基于线圈电流相轨迹-XGBoost的高压断路器故障诊断方法OA

To address issues in high-voltage circuit breaker fault identification,such as relatively sin-gular feature extraction and diagnostic algorithms'dependency on parameter selection,a fault diagnosis method based on coil current phase trajectory and XGBoost is proposed.Firstly,the Lyapunov exponent of the opening/closing coil current is analyzed,revealing the chaotic variation characteristics of the coil current when a circuit breaker fault occurs.The delay time parameter for reconstruction is optimized using the av-erage mutual information calculation method,followed by phase space reconstruction of the current signal.Subsequently,the fault features are extracted based on the phase space reconstruction trajectory.A feature vector is formed,comprising the maximum abscissa value and maximum ordinate value of the coil current phase trajectory,the Euclidean distance from internal turning points to the origin,and the origin moment.This vector serves as the input for the XGBoost identification model for training and fault recognition,yielding accurate diagnostic results.Finally,comparative analyses with the peak-valley features,global features,and models such as SVM,KNN,RF and BP demonstrate the superiority of the phase trajectory-XGBoost fault identification method based on coil current signals for high-voltage circuit breaker fault di-agnosis.