基于改进变分模态分解与确定学习的单相接地故障早期诊断研究OA

To address the limitations of traditional threshold-based methods in diagnosing single-phase ground faults in distribution networks,specifically their reliance on manual experience and inadequate noise immunity,this paper proposes an adaptive threshold diagnosis method based on improved variational mode decomposition and deterministic learning.First,the osprey optimization algorithm optimizes the variational mode decomposition parameters to decompose the zero-sequence voltage signal.Significant intrinsic mode functions(IMFs)are selected based on the Pearson correlation coefficient between each IMF and the original signal,and noise reduction is achieved through signal reconstruction.Second,leveraging deterministic learning theory,local modeling and identification of fault dynamics are performed to extract dynamic trajectories encapsulating fault characteristics.By leveraging the morphological mutation characteristics of this trajectory before and after the fault,an adaptive detection threshold is constructed to rapidly capture the onset of the fault.PSCAD/EMTDC simulation and 10 kV distribution network true test data verification show that the proposed method can accurately identify the fault moment under complex working conditions and provide a reliable criterion for subsequent fault line selection and section positioning.