地下水系統抽水操作規則最佳化之研究

Abstract

台灣地區的地形因素造成台灣地區河川坡陡流短，每逢颱風過境或是暴雨產生之豐水季，易導致河川瞬時流量過大，而引發洪災的發生；而枯水季時則因河川坡陡流短導致蓄水不易，民生及工業用水短缺。近年來，都市化與工商業的迅速發展使得用水需求日趨增加，但亦因時間、空間等種種因素導致供水不足之窘境。因此，台灣部分地區轉而開發地下水資源。然而，不當的開發以及過量的抽取，導致地層下陷、海水倒灌…等等問題的產生。有鑑於此，如何在適量抽取地下水之前提下，有效地分配地面水以及地下水，是為現時相當重要之課題。 本研究藉由一優選模式所產生之優選結果進行資料採礦(Data Mining)，希望藉由資料採礦得知水文資料之間的因果關係。在本研究中使用決策樹(Decision Tree)演算法，此演算法可輸出一個類似二元樹的操作規則。本研究之地下水系統抽水操作規則以一考慮地面地下水聯合運用優選模式之結果訓練而得，故能繼承該優選模式之特性，在不超抽地下水之前提下指定一合理抽水量，以期達成供水需求，並進一步執行實際操作抽水量之模擬。 操作規則一經建立之後，在往後的時期只需進行水文資料之比對，即可得知對應之抽水量。相較於其他演算法繁複的演算過程，使用操作規則可省下大量的求解時間，在管理層面有相當大的便利性。

Because of many spatial and time factors, it is hard to save water from precipitation and river in Taiwan. Recently, the demand of water is growing faster and faster because of urbanization and development of Industry and Commerce. However, the hydrology conditions in Taiwan cause insufficiency of water supply. Therefore, groundwater resource becomes a considerable option. Nevertheless, improper development and over-pumping result in many problems, such as land subsidence and inundation by the sea. Consequently, it’s an important topic to distribute surface water and groundwater effectively with a premise of pumping groundwater applicably. In this research, an optimization would be resulted by an optimization model, and then executed data mining to discover the relationship between hydrology information. Decision tree algorithm would be used in this study, and a binary-tree-liked operation rule could be generated. Pumping operation rule in groundwater system is trained by an optimization result from an optimization model, which combines surface water and groundwater as a conjunctive use. Hence, the characters of optimization model would be succeeded to pumping operation rule. So that, pumping operation rule could assign a reasonable pumping to meet the requirement of water supply demand and prevent over-pumping, then simulate pumping operation in real case as an advanced step. After the operation rule is established, quantity of pumping could be known by matching hydrology data. Compared with other algorithms, using operation rule won’t waste much time in solving process. So, operation rule is more convenient then other algorithms.