CN111460727A - Method for predicting service life of transformer by using multiple parameters - Google Patents

Abstract

The invention discloses a transformer life prediction method using multi-parameters, comprising the following steps: S1, using transformer state data to evaluate the state of the transformer; S2, using the historical data of the transformer and the decommissioning data of the same type of transformer to construct an evaluation database; S3, using The Elman network prediction method predicts the evaluation parameters of the transformer to obtain the predicted remaining life. Transformer status data includes transformer electrical test item data, oil dissolved gas analysis data, transformer oil characteristic data and working condition data; oil dissolved gas analysis data includes hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide and carbon dioxide; Transformer oil characteristic data includes micro water content, acid value, breakdown voltage, and oil dielectric loss. The method can be used to predict the life of the transformer, accurately replace the transformer, and provide a technical method to ensure the healthy operation of the transformer.

Description

A Transformer Life Prediction Method Using Multiple Parameters

technical field

The invention relates to the field of life prediction of power transmission and transformation equipment, in particular to a transformer life prediction method using multiple parameters.

Background technique

In order to meet the needs of my country's economic development, the power system has been continuously developed and the technical level has been continuously improved. Power transformers are important electrical equipment in the power system and play a key role in power generation, transmission and distribution.

It is necessary to ensure the healthy operation of the transformer. The healthy operation life of the transformer is related to many factors, and the internal insulation is one of the most important factors. Transformer life evaluation is a complex technical problem. It is necessary to deeply study the internal insulation aging mechanism of the transformer, conduct on-site non-destructive testing and judgment on the aging degree of the transformer, and discuss the correct method to evaluate the aging status of the transformer insulation, so as to predict the risk and reliability of its operation. Therefore, effective maintenance and replacement strategies can be made in combination with economic management to reduce the accident rate.

Large-capacity transformers are required to be equipped with an online monitoring system, and a complete transformer file must be established in the power production management system, including the basic information of the transformer, as well as operating status, repair and maintenance, testing, defects and their elimination records. Comprehensive and effective use of this information, correct evaluation of the operation level of power transformers, and establishment of methods to accurately evaluate the operating life of transformers are of great significance for ensuring the efficient and stable operation of transformers.

SUMMARY OF THE INVENTION

The purpose of the present invention is to solve the problem that the existing technology is difficult to use comprehensive information to evaluate the operating life of the transformer, and proposes a multi-parameter transformer life prediction method, which can be used to predict the life of the transformer and accurately replace the transformer. Provide technical methods to ensure the healthy operation of transformers.

In order to achieve the above technical purpose, a technical solution provided by the present invention is a method for predicting the life of a transformer using multiple parameters, comprising the following steps:

S1. Use the transformer state data to evaluate the state of the transformer;

S2. Use the historical data of the transformer and the decommissioning data of the same type of transformer to construct an evaluation database;

S3. Use the Elman network prediction method to predict the evaluation parameters of the transformer to obtain the predicted remaining life.

In step S1,

S11. Record the transformer state data as a state vector:

Y=(y ₁ ,y ₂ ,...,y _i ,...,y _n )(n=17)

和注意值

并利用公式(1)计算得出各个状态数据的得分情况；S12, determine the initial value of each state quantity _yi

and attention value

And use formula (1) to calculate the score of each state data;

若状态量为负劣化，取

Among them, x _i is the score value of the i-th state quantity, and y′ _i is the deterioration value of the i-th state.

If the state quantity is negatively degraded, take

S13. Use AHP to determine the constant weight of each indicator: W=[w ₁ ,w ₂ ,...,w _n ];

S14. Since the real state of the transformer when a certain performance is seriously degraded cannot be reflected under the constant weight system, the state variable weight vector in the variable weight theory is used for the construction of the weight value:

The state variable weight vector S(X)=(S ₁ (X),...,S _n (X)) is:

According to formula (2), the variable weight coefficient can be obtained as:

S15, using the formula (5) to evaluate the transformer state according to the score of each state quantity and the variable weight coefficient corresponding to the state quantity;

The transformer state data includes transformer electrical test item data, dissolved gas analysis data in oil, transformer oil characteristic data and working condition data;

The electrical test item data includes insulation resistance, absorption ratio, leakage current and DC resistance unbalance coefficient;

The dissolved gas analysis data in the oil includes hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide and carbon dioxide;

The transformer oil characteristic data includes micro water content, acid value, breakdown voltage, and oil dielectric loss;

The working condition data includes the transformer load rate and the transformer operating environment level.

In state evaluation, state variable weight vector is introduced for weight processing. The specific steps are as follows:

A1. Calculate the single-phase score of the item by using the measured value, attention value and initial value of the state data;

A2. Use AHP to determine the initial weight of a single experimental item;

A3. Use the state variable weight vector and the initial weight to calculate the variable weight coefficient;

A4. Use the variable weight coefficient and item score to calculate the transformer state assessment result.

In step S2, the following steps are included:

S21. Record the ex-factory score g ₀ of the same type of transformer and the score g _end when the fault occurs frequently and when it is nearing retirement in such transformer files;

S22 , measuring the transformer state data regularly, and calculating the state evaluation score of the transformer in step S1 , recording it in the transformer file, and constructing a test transformer state evaluation database.

Step S3 includes the following steps:

S31. Construct an Elman' network with 3 inputs and 1 output;

S32. The state score of the transformer in the first 3 years is used as the input, and the state score of the last 1 year is used as the output;

S33. Divide the N-year historical data of the transformer to obtain N-3 groups of training sets, and train the network;

S34. Use the trained Elman network to predict the transformer state;

S35, using the state of the retired transformer to determine the threshold value, and using the predicted state to analyze the remaining life of the transformer.

Beneficial effects of the present invention: The present invention is a method for predicting the life of a transformer using multiple parameters. , so that the deterioration degree of the single state quantity changes with the index value, which can better reflect the barrel effect of the index in the aging process. Using the local feedback characteristics of the Elman network and the dynamic memory function, it is used for the time series prediction of the state quantity of the transformer, showing a high prediction accuracy, and can predict the remaining life of the transformer, which provides guidance for the operation and maintenance personnel to replace the transformer.

Description of drawings

FIG. 1 is a flow chart of a method for predicting the life of a transformer using multiple parameters according to the present invention.

FIG. 2 is a state quantity index system of a multi-parameter transformer life prediction method according to the present invention.

FIG. 3 is a topological diagram of a prediction network Elman network using a multi-parameter transformer life prediction method according to the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only the best of the present invention. The embodiments are only used to explain the present invention, and do not limit the protection scope of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present invention.

Embodiment: As shown in Figure 1, a flow chart of a transformer life prediction method using multiple parameters, the specific steps are as follows:

S1. Use the transformer state data to evaluate the state of the transformer.

S11. As shown in Figure 2, the transformer status data is composed of transformer electrical test item data, dissolved gas analysis data in oil, transformer oil characteristic data and working condition data, specifically: the electrical test item data is insulation resistance, absorption ratio, leakage Current and DC resistance unbalance coefficient; analytical data of dissolved gas in oil are hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, carbon dioxide; transformer oil characteristic data is micro-water content, acid value, breakdown voltage, oil dielectric loss; The working condition data is the load rate of the transformer and the operating environment level of the transformer; it is recorded as a state vector:

Y=(y ₁ , y ₂ ,...,y _i ,...,y _n ) (n=17).

和注意值

并利用公式(1)计算得出各个状态数据的得分情况：S12, determine the initial value of each state quantity _yi

and attention value

And use formula (1) to calculate the score of each state data:

若状态量为负劣化，取

In formula (1), x _i is the rating value of the i-th state quantity, and y′ _i is the deterioration value of the i-th state.

If the state quantity is negatively degraded, take

S13. Determine the constant weights of each indicator according to the AHP:

W=[w ₁ ,w ₂ ,...,w _n ].

S14. Under the constant weight system, the real state of the transformer when a certain performance is seriously degraded cannot be reflected, and the state variable weight vector in the variable weight theory is used for the construction of the weight value:

The state variable weight vector S(X)=(S ₁ (X),...,S _n (X)) is:

According to formula (2), the variable weight coefficient can be obtained as:

S15 , according to the score of each state quantity and the variable weight coefficient corresponding to the state quantity, use the formula (5) to evaluate the transformer state, and record it in the transformer file.

S2. Use the historical data of the transformer and the decommissioning data of the same type of transformer to construct an evaluation database.

S21 records the ex-factory score g ₀ of the same type of transformer and the score g _end when the fault occurs frequently and when it is nearing retirement in such transformer files.

S22. Regularly measure the state data of the transformer, and use step 1 to calculate the state score of the transformer at this time, record it in the transformer file, and construct a test transformer state evaluation database.

S3. As shown in Fig. 3, using the prediction method to predict the evaluation parameters of the transformer to obtain the predicted remaining life.

S31. Select the historical state score data G of the transformer from the state database

G=(g ₁ , g ₂ , . . . , g _n ).

S32. Divide it into training samples and test samples. Among them, the division rules of the training samples are: take _{g 1} ～g _N as the first sample, and (g ₁ , g ₂ ,…,g _N-1 ) are independent variables, and g _N is the objective function; g _N+1 is the second sample, and (g ₂ , g ₃ ,...,g _N ) is the independent variable, and g _N+1 is the objective function; by sampling in sequence, the following training set can be obtained:

S33 , constructing an Elman neural network with N-1 input and 1 output to obtain a predicted network topology.

S34 , for network training, the first N-1 rows of data in the first column of the training set are used as the network input, and the Nth row is used as the output. After training the i group of data, the trained network parameters and network model are obtained. Before training, normalize the input and output for better performance and stability of the network:

In formula (6), g _min represents the minimum value in the score _gi data; g _max represents the maximum value in the score _gi data.

S35, network test. Use the normalized data to test the network, and then normalize the actual output results to score values to obtain the prediction accuracy of the network.

S36 , taking the state score g _n of the transformer over the years as the input, and outputting the state score of the next time, and comparing it with the threshold value of the retired transformer to obtain the predicted life of the transformer.

The specific embodiment described above is a preferred embodiment of a multi-parameter transformer life prediction method of the present invention, and is not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to the specific embodiment. All equivalent changes made according to the shape and structure of the present invention are within the protection scope of the present invention.

Claims (6)

1. A method for predicting the service life of a transformer by using multiple parameters is characterized by comprising the following steps:

s1, performing state evaluation on the transformer by using the transformer state data;

s2, constructing an assessment database by using the historical data of the transformers and the retired data of the transformers of the same type;

and S3, predicting the transformer evaluation parameters by using an Elman network prediction method to obtain the predicted residual life.

2. The method of claim 1, wherein the method comprises the steps of: in the step S1, in the step S,

and S11, recording the transformer state data as a state vector:

Y＝(y₁,y₂,...,y_i,...,y_n)(n＝17)；

s12, determining each state quantity y_iInitial value of (2)

And attention value

Calculating the score condition of each state data by using a formula (1);

wherein x is_iIs the score value of the status quantity of the ith item, y'_iIs the deterioration value of the i-th state, if the state quantity belongs to a positive deterioration condition, i.e., the deterioration is about severe as the value of the state quantity is larger,

if the state quantity is negative deterioration, take

S13, determining the constant weight value of each index by adopting an analytic hierarchy process: w ═ W₁,w₂,...,w_n]；

S14, because the real state of the transformer can not be reflected when a certain performance is seriously reduced under a constant weight system, the state variable weight vector in the variable weight theory is used for the construction of the weight:

wherein the state-variable weight vector S (X) ═ S₁(X),...,S_n(X)) is:

the variable weight coefficient can be obtained according to equation (2) as:

s15, evaluating the transformer state by using a formula (5) according to the scores of the state quantities and the variable weight coefficients corresponding to the state quantities;

3. the method of claim 2, wherein the method comprises predicting the life of the transformer using multiple parameters

The transformer state data comprises transformer electrical test project data, analysis data of dissolved gas in oil, transformer oil characteristic data and working condition data;

the electrical test project data comprise insulation resistance, absorption ratio, leakage current and direct current resistance unbalance coefficient;

the analysis data of the dissolved gas in the oil comprises hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide and carbon dioxide;

the transformer oil characteristic data comprises micro-water content, acid value, breakdown voltage and oil dielectric loss;

the working condition data comprises the load rate of the transformer and the operating environment grade of the transformer.

4. The method of claim 2, wherein the method comprises the steps of: when the state evaluation is carried out, a state variable weight vector is introduced to carry out weight processing on the state variable weight vector, and the specific steps are as follows:

a1, calculating the single-phase score of the item by using the actual measured value, the attention value and the initial value of the state data;

a2, determining the initial weight of each test item by using an analytic hierarchy process;

a3, calculating a variable weight coefficient by using the state variable weight vector and the initial weight;

and A4, calculating the transformer state evaluation result by using the variable weight coefficient and the item score.

5. The method of claim 1, wherein the method comprises the steps of: in step S2, the method includes the steps of:

s21, obtaining the factory score g of the same type of transformer₀Score g of frequent and near-retirement faults_endRecording the data in the transformer file;

and S22, periodically measuring the transformer state data, calculating the state evaluation score of the transformer at the moment by using the step S1, recording the state evaluation score in a transformer file, and constructing a test transformer state evaluation database.

6. The method of claim 1, wherein the method comprises the steps of: step S3 includes the following steps:

s31, constructing an Elman network with 3 inputs and 1 output;

s32, using the state score of the transformer in the previous 3 years as input, and using the state score in the next 1 year as output;

s33, dividing historical data of the transformer in N years to obtain N-3 groups of training sets, and training a network;

s34, predicting the state of the transformer by using the trained Elman network;

and S35, determining a threshold value by using the state of the retired transformer, and analyzing the residual life of the transformer by using the predicted state.

Images