摘要
短期负荷预测是电力系统安全经济运行的基础。由于负荷变化受众多因素影响,选择合适的变量集对于提高预测性能至关重要。针对数据驱动型与预测误差驱动型变量选择方法各自的特点,以及传统变量选择方法在相关性度量指标与选择策略上存在的问题,该文提出基于正交化最大信息系数、特征协同与随机森林的变量选择方法。该方法将数据驱动与预测误差驱动进行两阶段融合,前者作为变量预筛选阶段,后者完成变量精选,实现选择质量与计算复杂度的平衡;选择过程中综合考虑变量间的相关度、冗余度与协同度,能有效提高短期负荷预测的性能;通过算例从选择的变量集、预测误差大小、预测误差稳定性等方面验证该方法相对于传统短期负荷预测变量选择方法的优势。
Short-term load forecasting is the basis for the safe and economic operation of power system.Since electrical load is affected by many factors,choosing proper variable set is critical to improving forecasting performance.Owing to the characteristics of data driven variable selection(VS)method and forecast error driven VS method and the problems of traditional VS methods in correlation metrics and selection strategies,this paper proposed VS of orthogonal maximal information coefficient,feature interaction and random forest(OMICFI-RFVS).The method combines data driven method and forecast error driven method in two stages.The former was used as the variable pre-screening stage,and the latter completes the further VS.The method achieves the balance between selection quality and computational complexity.The method comprehensively considers the relevancy,redundancy and synergy between variables,which can effectively improve the forecasting performance.The advantages of OMICFI-RFVS were verified by cases from the selected variable set,forecast error and forecast error stability.
作者
郑睿程
顾洁
金之俭
彭虹桥
蔡珑
ZHENG Ruicheng;GU Jie;JIN Zhijian;PENG Hongqiao;CAI Long(Research Center for Big Data Engineering and Technologies(Department of Electronic Information and Electrical Engineering,Shanghai Jiao Tong University),Minhang District,Shanghai 200240,China)
出处
《中国电机工程学报》
EI
CSCD
北大核心
2020年第2期487-500,共14页
Proceedings of the CSEE
基金
国家重点研发计划项目(2016YFB0900100)
上海市科委重大项目(18DZ1100303).
关键词
短期负荷预测
变量选择
数据驱动
预测误差驱动
最大信息系数
short-term load forecasting
variable selection
data driven
forecast error driven
maximal information coefficient
作者简介
郑睿程(1995),男,硕士研究生,主要研究方向为电力系统负荷预测、数据挖掘与优化运行等,zhengrcchn@foxmail.com;通信作者:顾洁(1971),女,副教授,硕士生导师,研究方向为电力市场、电力系统规划等,gujie@sjtu.edu.cn。
