融合日期类型的改进线性回归短期负荷预测模型  

Improved Multivariate Linear Regression Short-term Load Forecasting Model Integrating Date Types

作　　者：王凌谊 王志敏 钱纹 朱玥 顾洁[2] 彭虹桥 时亚军 WANG Lingyi;WANG Zhimin;QIAN Wen;ZHU Yue;GU Jie;PENG Hongqiao;SHI Yajun(Power Grid Planning and Construction Research Center of Yunnan Power Grid Co.,Ltd., Kunming, Yunnan 650011, China;Research Center for Big Data Engineering and Technologies, School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China)

机构地区：[1]云南电网有限责任公司电网规划建设研究中心,云南昆明650011 [2]大数据工程技术研究中心(上海交通大学电子信息与电气工程学院),上海200240

出　　处：《广东电力》2019年第5期43-50,共8页Guangdong Electric Power

基　　金：国家重点基础研究发展计划项目(2016YFB0900101)

摘　　要：为了提高短期电力负荷预测精度,综合分析气象、经济、节假日等影响因素,以K-means聚类方法定义节假日变量,并考虑日期类型与气象因素的交叉效应,提出融合日期类型与气象因素的多元线性回归短期负荷预测模型。该模型弥补了广义线性模型无法体现节假日因素影响的缺点,在一定程度上提高了预测精度。利用西南某省2013—2016年实际电力负荷数据对模型进行检验,结果表明:该模型通过对历史数据的深度挖掘以及对日期信息的合理利用,提高了短期负荷预测精度,拓展了多元线性回归模型在短期负荷预测中的适应性。To improve short-term load forecasting precision, this paper comprehensively analyzes influencing factors including weather, economy, holidays, and so on, defines holiday variables by using the K -means clustering method and proposes a multivariate linear regression short-term load forecasting model integrating date types and meteorological factors based on cross effect of date types and meteorological factors. This model has made up shortages of generalized linear models being unable to reflect influence of holidays and improved forecasting precision to a certain degree. According to actual load data from 2013 to 2016 of one southwestern province, it performs tests. The results indicate this model can improve short-term load forecasting precision by excavating historical data and reasonably using date information and expand adaptability of the multivariate linear regression model in short-term load forecasting.

关 键 词：负荷预测 线性回归模型 日期类型 气象因素 K-MEANS聚类 

分 类 号：TM715.1[电气工程—电力系统及自动化]