The Circuit Breaker Fault Diagnosis And Condition Assessment Based On Vibration Signal

Circuit breakers are the key equipments for power grid operation,and play an important role in ensuring the safe operation of the power grid.Once a circuit breaker fails,the losses caused are far beyond its own value.At present,mechanical faults account for a relatively large number of circuit breaker faults.Increasing the accuracy of circuit breaker mechanical fault diagnosis is of great significance for improving the reliability of circuit breakers and the safety of power grid operation.The vibration signal generated during the circuit breaker breaking process,and it contains abundant mechanical status information,which can facilitate the non-invasive evaluation of the circuit breaker and provide effective evidence for the circuit breaker condition monitoring.This paper introduced the main types of circuit breakers,compared the characteristics of different types of circuit breakers,selected ZN28-type circuit breaker as the experimental object,combined with the analysis of the circuit breaker operating flow,collected vibration signals with CA-YD-103 acceleration sensor,and preliminarily analysised the vibration signal in the vibration events.The change of the mechanical state of the circuit breaker can be characterized by the vibration signal,but how to extract the features hidden in the signal requires effective signal processing means.This paper compares a variety of feature extraction methods,and one single signal processing method has the deficiency of one-sided.A combined feature extraction method is used to obtain the composite features of the vibration signal and enhance the specificity of the features.In an example test,it is verified that the composite features can help to improve the diagnostic accuracy.In order to improve the quality of features and solve the problem of the subjective and blindness issues of primary features,weights of the extracted features are evaluated using the method of the largest decline of the Gini index.After analyzing the correlation between features and classification,the mutual information is used.The feature set was de-redundant,and according to the combined effects of features,a sequential back search method was used to reduce the size of the feature set.Through the test of multiple sets of classical sample sets and examples,it was proved that The characteristic dimension reduction method adopted in this paper can effectively reduce the computational complexity while maintaining or even improving the diagnostic accuracy.Finally,this paper selects random forests to classify and test the four states of the circuit breaker.The test results show that the accuracy and ease of operation of the random forest are obviously better than those of the support vector machine.This paper presents a new idea for circuit breaker fault diagnosis,status assessment and risk pre-control by analyzing the circuit breaker vibration signal feature extraction method,feature quality improvement method and pattern recognition method,and has wide applicability.