基于XGBoost和QRLSTM的超短期负荷预测方法  

Ultra-Short-Term Load Forecasting Method Based on XGBoost and QRLSTM

作　　者：张照贝 顾春华 温蜜 ZHANG Zhao-bei;GU Chun-hua;WEN Mi(College of Computer Science and Technology,Shanghai University of Electric Power,Shanghai 200090,China;University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海电力大学计算机科学与技术学院,上海200090 [2]上海理工大学,上海200093

出　　处：《计算机仿真》2022年第1期90-95,110,共7页Computer Simulation

基　　金：国家自然科学基金(61872230,U1936213,61702321)。

摘　　要：针对目前超短期电力负荷预测存在特征挖掘不足和难以准确反映电力负荷不确定性信息的问题,提出基于XGBoost和QRLSTM的超短期负荷预测方法。首先采用XGBoost算法挖掘重要特征并生成点负荷预测结果,将二者结合作为概率预测方法的输入特征;其次使用LSTM与弹球损失构造QRLSTM概率预测方法;最后通过核密度估计方法获取电力负荷的概率密度曲线。采用新西兰公共电力负荷数据集进行仿真,结果表明提出的方法不仅可挖掘重要特征,而且更加准确反映电力负荷的不确定性信息。Aiming at the problems of insufficient feature mining in current ultra-short-term power load forecasting and difficulty in accurately reflecting the uncertainty information of power load, an ultra-short-term load forecasting method based on XGBoost and QRLSTM is proposed. First, the XGBoost algorithm was used to mine important features and generate point load prediction results, and the two were combined as the input features of the probabilistic prediction method;Secondly, the QRLSTM probability prediction method was constructed by LSTM and pinball loss;Fnally, the probability Density curve of power load was obtained by the kernel density estimation method. Using New Zealand public power load data set for experimental simulation, the results show that the proposed method can not only mine important features, but also reflect the uncertainty information of power load more accurately.

关 键 词：超短期电力负荷预测 特征挖掘 长短期记忆网络 核密度估计 

分 类 号：TM715[电气工程—电力系统及自动化]