應用類神經網路於地面地下水聯合運用規劃之研究

Abstract

隨著人口增加與氣候變遷，部分地區的地面水資源已日漸不敷需求，進而選擇使用地下水資源，然而過量的抽取將會造成嚴重的地層下陷、海水倒灌等問題。有鑒於此，本研究建立了一個地面地下水資源聯合運用規劃之優選模式，此模式能夠在考慮地下水位下降限制下，最佳化分配地面水與地下水之供給量，達到兼顧用水需求與地下水位穩定的目標。 本研究之優選模式目標函數採用最大化供水量，決策變數為地面水與地下水在各需水點於各時刻之供給量。地下水模擬部分，本研究利用已建構完成的濁水溪沖積扇模擬模式來建立類神經網路，模擬抽水補注行為對地下水位造成之變化，並將其公式化納入限制式，使得優選模式能夠計算出抽水量對地下水位之影響，進而在不超量抽水的前提下最佳化分配地面水與地下水之供給量。 本研究的優選模式可直接使用電腦套裝計算軟體求解，因此不需要大量的程式撰寫與運算時間，相當適合應用在變數數量龐大的問題上，例如複雜的水資源系統或多時段的求解。本研究以雲林地區為模擬案例，其結果可做為雲林地區水利單位調配水資源之參考，有助於減輕相當嚴重的地層下陷問題。

Inasmuch as population growth and weather change. The water resources was not enough in some places. So they will pump groundwater for use. However, pumping too much will bring the land subsidence problem. In this paper, we build a conjunctive use of surface and groundwater model. It will flow the drawdown constrain to optimize the surface water supply and groundwater supply. Let us have water can use and don’t worry land subsidence. In this paper, the model’s objective function is to maximize the water supply. The decision variables are surface and groundwater supply in every demand point. We use neural networks to simulation the groundwater level change by pump and discharge. The developed neural networks are incorporated into the optimization model as constraints. This model can calculate by packaged software. It will save much time to find the answer in linear problem or nonlinear problem. In complicated problem, it will better than use genetic algorithm to find answer. In this paper, we have a case study of Yunlin region. The result may will help alleviate the land subsidence problem.