基于长短期记忆的实时电价条件下智能电网短期负荷预测  

Short-Term Load Forecasting of Smart Grid Based on Long-Short-Term Memory Recurrent Neural Networks in Condition of Real-Time Electricity Price

作　　者：李鹏[1] 何帅 韩鹏飞 郑苗苗 黄敏 孙健[2] LI Peng;HE Shuai;HAN Pengfei;ZHENG Miaomiao;HUANG Min;SUN Jian(School of Electrical and Electronic Engineering,North China Electric Power University,Baoding071003,Hebei Province,China;State Grid Beijing Electric Power Company,Xicheng District,Beijing100031,China)

机构地区：[1]华北电力大学电气与电子工程学院,河北省保定市071003 [2]国网北京市电力公司,北京市西城区100031

出　　处：《电网技术》2018年第12期4045-4052,共8页Power System Technology

基　　金：国家自然科学基金项目(51577068);国家电网公司科技项目(520201150012)~~

摘　　要：在电力市场改革与智能电网建设的大背景下,电力将逐渐回归商品属性,电价也将实时波动,并对负荷产生影响。通过分析得出电价与负荷具有相关性,因此在预测模型中考虑了实时电价的影响,并对考虑实时电价的负荷预测模型与价格型需求侧响应之间的关系进行了讨论。针对前馈型神经网络不能处理序列间关联信息与传统循环神经网络无法记忆久远关键信息的缺陷,提出了基于长短期记忆循环神经网络的负荷预测模型,使用自适应矩估计算法进行深度学习。最后通过美国某地区的实际负荷和电价数据,验证了所提模型具有更高的预测精度。In background of electricity market reform and smart grid construction,electricity will gradually return to commodity properties,and electricity price will also fluctuate in real time and have impact on load.In this paper,it is concluded that the electricity price and load are relevant, therefore,the impact of real-time electricity price is considered in forecasting model,and relationship between the load forecasting model considering real-time electricity price and price based demand side response is discussed.The feed forward neural network cannot deal with correlation information between sequences and traditional recurrent neural network cannot memorize long-term critical information.In view of this problem,a load forecasting model based on long-short-term memory(LSTM)is proposed and deep neural network is trained with adaptive moment estimation(Adam). With the actual load and electricity price data of a certain area in the United States,it is verified that the model proposed in this paper has higher accuracy.

关 键 词：负荷预测 长短期记忆 实时电价 需求侧响应 深度学习 

分 类 号：TM715[电气工程—电力系统及自动化]