CN-121980923-A - Method, system, equipment and medium for predicting residual life of breaker operating mechanism

CN121980923ACN 121980923 ACN121980923 ACN 121980923ACN-121980923-A

Abstract

The invention provides a method, a system, equipment and a medium for predicting the residual life of a breaker operating mechanism, wherein the method comprises the steps of collecting sensor data of the breaker operating mechanism at the current moment, inputting the sensor data into a pre-trained life prediction model, obtaining a residual life predicted value of the breaker operating mechanism based on the output of the life prediction model, wherein the life prediction model is obtained by training based on a preset basic model, the preset basic model is provided with a deep Bayesian neural network for modeling model uncertainty, the output layer of the deep Bayesian neural network is provided with an accidental node layer for modeling accidental uncertainty, and the life prediction model is obtained when the model uncertainty and the accidental uncertainty meet preset convergence conditions through historical sensor data and corresponding residual life labels of the breaker operating mechanism. The method and the device can remarkably improve the accuracy and the reliability of prediction.

Inventors

WANG YUEBIN
GUO BAODONG
LIU SHENG
Zhuang Qinghan
YANG BIN
YANG FENG
SU CHAOHUI
KONG QINGCHUN
GU PENGCHENG
HAO WEIHAN
CHENG YANJIE
LI ZHANG
KANG HAIDONG
HUANG YANG
DONG YIHAN
LIU PENGCHENG

Assignees

三峡新能源阳江发电有限公司
广东阳江市创源海上风电综合投资有限公司
中国长江三峡集团有限公司
中国能源建设集团广东省电力设计研究院有限公司

Dates

Publication Date: 20260505
Application Date: 20251229

Claims (10)

1. A method for predicting remaining life of a circuit breaker operating mechanism, comprising: Collecting sensor data of a breaker operating mechanism at the current moment; Inputting the sensor data into a pre-trained life prediction model, and obtaining a residual life prediction value of the breaker operating mechanism based on the output of the life prediction model; The life prediction model is obtained by training a preset basic model based on a preset basic model, the preset basic model is configured with a deep Bayesian neural network for modeling model uncertainty, an output layer of the deep Bayesian neural network is configured with an accidental node layer for modeling accidental uncertainty, and the life prediction model is obtained when the model uncertainty and the accidental uncertainty meet preset convergence conditions through historical sensor data of a breaker operating mechanism and corresponding residual life labels.
2. The method for predicting remaining life of a circuit breaker operating mechanism according to claim 1, wherein the prediction distribution of the life prediction model is expressed as: ; Wherein, the For the purpose of the uncertainty of the model, In order to account for the occasional uncertainty, And X is a set of the historical sensor data, Y is a residual life label corresponding to the historical sensor data, X is a historical sensor data variable, Y is a residual life variable, and alpha and beta are two intermediate node parameters in the accidental node layer.
3. The method for predicting the remaining life of a circuit breaker operating mechanism of claim 1, wherein the training method for the life prediction model comprises: Acquiring a basic model, historical sensor data and a residual life label; performing iterative optimization on parameterized distribution of the basic model by using the historical sensor data and the residual life labels; and obtaining a life prediction model when the relative relation between the parameterized distribution subjected to iterative optimization and the real posterior distribution meets a preset condition.
4. A method for predicting the remaining life of a circuit breaker operating mechanism as recited in claim 3 wherein said method for predicting the remaining life further comprises, after said iteratively optimizing the parameterized distribution of said base model: performing Monte Carlo sampling on the neural network weight distribution of the basic model; Sampling based on the parameter distribution of the Monte Carlo sampling result to obtain discrete prediction distribution; And calculating prediction expectations and prediction variances of the discrete prediction distribution, wherein the discrete prediction distribution, the prediction expectations and the prediction variances are used for quantifying the model uncertainty and the accidental uncertainty.
5. The method for predicting remaining life of a circuit breaker operating mechanism of claim 1, further comprising, prior to said iteratively optimizing the parameterized distribution of the base model: And constructing an objective function through an evidence lower bound algorithm, and determining the preset condition according to the objective function, wherein the objective function is constructed according to the expectation of a log likelihood function under parameterized distribution and KL divergence between the parameterized distribution and a true posterior distribution.
6. The method for predicting the remaining life of a circuit breaker operating mechanism according to claim 1, wherein said obtaining a predicted value of the remaining life of the circuit breaker operating mechanism based on the output of the life prediction model comprises: Acquiring historical output data of the life prediction model, acquiring a preliminary predicted value based on the sensor data, wherein the historical output data comprises life prediction distribution of a plurality of time steps; calculating a weighted average of life prediction distributions for a plurality of time steps; And carrying out fusion processing by adopting particle filtering based on a weighted average result and the preliminary predicted value through a preset updating function to obtain the residual life predicted value.
7. The method for predicting remaining life of a circuit breaker operating mechanism of claim 6, wherein said predetermined update function is expressed as: ; Wherein, the For the weight of the ith particle at the kth moment, For the weight of the ith particle at time k-1, As the weight of the jth particle at time k-1, The lifetime prediction distribution based on the sensor data x 1:k from 1 to k is represented, y k represents the lifetime prediction value at k, and N s represents the particle number.
8. A residual life prediction system of a breaker operating mechanism is characterized by comprising a data acquisition module and a prediction module, wherein, The data acquisition module is used for acquiring sensor data of the breaker operating mechanism at the current moment; the prediction module is used for inputting the sensor data into a pre-trained life prediction model and obtaining a residual life prediction value of the breaker operating mechanism based on the output of the life prediction model; The life prediction model is obtained by training a preset basic model based on a preset basic model, the preset basic model is configured with a deep Bayesian neural network for modeling model uncertainty, an output layer of the deep Bayesian neural network is configured with an accidental node layer for modeling accidental uncertainty, and the life prediction model is obtained when the model uncertainty and the accidental uncertainty meet preset convergence conditions through historical sensor data of a breaker operating mechanism and corresponding residual life labels.
9. A terminal device comprising a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, the processor implementing the method of predicting remaining life of a circuit breaker operating mechanism according to any one of claims 1 to 7 when executing the computer program.
10. A computer readable storage medium, characterized in that the computer readable storage medium comprises a stored computer program, wherein the computer program, when run, controls a device in which the computer readable storage medium is located to perform the method of predicting the remaining life of a circuit breaker operating mechanism according to any one of claims 1 to 7.

Description

Method, system, equipment and medium for predicting residual life of breaker operating mechanism Technical Field The present invention relates to the field of equipment life prediction, and in particular, to a method, a system, an apparatus, and a medium for predicting a remaining life of a breaker operating mechanism. Background In a power system, a breaker operating mechanism is used as a core power component of a high-voltage breaker, and the prediction of the residual life of the breaker operating mechanism is a key for guaranteeing the safe and stable operation of the power system and making a scientific maintenance plan according to conditions. With the wide application of deep learning technology in the industrial field, the deep learning technology has great potential in the field of equipment life prediction by virtue of strong nonlinear data processing capability. At present, part of researches on the prediction of the residual life of the operating mechanism of the circuit breaker adopt a traditional deep learning model (such as GRU and Bi-LSTM) to analyze the state monitoring data of the operating mechanism, and the life prediction is realized by extracting the characteristics of current, travel and the like. However, the existing deep learning model multidimensional deterministic model can only output a single-point prediction result, cannot quantify uncertainty in the application process, and if the model is directly applied to the prediction of the residual life of a breaker operating mechanism and guidance of operation and maintenance decision, the problem of low accuracy of the prediction result exists. Disclosure of Invention The invention provides a method, a system, equipment and a medium for predicting the residual life of a circuit breaker operating mechanism, which are used for solving the technical problem of how to improve the prediction accuracy of the residual life of the circuit breaker operating mechanism. In order to solve the above technical problems, the present invention provides a method for predicting remaining life of a breaker operating mechanism, including: Collecting sensor data of a breaker operating mechanism at the current moment; Inputting the sensor data into a pre-trained life prediction model, and obtaining a residual life prediction value of the breaker operating mechanism based on the output of the life prediction model; The life prediction model is obtained by training a preset basic model based on a preset basic model, the preset basic model is configured with a deep Bayesian neural network for modeling model uncertainty, an output layer of the deep Bayesian neural network is configured with an accidental node layer for modeling accidental uncertainty, and the life prediction model is obtained when the model uncertainty and the accidental uncertainty meet preset convergence conditions through historical sensor data of a breaker operating mechanism and corresponding residual life labels. Preferably, the prediction distribution of the life prediction model is expressed as: ; Wherein, the For the purpose of the uncertainty of the model,In order to account for the occasional uncertainty,And X is a set of the historical sensor data, Y is a residual life label corresponding to the historical sensor data, X is a historical sensor data variable, Y is a residual life variable, and alpha and beta are two intermediate node parameters in the accidental node layer. Preferably, the training method of the life prediction model includes: Acquiring a basic model, historical sensor data and a residual life label; performing iterative optimization on parameterized distribution of the basic model by using the historical sensor data and the residual life labels; and obtaining a life prediction model when the relative relation between the parameterized distribution subjected to iterative optimization and the real posterior distribution meets a preset condition. Preferably, after the iterative optimization of the parameterized distribution of the base model, the residual life prediction method further includes: performing Monte Carlo sampling on the neural network weight distribution of the basic model; Sampling based on the parameter distribution of the Monte Carlo sampling result to obtain discrete prediction distribution; And calculating prediction expectations and prediction variances of the discrete prediction distribution, wherein the discrete prediction distribution, the prediction expectations and the prediction variances are used for quantifying the model uncertainty and the accidental uncertainty. Preferably, before the iterative optimization of the parameterized distribution of the base model, the method further includes: And constructing an objective function through an evidence lower bound algorithm, and determining the preset condition according to the objective function, wherein the objective function is constructed according to the expectation of a log likelihood