基于分形特性修正气象相似日的节假日短期负荷预测方法  

Holiday Short-Term Load Forecasting Based on Fractal Characteristic Modified Meteorological Similar Day

作　　者：李滨[1] 黄佳[1] 吴茵[2] 覃芳璐 

机构地区：[1]广西电力系统最优化与节能技术重点实验室(广西大学),广西壮族自治区南宁市530004 [2]广西电网公司调度控制中心,广西壮族自治区南宁市530023

出　　处：《电网技术》2017年第6期1949-1955,共7页Power System Technology

基　　金：国家自然科学基金项目(51407036);国家重点基础研究发展计划项目(973项目)(2013CB228205)~~

摘　　要：节假日负荷易受气象信息、国家调休政策等影响,预测精度较低。为解决上述问题,提出了基于分形特性修正气象相似日的节假日短期负荷预测方法。采用归一化处理和日期适当调整,解决了数据值之间的差异性、经济增长率和负荷变化趋势不一致的问题;将分形特性的自相似性引入节假日短期负荷预测,剔除气象突变带来的不良影响,在海量历史样本集中精准确定相似日查找范围;依据相似性原理,综合考虑气象、日类型等影响因素,建立曲线辨析函数计算负荷的差异系数,在特定范围内查找与待预测节假日气象最相似的一天。以南方某电网数据为实际算例进行仿真,结果表明所提出方法满足工程实际的需求,其中2015年春节期间96点日负荷准确度达97.63%。Easily affected by meteorological information and national policies, daily load forecasting accuracy in holidays is low and vulnerable in assessment. To solve this problem, this paper proposes a method based on fractal characteristic modified meteorological similar day to forecast holiday short-term load. Normalization and date adjustments are adapted to solve differences among data values and inconsistency between economic growth rate and load variation trend. Self-similarity of fractal characteristics is introduced to holiday short-term load forecast to remove adverse effects of unsettled meteorology, so that search range of similar days can be focused on precisely＇ among massive historical samples. According to similarity principle, factors such as meteorology and type are taken into comprehensive consideration. Curve discrimination function is established to calculate load difference coefficient, and the day with its weather most similar to that of festivals to be predicted can be found within specific range. Taking data of Southern Power Grid as a practical example, simulation results show that the proposed method meets practical needs. Moreover, load curve accuracy is up to 97.63% during Spring Festival in 2015.

关 键 词：节假日短期负荷预测 气象相似日 分形自相似性 

分 类 号：TM715[电气工程—电力系统及自动化]