基于变分模态分解与空洞卷积神经网络的配电网故障选线方法

小电流接地系统发生单相接地故障时，零序电流故障特征微弱且繁杂多变，传统选线方法可靠性有待提高。提出了一种基于变分模态分解（variational mode decomposition， VMD）与空洞卷积神经网络的配电网故障选线方法。首先，分析配电网健全线路和故障线路的电气特征，采用零序电流作为故障特征信号，为选线模型的输入量提供理论依据；其次，通过变分模态分解把零序电流序列分成不同频率的固有模态函数，提高故障信号特征的平稳性和差异性；然后，采用空洞卷积神经网络作为选线网络，以增大卷积操作感受野的方式增强模型的自适应分类能力；最后，在MATLAB/Simulink中构建10 kV配电网进行算例分析，结果表明，该方法在不同故障场景条件下均有较高的选线效果，验证了所提方法的鲁棒性与准确性。

A Method of Fault Line Selection for Distribution Networks Based on Variational Mode Decomposition and Dilated Convolutional Neural Network

When a single-phase grounding fault occurs in the low-current grounding system， the fault characteristic of the zero-sequence current is weak and highly complex， therefore the reliability of the traditional line selection method is badly in need of improvement. To address this issue， a fault line selection method for distribution networks based on variational mode decomposition （VMD） and dilated convolutional neural network is proposed in this paper. Firstly， the electrical characteristics of healthy lines and faulty lines are analyzed， and zero-sequence current is used as fault characteristic signal to provide theoretical basis for the input of the line selection model. Secondly， the zero-sequence current sequence is decomposed into different frequency intrinsic mode functions （IMF） using the VMD technique to enhance the stability and distinctiveness of the fault signal features. Secondly， dilated convolutional neural network is used as line selection network to improve the adaptive classification ability of the model by enlarging the receptive field of convolutional operation. Finally， a 10 kV distribution network is constructed in MATLAB/Simulink for the case study analysis. The results suggest that the proposed method yields favorable conductor selection outcomes across various fault scenarios， thus verifying the robustness and accuracy of the proposed approach.