Application of Neural Network

This project presents an application of artificial neural network (ANN) approach to the simulation and prediction of acid deposition conditions in the United States. The previous research is mainly focused on the effect of emission. However, with the increase of NOx emissions, acid deposition problem could become more serious in some areas despite stricter controls on SO2 pollution from coal-burning power plants. Therefore, a more systematic approach that focuses on both SO2 and NOx emissions is needed for the longterm simulation and prediction of the acid deposition condition. In this study, acid deposition data from various monitoring sites in U.S. were collected. Feedforwrd backpropagation ANN models were constructed. Based on SO2 and NOx emissions from power plants and field data on precipitation chemistry, the networks were trained and tested. Then, the networks were used to predict SO2 and NOx emissions and hydrogen ion concentration according to the projected acid precipitation ion concentrations. The results obtained from the ANN modeling approach indicate that although it is promising to bring the precipitation pH value up to 4.8, it is difficult to increase the pH to a normal value of 5.0 based on the current acid deposition data patterns. A higher pH value of 6.0, which is necessary for full ecosystem recovery, would be a more challenging goal for 2010. Based on the prediction results for SO2 and NOx emissions, it is critical that the reduction goal for NOx is added to the acid deposition related policy.