An enhanced interval PM2.5 concentration forecasting model based on BEMD and MLPI with influencing factors. (15th February 2020)

• Record Type: Journal Article
• Authors: Wang, Zicheng; Chen, Liren; Ding, Zhenni; Chen, Huayou
• Is Part Of: Atmospheric environment. Volume 223(2020)
• Publication Date: 2020-02-15
• Subjects: Bivariate empirical mode decomposition -- Interval forecasting -- PM2.5 concentration -- Interval multilayer perceptron -- Mode reconstruction; Air -- Pollution -- Periodicals; Air -- Pollution -- Meteorological aspects -- Periodicals; 551.51
• DOI: 10.1016/j.atmosenv.2019.117200
• Languages: English
• ISSNs: 1352-2310

Abstract

Abstract: In order to protect public health by providing an early warning of harmful air pollutants, various forecasting models are proposed to forecast the average values of daily pollutant concentrations. In fact, even on the same day, the concentration of pollutants will fluctuate greatly during different time periods, point-based models can not reflect the variability well. Thus, an enhanced interval PM2.5 concentration forecasting model is developed in this paper, which is based on interval decomposition ensemble and considering influencing factors. For the purpose of obtaining main influencing factors, interval grey incidence analysis (IGIA) is used to select input variables for model. The interval-valued time series (ITS) of PM2.5 concentration and its influencing factors are decomposed into a finite number of complex-valued intrinsic mode functions (IMFs) and one complex-valued residual by bivariate empirical mode decomposition (BEMD) algorithm. Considering the different amounts of various IMFs, the complex-valued IMFs and residual are clustered into fewer classes by reconstruction technique. Then, interval multilayer perceptron (MLP I ) is employed to fit the lower and upper bound simultaneously of all classes to obtain the corresponding forecasting results, which are combined to generate the aggregated interval-valued output by a simple addition approach. The model is tested by the dataset collected from three environmental monitoring stations in Beijing, China. Experimental results show that the enhanced model outperforms other considered models by means of forecasting accuracy and stability. Highlights: Model is applied to interval-valued PM2.5 concentration forecasting. BEMD and MLP I model the lower and upper bound of ITS simultaneously. A new interval-valued reconstruction technique is proposed. The proposed model takes into account influencing factors. The enhanced model improves accuracy and stability.