基于误差校正的能源消费总量预测方法

Prediction method of total energy consumption based on error adjustment

鉴于能源消费量具有趋势性、非平稳性等特点,而差分自回归移动平均模型(ARIMA)只能反映能源消费量的部分信息,预测结果不太理想。为了提高预测精度,提出了基于二次误差校正ARIMA模型的能源消费预测方法:首先采用ARIMA模型对能源消费总量进行初步预测,然后构建偏最小二乘回归支持向量机模型(PLS-SVM)对残差序列数据中未被解释的部分进行分析和拟合,并对未来的残差进行预测。最后利用所得残差预测值对能源消费总量预测值进行校正。对福建省1978—2012年的能源消费总量数据进行仿真,实验结果表明,与ARIMA等方法相比,本文提出的方法获得了较好的预测结果,是一种有效的能源消费量预测方法。

Autoregressive integrated moving average model（ ARIMA） can only reflect part of the information of energy consumption with the characteristics development of trends, non-stabilities and unsatisfactory predictions. In order to improve the prediction accuracy,the prediction method of energy consumption is proposed based on the twice error adjustment ARIMA model. Firstly,ARIMA model has a preliminary forecast on total energy consumption. Then partial least squares regression support vector model（ PLS-SVM） is built for unexplained residual sequence data analysis and fitting of parts,to predict the residual error of the future. Finally,total energy consumption will be corrected through the prediction value obtained by the residual error. The total energy consumption data simulation results from 1978 to 2012 of Fujian Province show that the proposed method obtains a better prediction. Comparing with other usual prediction methods such as ARIMA,we find it is an effective prediction method of energy consumption.