Active Distribution Network Fault Identification Method Based on ABLSTM

Active distribution network mostly adopts neutral point non Earthing system, so when short circuit fault occurs, the fault characteristics are weak, and traditional over-current relay is difficult to detect, which also causes great difficulties for fault identification. This article proposes an LSTM fault recognition algorithm called Attention Based LSTM (ABLSTM) with embedded attention mechanism. To verify the performance of the algorithm, it first establishes an IEEE119 node active distribution network simulation model with three distributed power sources in MATLAB/SIMULINK. Then, fault data of different feeder faults are obtained at the measurement points through simulation. Finally, the obtained data is used as input for fault identification using ABLSTM. The results indicate that the classification accuracy of the algorithm is higher than other methods.