CN-122020412-A - Transformer risk assessment method and device based on multi-index fusion

CN122020412ACN 122020412 ACN122020412 ACN 122020412ACN-122020412-A

Abstract

The invention discloses a transformer risk assessment method and device based on multi-index fusion, relates to the technical field of electrical performance testing and electrical fault detection, and mainly aims to solve the problem of low accuracy of risk assessment of an existing transformer. The method mainly comprises the steps of respectively calculating comprehensive risk values of different transformers according to static risk indexes of the transformers, determining static risk levels according to the comprehensive risk values, acquiring real-time monitoring data in the running process of the transformers according to data monitoring frequencies determined by the static risk levels, identifying first state evaluation results of the transformers according to the real-time monitoring data through a pre-built Bayesian classifier, identifying second state evaluation results of the transformers through a pre-built fuzzy evaluation model, and carrying out collaborative verification according to the first state evaluation results and the second state evaluation results of the transformers for any one transformer to obtain risk evaluation results of the transformers. The method is mainly used for evaluating the risk of the transformer.

Inventors

FENG JIAN
LI DIANYANG
WANG BOWEN
ZHANG BOWEN
LI YAJING

Assignees

东北大学

Dates

Publication Date: 20260512
Application Date: 20251209

Claims (10)

1. The transformer risk assessment method based on multi-index fusion is characterized by comprising the following steps of: Respectively calculating comprehensive risk values of different transformers according to static risk indexes of all transformers in a target power supply area, and determining static risk grades of different transformers according to the comprehensive risk values; Responding to a transformer risk assessment instruction of the target power supply area, acquiring data monitoring frequencies corresponding to different transformers according to the static risk level, and acquiring real-time monitoring data in the running process of the transformers according to the data monitoring frequencies; Identifying a first state evaluation result of each transformer through a pre-constructed Bayesian classifier according to the real-time monitoring data, and identifying a second state evaluation result of each transformer through a pre-constructed fuzzy evaluation model; And aiming at any transformer, carrying out cooperative verification according to a first state evaluation result and a second state evaluation result of the transformer to obtain a risk evaluation result of the transformer.
2. The method of claim 1, wherein the static risk indicator comprises a plurality of equipment importance indicators and a plurality of equipment hidden danger indicators; The calculation process of the comprehensive risk value of any transformer comprises the following steps: Weighting and summing all equipment importance indexes of the transformer to obtain equipment importance, wherein the equipment importance indexes comprise power supply area attributes, economic cost, voltage level and associated scale; Carrying out weighted summation on various equipment hidden danger degree indexes of the transformer to obtain equipment hidden danger degree, wherein the equipment hidden danger degree indexes comprise health degree, fault influence degree and maintenance frequency, and the health degree is determined based on local discharge capacity and winding leakage current; calculating a static risk value of the transformer according to the equipment importance, the equipment hidden trouble degree and the fault result severity; And calculating the comprehensive risk value of the transformer according to the relative importance coefficient of the static risk value and the transformer.
3. The method of claim 2, wherein the determining of the relative importance coefficients comprises: for any transformer, carrying out weighted summation on all importance indexes of the transformers, and calculating the ratio of weighted summation results of every two transformers to obtain a comprehensive importance ratio; Constructing a relative importance matrix of the global transformer by taking the comprehensive importance ratio as a matrix element; solving the maximum eigenvalue of the relative importance matrix and the eigenvector corresponding to the maximum eigenvalue, and carrying out normalization processing on each component in the eigenvector to obtain the relative importance coefficients of different transformers.
4. The method according to claim 1, wherein the pre-constructed bayesian classifier includes probability density functions and prior probabilities corresponding to different preset operating states respectively; identifying, for any one of the transformers, a first state evaluation result of the transformer by a pre-constructed bayesian classifier, including: Extracting multidimensional dynamic characteristic parameters from the real-time monitoring data, wherein the multidimensional dynamic characteristic parameters comprise oil chromatography gas content, winding temperature, local discharge capacity and winding leakage current ratio; substituting the multidimensional dynamic characteristic parameters into a probability density function of the preset running state for any preset running state, calculating to obtain conditional probability, and calculating posterior probability of the preset running state according to the conditional probability and prior probability of the preset running state; taking a preset running state with the maximum posterior probability as a first state evaluation result of the transformer; The probability density function constructing process comprises the steps of determining prior probabilities corresponding to different preset operation states according to historical fault statistical data and preset electrical parameter thresholds, extracting characteristic mean value vectors and characteristic covariance matrixes in different preset operation states from normal distribution by assuming that dynamic characteristics in each preset operation state obey multidimensional normal distribution, and constructing probability density functions corresponding to different preset operation states according to the characteristic mean value vectors and the characteristic covariance matrixes.
5. The method of claim 1, wherein the pre-constructed fuzzy evaluation model comprises a pre-constructed trapezoidal membership function and an initial fuzzy relation matrix; for any one of the transformers, identifying a second state evaluation result of the transformer through a pre-constructed fuzzy evaluation model, wherein the second state evaluation result comprises the following steps: extracting real-time parameters of preset evaluation indexes from the real-time monitoring data, wherein the preset evaluation indexes comprise an acetylene exceeding rate in oil, a winding temperature rise rate, a partial discharge exceeding rate and an oil quality degradation degree; Inputting the real-time parameters into the trapezoidal membership functions, and calculating the real-time membership degree of each preset evaluation index relative to different preset running states; updating the initial fuzzy relation matrix through the real-time membership to obtain an updated fuzzy relation matrix, wherein the initial fuzzy relation matrix comprises membership degrees of different preset evaluation indexes to different preset running states; according to the index weights of different preset evaluation indexes and the updated fuzzy relation matrix, fuzzy synthesis is carried out through a weighted average fuzzy operator, and a comprehensive membership vector is obtained; and taking the preset running state with the largest membership degree in the comprehensive membership degree vector as a second state evaluation result of the transformer.
6. The method of claim 1, wherein for any transformer, performing a collaborative check according to the first state evaluation result and the second state evaluation result of the transformer to obtain a risk evaluation result of the transformer, comprising: If the first state evaluation result is consistent with the second state evaluation result, taking the first state evaluation result or the second state evaluation result as a risk evaluation result; If the first state evaluation result is inconsistent with the second state evaluation result, invoking a static risk level of the transformer, and determining a target iteration number and a target feedback correction strategy according to the static risk level; Correcting the prior probability of the Bayesian classifier and/or the index weight of the fuzzy evaluation model through an error feedback method according to the target feedback correction strategy, repeatedly executing the evaluation process of the first state evaluation result and the second state evaluation result based on the corrected prior probability and/or the corrected index weight, so as to iteratively update the first state evaluation result and the second state evaluation result until the updated first state evaluation result is consistent with the updated second state evaluation result or the iteration times reach the target iteration times, and generating a risk evaluation result according to the first state evaluation result and the second state evaluation result which are output by the last iteration.
7. The method of claim 6, wherein the static risk level comprises a high risk level, a medium risk level, and a low risk level; Determining a target iteration number and a target feedback correction strategy according to the static risk level, wherein the method comprises the following steps: Under the condition that the static risk level is a high risk level, configuring a target iteration number to be a first preset value, and determining a first correction strategy to be a target feedback correction strategy, wherein the first correction strategy comprises synchronously correcting the index weight of the fuzzy evaluation model and the prior probability of the Bayesian classifier; Under the condition that the static risk level is the risk level, configuring the target iteration times to the second preset value, and determining a second correction strategy to be a target feedback correction strategy, wherein the second correction strategy comprises the steps of determining a model with lower output confidence level in a Bayesian classifier and a fuzzy evaluation model as a target correction model, and correcting the target correction model; and under the condition that the static risk level is a low risk level, determining a third correction strategy as a target feedback correction strategy, wherein the third correction strategy corrects the weight of the fuzzy evaluation model, and when the corrected first evaluation state result and the corrected second evaluation state result are not agreed, the corrected first evaluation state result is used as the risk evaluation result of the transformer.
8. The utility model provides a transformer risk assessment device based on multi-index fuses which characterized in that includes: The static risk assessment module is used for respectively calculating comprehensive risk values of different transformers according to static risk indexes of all transformers in a target power supply area and determining static risk grades of different transformers according to the comprehensive risk values; the acquisition module is used for responding to the transformer risk assessment instruction of the target power supply area, acquiring data monitoring frequencies corresponding to different transformers according to the static risk grade, and acquiring real-time monitoring data in the running process of the transformers according to the data monitoring frequencies; the running state evaluation module is used for identifying a first state evaluation result of each transformer through a pre-built Bayesian classifier according to the real-time monitoring data, and identifying a second state evaluation result of each transformer through a pre-built fuzzy evaluation model; And the collaborative verification module is used for carrying out collaborative verification on any transformer according to the first state evaluation result and the second state evaluation result of the transformer to obtain a risk evaluation result of the transformer.
9. A storage medium having stored therein at least one executable instruction for causing a processor to perform operations corresponding to the transformer risk assessment method according to any one of claims 1 to 7.
10. The terminal is characterized by comprising a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus; the memory is configured to store at least one executable instruction, where the executable instruction causes the processor to perform operations corresponding to the transformer risk assessment method according to any one of claims 1 to 7.

Description

Transformer risk assessment method and device based on multi-index fusion Technical Field The invention relates to the technical field of electrical performance testing and electrical fault detection, in particular to a transformer risk assessment method and device based on multi-index fusion. Background The power transformer is used as core equipment of a power transmission and distribution link of a power grid, and the risk state of the power transformer is directly related to safe and stable operation of a power system. The risk of the transformer is accurately estimated, so that not only is the hidden trouble of the equipment self identified, but also the structural effect of the transformer in the power grid topology is considered, namely the same fault occurs in a hub transformer substation and a terminal transformer substation, and the influence degree of the same fault on the system is obviously different. Therefore, the scientific risk assessment system needs to take the inherent attribute of equipment, the correlation characteristic of the system and the dynamic electricity physical quantity monitoring into consideration, so that an effective decision basis can be provided for the operation and maintenance of the power grid. When a current power system performs risk assessment of a transformer, a single risk assessment model or a fixed-threshold monitoring means is generally adopted. The method has the problem that evaluation dimension is one-sided, early warning sensitivity and resource consumption are difficult to consider. The method is characterized in that on one hand, for a large number of operation equipment, a 'one-cut' monitoring frequency and an evaluation strategy are adopted, so that early fault characteristics of high-risk equipment are difficult to capture in time, excessive monitoring and calculation resource waste exist in low-risk equipment, on the other hand, a single evaluation model is limited by inherent assumption or data dependency, misjudgment or missed judgment easily occurs under complex operation conditions, and stability and reliability of an evaluation result are insufficient. Disclosure of Invention In view of the above, the present invention provides a transformer risk assessment method and device based on multi-index fusion, which mainly aims to solve the problem of insufficient stability and reliability of the existing transformer assessment result. According to one aspect of the invention, there is provided a transformer risk assessment method based on multi-index fusion, comprising: Respectively calculating comprehensive risk values of different transformers according to static risk indexes of all transformers in a target power supply area, and determining static risk grades of different transformers according to the comprehensive risk values; Responding to a transformer risk assessment instruction of the target power supply area, acquiring data monitoring frequencies corresponding to different transformers according to the static risk level, and acquiring real-time monitoring data in the running process of the transformers according to the data monitoring frequencies; Identifying a first state evaluation result of each transformer through a pre-constructed Bayesian classifier according to the real-time monitoring data, and identifying a second state evaluation result of each transformer through a pre-constructed fuzzy evaluation model; And aiming at any transformer, carrying out cooperative verification according to a first state evaluation result and a second state evaluation result of the transformer to obtain a risk evaluation result of the transformer. Further, the static risk index comprises a plurality of equipment importance indexes and a plurality of equipment hidden danger indexes; The calculation process of the comprehensive risk value of any transformer comprises the following steps: Weighting and summing all equipment importance indexes of the transformer to obtain equipment importance, wherein the equipment importance indexes comprise power supply area attributes, economic cost, voltage level and associated scale; Carrying out weighted summation on various equipment hidden danger degree indexes of the transformer to obtain equipment hidden danger degree, wherein the equipment hidden danger degree indexes comprise health degree, fault influence degree and maintenance frequency, and the health degree is determined based on local discharge capacity and winding leakage current; calculating a static risk value of the transformer according to the equipment importance, the equipment hidden trouble degree and the fault result severity; And calculating the comprehensive risk value of the transformer according to the relative importance coefficient of the static risk value and the transformer. Further, the determining process of the relative importance coefficient includes: for any transformer, carrying out weighted summation on all importance indexes of the