查詢結果分析
來源資料
頁籤選單縮合
題名 | 倒傳遞神經網路應用於嘉南灌區作物蒸發散量之推估=Application of Backpropagation Neural Network to Estimate Crop Evapotranspiration for ChaiNan Irrigated Area |
---|---|
作者 | 郭勝豐; 程澄元; 劉振宇; Kuo, Sheng-feng; Cheng, Cheng-yuan; Liu, Chen-wuing; |
期刊 | 農業工程學報 |
出版日期 | 20060300 |
卷期 | 52:1 民95.03 |
頁次 | 頁24-34 |
分類號 | 434.287 |
語文 | chi |
關鍵詞 | 倒傳遞神經網路; 蒸發散量; 農業水資源; Backpropagation neural network; Evapotranspiration; Agricultural water resource; |
中文摘要 | 作物蒸發散量為農田水利會推估農業用水量以制訂年度灌溉計畫之基本資料,作物蒸發散量之精確推估可提升農業用水管理技術,有效率地使用農業水資源,建立永續生態環境。 倒傳遞神經網路模式應用於建立農業氣象因子與蒸發散量間的關係,藉以預測嘉南地區蒸發散量,嘉南農田水利會灌溉技術推廣中心農業氣象站之氣象資料作為輸入層,其氣象因子包括:(1)最高溫度;(2)最低溫度;(3)平均溫度;(4)相對溼度;(5)風速;(6)日照時數;(7)日輻射量;(8)露點;(9)上午地溫(10)下午地溫等。研究結果得知以隱藏層神經元數為10及Penman公式計算之蒸發散量為輸出層於預測83年度蒍發散量之相關係數達0.993,實際蒸發散量為911.6公厘,神經網路模式預測值為864.4公厘,誤差百分比為1.67%。另以隱藏層神經元數嵨0及蒸發?實測蒸發量為輸出層於預測83年度蒸發散量之相關係數為0.708,蒸發?實測蒸發量為1674.1公厘,神經網路模式預測值為1451.7公厘,誤差百分比為13.23%。 |
英文摘要 | Crop evapotranspiration is the essential data to efficiently estimate the agricultural water requirements for promoting the technical of agricultural water management and drawing up the irrigation plan for irrigation association. Backpropagation artificial neural network was applied to establish the relationship between agricultural climate factors and evapotranspiration to estimate the crop evapotranspiration for ChiaNan area. The ten meteorological data, recorded by Irrigation Technical and Promotion Center (ITRC) of ChiaNan Irrigation Association, were used as input data for netural network model as follows: (1)maximum temperature, (2)minimum temperature ,(3)mean temperature ,(4)relative humidity ,(5)wind speed ,(6) sunshine hours, (7)solar radiation, (8)dew point, and (9) soil temperature on morning, (10) soil temperature on afternoon. Based on the module of ten hidden layers and the output data of evapotranspiration ws calculated by Penman equation with the meterological data from ITRC in 1994, the simulation results demonstrate that the correlation coefficient for input and output layers was higher to 0.993; furthermore, the prediction error was only 1.67% based on he actual and predict evapotranspiration were 911.6 mm and 896.4 mm, respectively. On the other hand, the module with twenty hidden layers and output data of evaporation was recoded by pan evaporation, the results show that the correlation coefficient related to input and output layers was lower to 0.708; furthermore, the prediction error was 13.23% based on he actual and predict evaporation were 1673.1 mm and 1451.7 mm, respectively. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。