LSTM对配电台区短期负荷预测的适用性研究  

Research on Applicability of LSTM to Short-term Load Forecasting in Distribution Station Area

作　　者：王繁 王果[1,2] 周子轩 乔智 牛晨 WANG Fan;WANG Guo;ZHOU Zixuan;QIAO Zhi;NIU Chen(College of Automaton and Electrical Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China;Key Laboratory of Opto-Technology and Intelligent Control of Ministry of Education,Lanzhou Jiaotong University,Lanzhou 730070,China;Rail Transit Electrical Automation Engineering Laboratory of Gansu Province,Lanzhou Jiaotong University,Lanzhou 730070,China)

机构地区：[1]兰州交通大学自动化与电气工程学院,兰州730070 [2]光电技术与智能控制教育部重点实验室(兰州交通大学),兰州730070 [3]甘肃省轨道交通电气自动化工程实验室(兰州交通大学),兰州730070

出　　处：《电力系统及其自动化学报》2021年第8期41-48,共8页Proceedings of the CSU-EPSA

基　　金：国家自然科学基金资助项目(51867012,51767013);兰州交通大学“百名青年优秀人才培养计划”基金资助项目;甘肃省科技计划资助项目(17JR5RA083)。

摘　　要：为研究长短期记忆LSTM(long-short termmemory)神经网络对不同类型配电台区短期负荷预测的适用性,以某市多个配电台区为对象,构建了LSTM短期负荷预测模型并进行适用性分析。采集各台区的负荷数据,通过K均值聚类算法、台区容量和用电类别对台区进行分类,标记并修正不良数据。考虑工作日和季节因素,采用LSTM建立配电台区负荷预测模型,分析不同类型台区的预测结果。研究结果表明,平均负荷和缺失值占比对预测精度影响较大,且LSTM更适用于平均负荷在40 kW以上的配电台区短期负荷预测,而对于平均负荷小于40 kW的配电台区的预测,效果随平均负荷的减小而降低。To study the applicability of long-short term memory(LSTM)neural network to the short-term load forecast⁃ing in different types of distribution station area,a short-term load forecasting model of LSTM is constructed with multi⁃ple distribution station areas in one city as objects and its applicability is analyzed.The load data in each station area is collected,and then the station areas are classified by the K-means clustering algorithm,station area capacity and pow⁃er consumption category.The bad data is marked and corrected,and the load forecasting model for distribution station area is established using LSTM,with the consideration of working days and seasonal factors.The prediction results of different types of station area are analyzed.Results show that average load and the ratio of missing value have a great in⁃fluence on the prediction accuracy.In addition,LSTM is more suitable for the short-term load prediction in distribution station areas with average load above 40 kW,while the prediction effect in those with average load less than 40 kW de⁃clines with the decrease in average load.

关 键 词：配电台区 短期负荷预测 长短期记忆神经网络 平均负荷 

分 类 号：TM714[电气工程—电力系统及自动化]