High Impedance Arc Fault Detection in Distribution Networks Based on Instantaneous Parameters Characteristic in Time-frequency Domain

With the development of the scale and intelligence level, the requirement for rapid and accuracy fault detection is highly increasing. High impedance arc grounding fault (HIAF) detection has always been one of the difficulties in distribution network fault detection due to the randomness and nonlinear characteristic represented in fault current and voltage. This paper proposed a novel approach for HIAF detection via features extracted in time-frequency domain, which can avoid the problem that common-used features are easily affected by voltage level and current level. To obtain the significant features in time-frequency, Variational Mode Decomposition (VMD) has been applied to decompose fault signal to Intrinsic Mode Function (IMF) with mutually independent central frequency bands. Meanwhile, Hilbert transform (HT) has been used to analyze the instantaneous frequency and phase, and the frequency and phase characteristics of typical components are used as features for identifying high resistance grounding. To realize rapidly and accuracy arc detection, a long- and short-term memory (LSTM) based method has been used to classify HIAF and other faults. Though the actual HIAF data from distribution network, performance of the method has been verified.