A CNN Adaptive Model to Estimate PM10 Monitoring

In this work we introduce a model for studying the distribution and control of atmospheric pollution from PM10. The model is based on the use of a cellular neural network (CNN) and more precisely on the integration of the mass-balance equation; at the same time it simulates the scenario regarding a planar grid describing the whole studied area (the city of Palermo) by means of a CNN and a set of Bayesian networks. The CNN allows us to define a grid system whose dynamic evolution is a redefinition of the diffusion equation that considers contributions coming from near cells for each element of the grid. Dynamics of each cell is influenced by meteorological effects and by parameters related to topology and urban structure of the studied micro-zone (a single cell of the whole grid). These latter define the cell state and their effects are weighted by several other parameters in a polynomial function. The process of identification of these parameters is done by the minimization of an error index that involves estimated and forecasted data with the use of Bayesian networks. Results we obtained are encouraging and the proposed model seems interesting since it integrates two different paradigms: the forecasting with the simulation of a cellular system.