透過您的圖書館登入
IP:18.222.119.148
  • 學位論文

適應性模糊推論系統與倒傳遞類神經網路應用於桃園地區光化學污染物之預測

Predicting Photochemical Pollutants in Taoyuan Area Using Adaptive Network Based Fuzzy Inference System and Backpropagation Neural Network

指導教授 : 白子易 張迪惠
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


本研究應用倒傳遞類神經網路(Back-Propagation Neural Network, BNN)及適應性模糊推論系統(Adaptive Network Based Fuzzy Inference System, ANFIS),建立桃園地區空氣品質之預測模式,變數因子分別為臭氧(ozone)、大氣溫度(TEMP)、風向(WIND DIREC)、風速(WIND SPEED)、二氧化氮(Nitrogen dioxide, NO2)、氮氧化物(Nitrogen oxides, NOX)、二氧化硫(Sulfur dioxide, SO2),以2011年1月~11月為網路輸入参數建立最適化網路,對2011年12月之O3及NOx作預測,BNN研究結果顯示O3預測部份以3個影響參數對應1個輸出參數的平均絕對百分比誤差值(Mean Absolute Percentage Error, MAPE)14.92%為最佳,NOx訓練部份以4個影響參數對應1 個輸出參數MAPE值27.91%為最佳,ANFIS研究結果顯示O3測試部份鐘形隸屬函數3個影響參數對應1個輸出參數的平均絕對百分比誤差值17.55 %為最佳;NOx訓練部份鐘形隸屬函數以4個影響參數對應1 個輸出參數MAPE值23.32%為最佳;O3 之ANFIS預測結果以三角型為最好、高斯次之、最後分別為鐘形以及梯型之隸屬函數,NOX預測結果以高斯為最好、鐘型次之、最後分別為三角形以及梯型之隸屬函數。BNN及ANFIS預測之結果良好,顯示BNN及ANFIS模型對於掌握預測濃度及變動趨勢效果良好。

並列摘要


This study employed adaptive network based fuzzy inference system (ANFIS) and Back-Propagation Neural Network (BNN) method to establish an air quality prediction model of Taoyuan area .Variable factors used Ozone, Temp, Wind direc, Wind speed, Nitrogen oxides, Nitric oxide, Nitrogen dioxide, Sulfur dioxide. We input data between January and November, 2011 as parameters to establish an optimizing network to predict the air quality of O3 and NOx on December, 2011. BNN research shows that the best mean absolute percentage error (MAPE) 14.92% by using three input parameters to compare one output parameter in O3. In predict part, the best result will be MAPE 27.91% by using four input parameters to compare one output parameter in NOx.ANFIS research shows that the best mean absolute percentage error (MAPE) 17.55% by using gbellmf three input parameters to compare one output parameter in O3. In predict part, the best result will be MAPE 23.32% by using four input parameters to compare one output parameter in NOx.According to training and prediction of O3 and NOx figure, the simulation result is better. O3 and NOX could be achieved using different types of ANFIS.And utilize four kinds trimf, trapmf, gbellmf and gaussmf in ANFIS to four the Membership function and predict,O3 of prediction result had better regard trimf as all, gaussmf take second, under the Membership function for gbellmf and trapmf respectively finally.NOX of prediction result had better regard gaussmf as all, gbellmf take second, under the Membership function for trimf and trapmf respectively finally.Therefore, ANFIS and BNN mode makes a conclusion that predicts effect perfects.

並列關鍵字

BNN Nitrogen oxides Ozone ANFIS

參考文獻


41. 行政院環境保護署 http://www.epa.gov.tw/
31. 高敬棠,「應用類神經網路預測雲林地區光化學污染物之研究」,碩士論文,朝陽科技大學環境工程與管理系,(2011)。
34. 許育誌,「適應性類神經模糊系統於二足機器人ZMP之應用」,碩士論文,國立中央大學光機電工程研究所,(2007)。
39. 蘇博聖,「利用類神經網路探討中部地區臭氧之減量策略」,碩士論文,中興大學環境工程學系所,(2006)。
1. Acuna G., H. Jorquera, R. Perez, ”Neural network model for maximum ozone concentration prediction., ”Lecture Notes in Computer Science, No.1112, pp.263-268, (1996).

延伸閱讀