Art fault detection method based on variational mode decomposition energy entropy hybrid time domain feature and random forest
DONG Zhiwen1,2, SU Jingjing1,3
1. College of Computer and Control Engineering, Minjiang University, Fuzhou 350118; 2. School of Electronic, Electrical Engineering and Physics, Fujian University of Technology, Fuzhou 350108; 3. Zhejiang Institute of Mechanical & Electrical Engineering Co., Ltd, Hangzhou 310051
Abstract：In household appliances, nonlinear load devices continue to grow, which makes the limitations of traditional fault current detection methods based on "zero crossing" characteristics become evident. This paper introduces a novel approach for fault arc identification, combining signal time domain features with variational mode decomposition (VMD) intrinsic modal energy entropy. The analysis focuses on line current, initially extracting time-frequency features. Subsequently, the VMD algorithm decomposes the fault arc current into intrinsic mode function (IMF), and their energy entropy is computed. By leveraging both time-domain and energy entropy features, a multidimensional feature vector is employed for fault arc identification within a random forest model. Comparative experiments indicate that this method achieves an impressive fault arc identification accuracy of up to 99% and is suitable for diverse low-voltage distribution scenarios including various load types and nonlinear loads.
董志文, 苏晶晶. 基于变分模态分解能量熵混合时域特征和随机森林的故障电弧检测方法[J]. 电气技术, 2024, 25(1): 1-7.
DONG Zhiwen, SU Jingjing. Art fault detection method based on variational mode decomposition energy entropy hybrid time domain feature and random forest. Electrical Engineering, 2024, 25(1): 1-7.
2024, Vol. 25 
