基于CNN-LSTM算法的分布式电力系统故障识别模型研究  

Research on Fault Identification Model of Distributed Power System Based on CNN-LSTM Algorithm

作　　者：徐茂林 王云涛 祁婧一 XU Maolin;WANG Yuntao;QI Jingyi(Mengcun Hui Autonomous County Power Supply Branch,State Grid Hebei Electric Power Co.,Ltd.,Cangzhou,Hebei Province,061400 China;Ultra High Voltage Branch of State Grid Jibei Electric Power Co.,Ltd.,Beijing,102488 China)

机构地区：[1]国网河北电力有限公司孟村回族自治县供电分公司,河北沧州061400 [2]国网冀北电力有限公司超高压分公司,北京102488

出　　处：《科技资讯》2025年第4期102-104,共3页Science & Technology Information

摘　　要：随着现代电力系统的复杂性和规模不断增加,分布式电力系统中故障识别与继电保护的难度也逐步提升。鉴于此,提出了一种基于卷积神经网络、长短期记忆网络、通道自注意力机制的故障识别与继电保护模型,利用其对分布式电力系统中的故障类型进行精确识别,以达到继电保护的作用。测试结果表明,信噪比为2 dB时,改进模型的检测准确率为0.96,相较于传统卷积神经网络提高了31.51%。结果表明,该模型在不同信噪比下的识别准确率显著优于传统方法,为分布式电力系统的安全运行提供了技术支持。With the increasing complexity and scale of modern power systems,the difficulty of fault identification and relay protection in distributed power systems is gradually increasing.In view of this,a fault recognition and relay protection model based on convolutional neural network,long short-term memory network,and channel self attention mechanism is proposed to accurately identify fault types in distributed power systems to achieve relay protection.The test results show that when the signal-to-noise ratio is 2dB,the detection accuracy of the improved model is 0.96,which is 31.51%higher than that of traditional convolutional neural networks.The results show that the model significantly outperforms traditional methods in recognition accuracy under different signal-to-noise ratios,providing technical support for the safe operation of distributed power systems.

关 键 词：卷积神经网络 长短期记忆网络 电力系统 故障识别 继电保护 

分 类 号：TM726[电气工程—电力系统及自动化]