On Prediction of Air Pollution Using Piecewise Affine Models

Since air pollution affects both public health and economic growth, the issue has received more attention recently. Model-based early warning systems or pollution management tactics can be used to assist in combating dangerous air pollutants if accurate prediction models are available. This paper presents an approach to forecasting air contaminants using a piecewise affine model, which has a high prediction power. To identify the piecewise affine model, this study adopts effective clustering to identify the model. The proposed hierarchical clustering method improves the widely used BIRCH (Balanced Iterative Reducing and Clustering using Hierarchies) by adding a refining step to handle clusters with arbitrary geometries. Additionally, an optimization strategy like GA (Genetic Algorithm) is used to jointly estimate the model order and parameters. Measurements of Shenyang's air quality are used to illustrate the proposed approach, and the outcomes reflect the method's good prediction ability.