基于WOA-AM-GRU的短期电力负荷预测方法的研究  

Research on Short-term Electric Load Forecasting Method Based on WOA-AM-GRU

作　　者：韩明冲 韩杰[2] 姜超 苏本勇 韩冰 HAN Mingchong;HAN jie;JIANG Chao;SU Benyong;HAN Bing(State Grid Caoxian Electric Power Supply Company,Heze 274400,China;Beijing Normal University,Beijing 100875,China;State Grid Jining Electric Power Supply Company,Jining 272008,China)

机构地区：[1]国网山东省电力公司曹县供电公司,山东菏泽274400 [2]北京师范大学,北京100875 [3]国网山东省电力公司济宁供电公司,山东济宁272008

出　　处：《山东电力技术》2024年第12期27-33,共7页Shandong Electric Power

基　　金：恩施州科技计划项目“台区计量及安防治理大数据分析”(D20180017)。

摘　　要：当代社会发展迅速,电力系统愈加复杂,为确保电力系统经济安全高效可靠运行,短期电力负荷预测方面的研究必不可少。针对短期电力负荷数据具有的复杂性、多样性和一定规律性的特点,提出一种基于鲸鱼优化算法-注意力机制-门控逻辑单元(whale optimization algorithm-attention mechanism-gated recurrent unit,WOA-AM-GRU)模型的短期电力负荷预测模型。首先,门控逻辑单元(gated recurrent unit,GRU)可挖掘历史负荷数据间的隐藏规律,为实现负荷预测打下基础;其次,模型融合注意力机制,可发现数据间隐藏的关联程度,得到特征信息的贡献比重,以此对数据进行加权处理,凸显影响大的特征值,提高模型运行效率;再次,针对注意力机制-门控逻辑单元(attention mechanism-gated recurrent unit,AM-GRU)模型超参数选择困难问题,利用鲸鱼优化算法对其进行参数寻优,提高模型的预测精度。最后,构建基于反向传播(back propagation,BP)神经网络、长短期记忆神经网络(long short-term memory,LSTM)、GRU、AM-GRU和WOA-AM-GRU的短期负荷预测模型,结合实例进行对比试验分析,结果证实了基于WOA-AM-GRU的短期电力负荷预测模型的优越性。The rapid development of contemporary society has led to increasingly complex electricity systems.To ensure the economic security,efficiency,and reliability of electricity systems,research in short-term electricity load forecasting is indispensable.Considering the complexity,diversity,and certain regularities inherent in short-term electricity load data,we propose a short-term electricity load forecasting model based on the whale optimization algorithm-attention mechanism-gated recurrent unit(WOA-AM-GRU)model.Firstly,the gated recurrent unit(GRU)can excavate the hidden laws among the historical load data,laying the foundation for realizing load forecasting.Secondly,the model integrates the attention mechanism,which can discover the degree of hidden correlation among the data and get the contribution proportion of the feature information so as to weight the data,highlight the feature values with large influence,and improve the model operation efficiency.Furthermore,for the problem of difficult selection of hyperparameters of the attention mechanism-gated recurrent unit(AM-GRU)model,the whale optimization algorithm is used to optimize its parameters and improve the prediction accuracy of the model.Finally,the short-term load forecasting model based on back propagation(BP),long short-term memory(LSTM),GRU,AM-GRU,and WOA-AM-GRU is constructed,and the comparative experimental analysis is carried out in combination with examples,and the results confirm the superiority of the short-term power load forecasting model based on WOA-AM-GRU.

关 键 词：短期电力负荷预测 门控循环单元 注意力机制 鲸鱼优化算法 WOA-AM-GRU 

分 类 号：TM732[电气工程—电力系统及自动化]