運用懲罰機制遺傳演算法於水庫颱洪操作之規劃

Reservoir Flood Operation Planning Using Penalty Guided Genetic Search

王國威(Kuo-Wei Wang)

指導教授 ： 張麗秋

淡江大學/工學院/水資源及環境工程學系碩士班/碩士(2006年)

https://doi.org/10.6846/TKU.2006.00175

摘要

在人口數量快速成長，對於用水的需水量愈來愈大，而在水庫因日漸淤積造成有效蓄水量逐漸減少的情況下，對於如何提升現有水庫操作之效能性與效率性，達到水庫蓄水利用之最大利益，應是水庫經營管理的最重要課題。在颱洪時期是台灣地區水庫儲蓄水資源的重要時機，在此時期水庫的運轉操作除了必須保護水庫壩體結構之安全與降低壩頂溢流之風險外，對於防洪與蓄水也是防洪運轉操作的重要考量因素。
最佳化搜尋方法發展至今已有許多不同演算法應用於水庫的操作與經營管理，卻經常面臨到決策模式所求得之最佳結果與實際需求有一段差距的窘境，為了加強水庫合理操作效率，本研究以水庫防洪操作運轉辦法及其相關軟硬體設施限制來建立防洪操作運轉模式，並運用兩種不同最佳化搜尋法－簡形法（simplex method）與限制型遺傳演算法（constrained genetic algorithm）搜尋水庫最佳洩放歷程，以探討水庫合理的防洪運轉操作之決策。其中，先利用簡形法搜尋，必須先假定洩放流洪峰發生位置，方能使用此法搜尋最佳解；而遺傳演算法使用在有眾多的限制式的最佳化問題，必須先設定適當的懲罰函數以提昇限制型遺傳演算法之搜尋效能。因此，本研究以石門水庫防洪運轉操作為例，分別以簡形法與限制型遺傳演算法來搜尋洩放歷程，結果顯示兩種方法皆能有效地消減洪峰並儲存洪水，而限制型遺傳演算法所提供之洩放歷程較簡形法在實際操作上更為合理且具實用性。

關鍵字

水庫颱洪操作 ； 簡形法 ； 限制型遺傳演算法 ； 懲罰函數

並列摘要

Constant population growth and dramatic economic development have resulted in a tremendous demand for water resources. With the gradual decrease of the reservoir capacities, attention must focus on improving the operational effectiveness and efficiency of existing reservoirs for maximizing the beneficial use of water storage. In Taiwan, typhoon could bring abundant water resources for future use while cause downstream flood. Consequently, a real-time flood control of a reservoir should consider decreasing peak flood stage downstream and storing floodwaters for future usage.
Various optimization strategies (techniques) have been developed for reservoir system management and operation; however, there still exists a gap between academic research and real-world implementations. To provide information for enhancing rational operating decisions, we proposed a reservoir flood operation optimization model with flood control requirements and existing operational regulations and the simplex method and Genetic Algorithm (GA) as searching engines. However, there are some requirements before using these two search methods. For simplex method, the time of release peak is assumed at the stage of after the flood peak. For genetic algorithm, since they are usually designed for finding the optimum solution to unconstrained problems, they have to be adapted to handle constraints for constrained optimization problems. Due to many constraints and flood control requirements, it is difficult to reach a feasible solution without violation of constraints. To tackle this bottleneck, the proper penalty strategy for each parameter was proposed to guide the GA searching procedure. The two searching methods were applied to the problem of finding the optimal release and desired storage, taking the Shihmen reservoir in Taiwan as a case study. The results demonstrated that these two methods could effectively reduce flood damages and storage floodwaters after the flood operation; moreover, GA can provide an adequate and robust way for searching the rational flood operating hydrograph.

並列關鍵字

Reservoir Flood Operation ； Simplex Method ； Constrained Genetic Algorithm ； Penalty Function

參考文獻

Archibald, T.W., McKinnon, K.I.M. and Thomas, L.C. 1996. An aggregate stochastic dynamic programming model of multireservoir systems. Water Resources Research 33(2):333-340.

Bhave, P. and Sonak, V. 1992. A critical study of the linear programming gradient method for optimal design of water supply networks. Water Resour. Res. 28(6): 1577-1584.

Chang, L.C. and Chang, F.J. 2001. Intelligent control for modeling of real time reservoir operation. Hydrological Processes 15:1621-1634.

Chang, F.J., Hui, S.C. and Chen, Y.C. 2002. Reservoir operation using grey fuzzy stochastic dynamic programming. Hydrological Processes 16:2395-2408.

Chang, J.X., Huang, Q. and Wang, Y.M. 2005. Genetic algorithms for optimal reservoir dispatching. Water Resour. Manage. 19:321-331.