能源互联网终端用户异常负荷数据辨识与修正  

Identification and correction of abnormal load data of energyInternet terminal users

作　　者：张凯 郭威 魏新杰 杨小龙 罗欣 ZHANG Kai;GUO Wei;WEI Xinjie;YANG Xiaolong;LUO Xin(State Grid Hebei Electric Power Co.,Ltd.,Shijiazhuang 050022,China;Marketing Service Center,State Grid Hebei Electric Power Co.,Ltd.,Shijiazhuang 050022,China;Information and Communication Branch,State Grid Hebei Electric Power Co.,Ltd.,Shijiazhuang 050022,China;Beijing Qingsoft Innovation Technology Co.,Ltd.,Beijing 100080,China)

机构地区：[1]国网河北省电力有限公司,石家庄050022 [2]国网河北省电力有限公司营销服务中心,石家庄050022 [3]国网河北省电力有限公司信息通信分公司,石家庄050022 [4]北京清软创新科技股份有限公司,北京100080

出　　处：《电测与仪表》2024年第11期61-67,共7页Electrical Measurement & Instrumentation

基　　金：国家电网有限公司总部科技项目(1300-2020 13387A-0-0-00)。

摘　　要：随着能源互联网的快速发展,终端用户侧用能数据呈现爆炸式增长。采集到的海量数据因自身设备故障或者外部环境因素影响会出现大量的异常负荷数据。文章提出了基于PSO-BiLSTM神经网络的能源互联网异常负荷数据检测与修正方法。所提方法首先通过大量正常负荷数据训练双向LSTM模型,并选择粒子群(particle swarm optimization, PSO)优化算法对预测模型的参数进行寻优,将经过参数调优的双向LSTM模型用于负荷预测。基于负荷预测结果,采用误差分析和异常值判定准则来检测负荷曲线中的异常负荷数据,最后将检测出的异常负荷数据应用预测结果对其进行修正。实验证明,所提方法具有较好的异常负荷数据检测效果,且易于训练,异常负荷数据检测错误率较低。With the rapid development of energy Internet,the energy consumption data of terminal users presents explosive growth.A large amount of abnormal load data may occur in the collected massive data due to device faults or external environmental factors.This paper presents a method for detecting and correcting abnormal load data of energy Internet based on PSO-BiLSTM neural network.Firstly,a large amount of normal load data is used to train the bidirectional LSTM model,and PSO optimization algorithm is selected to optimize the parameters of the prediction model,and the optimized bidirectional LSTM model is used for load prediction.Secondly,based on the load prediction results,the abnormal load data in the load curve are detected by error analysis and outlier judgment criterion.Finally,the abnormal load data detected are corrected by the prediction results.Experimental results show that the proposed method has good effect on abnormal load data detection,and it is easy to train,and the error rate of abnormal load data detection is low.

关 键 词：能源互联网 异常负荷检测 异常负荷修正 双向LSTM 误差分析 

分 类 号：TM714[电气工程—电力系统及自动化]