Detection Method of Series Arc Fault in Three-Phase Frequency Converter Load Circuit Under Unknown Working Conditions
H. Gao, Kunyuan Wang, Zhiyong Wang, Jiacheng Cai, Yuze Lv
Abstract
Accurately detecting the series arc fault (SAF) in power lines and realizing the protection of the circuit are of great significance for effectively preventing electrical fire accidents. However, existing research results are difficult to effectively detect SAF that occurred in three-phase frequency converter load circuit under unknown working conditions. To solve the above problems, a new approach for SAF detection based on real-time training, updating prediction models and predicting residuals was proposed in this paper. First, a prediction model based on dung beetle optimization algorithm optimized extreme learning machine was proposed. Second, SAF experiments in three-phase frequency converter load circuit were carried out under commercial power supply condition. The model was trained and updated by using two cycle current signals, so as to predict the next cycle current signal. The SAF detection was realized by combining the prediction of residual mean square error and a threshold. Finally, the detection performance of the proposed method was tested under three unknown working conditions, which are different power harmonics, operating parameters of the frequency converter, and operating current, respectively. The results indicated that the proposed method can effectively detect the SAF occurred in the frequency converter load circuit under three unknown working conditions.