Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/5295
標題: 利用類神經網路預測台中都會區臭氧趨勢之研究
An Application of Artificial Neural Networks to the Predication of the Trend of Ozone Situations in Taichung City
作者: 黃宗仁
Huang, T.R.
關鍵字: Ozone;臭氧;Predication;Artificial neural network;Time series;Multiple Regression;預測;類神經網路;時間序列法;複迴歸分析法
出版社: 環境工程學系
摘要: 
根據環保署1994~2000年資料顯示,中部地區正面臨日漸嚴重的臭氧污染問題,如能有效建立臭氧生成趨勢之預測方法,則將可做為預警系統之參考。而由於臭氧的生成乃是極複雜的非線性機制,所以本研究在此利用擅於處理非線性問題的類神經網路,來嘗試建立一套台中市臭氧趨勢的預測模式,並將其預測結果與使用時間序列法及複迴歸分析法所得之結果相比較,以評斷模式之優劣。
本研究先利用1996~1999年的臭氧資料來進行最適化模式之建立,之後再應用於預測2000年之臭氧變化趨勢。結果顯示,三種方法中以類神經網路表現最好,其對於短期(1小時及4小時)臭氧濃度的預測效果較好,而對於長期的預測則不盡理想;在時間序列方面,其僅能預測濃度的變動趨勢,對於臭氧極值(極低或極高)則無法正確模擬;而在複迴歸分析法方面,雖然其對於高臭氧濃度(>60ppb)的預測會有低估的現象,但對於低臭氧濃度(<60ppb)的預測結果堪稱理想。

This research used artificial neural networks (ANN) to develop a model to predict the trend of ozone situations in Taichung City. The optimal network was trained by using the monitoring data during the period from 1994 to 1999, and then the network is applied to predict the ozone situations of year 2000. Time Series and Multiple Regression were also used in this research, and their results were compared with ANN's.
The results showed that among these three methods, ANN appears to have the best performance. However, although ANN's short-term predictions are quite effective, its long-term predictions are still need to be improved in this research. On the other hand, Time Series can only predict the trend of ozone concentration, and its prediction of the critical values appears imprecise. For Multiple Regression Analysis, the predicted ozone concentrations were usually underestimated when they are above 60 ppb, but effective when under 60 ppb.
URI: http://hdl.handle.net/11455/5295
Appears in Collections:環境工程學系所

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.