改进型DNN的配电网接地故障选线方法  

Improved DNN method for grounding fault line selection in distribution network

作　　者：姜学文 罗冰 李豪 刘伟 李奔 陈亮 Jiang Xuewen;Luo Bing;Li Hao;Liu Wei;Li Ben;Chen Liang(Ezhou Power Supply Company,State Grid Hubei Electric Power Company,Ezhou 436000,Hubei,China)

机构地区：[1]国网湖北鄂州市供电公司,湖北鄂州436000

出　　处：《电测与仪表》2023年第9期145-150,共6页Electrical Measurement & Instrumentation

基　　金：国家电网公司科技项目(5215F02000DA)。

摘　　要：针对配电网单相接地故障特征信息不清晰且现有选线方法易受故障条件和环境噪声的影响,根据配电网暂态故障特征和稳态故障特征,提出了一种改进的深度神经网络用于故障选线。对深度学习网络的损失函数和学习率进行优化,进一步提高选线的效率和准确性。通过仿真验证了该方法的可行性。结果表明,与改进前相比,改进后的训练迭代次数由86次降低到30次,训练效率提高了65.12%,故障判断的准确率由95%提高到99%,具有较好的抗干扰能力,有一定的参考价值。Aiming at the unclear characteristic information of single-phase grounding fault in distribution network and the existing line selection methods are easy to be affected by fault conditions and environmental noise,according to the transient fault characteristics and steady-state fault characteristics of distribution network,an improved deep neural network is proposed for fault line selection.The loss function and learning rate of the deep learning network are optimized to further improve the efficiency and accuracy of line selection.The feasibility of the proposed method is verified by simulation.The results show that the number of training iterations after the improvement is reduced from 86 times to 30 times,the training efficiency is improved by 65.12%,and the accuracy rate of fault judgment is improved from 95%to 99%,which has good anti-interference ability and has certain reference value.

关 键 词：配电网 单相接地故障 故障选线 深度神经网络 暂态故障特征 稳态故障特征 

分 类 号：TM711[电气工程—电力系统及自动化]