Water resources decision making using meta-heuristic optimization methods

Abstract

This dissertation work is part of a larger research effort involving soil-aquifer treatment (SAT). The dissertation's focus was to investigate meta-heuristic (global) optimization methods suitable for developing water resources decision support system (DSS), particularly to optimally design and operate groundwater storage and recovery projects. The effort included developing an integrated simulation-optimization management model for complex aquifer recharge/extraction operation considering water quality transformation. The research demonstrated successful integration of three-dimensional hydraulic, water quality, and particle tracking models with shuffled complex evolution (SCE) optimization algorithm. It also included developing the shuffled frog leaping algorithm (SFLA), a meta-heuristic optimization technique for solving discrete/combinatorial problems, and its application to aid decision making in water supply and distribution system optimization issues. SFLA is a memetic, meta-heuristic population-based cooperative search metaphor inspired by natural memetics. SFLA was developed by extending the logic of two existing global optimization techniques for continuous optimization problems. The local search is completed using an extension of the particle swami optimization (PSO) method, and the global exploration is performed by a technique similar to that used in the shuffled complex evolution (SCE) algorithm. SFLA was tested favorably on several literature test functions and engineering problems that present difficulties to many global optimization problems. The effectiveness and suitability of this algorithm has also been demonstrated by applying it to a groundwater model calibration problem and several water distribution system design problems that are considered as benchmark problems in the literature. The comparison of SFLA with other existing global optimization methods, such as genetic algorithms (GA), in terms of the likelihood and efficiency of converging to a global optimal solution, suggests that SFLA can be an effective algorithm for solving discrete/combinatorial optimization problems.