基于逐步回归和SVR方法的上海夏季日最大电力负荷的模拟研究

Simulation Study on the Summer Daily Maximum Electric Load in Shanghai Based on Stepwise Regression Analysis and SVR Method

利用20132017年上海夏季(69月)逐日最大电力负荷和同期气象资料,对上海夏季的气温、炎热累积效应等气象因子与用电负荷进行相关分析,筛选建模自变量。基于逐步回归及SVR支持向量回归方法,分别建立了夏季日最大电力负荷预测模型并进行模拟误差对比。其中,以气象要素为自变量构建了逐步回归模型“方案1”,在方案1基础上加入前一天气象负荷构建了逐步回归模型“方案2”,并利用拓展后的特征值构建了SVR支持向量回归模型。对比两种方案的逐步回归模型和SVR支持向量回归模型的模拟结果可知,SVR支持向量回归模型的模拟误差最小,为3.3%,逐步回归模型的模拟误差分别为4.5%(方案1)和3.8%(方案2);从逐月的模拟效果来看,6月和9月逐步回归模型的模拟效果均优于SVR回归模型的模拟结果,而7、8月SVR支持向量回归模型的模拟效果又明显优于逐步回归模型的模拟结果,表明SVR方法能很好地模拟上海夏季高温条件下的极端负荷,在今后的实际负荷预测业务中可综合利用两类模型开展工作。

Using the daily maximum electric power load and meteorological data in Shanghai during the summer(JuneSeptember)of 20132017,this paper analyzes the correlations between the maximum electricity load and various meteorological factors,such as temperature and heat accumulation effect in summer in Shanghai,and screens independent variables for modeling.In addition,based on the stepwise regression and SVR support vector regression methods,the summer daily maximum electric power load forecasting models are established,and the simulation errors are compared.Among them,the stepwise regression model“Scheme 1”is constructed with meteorological elements as independent variables,and the stepwise regression model“Scheme 2”is constructed by adding the meteorological load of the previous day on the basis of Scheme 1,and the SVR support vector regression model is constructed using the expanded eigenvalues.Comparing the simulation results of the stepwise regression model and the SVR support vector regression model of the two schemes,we can see that the simulation error(3.3%)of the SVR method is the smallest,while the simulation errors of the stepwise regression model are 4.5%(Scheme 1)and 3.8%(Scheme 2),respectively.Judged from the monthly simulation results,the simulation of the stepwise regression model for June and September are better than the SVR regression model,but the simulation results of the SVR method for July and August are significantly prior to the results of stepwise regression model.These results show that the SVR method can well simulate the extreme load under the summer high temperature conditions in Shanghai,and the two types of models can be comprehensively used in the actual load forecasting operation in the future.