电力系统及其自动化学报2025，Vol.37Issue(9)：64-70,7.DOI:10.19635/j.cnki.csu-epsa.001547
基于注意力机制的CNN-BiGRU超短期省间现货购电需求预测
Ultra-short-term Inter-provincial Spot Electricity Purchase Demand Forecasting Based on Attention Mechanism with CNN-BiGRU
摘要
Abstract
To solve the problems in ensuring electricity supply in the inter-provincial spot market.,it is necessary for electricity marketing units to conduct in-depth research on electricity demand within the market and develop scientific purchasing strategies.In this paper,an ultra-short-term inter-provincial spot electricity purchase demand forecasting model is proposed,which combines convolutional neural networks,bidirectional gated recurrent unit(BiGRU)net-works and an Attention mechanism.First,the least absolute shrinkage and selection operator(Lasso)coefficient meth-od is used to select features that affect the inter-provincial spot electricity purchase demand.Second,the convolutional neural networks(CNN)are utilized to extract local features from the time series of inter-provincial spot electricity pur-chase demand,while the BiGRU is employed to capture the long-term dependencies in the time series.In addition,the Attention mechanism focuses on important time steps to improve the prediction accuracy.The results of a simulation ex-periment based on the actual measured data of inter-provincial spot electricity purchase demand show that the proposed model has a high accuracy in ultra-short-term inter-provincial spot electricity purchase demand forecasting,and it sig-nificantly outperforms single models and other combined forecasting models.关键词
电力现货市场/需求预测/双向门控循环单元网络/注意力机制/组合模型Key words
electricity spot market/demand forecasting/bidirectional gated recurrent unit(BiGRU)network/Atten-tion mechanism/combined model分类
信息技术与安全科学引用本文复制引用
杨世海,薛冰,李磊,周瑶..基于注意力机制的CNN-BiGRU超短期省间现货购电需求预测[J].电力系统及其自动化学报,2025,37(9):64-70,7.基金项目
国网江苏省电力有限公司科技项目(J2023172). （J2023172）
