Identification Method of Low Voltage Load Switch Card Based on EMWPE-Relief-LSTM

Low-voltage load switches are difficult to filter because of their high dimensionality and noise, which results in low accuracy of fault identification. In order to solve this problem, this paper proposes a fault recognition method based on the combination of empirical multi-scale weighted permutation entropy (EMWPE) and LSTM network. The time series data of sensor output is decomposed into intrinsic modal components (IMFs) by set empirical modal decomposition (EEMD). Based on the kurtosis and variance criteria, the suitable IMFs are selected, and the multi-scale weighted permutation entropy of these IMFs is calculated as eigenvector. Relief algorithm is used to reduce the dimensionality of fault feature vectors, and the reduced feature vectors are input into the LSTM classifier to output the eigenmodal components of the fault. The experimental results show that the method can effectively identify the stuck faults of low-voltage load switches with an accuracy rate of 99.3%, which provides a strong guarantee for the reliability and security of low-voltage distribution system.