摘要
针对智能电网条件下用户异常用电行为问题,提出了一种基于主成分分析和深度循环神经网络(PCA-RNN)的异常用电行为检测方法。该方法首先利用核主成分分析对电力负荷数据进行降维处理,生成主成分特征子集,然后基于长短记忆网络(LSTM)和门控循环单元(GRU)构建深度循环神经网络(RNN)模型,检测异常用电行为。实验结果表明,该方法能够有效检测出异常用电行为,且具有较高的准确率和鲁棒性。
In order to solve the problem of abnormal power consumption behavior of users in smart grid,a detection method of abnormal power consumption behavior based on principal component analysis and deep cyclic neural network(PCA-RNN)is proposed.Firstly,kernel principal component analysis is used to reduce the dimension of power load data,and principal component feature subset is generated.Based on long short memory network(LSTM)and gating cycle unit(GRU),a deep cycle neural network(RNN)model is constructed to detect abnormal power consumption behavior.The experiment results show that the method can effectively detect abnormal electrical behavior,and has high accuracy and robustness.
作者
赵玉谦
赵彩霞
张倚天
ZHAO Yu-qian;ZHAO Cai-xia;ZHANG Yi-tian(State Grid Henan Skills Training Center,Zhengzhou 450051,China)
出处
《信息技术》
2021年第8期127-132,共6页
Information Technology
关键词
智能电网
电力数据
异常行为检测
主成分分析
深度循环神经网络
smart grid
power data
abnormal behavior detection
principal component analysis
deep cyclic neural network
