Accurate Fault Location in Active Distribution Network Based on Continuous Wavelet Transform and Deep Learning

Some existing traditional traveling wave distribution network fault location methods are overly dependent on the local wave head information of time domain or frequency domain signals, and the access of distributed power sources will affect the fault traveling wave. This paper describes the difficulty of extracting single-phase grounding fault features in distribution networks, analyzes the problem of low accuracy of existing fault location methods, and proposes a method for accurately locating fault points based on continuous wavelet transform and LSTM. This method obtains the fault voltage of each feeder, first decouples the fault voltage waveform, obtains the line mode component, and then uses continuous wavelet transform to obtain the corresponding CWT coefficient matrix, and then generates an RGB image from the coefficient matrix, uses the LSTM layer to extract the corresponding relationship between the time-frequency features of the mapped RGB image and the fault distance, and finally outputs the fault distance through the SoftMax layer mapping.