A fault classification method for distribution lines based on convolutional neural network

The structure, equipment environment and operating conditions of the distribution network are becoming more and more complex, and the difficulty of fault diagnosis and location in active distribution network is also increasing. Distribution network fault detection and classification can help uncover potential safety hazards and prevent grid failures from causing damage to electricity consumers and equipment. In this paper, we propose a fault classification method based on convolutional neural networks (CNN) model based on time series imaging. In this paper, a 10KV distribution network model with different photovoltaic integration locations and grid fault events is built in EMTDC/PSCAD environment to obtain large number of fault data of the distribution network, such as ground faults, short-circuit faults, lightning strikes, etc. In this method, the three-phase voltage and current signals of the time series are encoded into two-dimensional images using the Gramm Angle Field (GAF), and then the convolutional neural network model is used for training, and at last the classification performance of the method is verified according to various data. The results show that the proposed method has higher accuracy than other algorithms in fault classification of new distribution networks with PV access, and has better robustness in the presence of bad data.