A fault location method for distribution networks Based on 3D-CNN and Multi-device data fusion

In response to the problems of weak anti-interference ability, low fault tolerance and large positioning error caused by the reliance on local node data in the existing fault location methods for multi-terminal transmission lines, a fault location method for distribution networks based on 3D convolutional neural network (CNN) and multi-device data fusion is proposed. This method first processes the travelling wave (TW) signals from multiple measurement points through time-frequency analysis to generate a time-frequency spectrogram, and then stacks it in space-time to construct a three-dimensional data fusion body; finally, a deep 3D-CNN model is used to learn and process the fusion body to achieve precise fault location. The simulation results show that the proposed positioning framework has excellent performance. The average relative positioning error is stable at below 1.5%, the recognition accuracy for high-resistance faults reaches 94%, and the absolute error is less than 60 m. Even in harsh conditions with strong noise interference (20 dB signal-to-noise ratio), this method can still maintain stable operational performance.