基于数据融合的中长期概率性负荷预测方法研究  

Research on medium and long-term probabilistic load forecasting method based on data fusion

作　　者：龙禹 阮文骏 刘梅 周雨奇 LONG Yu;RUAN Wenjun;LIU Mei;ZHOU Yuqi(State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing 210024,China;Beijing Tsingsoft Technology Co.,Ltd.,Beijing 100085,China;Marketing Service Center of State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing 210019,China)

机构地区：[1]国网江苏省电力有限公司,南京210024 [2]北京清软创新科技股份有限公司,北京100085 [3]国网江苏省电力有限公司营销服务中心,南京210019

出　　处：《电力需求侧管理》2024年第1期9-15,共7页Power Demand Side Management

基　　金：国家电网有限公司科技项目(5108-202218280A-2-0-XG)。

摘　　要：月度负荷预测是电力系统中长期运行和营销工作开展的基础,概率性电力负荷预测能够刻画中长期不确定性,更好地支撑新型电力系统负荷评估和调控策略制定。在此背景下,以系统负荷作为研究对象开展中长期概率性预测方法研究,提出了基于细粒度数据融合的中长期概率性预测方法。首先,根据影响因素建立小时级的多元线性回归模型对细粒度的负荷进行建模,再根据影响因素的不同预测值生成未来不同场景下的细粒度预测结果。其次,根据“自下而上”的时间层级协调策略,对每一个场景均进行月度聚合,生成不同层级地区的月度负荷预测结果,形成概率性预测结果。最后,以中国东部某区域及其下辖地区的负荷数据为例,验证了方法的有效性。Monthly load forecasting is the basis for medium and long-term operation of power system and development of marketing work,and probabilistic power load forecasting can portray medium and long-term uncertainty,and better support the new type of power system load assessment and regulation strategy development.In this context,the medium and long-term probabilistic forecasting method is stud⁃ied with the system load as the research object,and the medium and long-term probabilistic forecasting method based on fine-grained data fusion is proposed.Firstly,an hourly multiple linear regression model is established to model the fine-grained loads based on the influenc⁃ing factors,and then the fine-grained forecasts under different scenarios are generated based on the different predicted values of the influ⁃encing factors.Secondly,according to the“bottom-up”temporal hierarchy coordination strategy,monthly aggregation is performed for each scenario,and monthly load forecasts are generated for different hierarchical regions to form probabilistic forecasts.Finally,the effectiveness of the method is verified by taking load data of a region in eastern China and its subordinate areas as an example.

关 键 词：中长期负荷预测 细粒度 数据融合 概率性预测 

分 类 号：TM715[电气工程—电力系统及自动化]