基于LSTM的变电站电气设备运行状态自动化预测方法  

Automatic Operation Status Prediction Method of Substation Electrical Equipment Based on LSTM

作　　者：张懿 吴艳 Zhang Yi;Wu Yan(State Grid Zhejiang Electric Power Company Hangzhou Power Supply Company,Hangzhou,China;China United Engineering Co.,Ltd.,Hangzhou,China)

机构地区：[1]国网浙江省电力公司杭州供电公司,浙江杭州 [2]中国联合工程有限公司,浙江杭州

出　　处：《科学技术创新》2025年第2期39-42,共4页Scientific and Technological Innovation

摘　　要：针对变电站电气设备运行状态传统监测方法的不足,本文提出了一种基于长短期记忆(LSTM)网络的变电站电气设备运行状态自动化预测方法,利用LSTM网络处理时间序列数据的优势,实现了对设备运行状态的高精度预测。研究中,首先通过高斯滤波器对采集的声波信号进行去噪处理,随后采用小波包分解技术提取信号特征,构建特征向量,进一步结合卷积神经网络(CNN)和LSTM构建混合模型,优化特征提取和时间序列预测过程。实验结果表明,所提方法在平均误差幅度和预测时间上均优于数字孪生技术(DTSE)和自适应概率神经网络(APNN)方法,验证了该方法在提高电力系统预测准确性和运维效率方面的潜力。In view of the shortcomings of traditional monitoring methods of substation electrical equipment operating status,this paper proposes an automatic prediction method of substation electrical equipment operating status based on long short-term memory(LSTM)network.By taking advantage of LSTM network to process time series data,the equipment operating status can be predicted with high precision.In the research,the acquired acoustic signal is de-noised by Gaussian filter first,and then the signal features are extracted by wavelet packet decomposition technology and the feature vector is constructed.The mixed model is further constructed by combining convolutional neural network(CNN)and LSTM to optimize the process of feature extraction and time series prediction.The experimental results show that the proposed method outperforms the digital twin technique(DTSE)and the adaptive probabilistic neural network(APNN)method in terms of average error margin and prediction time,which verifies the potential of the proposed method in improving the prediction accuracy and operation and maintenance efficiency of power system.

关 键 词：长短期记忆网络 变电站电气设备 信号去噪 小波包分解 状态预测 

分 类 号：TM63[电气工程—电力系统及自动化]