基于分段多目标相似日选取法的短期负荷预测  

Short-term load forecasting based on piecewise similar selecting method with multiobjective

作　　者：王剑锋[1] 向铁元[1] 徐富祥[1] 王亮[1] 

机构地区：[1]武汉大学电气工程学院,湖北武汉430072

出　　处：《武汉大学学报（工学版）》2016年第3期435-440,共6页Engineering Journal of Wuhan University

基　　金：国家科技支撑计划项目(编号:2013BAA02B00)

摘　　要：提出了一种基于分段多目标相似日选取算法.该方法将负荷进行分段处理,将一天的负荷根据负荷波动规律分为5段,每段分别选取相似日,可以很好地克服选择的相似日只有部分相似的情况,并提出了虚拟相似日的概念.采用负荷曲线形状相似度最大和曲线差异度最小的多目标粒子群算法,可以保证选择的相似日的负荷曲线与预测日的负荷曲线在形状上和数量上的差别最小,从而可以提高负荷预测的精度,根据该算法得到的特征系数可以很好地判断出影响该段负荷变化的主导因素.将该相似日选择算法结合改进灰色预测算法应用到某地实际负荷预测中,结果表明该算法在相似日选择和负荷预测中均具有较高的精度.A piecewise similar selecting method with multiobjective is proposed. According to the laws of load fluctuation,the load in one day is divided into 5 segmentations. Different similar days are selected for every segmentation, which can overcome the disadvantages of selecting one similar day. And the concept of virtual similar day is proposed. The multiobjective particle swarm optimization algorithm, which is based on the maximum similar degrees of load curve in shape and the minimum different degrees of load curve in quantity,is used to improve the accuracy of load forecasting and judge the dominant factors of load change in every segmentation. The proposed similar day selecting algorithm and the improved grey prediction al- gorithm are applied to actual load forecasting. The results show that both in the selection of similar day and load forecasting, the proposed algorithms have higher accuracies.

关 键 词：短期负荷预测 相似日 多目标粒子群 灰色预测 

分 类 号：TM715[电气工程—电力系统及自动化]