Newly-Designed DC-Arc Fault Detection Strategy Via EMD Signal Analysis and GRU Neural Network for Solar PV Power Generation System

In this study, an intelligent arc-fault detection algorithm for solar photovoltaic (PV) power generation systems is investigated based on the empirical mode decomposition (EMD) and the gate recurrent unit neural network (GRU-NN). The proposed algorithm uses the EMD to extract valuable information from current signals and then sequences the statistical indexes of each mode from the EMD according to modal orders. Moreover, the GRU-NN captures the features and variation trends among different modes and eventually detects the dc-arc fault. Through experimental verifications, the average detection accuracy is over 98.7%, with an 8ms execution time under all examined conditions. In addition, an online updating method is also introduced in this study to ensure the adaptability of the proposed algorithm. Combined with this online-updating method, the proposed EMD-GRU-NN scheme could quickly modify the model and ensure the accuracy of the arc fault identification, even for different PV stations. The online updating performance will also be verified by experiments in Taiwan and China PV stations to improve over 25% detection accuracy to previous research.