| Summary: | 碩士 === 國立中興大學 === 環境工程學系所 === 107 === Many epidemiological studies indicate that suspended particles will result in cardiovascular and respiratory diseases, therefore it is very important to grasp the changing trends of PM10 and PM2.5 concentrations. The generation reactions of PM10 and PM2.5 are very complexly nonlinear mechanisms. Artificial neural network (ANN) has the ability to deal with nonlinear problems and has been widely used to predict ozone concentrations. Therefore, this study employs ANN and multiple linear regression (MLR), which is traditionally used, combined with principal component analysis (PCA) to establish centration prediction models for PM10 and PM2.5 of Taichung City.
The goal of this study is to predict the next day’s PM10 and PM2.5 concentrations of Chungming, Hsitun and Shalu air quality monitoring stations of Taichung City. Meteorological and pollutants monitoring data between 2015 and 2017 are used. The results show that each prediction models all exhibit qualified performance, and ANN has the best prediction results. Among the three stations, the prediction of Chungming is the most accurate. For ANN, the annual predictions are better than the seasonal predictions. However, seasonal predictions of MLR in Hsitun stations are better than annual ones. But, for Chungming and Shalu stations with MLR the annual predictions are better than the seasonal The prediction performance ANN or MLR are both worsen when combined with PCA. Both the ANN and MLR can predict the PM2.5 concentration levels, and about 80% of the preduction errors are within 1 level.
|
|---|
