智慧型區域水資源調配管理與休耕決策系統

Intelligent regional water resources management and fallow decision-making system

王國威(Kuo-Wei Wang)

指導教授 ： 張斐章

國立臺灣大學/生物資源暨農學院/生物環境系統工程學研究所/博士(2011年)

https://doi.org/10.6342/NTU.2011.02909

摘要

水資源管理包含了很多面向，例如長期的最佳水庫操作、水資源分配等問題。長期的水庫操作是一個具有相當多變數的複雜問題，在進行最佳化搜尋時難以有效地找到最佳解。此外，由於農業用水容忍度較高，在乾旱期間用水遭遇不足時，往往向農業尋求調整支援，多以停灌休耕方式來解決用水不足的窘境。有鑑於此，本研究分別研擬(1)最佳化搜尋遇到變數過多時的變數拆解策略，以及(2)乾旱時期停灌休耕的管理系統。
本研究首先提出多階層交互遺傳演算法（MIGA），其能將複雜的系統拆解成許多子系統，並提供關鍵訊息於各階層之間傳遞以銜接子系統之間的關聯。本研究以台灣北部之石門水庫為案例，並採用水庫操作規線模擬結果及單一遺傳演算法搜尋結果一同與MIGA搜尋結果進行比較，結果顯示MIGA不論在效率或效能上皆優於單一遺傳演算法之搜尋成果，MIGA相較水庫操作規線模擬於20年水庫操作的案例顯示，效能之改善率超過15%，並同案例中MIGA相較單一遺傳演算法效能之改善率超過25%，且節省80%的時間。
此外，北台灣因用水量逐年成長導致用水壓力亦相對增加，復加氣候變遷使得缺水情況日趨嚴重；援此，本研究研擬架構停灌休耕的決策過程，並包含三個停灌休耕策略：(1) 缺水門檻的制訂；(2) 適當之停灌休耕比例；(3) 宣布實行停灌休耕的發佈時機。藉由現有庫容及未來入流量資訊進行系統模擬，提供初步缺水資訊及制訂乾旱門檻。當乾旱情況確認之後，使用模糊類神經網路推估不同缺水情勢以分析適合的缺水率，同時反應公共及農業用水之間配水交互影響。最後，停灌休耕發佈時機由M-5規線模擬及GA搜尋成果進行研擬。結果顯示本研究提出之智慧型決策停灌休耕系統係為相當實用的工具，並且對於水資源永續利用能有所助益。本研究成果包含以水庫庫容資訊之缺水門檻，分別於案例 I及II當中呈現低於Q70 – Q60（輕微缺水）與Q90 – Q80（嚴重缺水）兩種不同缺水等級情況，提供水資源決策管理者初步判斷是否發生嚴重缺水情況，並建議愈早發佈所造成之損失較少（以2002年與2004年為例，損失減少約22%）。
本研究主要探討水資源管理當中相當重要之長期最佳水庫操作方法以及水資源不足時之調配決策問題，本研究之MIGA成功地解決長期水庫操作所面臨到的困難，以及智慧型停灌休耕決策系統能建議水資源決策管理者面臨乾旱時期採行之策略。

關鍵字

最佳化 ； 水庫操作 ； 分解 ； 停灌休耕 ； 決策 ； 多階層交互遺傳演算法 ； 調適性模糊類神經網路

並列摘要

Water resources management, such as long-term optimal reservoir operation or water allocation, is still a difficult task for decision makers. Complicated systems with high dimensional variables such as long-term reservoir operation usually prevent methods from reaching optimal solutions. In addition, the tolerance of water shortage for irrigation use is higher than that of other water uses such as domestic and industrial uses during drought periods. Therefore, in drought periods, the saved water from fallow areas requires to be reallocated for meeting the water demands of public use. This study proposes a variable decomposition strategy and a fallow management system for long-term reservoir operation and fallow management, respectively.
First, a multi-tier interactive genetic algorithm (MIGA) is proposed in this study which decomposes a complicated system (long series) into several small-scale sub-systems (sub-series) where Genetic Algorithm (GA) is applied to each sub-system and the multi-tier (key) information mutually interacts among individual sub-systems so that obtain the optimal solution to long-term reservoir operation. The Shihmen Reservoir in Taiwan is used as a case study. In addition, three long-term operation cases are implemented with the MIGA search, a sole GA search and a simulation based on the M-5 rule curves (the operation guideline of the Shihmen Reservoir) for comparison purpose. The improvement rate of fitness values for MIGA over the M-5 rule curves increases more than 15%. The improvement rate of fitness values for MIGA over sole GA increases more than 25%, and the computation time dramatically decreases 80% in a 20-year long-term operation case. Results demonstrate that MIGA is far more efficient than the sole GA and can successfully and efficiently increase the possibility of achieving an optimal solution.
Second, in the catchment of northern Taiwan, water scarcity becomes more and more serious and water stress and a deduction of water supply to irrigation use due to climatic change becomes major problems as well. This study proposes a framework of strategic pre-fallow decision making processes and addresses three important fallow issues: (1) the threshold of water shortage; (2) the appropriate fallow ratio; and (3) the timing for publicizing the fallow ratio. The first crop of paddy in northern Taiwan is used as a case study. A great number of system simulations based on current reservoir storage and upcoming three-month inflows are used to provide the preliminary judgment of water shortage and then to assign the drought threshold. After the drought situation is substantiated, a neuro-fuzzy network is used to determine the suitable fallow ratio and provide both water shortage information and the impacts of restrictions on irrigation water. Finally, the timing to publicize the fallow ratio is determined through the analysis of current M-5 operation rule curves simulation and genetic algorithms (GAs) search. Results demonstrate that the proposed intelligent fallow decision-making system can be a very useful tool and beneficial for the sustainable use of water resources. Drought thresholds of reservoir storage should be Q70 to Q60 (slight drought) and Q90 to Q80 (severe drought), respectively, when considering an early fallow strategy (Case I) and a late fallow strategy (Case II). The earlier the timing to publicize the fallow ratio is, the less the amount of fallow compensation is (approximately 22% less in this case study).
In sum, the problems of long-term reservoir operation can be successfully solved by MIGA and decision-makers can refer to the decision-making system for fallow strategies proposed in this study when allocating water resources during drought periods.

並列關鍵字

Optimization ； Reservoir operation ； Decomposition ； Fallow ； Decision-making ； Multi-tier interactive genetic algorithms (MIGA) ； Adaptive neuro-fuzzy inference system (ANFIS)

參考文獻

Anelli G., Broggi A., Destri G., 1998. Decomposition of arbitrarily shaped binary morphological structuring elements using genetic algorithms. IEEE Transactions on Pattern Analysis and Machine Intelligence 20(2): 217–224.

Atiya A.F., El-Shoura S.M., Shaheen S.I., El-Sherif M.S., 1999. A comparison between neural-network forecasting techniques－case study: river flow forecasting. IEEE Trans. on Neural Networks 10(2): 402–409.

Cai X., McKinney D.C., Lasdon L.S., 2001. Solving nonlinear water management models using a combined genetic algorithm and linear programming approach. Advances in Water Resources 24(6): 667–676.

Cai X., McKinney D.C., Lasdon L.S., 2002. A framework for sustainability analysis in water resources management and application to the Syr Darya Basin. Water Resources Research 38(6): 21/1–21/14.

Chandramouli V., Raman H., 2001. Multireservoir modeling with dynamic programming and neural network. Journal of Water Resources Planning and Management 127(2): 89–98.