Research And Application On Hybrid Evolutionary Multi-Objective Algorithm Based On Decomposition

The thought of evolutionary algorithm is based on Darwin’s evolution theory. It is an artificial intelligence technology which is similar to the process of biological natural selection and evolution. Because of its high robustness, It has better efficient solution in solving highly complex nonlinear problem. It codes the entire parameter space before dealing with problems, start searching from a set of nodes. In the process of search, do not need to use derivatives or other method to modify the objective function value of the information. Because of evolutionary algorithm has good versatility, high nonlinearity, parallelism, etc, so it can quickly and efficiently get the Pareto optimal solution of multi-objective problem is obtained.This thesis first introduces procession of multi-objective optimization problem, the background and research status, the concept of evolutionary algorithm principle and the typical evolutionary algorithm, then presents a decomposition multi-objective evolution-nary algorithm based on distribution estimation.According to the optimal solution of the problem diversity, uniformity and efficiency of algorithm are studied. The numerical analysis and experiment show that the improved new algorithm not only possesses the advantages of the original algorithm, but also improve the running speed of the algorithm. Applied to the scheduling problem of water supply, under the premise of without any increase in operating costs, improve the efficiency of pumping station and saving some time.The thesis contains following tasks:1. This thesis briefly introduces the multi-objective optimization problem and research significance of multi-objective evolutionary algorithm.2. Briefly introduces the multi-objective optimization problem, the basic principle of evolutionary algorithm, it also introduces several kinds of typical algorithm flow, analysis the advantages and disadvantages of algorithm.3. Briefly introduces the basic principle and research status of MEDA and MOEA/D.4. According to the characteristic of MEDA and MOEA/D, in order to improve the diversity and uniformity of the algorithm, and reduce the computing time, a new multi-objective evolutionary algorithm based on distribution estimation of decomposition is proposed. It contains the advantages of MEDA and MOEA/D.Using the idea of decomposition algorithm, It decomposed a multi-objective into a number of scalar optimization sub-problems, next using ideas of MEDA, establishing a probability model for sub-problem and sampled from the model. The data of experiments shows superiority of the algorithm.5. The new algorithm is applied to water supply scheduling problem, based on date analysis and experimental results show the superiority of the improved algorithm.