基于X-12-ARIMA季节分解与年度电量校正的月度电量预测

Abstract

月度电量预测是电力计划部门安排运行计划与制定购售电计划的基础.提出一种综合考虑多种经济因素的月度电量预测方法.首先,采用X-12-ARIMA模型对月度电量和多种经济因素进行季节分解,并利用逐步回归分析研究各经济量与用电量的关联关系和回归模型,获得初步预测结果;然后,利用多项式拟合进行年度电量预测,并对已有月度电量预测结果进行调整;最后,采用自回归积分滑动平均模型(autoregressive integrated moving average model, ARIMA)对受气象与节假日因素影响较大的月份进行分季节预测修正,获得精度良好的月度电量预测模型.该文采用广东省2009年3月至2014年4月的经济数据与电量数据对2014年5月至2015年4月的电量数据进行预测.预测结果的平均预测精度为97.78％,验证了预测模型的有效性.

Monthly electricity forecast is the basis for the planning department of power grid corporation to arrange the operation and purchase plan. This paper presents a method for forecasting monthly electricity consumption considering various economic factors. First, we adopt the X-12-ARIMA model to decompose the economic data and monthly electricity data into seasonal part, and adopt stepwise regression analysis to study the correlation and regression model between the economic factors and electricity consumption and get the primary forecast results. Then, we use polynomial fitting to forecast the annual electricity, and use the result of annual electricity to adjust the monthly electricity consumption. Finally, we adopt autoregressive integrated moving average (ARIMA) model to correct the seasonal forecast for the months which are obviously influenced by meteorological factors and holidays, and obtain the monthly electricity forecasting model with good precision. In this paper, monthly electricity consumption from May 2014 to April 2015 in Guangdong Province are forecasted by economic data and monthly electricity consumption from March 2009 to April 2014 to obtain the average prediction accuracy of 97. 78%, which verifies the effectiveness of the prediction model.