基于改进人工免疫网络的配电网单相接地故障辨识方法

配电网故障辨识有利于线路的全面分析与后续检修,但过渡电阻较高时单相接地故障稳态特征不明显,辨识正确率受到影响.提出了一种基于改进人工免疫网络的单相接地故障辨识方法,利用希尔伯特-黄变换提取故障暂态过程中零序电压的固有模态分量,得到反映波形变化程度的频域特征,进而构造特征向量;采用记忆细胞方差自适应调整环节对人工免疫网络进行改进,以生成能够体现原有特征且均匀分布的记忆细胞来进行故障判断,可以克服训练数据样本少、分布不均匀的问题,并且网络训练速度快.在MATLAB/Simulink中建立仿真模型,测试集的验证结果表明本文辨识方法在小样本训练的情况下能对故障与正常扰动进行识别,训练时间短,不需要人为设置阈值;同时,在中性点接地方式变化、信号存在噪声以及系统拓扑结构变化时,该方法适应性依然良好.

Single Phase Earth Fault Identification Method in Distribution Network Based on Improved Artificial Immune Network

The fault identification of distribution network is beneficial to the comprehensive analysis and maintenance, but the steady-state characteristics of single-phase earth fault are not obvious when the transition resistance is high, and the identification accuracy is affected. An improved artificial immune network based method for single phase earth fault is proposed. After taking disturbance into account and analyzing the fault transient process, the mode components of zero sequence voltage were extracted by Hilbert Huang transform, and the frequency domain characteristics reflecting the waveform were obtained, and the identification eigenvector was constructed. Using artificial immune network to generate memory cells with original characteristics and uniform distribution for fault type judgment can overcome small training data, and increase the adaptive adjustment of memory cell variance to improve the training speed. The simulation model is established in MATLAB/SIMULINK, and the results of test set show that the identification method can identify fault and normal disturbance in the case of small sample training, and the training time is short, and there is no need to set the threshold manually. At the same time, the adaptability of the method is still good when the neutral grounding mode, signal noise and system topology change.