Application Study Of Swarm Intelligence For Continuous Optimization Problems

There are a lot of NP-hard optimization problems in the engineering application, which are difficult to solve by traditional mathematical techniques. These years, as a population-based heuristic random optimization method, Swarm Intelligence algorithm attracted more and more attention of researchers. Unlike other classical mathematical methods, Swarm Intelligence is useful in dealing with a high number of dimensions and problems where problem specific information is non-existent. Its rapid convergence and parallelism make it a good candidate for solving engineering optimization problems. Ant colony optimization (ACO) and particle swarm optimization (PSO) are two new paradigm of Swarm Intelligence, which are based on population evolution to reach the optimum. This thesis discusses the basic theory and characteristic of the two methods, and proposes improving scheme accordingly. Each improving scheme is examined on several test cases. The achievements in the research work of this dissertation include:1. The biological mechanism, the development and the character of the ACO algorithm are outline, and the key of its surperiority in solving combinatorial optimization problems is concluded. Moreover, the shortage in solving the continuous space optimization problems are presented. Based on analysis of the difficulties of solving NP-hard problems, a new general model of continuous ant colony algorithm is proposed, which includes rapid iterative searching and simple pheromone communication scheme. The efficiency of the method is tested and compared with other methods. The simulation results show that the new method outperform the other ACO and PSO algorithm, and equal with improved Genetic Algorithm.2. The background, the procedure and structure and the character of the PSO algorithm are reviewed. The shortage of basic PSO is analyzed and the cause of its premature convergence is concluded. An adaptive mutation scheme is proposed to add in the basic PSO, which inducts global information into the optimizing procedure and helps the algorithm to jump out of the local optima. Further more, for improving the precision of the optimization, a local searching is implemented around the optimum after the improved PSO algorithm tends to be convergent. The improving method is tested and compared with other methods and the efficiency of the improved method is guaranteed.3. The function model of Economic Load Dispatch (ELD) in power system and its characteristic is discussed. The techniques used to solving the ELD problem are failed to find the feasibility solution efficiently. A kind of infeasibility repair scheme is proposed, which is easy and robust to implement. Combined with the improving PSO proposed in the paper, the whole technique is applied in the ELD problem and reaches a better result.Finally, the work of this dissertation is summarized and the prospective of further research on optimization technique is discussed.