基于潜在需求响应特征优化的短期负荷预测  

Short-Term Load Forecasting Based on the Optimization of Latent Demand Response Features

作　　者：田毅 黄冬梅 孙锦中 TIAN Yi;HUANG Dongmei;SUN Jinzhong(School of Electrical Engineering,Shanghai University of Electric Power,Shanghai 200090,China;School of Electronics and Information Engineering,Shanghai University of Electric Power,Shanghai 200090,China)

机构地区：[1]上海电力大学电气工程学院,上海200090 [2]上海电力大学电子与信息工程学院,上海200090

出　　处：《上海电力大学学报》2025年第2期106-111,共6页Journal of Shanghai University of Electric Power

摘　　要：为提升负荷峰值预测的准确率,并优化需求响应(DR)在负荷预测中的效果,提出了一种基于潜在DR特征优化的预测模型。首先,在分时电价下建立基于消费者心理学的DR模型,通过负荷转移率得出潜在DR信号。其次,在考虑负荷时序性的基础上,对量化的DR信号进行聚类分析,挖掘DR信号的内在联系,同时进行相似序列筛选,提取相似序列并划分电价,提升电价划分的有效性。最后,以滑动窗口方法划分数据并输入长短期记忆网络,进行预测值修正。算例表明,在考虑DR特征的基础上提取相似序列,能充分发挥DR的作用,提升预测精度。To improve the accuracy of peak load forecasting and optimize the effectiveness of demand response(DR)in load forecasting,a prediction model that optimizes latent DR features is proposed in this paper.Firstly,under time-of-use pricing,a DR model based on consumer psychology is established to derive potential DR signals through load shifting rates.Secondly,considering the temporal characteristics of load,the quantified DR signals are subjected to cluster analysis to uncover the intrinsic relationships of the DR signals.Simultaneously,similar day sequences are selected to extract similar sequences and segment the pricing,enhancing the effectiveness of price segmentation.Finally,the data is divided using a sliding window approach and input into a long short-term memory network for prediction and correction.Case studies indicate that extracting similar sequences based on DR features can better utilize the role of DR,thereby improving forecasting accuracy.

关 键 词：短期负荷预测 需求响应 聚类划分 相似序列 动态电价 预测值修正 

分 类 号：TM715[电气工程—电力系统及自动化]