Detection of fault location in branching power distribution network using deep learning algorithm

An immediate response is desired when any failure on overhead power distribution systems has occurred, and the TDR(Time Domain Reflecting) method has been introduced to detect fault detection locations. However, accurate detection of the fault point is difficult due to waveform distortion and decrease in amplitude of TDR pulse in complex networks with multiple branches and electric power distribution equipment such as transformers and switchgear. A method for detecting fault points from TDR waveforms using deep learning is proposed in this study. The proposed method can be applied to detect fault locations and fault types in branching power distribution networks where multiple reflected waves are observed. Since the deep learning algorithm requires a large amount of waveform data, we developed a fast simulation method to create the data. To simulate the circuit rapidly, the power distribution line was treated as a transmission line, thereby deriving the fundamental matrix of the transmission line. Additionally, the equivalent circuit model of the power distribution network was represented by cascading the fundamental matrix. The TDR waveform data was obtained rapidly by calculating the equivalent circuit using MATLAB. We used this to create many TDR waveform data of an overhead distribution network model with multiple branches and performed fault locations detection using a deep learning algorithm. As a result, it was shown that the location of accident was identified with 96.8% accuracy.