长三角电力消费量特征提取及预测

Feature Extraction and Prediction of Electricity Consumption in the Yangtze River Delta

首先构建小波重构—支持向量机混合模型,将各个特征数据集分离出来,并应用支持向量机分别进行预测。选用结构时间序列模型进行对比研究。其次,应用长三角月度电力消费量数据进行实证分析。最后,使用ERR和MAPE检验两个模型的预测性能。误差结果证明,所构建的混合模型误差相对较小,预测效果较好。

This paper constructs a hybrid model of wavelet reconstruction and support vector machine to extract each feature data set,and predicts the electricity consumption by support vector machine. And then,the structural time series model is employed to make the comparative analysis,because the structural time series model can predict the components of seasonal,cyclical,trend and irregular factors directly. And then,it makes the empirical analysis based on the Yangtze River Delta monthly electricity consumption data. Finally,it test the predictive performance of the two models by ERR and MAPE. The error results show that the proposed hybrid model has the relatively small error and the better prediction performance.