摘要
为了提高发电厂继电保护监测的准确率,提出融合卷积神经网络(CNN)与长短期记忆(LSTM)网络的发电厂继电保护自动监测方法。卷积神经网络用于综合处理故障样本中的抽象特征,并通过卷积核提取局部特征,补全缺失数据,提高数据完整性。长短期记忆网络通过遗忘门、隐藏层和传输门处理削减后的数据,建立时序模型,捕捉输入数据与故障特征的依赖关系。实验结果显示,该方法的监测准确率高达96.5%以上。
features in fault samples,extract local features through convolutional kernels,complete missing data,and improve data integrity.The LSTM network processes the reduced data through forget gates,hidden layers,and transmission gates,establishes a temporal model,and captures the dependency relationship between input data and fault features.The experimental results show that the monitoring accuracy of this method is over 96.5%.
作者
魏卿
王天阔
于龙飞
王迪扬
黄佳
江兴旺
邵玲
WEI Qing;WANG Tiankuo;YU Longfei;WANG Diyang;HUANG Jia;JIANG Xingwang;SHAO Ling(Huadian Electric Power Research Institute Co.,Ltd.,Hangzhou,Zhejiang 310030,China;Mamaya Branch of Guizhou Beipanjiang Electric Power Co.,Ltd.,Anshun,Guizhou 561301,China)
出处
《自动化应用》
2025年第16期60-62,共3页
Automation Application
