摘要
针对分布式电力交易系统中存在对海量数据分析能力差的问题,提出了一种基于深度学习的电能消纳供需预测模型。构建了分布式电力交易系统的技术和功能架构,并提出了相应的电价预测算法,采用经验模态分解(EMD)算法对历史电价、供电和用电数据进行预处理。在传统长短期记忆网络(LSTM)的基础上引入注意力(Attention)机制对输入数据进行特征提取,实现日前电价的预测。通过对贵州省电力交易中心的历史数据进行的仿真测试结果表明,相较于传统LSTM算法,文中所提算法预测得到的电价曲线更贴近实际曲线,预测准确性更高。在实际应用中,算法稳定性良好且应用效果理想,能够辅助电力交易主体进行决策,提高其市场的竞争力。
Aiming at the problem of poor ability to analyze massive data in distributed power trading system,a power consumption supply and demand prediction model based on deep learning is proposed.The technical and functional architecture of distributed power trading system is constructed,and the corresponding price prediction algorithm is proposed.The Empirical Mode Decomposition(EMD)algorithm is used to preprocess the historical price,power supply and consumption data.Based on the traditional Long Short⁃Term Memory(LSTM),Attention mechanism is introduced to extract the features of the input data to realize the prediction of day ahead electricity price.Through the simulation test of the historical data of Guizhou Power Trading Center,the results show that compared with the traditional LSTM algorithm,the price curve predicted by the proposed algorithm is closer to the actual curve and the prediction accuracy is higher.In practical application,the algorithm has good stability and ideal application effect.It can assist the power trading subject in decision⁃making and improve its market competitiveness.
作者
代江
田年杰
朱思霖
姜有泉
赵倩
卢孟林
DAI Jiang;TIAN Nianjie;ZHU Silin;JIANG Youquan;ZHAO Qian;LU Menglin(Power Dispatching Control Center,Guizhou Power Grid Co.,Ltd.,Guiyang 550002,China)
出处
《电子设计工程》
2023年第9期63-67,共5页
Electronic Design Engineering
基金
贵州电网公司科技项目(066500KK52190008)。
关键词
电价预测
深度学习
长短期记忆网络
注意力
electricity price forecast
deep learning
Long Short⁃Term Memory
attention
