A monthly electricity consumption forecasting method based on vector error correction model and self-adaptive screening method

Economic growth has greatly fluctuated around the world in recent years, and external economic factors (EEFs) have imposed more obvious effects on electricity consumption. To improve the accuracy and applicability of mid-term, especially monthly, electricity consumption forecasting, a novel monthly electricity consumption forecasting framework (denoted as SAS-SVECM for short) based on vector error correction model (VECM) and self-adaptive screening (SAS) method is proposed in this paper, which fully explores and integrates the potential impacts from and relationships between EEFs. The SAS-SVECM firstly implements X-12-ARIMA to extract seasonal peaks from the electricity consumption and EEF time series. Second, a VECM is used to address correlations and time lag effects between electricity consumption and EEFs. And a SAS method is proposed to identify the most possible influential EEF self-adaptively, which appropriately addresses the contradiction between data quantity and data length. The SAS-SVECM achieves significant forecasting accuracy enhancement and good adaptability. Finally, an empirical example, using real monthly electricity consumption and macroeconomic data of China (2000-2014), was studied to verify the effectiveness of SAS-SVECM.