Dual Similarity-based Method for Proactive Fault Detection in Power Distribution Networks

The fault detection of power distribution networks is of paramount importance to ensure power supply safety and reliability as well as efficient system maintenance. However, the inadequacy of measurements and complex operational conditions bring about difficulties for conventional logic-based and model-based fault detection methods. This paper proposed a dual similarity-based method (DSM) for power distribution fault detection. It contains the operation mode similarity and temporal similarity. The former reconstructs the operation mode by matrix profile based on k-means, and Euclidean distance is used to measure the similarity of the operating modes. The latter models the temporal similarity, and the LSTM is adopted to take the advantage of high-resolution continuous measurement for fault detection. Through the dual similarity-based method, the proposed solution can effectively identify the typical anomalies in the power distribution network with high accuracy and recall performance. The proposed solution is extensively assessed through experiments for a range of tripping events in 75 distribution transformers based on realistic datasets. The numerical results confirmed that it outperforms the conventional fault detection solutions.