Study On Fault Diagnosis Method Of High Voltage Circuit Breaker Based On Multi-Source Data

High voltage circuit breakers play an important role in control and protection in power system.With the rapid development of online monitoring technology,the fault monitoring and detection means of high voltage circuit breakers are gradually improved,and the types of monitoring signals are more diverse.Multi-source monitoring signals such as vibration signal,opening and closing coil current,moving contact travel and motor current are gradually applied to the fault diagnosis of high voltage circuit breakers.However,under the background of diversified monitoring signals,the problems of high voltage circuit breakers include an unbalanced number of samples,difficult adaptive extraction of fault feature and low reliability of single signal fault diagnosis,and inability to comprehensively use multiple types of monitoring signals.Therefore,it is of great significance to effectively use various monitoring signals,and study the fault diagnosis algorithm of high voltage circuit breakers suitable for different quantities and types of monitoring signals.This thesis further processes and analyzes various monitoring signals,and proposes a fault diagnosis method of high voltage circuit breakers based on multi-source data to help field operation and maintenance personnel quickly judge the fault type of high voltage circuit breakers.Firstly,through the data generation and data screening,this thesis solves the problem of sample imbalance.For a small number of samples,the time-frequency graph of the actual sample is used to train the Generative Adversarial Networks,so that it can generate the timefrequency graph similar to the actual sample.The synthetic sample can be obtained by inverse transformation of the time-frequency graph.For a large number of samples,the redundant data is screened through three stages: eliminating noise samples,retaining boundary samples and screening non-boundary samples,so as to reduce the number of samples on the basis of ensuring the reliability of fault diagnosis.Aiming at the problem that it is difficult to extract the fault feature of high voltage circuit breakers adaptively,the sample entropy is used to judge the complexity of monitoring signal.The monitoring signal with low sample entropy has low time domain complexity,and can be diagnosed only by time features.However,the monitoring signals with higher sample entropy have more complex characteristics in time domain and frequency domain,so the timefrequency characteristics need to be extracted.Further,the Stacked Autoencoder Network is used to mine the potential correlation between the time domain or time-frequency characteristics and the state of circuit breakers.Finally,the vibration signal and coil current are used to verify the method.The results show that the proposed method has an excellent diagnostic effect for different monitoring signals.Aiming at the low reliability of single signal fault diagnosis,this thesis considers the incomplete phenomenon of data loss,omission and fragmentary of monitoring signals,and uses multi-source monitoring signals as the input of fault diagnosis model.Through the multiinput convolution network,the feature of different monitoring signals are deeply mined and information fusion is carried out to complete the fault diagnosis;Finally,the incremental learning method is used to adjust the trained multi-input convolution network,so that it can diagnose the new fault types effectively.The experimental results show that the proposed method can reliably complete the fault diagnosis even if there are a large number of incomplete data.Finally,based on the above achievements,this thesis designs a high voltage circuit breaker diagnosis system which can adaptively extract various signal characteristics and improve the fault diagnosis model.Meanwhile,this thesis makes a detailed plan for its structure and function,hardware selection,display interface and so on.Theoretical research and example analysis show that the fault diagnosis method of high voltage circuit breakers based on multi-source data can be applied to high voltage circuit breakers with different quantities and types of monitoring signals,and the fault diagnosis effect is accurate and reliable,which can provide some guidance for on-site operation and maintenance personnel to make maintenance decisions.