摘要
GM(1,1)模型是中长期负荷预测的一种常用方法。为了解决残差GM(1,1)模型在残差预测值符号判断中存在的问题,提出一种基于朴素贝叶斯法的改进残差GM(1,1)模型。该模型根据历史负荷增长率区间,对负荷增长状态进行划分,统计各个状态下残差正负号出现的个数和各状态前一年份残差正负号个数,然后利用朴素贝叶斯法建立分类器,判断残差预测值符号。将改进模型应用于某县用电量预测中,算例结果表明改进模型能够有效地提高中长期负荷预测的精度,具有一定的实用价值。
GM(1,1) model is a kind of common method for medium and long term load forecasting. In order to solve problems of the residual error GM(1,1) model in judging the symbol of residual error predicted value,this paper presents a kind of improved residual error GM(1,1) model based on nave Bayes method. According to historical load growth rate interval,this model can divide load growth states and count numbers of residual error sign under different states as well as numbers of residual error sign a year ago of each state. Then it applies nave Bayes method in building the classifier and judging the symbol of residual error predicted value. The improved model was applied in electricity consumption forecasting in a county. Example result indicates that this improved model can effectively improve precision of medium and long term load forecasting which means certain practical value.
出处
《广东电力》
2017年第9期81-85,共5页
Guangdong Electric Power
