GIS Fault Rate Prediction Method Based on Convolutional Neural Network

GIS equipment is critical to the stable operation of the power system, through calculating the fault probability taking into consideration of the insulation defect category, better risk evaluation in respect of equipment status and system operation can be realized. Currently, research on partial discharge data primarily focuses on diagnosis of fault type after data processing, on this basis and through utilizing the on-site measured data of substation, this paper proposes a GIS equipment fault rate calculation method based on convolutional neural network (CNN). This method can give effective evaluation of the seriousness of the equipment defect, and has high engineering application and robustness. Using design defect classification module and fault binary classification module and through utilizing multi-layer convolutional network, calculation of the fault probability of each type of defect can be realized. Through comparing with other classification models, the probability calculation result of the model designed by this paper has high accuracy.