Improvement Of Water Circulation Algorithm And Gravity Search Algorithm

Swarm intelligence optimization algorithm is a kind of stochastic optimiza-tion algorithms, which simulate the process of living states of creature in nature by optimizing at unconscious behavior. Due to its simplicity, easy operation and strong practicability, it has gradually attracted wide attention in various field-s and becomes an effective tool to solve many complex optimization problem-s. As new swarm intelligence optimization algorithms, water cycle algorithm and gravitational search algorithm have better performance in solving complex optimization problems because of less parameters and fast convergence speed. However, they are easily trapped into local optima during late stage. In order to enhance their performance, this thesis improves two algorithms respectively.1. Noticing that the flow of water influenced by gravitational force and strong global search ability of gravitational search algorithm, a gravitation-based chaos water cycle algorithm is presented by suitably integrating gravitational search into water cycle algorithm. In new algorithm, information of particles is fully shared and communicated in groups, and then transmitted effectively through groups under the frame of water cycle algorithm. Thus the proposed method enhances the search ability. In addition, to ensure the algorithm has better population diversity during iteration and avoid premature convergence, a group of chaotic formula for stream and river are adopted by defining a new chaotic mapping.2. To enhance convergence speed and accuracy of the algorithm, a hybrid gravitational search algorithm with information entropy is proposed by intro-ducing the concept of global optimum in particle swarm optimization to modify the updating formula, and constructing information entropy model to describe the degree of search ability. The new algorithm improves search ability through adding memory of particles, and choosing different weights according to differ-ent threshold of the information entropy. So the new algorithm balances the abil-ity of global search and local search.3. To improve performance of the algorithm, a gravitational search algo-rithm based on chaotic reproduce strategy is proposed by constructing chaotic reproduce strategy and adaptive t distribution mutation strategy. The solutions generated by the algorithm can traverse the entire space by using advantage of chaotic map, which increases the population diversity. Moreover, adaptive t dis-tribution mutation strategy improves the ability of exploration and exploitation.Numerical experiments demonstrate that the proposed algorithms have bet-ter convergence speed accuracy and speed.