Abstract:

To solve the problem of the randomness and instability of fault arc signal,a DC series arc fault detection method based on sparrow search algorithm extreme learning machine (SSA-ELM) is proposed.Firstly,the experimental platform is built according to UL 1699B standard for data collection.Secondly,the DC arc fault signal is decomposed into the intrinsic mode components (IMF) by the complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN).On the basis of the correlation coefficient and energy distribution of each IMF,the sample entropy as the feature quantity is extracted.Finally,SSA-ELM is used to learn the feature quantity,and the model is used for DC series arc fault detection.The experimental results show that CEEMDAN method is not sensitive to signal interference,SSA-ELM has a fast learning speed and is more accurate and sensitive for DC series arc fault detection and recognition.

Key words: DC series arc, empirical mode decomposition, arc fault, sparrow search algorithm(SSA), extreme learning machine(ELM)