Abstract

The purpose of this study is to develop an efficient optimization model to determine storage strategies for water supply purpose in multi-reservoir systems. The specific goals in designing a multi-reservoir system for water supply purpose is to optimize the reservoir sizes and configurations to satisfy demands for water supply at the minimum cost. Four models are developed to optimize the configuration of multi-reservoir systems for water supply purposes. These models apply optimal control theory (OCT) and penalty successive linear programming (PSLP) as the most promising techniques to optimize large and complex water resources systems. Three of these models are based on the OCT. They have, however, different approaches to join the cost function to other objectives. The fourth model employs a new composite optimization algorithm, which is introduced in this study. This is called PSLP-OCT model and consists of two OCT and PSLP modules. These two modules interactively share their results during the optimization iterations. Multi-objective programming methods are implemented in the four models in order to consider the two non-commensurate objectives of minimizing cost and water deficit. The weighting and epsilon constraint methods are used as the most suitable generating techniques to incorporate the problem objectives. The comparative performances of the design models on several case studies showed that the design models based on the OCT algorithm fail to design the multi-reservoir system optimally. However, The PSLP-OCT performance indicated that it is a very promising optimization method to design multi-reservoir systems regardless of their size. The PSLP-OCT model is the first model of its type that applies the proposed composite algorithm and incorporates multi-objective programming into the multi-reservoir design problems. Due to the inherent characteristics of the optimal control theory, the control variables in the OCT, module is not sensitive to the initial solution. The model structure is adapted such that the PSLP module is independent of the design period length. Therefore, using large hydrologic data does not affect its problem dimension.