A novel hybrid air quality early-warning system based on phase-space reconstruction and multi-objective optimization: A case study in China

Abstract With the haze pollution occurred frequently in recent years, establishing an effective air quality early-warning system has become the top priority. Nowadays, researchers have provided numerous methods for air quality early-warning. However, the majority of studies ignores the significance of data preprocessing and air quality evaluation while designing an early-warning system, which leads to poor forecasting performance and insufficient information. A novel hybrid air quality early-warning system that consists of three modules: data preprocessing, forecasting and air quality evaluation module is designed in this paper. Aiming at extract chaotic characteristics of raw data, a new hybrid data preprocessing strategy is firstly developed to construct a more stable series of pollutant data for forecasting. Then a multi-objective grasshopper optimization algorithm is adopted in order to enhance the forecasting capability of accuracy and stability in the forecasting module. Moreover, a fuzzy evaluation module of air quality is proposed to provide comprehensive results for the system. Through the eighteen data sets’ experimental process, the results and discussions indicate that not only the forecasting method achieves higher accuracy and stronger stability than other comparison models, but also the evaluation module provides sufficient air quality information, which forms a scientific guidance to decision-makers against air pollution.