

Along with attributes such as the transformer's installation location, installation position, manufacturer, and capacity, the measured value of the average degree of polymerization and the remaining lifespan as a result of the digital twin simulation are displayed, along with the average value and 95% and 99% confidence intervals in the probability distribution of the average degree of polymerization. If the judgment result is abnormal, an abnormality is displayed. The average value in the probability distribution of the average degree of polymerization is MDt.

Embodiment 2.
FIG. 4 shows an example of the probability distribution of the average degree of polymerization and the abnormal threshold distribution.

The degradation determination unit 33 uses the obtained probability distribution of the average degree of polymerization and the abnormality threshold distribution for determining whether the average degree of polymerization is abnormal to calculate the log-likelihood ratio LR, which is expressed as the logarithm of the ratio between the probability P0(x) that the average degree of polymerization is normal and the probability P1(x) that it is abnormal. The probability distribution of the average degree of polymerization is obtained from Dt(1) to Dt(L).

The degradation determination unit 33 can determine that the transformer 2 is abnormal if the log-likelihood ratio LR exceeds a threshold value.

Variant.
The present disclosure is not limited to the above-described embodiment. For example, the following modifications are also possible.

(1) Kalman Filter In the above embodiment, a particle filter is used as a method for estimating the posterior probability density distribution in the likelihood estimation process, but this is not limiting. For example, a Kalman filter may also be used.

(2) Learning of Initial Values The initialized values of D0 and R0 in step S101 of the flowchart in Figure 2 may be changed based on the remaining life estimated by the processing in the flowchart in Figure 2 and the actual life of the transformer.

For example, when D0 = 1100 and R0 = 18000 are set, if step S104 is judged as YES at time t = 3, and transformer 2 has not actually reached the end of its life, the abnormality determination device may set the value of D0 to a value smaller than 1100 and the value of R0 to a value smaller than 18000.

(3) Deep Learning The abnormality determination device may learn the time-series change of the average polymerization degree by learning or the like and predict the future average polymerization degree. The abnormality determination device may then use the predicted future average polymerization degree instead of the measured value to estimate the future remaining life.

The learning algorithm can be deep learning, which learns to extract features themselves, or machine learning can be performed using other known methods, such as genetic programming, functional logic programming, or support vector machines.

(4) The abnormality determination device 3 described in the embodiment can be configured with corresponding operations implemented using digital circuit hardware or software. When the functions of the abnormality determination device 3 are realized using software, the abnormality determination device 3 can include, for example, a processor 1003 and memory 1002 connected by a bus 1001, as shown in FIG. 5, and the processor 1003 can execute a program stored in the memory 1002.

(5) Supplementary Note (Supplementary Note 1)
an analyzer for measuring the average degree of polymerization of insulating paper in a transformer based on the amount of dissolved gas of a specific component dissolved in insulating oil in the transformer;
A transformer diagnostic device comprising: a state equation representing the relationship of time change of a state variable, which is the remaining life of the transformer based on the life loss of the transformer; an observation equation representing the relationship between an observation variable, which is the average degree of polymerization, and the state variable; and an abnormality determination device that determines an abnormality in the transformer based on a measurement value of the average degree of polymerization measured by the analysis device.

(Appendix 2)
The abnormality determination device uses Equation (A1) as the state equation and Equation (A2) as the observation equation,

where Rt is the remaining life at time t, Dt is the average degree of polymerization at time t, Vt is the life loss at time t, ut is the system noise at time t, and vt is the observation noise at time t. A diagnostic device for a transformer as described in Appendix 1.

(Appendix 3)
3. The transformer diagnosis device according to claim 1, wherein the abnormality determination device calculates a life loss of the transformer based on a hot spot temperature of the transformer.

(Appendix 4)
4. The transformer diagnosis device according to claim 1, wherein the abnormality determination device estimates the state variables using particle filtering.

(Appendix 5)
measuring an average degree of polymerization of insulating paper in the transformer from the amount of dissolved gas of a specific component dissolved in insulating oil in the transformer;
A diagnostic method for a transformer, comprising: a step of determining an abnormality in the transformer based on a state equation representing the relationship of time change of a state variable, which is the remaining life of the transformer based on the life loss of the transformer; an observation equation representing the relationship between an observation variable, which is the average degree of polymerization, and the state variable; and a measured value of the measured average degree of polymerization.

The embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The scope of the present invention is indicated by the claims, not the above description, and is intended to include all modifications that are equivalent in meaning to and within the scope of the claims.

1 Transformer diagnostic device, 2 Transformer, 3 Abnormality determination device, 4 Display device, 5 Oil-dissolved gas analyzer, 6 Temperature sensor, 7 Measurement database, 8 Transmission device, 31 Prediction unit, 32 Likelihood estimation unit, 33 Deterioration determination unit, 1001 Bus, 1002 Memory, 1003 Processor.

Claims (5)

an analyzer for measuring the average degree of polymerization of insulating paper in a transformer based on the amount of dissolved gas of a specific component dissolved in insulating oil in the transformer;
and an abnormality determination device that calculates a probability distribution of the remaining life and a probability distribution of the average degree of polymerization based on a state equation that represents the relationship between the time change of a state variable, which is the remaining life of the transformer based on the life loss of the transformer, and an observation equation that represents the relationship between an observation variable, which is the average degree of polymerization , and the state variable, calculates a likelihood of the average degree of polymerization using a measurement value of the average degree of polymerization measured by the analysis device and the probability distribution of the average degree of polymerization, and estimates the remaining life based on the likelihood of the average degree of polymerization and the probability distribution of the remaining life, thereby determining an abnormality in the transformer. The abnormality determination device uses Equation (A1) as the state equation and Equation (A2) as the observation equation,

2. The transformer diagnostic device of claim 1, wherein Rt is the remaining life at time t, Dt is the average polymerization degree at time t, Vt is the life loss at time t, ut is the system noise at time t, and vt is the observation noise at time t. The transformer diagnostic device of claim 2, wherein the abnormality determination device calculates the transformer's loss of life based on the transformer's hot spot temperature. The transformer diagnostic device of claim 3, wherein the abnormality determination device estimates the state variables using particle filtering. measuring an average degree of polymerization of insulating paper in the transformer from the amount of dissolved gas of a specific component dissolved in insulating oil in the transformer;
A method for diagnosing a transformer, comprising the steps of: calculating a probability distribution of the remaining life and a probability distribution of the average degree of polymerization based on a state equation that represents the relationship between the time change of a state variable, which is the remaining life of the transformer based on the life loss of the transformer, and an observation equation that represents the relationship between an observation variable, which is the average degree of polymerization , and the state variable; calculating a likelihood of the average degree of polymerization using a measured value of the average degree of polymerization and the probability distribution of the average degree of polymerization; and determining an abnormality in the transformer by estimating the remaining life based on the likelihood of the average degree of polymerization and the probability distribution of the remaining life.