An adaptive decomposition and ensemble model for short-term air pollutant concentration forecast using ICEEMDAN-ICA

Precise short-term atmospheric pollutant concentration forecasting is significant for providing early warning information against harmful pollutants. Many studies on pollutant concentration prediction have proven the excellence of decomposition and ensemble models. However, in most of those studies, the training and test sets are divided based on the decomposition results rather than the original time series. In such decomposition and ensemble framework, future information is used for prediction, which is impractical. Furthermore, a significant boundary effect in the decomposition results is also a serious problem. Thus, this study develops an adaptive forecasting scheme aiming at ensuring the model practicality and adapting to the boundary effect. This study also introduces independent component analysis (ICA) to help extract the hidden information of the original series and improves the ability to screen influential variables. Finally, an adaptive decomposition and ensemble model combined with ICA is developed. Using data collected from Beijing Shunyi station, a case study and two comparative experiments are conducted, through which the contribution of the methods used in the proposed model and the superior performance of the model are demonstrated.

Automated summary

This study highlights the importance of precise short-term atmospheric pollutant concentration forecasting and addresses the boundary effect in decomposition results. By proposing an adaptive forecasting scheme and introducing ICA, the study develops an adaptive decomposition and ensemble model that demonstrates superior performance in predicting pollutant concentrations.