Mechanical defect diagnosis of high voltage circuit breakers based on the combination of stroke curve and current signal

In practical engineering, the mechanical state of circuit breaker is different and the mechanical characteristic curve is similar, so it is easy to be misdiagnosed using single signal to diagnose the defects of high voltage circuit breakers. In view of this problem, this paper proposes a defect diagnosis method based on the combination of stroke curve current signal of high voltage circuit breakers optimized by random forest feature optimization. Firstly, the characteristic values of stroke curve and current signal are extracted according to the mechanical characteristic signals under different states. Secondly, redundant features are eliminated by random forest algorithm combined with support vector machine model to determine the optimal feature subset database. Finally, grey wolf optimization-support vector machine model is used to further improve the comprehensive performance of circuit breaker defect diagnosis. The diagnostic results show that compared with the single signal diagnosis method, the accuracy of the combined diagnosis method can be improved by up to 22.5%. In addition, the defect diagnosis accuracy of the high voltage circuit breaker combined with the line signal can be further improved by 2.5% through feature selection and support vector machine model optimization.