Distribution grid fault diagnosis based on SR-GRU and fault indicator

Efficient utilization of massive recorded data from fault indicators in distribution grid is beneficial to improve fault diagnosis. In this paper, the state detection and fault diagnosis of distribution grid based on fault indicator data were realized by combining high-dimension statistical analysis and artificial intelligence methods. Spectral residual algorithm was used to obtain saliency map of fault recorded data in rolling time windows, which can filter out useless background information. By training the distribution grid fault diagnosis model based on saliency map, a GRU network with high real-time and accuracy was obtained, and the overall accuracy of fault diagnosis reached 98%. The distribution grid fault diagnosis based on SR-GRU can not only make efficient use of massive fault indicator data, but also improve the universality and intelligence of the distribution grid fault diagnosis.