Differential neural network approach in information process for prediction of roadside air pollution by peat fire

Abstract

The paper presents a novel differential neural network model estimating the dispersion of CO emissions from a peat fire near a highway. We have developed approaches for the optimization of the model on the base of simulated and experimental measurements of CO concentrations in the area of dispersion of the smoke cloud. The numerical solutions of the problem are presented in the form of neural network approximations by the Gaussian model and in the form of neural network approximate solutions of partial differential equations. The trained neural network model can be used for the prediction of emergency when wind speed and direction and other fire parameters are changing. The method is also recommended for the development of air quality monitoring and predicting information systems.