Methods And Applications Of Robust Multi-Objective Optimization Based On Swarm Intelligence

A number of optimization problems exist in the science and engineering fields. Traditional optimization methods have some advantages of complete theory, high efficiency and robustness. But traditional optimization methods are hard to solve these problems effectively while the problems are getting more and more complex. As population-based global search method, swarm intelligence is fit to solve the complicated problems including simple objective optimization problems, multi-objective optimization problems and constraint optimization problems. Therefore, the research of swarm intelligence has received more and more attentions. Now, it has become a hot research field in evolutionary computation.The aim of this thesis is to explore and research particle swarm optimization and to design effective algorithms and strategies for single-objective constrained optimization problems and multi-objective optimization problems, and to do the corresponding theoretic and experimental analysis. And on this basis of pursuing the quality of solution, a robust single-objective optimization method is proposed and used in multi-objective optimization. The corresponding theoretic and experimental analysis is also done here. In the end, all of these research results are applied to practical problems. The main research works in this thesis consist of 4 aspects.1. For single objective constrained optimization, a modified constrained particle swarm optimization algorithm is proposed. This algorithm effectively combines two finesses with the objective function value and the violation value of constraint functions to estimate the particles in a dynamic way. Meanwhile, the adaptive weight function is adopted in the calculation of the violation value. Then, the strategy of keeping an adaptive relation of weight coefficients is proposed, and the strategy of swarm tournament selection and mutation are improved. Experimental results show that the convergent speed of the algorithm is fast and the result is valid.2. For multi-objective constrained optimization, a modified multi-objective particle swarm optimization algorithm was proposed, based on maximin fitness function. This algorithm introduces the computation of maximin fitness function value. Besides, the conception ofε-dominance was imported and a modified computation method and an alterableε-dominance strategy were put forward, which effectively improved the convergence speed of the algorithm and the diversity of the particles. Experimental results show that the convergent speed of the algorithm is fast and the result is valid.3. For robustness of the results of optimization problems, a robust single objective optimization method is presented. Disturbed by random selection within a certain number of particles in the sample, calculate the sample mean and variance of the target space, which is proposed for the maximization and minimization problems of the robust model ofa single target, and extended to multiple targets.4. For the application of these research results, the modified constrained particle swarm optimization algorithm is applied to solve energy optimization problems of the urban water supply process, the modified multi-objective particle swarm optimization algorithm was applied to starting performance optimization of DC inverter compressor, and the robust optimization method is applied to optimization of the parameters of the particle swarm optimization algorithm. The results have been verified through these examples.