基于regARIMA模型的月度负荷预测效果研究

Study on the Monthly Power Load Forecasting Performance Based on regARIMA Model

摘要: 为探究离群值对月度负荷预测效果的影响，建立计及离群值影响的季节性ARIMA月度负荷预测模型（regARIMA），选择1999—2017年北京、甘肃等5省（市）的实际月度负荷数据，对预测效果进行比较研究。结果表明，与普通ARIMA模型相比，考虑了离群值影响的regARIMA模型的3年样本内平均预测误差得到明显改善；应用regARIMA模型进行提前12期的样本外预测，预测精度获得不同程度的提升。

Abstract: In order to explore the impact of outliers on the monthly power load forecasting performance, a seasonal ARIMA model considering the impact of outliers (regARIMA) is established. The actual monthly power load data series of 5 provinces recorded from January 1999 to December 2017 are used to verify the accuracy of power load forecasting. The empirical results show that the forecasting error of the regARIMA model considering the outliers impact is significantly reduced within samples for last 3 years. The forecasting accuracy of the regARIMA out of samples for 12 steps ahead is also improved to some extent.