Low voltage series arc fault identification method based on VMD and RVFL neural network

Abstract

Abstract In low-voltage distribution system, series arc fault is one of the main causes of fire accidents. The series arc current characteristics of load are similar to its normal working current, and the ordinary current protection device cannot detect it effectively. In this paper, an arc recognition method based on variational modal decomposition (VMD) and random vector functional link (RVFL) neural network is proposed. The series arc current data of different loads are collected by the low-voltage series arc experimental platform, and the arc current characteristics are analyzed. Four different intrinsic mode functions (IMF) are obtained by variational modal decomposition, and their fuzzy entropy is calculated respectively to form an eigenvector data set, which is input into RVFL neural network for training and identification. In this paper, approximate entropy (ApEn), sample entropy (SampEn) and fuzzy entropy (FuzzyEn) are extracted as feature vectors for each IMF of VMD, their discrimination for different loads is analyzed, and the recognition accuracy of RVFL neural network compared with SVM and BP neural network algorithm is compared. It is verified that the arc recognition method proposed in this paper has better recognition effect.