Adaptive wavelet neural network for prediction of hourly NOX and NO2 concentrations
WSC '04 Proceedings of the 36th conference on Winter simulation
Hi-index | 0.00 |
SO/sub 2/ air pollution is still an important environmental problem in Slovenia, especially around big thermal power plants. A modern approach involving a multilayer perceptron neural network based short term air pollution prediction model is explained. Neural network based models are rarely used in the field of air pollution. The models were developed for the thermal power plant. An extensive database is available for this site. It includes meteorological data, ambient concentrations and emission data for a four year period. A model that predicts SO/sub 2/ concentration for one averaging interval (half an hour) in advance is explained in detail. Development of the model should solve several problems, including selection of an appropriate structure (number of neurones), selection of features, pattern selection and determination of suitable training algorithm parameters. Results are very encouraging and show that the method is worthy of further research.