An air pollution early-warning system with the mechanism of dynamic evaluation

Due to the increasing energy consumption, the issue of air pollution would continue to attach attention. Although technological progress has improved the relative efficiency of energy, the task of air pollution monitoring cannot be ignored. The existing air pollution early-warning methods based on univariate time series forecasting suffer from multiple limitations. To take measures to monitor air quality status effectively, an air pollution early-warning framework is built based on statistical comprehensive evaluation and multiscale forecasting. It can settle the problems, such as accuracy improvement bottlenecks and diminishing spatial comparability. Through analyzing pollutant characteristics, the original time series is transformed into fuzzy sets, through which practical air quality levels and pollution condition ranking would be obtained. Only when the air quality classes meet the varying patterns of multivariable time series does the improvement of the forecasting accuracy make sense for air pollution dynamic monitoring. So, a multiscale forecasting method-oriented data characteristic identification is introduced to enhance the performance of the forecast engine. By recognizing the different component features, the optimal forecast allocation strategy is selected, and the data prediction errors would not increase. The results of comparative experiment indicate that the proposed model can realize dynamic evaluation of air quality and reliable air pollution early-warning performance.