| Reservoir operation plays an important role in managing a regional water distribution system. During normal periods of operation, when inflows are plentiful, it is not difficult to distribute the available stored water from different reservoirs to meet the planned demands imposed by competing users. However, during periods of drought, or when anticipating a drought, the planned demands cannot be met completely, and a water shortage occurs. Therefore, a standard policy is needed to provide the rules for operating the reservoir to efficiently utilize the water. As a result, the rule curves and hedging rules are introduced as guidelines for the reservoir management and operation.; This study develops a mixed-integer linear programming (MILP) model that simultaneously considers both the rule curves and hedging rules to optimize the management and operation for a multi-purpose, multi-reservoir system. To minimize the impact of drought, the hedging rules effectively reduce the ongoing water supply to balance with the target storage requirement.; A mixed-integer nonlinear programming (MINLP) model is then developed based on the proposed MILP model to evaluate and modify the current rule curves and hedging rules. Due to the mathematical complexity of this MINLP model, a number of methodologies are proposed to solve this model. These methodologies are: Generalized Benders Decomposition, Penalty Coefficient Methods, Pseudo-integer Method, Simulated Annealing, and one mathematical transformation technique. The results of these methodologies are explored and investigated. After examining the performance of these methodologies, the transformation technique seems best suited to solving this proposed MINLP model.; A multi-reservoir system in the southern region of Taiwan is used as an actual case study to demonstrate the capability and utility of the proposed models. The results obtained indicate that the MILP model can efficiently utilize the water resources when incorporating the current rule curves and hedging rules. In addition, the new set of rule curves and hedging rules obtained by the MINLP model can successfully improve the performance of the current system operation. |
