以類神經網路推估蒸發量

Estimating Evaporation by Artificial Neural Networks

吳國儒(Gwo-Ru Wu) ； 高慧珊(Huey-Shan Kao) ； 鍾昌翰(Chang-Han Chung) ； 何宜樺(Yi-Hua Ho) ； 張斐章(Fi-John Chang)

《農業工程學報》 54卷3期 (2008/09) Pp. 1-13

https://doi.org/10.29974/JTAE.200809.0001

摘要

蒸發現象係水文循環中重要的一環，蒸發量的估算爲水資源規劃與農業灌溉營運之重要參考依據。蒸發量的估算可籍由蒸發皿直接量測，惟若無發皿量測資料，一般都以傳統經驗式利用氣象因子推估：因參數不易推估而影響其精確性及使用性。本研究擬以具有高度訊息處理與非線性映射能力的類神經網路，並利用氣象量測資料，如風速、溫度、日照等，建立蒸發量推估模式。本研究選用倒傳遞類神經網路(BP)、自組特徵映射網路(SOM)及調適性網路模糊推論系統(ANFIS)等三種類神經網路架構來建立推估模式，並與Modified Penman經驗式及Penman-Monteith經驗式的推估結果進行比較。本研究以中央氣象局恆春站2002-2004年的日氣象觀測資料共計1035筆，作爲模式訓練與測試用，結果顯示三種類神經網路推估日蒸發量之結果皆相當準確，其均方誤差約爲1.1 mm/day，較傳統經驗公式推估結果低0.1 mm/day，研究結果可爲往後相關研究之參考依據。

關鍵字

蒸發量 ； 倒傳遞類神經網路 ； 自組特徵映射網路 ； 調適性網路模糊推論系統

並列摘要

The phenomenon of evaporation is an important component of the hydrological circle, and it's an important basis of reference for planning the use of the outer resources and managing the agricultural irrigation. Estimation of evaporation could he measured by evaporimeter directly; in case of no evaporimeter, it is generally estimated by tranditional evaporation formulas. However, the formulas indeed is complex with many paraments which are difficult to property estimate, in this study, we plan to establish an applicable model fun precisely estimating evaporation based on climate variabies, such as wind speed, temperature, aunt radiation, by using artificial neural networks (ANNs). Three different kinds of ANN, i.e. Bach Propagation Neural Network, Self Organizing Map, and Adaptive Network-Based Fuzzy Inference System, are used. The data obtained from Central Weather Bureau in HengChuen station during the period of 2002-2004 are used to train and verify tine networks. The results are compared with the outcome of Modified Penman formulas and Penman-Monteith formulas, and we find that estimation of evaporation by mixing the proposed artificial neural networks is much undue precise than tine tranditional evaporation formulas. The ANNs can be a potential tool for estimating the evaporation effectively from limited climatic data.

並列關鍵字

Evaporation ； Back Propagation Neural Network ； Self-Organizing Map ； Adaptive Network-Based Fuzzy Inference System