Identification Method of Fault Hidden Danger of Transmission Line for Reclose by Mountain Fire Based on ARMA Model Fitting

Abstract

The transmission line fault caused by mountain fire is a fault state between transient fault and permanent fault. For the reclosing after the transmission line fault caused by mountain fire, the randomness and nonlinearity of the transition impedance often lead to the reclosing failure, and even lead to the burning of the circuit breaker. If the potential fault of transmission line fault can be found as soon as possible, the circuit breaker can be controlled to trip to avoid damage. In this paper, a feature extraction method based on autoregressive moving average model(ARMA) fitting is proposed. The characteristic of fast and smooth signal of ARMA is used to select the appropriate order for the half cycle voltage signal of the transmission line to perform coefficient fitting,and to complete the feature extraction of the potential fault signal. The feature vectors are input into support vector machine (SVM) and K-nearest neighbor classifier (KNN) respectively for identification potential fault, and it is found that SVM has higher identification accuracy. Compared with other sequential signal identification algorithms with similar identification accuracy, the complexity of this algorithm is greatly reduced. The algorithm is verified based on the existing recording voltage wave of reclosing failure by mountain fire fault. The calculation results show that the recall rate, precision rate and F1 index are 98.5%, 98.4% and 98.4% respectively by combining the feature engineering composed of ARMA model fitting coefficient with support vector machine classifier.