Application of Swarm Intelligence and Evolutionary Computation Algorithms for Optimal Reservoir Operation

摘要

Real-world problems often contain complex structures and various variables, and classical optimization techniques may face difficulties finding optimal solutions. Hence, it is essential to develop efficient and robust techniques to solve these problems. Computational intelligence (CI) optimization methods, such as swarm intelligence (SI) and evolutionary computation (EC), are promising alternatives to conventional gradient-based optimizations. SI algorithms are multi-agent systems inspired by the collective behavior of individuals, while EC algorithms implement adaptive search inspired by the evolution process. This study aims to compare SI and EC algorithms and to compare nature-based and human-based algorithms in the context of water resources planning and management to optimize reservoir operation. In this study four optimization algorithms, including particle swarm optimization (PSO), teaching-learning based optimization algorithm (TLBO), genetic algorithm (GA), and cultural algorithm (CA), were applied to determine the optimal operation of the Aydoghmoush reservoir in Iran. This study used four criteria, known as objective function value, run time, robustness, and convergence rate, to compare the overall performances of the selected optimization algorithms. In term of the objective function, PSO, TLBO, GA, and CA achieved 2.81 x 10(-31), 1.66 x 10(-24), 4.29 x 10(-4), and 1.44 x 10(-2), respectively. The results suggested that although both SI and EC algorithms performed acceptably and provided optimal solutions for reservoir operation, SI algorithms outperformed the EC algorithms in terms of accuracy of solutions, convergence rate, and run time to reach global optima.