遺傳演算法於多水庫最佳操作規線優選之應用

Abstract

在水資源營運系統中，各水庫有其既定之操作方式，一般而言，水庫為方便操作，均以操作規線(rule curve)為操作之方式，因此操作規線的訂定對水庫系統的操作效率影響甚大，惟以往操作規線的決定往往以模擬或其他間接的方法決定，傳統的優選方法並不易直接應用在最佳規線的求得。 在本研究中，應用遺傳演算法結合HEC-5模式，以建構出以規線操作為基礎之最佳水資源調配模式。在HEC-5模式中，除各水庫本身之操作規線外，並以"指標平衡"及"對等水庫"等聯合營運的操作原則決定個別水庫之應放水量，並無優選規線之特性，經HEC-5模式模擬之結果，乃是在給定之規線下一合理之操作方式，因此本研究進一步以系統內各水庫之操作規線為決策變數，缺水指標（S.I Index）為優選之目標，經遺傳演算法結合HEC-5模式求得以規線操作為基礎並考量系統聯合營運之各水庫最佳操作規線。 本模式應用之區域為台灣南部地區，曾文溪及高屏溪流域之水資源營運系統，經本模式優選後之結果發現「在灌溉用水不需打折之情況下，以優選後之操作規線操作所得之缺水指標，相較於在灌溉用水需打折情況下操作，以各水庫現行操作規線操作所得者更佳，由此可證明，本模式確實適用於區域聯合營運下水庫操作規線之優選。

Reservoirs, the most important hydraulic facilities in a water resources system, can significantly impact regional water conservation. In practice, 'rule curves' generally regulate reservoir operations, accounting for why determining the optimal rule curves is the critical task in reservoir management. Rule curves are characterized by a set of operating zones and operating guidelines associated with each zone. Rule curves, including the operating zones, are conventionally determined by summing up the repeated simulation results by using historical or synthetic hydrological data. However, for a complicated multi-reservoirs system, adequate rule curves may be difficult to obtain using only simulations. Hence, this work presents a novel optimization model capable of computing the optimal operating zones for each reservoir in a complicated multi-reservoirs system. The proposed model integrates the Genetic Algorithm (GA) and the HEC-5 model, i.e., a widely used reservoir simulation model. To fulfill the total demand, the HEC-5 distributes the releases among reservoirs based on the rule curve of each reservoir and the principle of 'index balance'. However, the HEC-5 is only a simulation model incapable of computing the optimal rule curves. Therefore, by embedding the HEC-5 model into the Genetic Algorithm, this work develops a rule curve based management model (GHEC-5). In GHEC-5, a chromosome represents a potential combination of the operating zones, i.e., a realization of the decision variables, for all the reservoirs. GHEC-5 then computes the fitness value for all the chromosomes by using HEC-5 and improves the chromosomes by GA. The shortage index (SI) is the objective function used herein. The GHEC-5 has been applied to the water resources system in southern Taiwan, including the Tsengwen River and Kaopin River watersheds. Major components in the system include three reservoirs, four river dams for water intake and two demand areas. Numerical studies demonstrate that the optimal rule curves obtained by GHEC-5 can significantly improve the system capacities than the existing rule curves. Therefore, GHEC-5 can facilitate decision makers in obtaining an optimal policy for regional water resource management. (Keywords: Reservoir Operation, Rule Curves, Genetic Algorithm, HEC-5)