Optimal usage of water resources systems and increasing the efficient planning of water are ways to contend with water scarcity. The constantly increasing demand for a sufficient quantity and quality of water, has forced engineers and planners to contemplate and propose ever more comprehensive, complex and ambitious plans for water resources systems. The application of systems methods such as optimization and simulation (S/O) can significantly aid in the definition, evaluation and selection of water resources investments, designs and policies. Optimization techniques are strategies to access desired choices in water resources investments. In this research, Harmony Search (HS) algorithm as a new meta-heuristic optimization method is introduced and its applications to determine the operational rule curves from multi-reservoirs systems under different climatic conditions is investigated. The new HS meta-heuristic algorithm was derived based on natural musical performance processes that occur when a musician searches for a better state of harmony. Using the memory for determination of harmony vector, simple mathematics and high speed calculation are the special features of HS. The aims of this study are presenting the concepts of HS and examining its capabilities for multi-reservoir systems optimization together with consideration of climate variability for reservoirs operation using an integrated S/O model. The performance of the model is investigated in three applications. In the first application the typical four-reservoir problem is tested for validation of formulation. In the second application the operational rule curves of Zayandeh-rud multi-reservoir system using HS is determined. Comparison of results with Genetic algorithm reveals better computational performances of HS. Introducing an adaptive operational model under different climatic conditions is the third application. In this application, the impact of three different climatic conditions, namely dry, normal and wet periods, on the operational performance was evaluated. A real-time hybrid model consisting of a data-driven Support Vector ltr"
