Fault diagnosis of power cables based on wavelet transform and CNN-informer

Abstract

Accurate identification of early-stage cable faults is a necessary prerequisite for timely elimination of fault hazards. This article proposes a method for power cable fault diagnosis using wavelet transform and CNN-Informer, which can accurately identify early-stage cable faults from overcurrent signals caused by constant impedance faults, capacitor switching, and excitation inrush currents. The method extracts features from the overcurrent signals using wavelet transform and then constructs a CNN-Informer network fusion model. The model is trained by adjusting the network parameters to establish the mapping relationship between input features and class encoding. Simulation results show that the proposed method can effectively classify overcurrent signals and accurately identify early-stage cable faults, demonstrating high practical value in engineering applications. Furthermore, comparative simulation experiments with similar methods conclude that the proposed method in this article achieves higher accuracy and precision.