Improved Multi-objective Evolution Algorithms And Their Applications

Multi-objective optimization problems are quite common in engineering practice.Multi-objective evolutionary algorithms have become a powerful tool to solve the multi-objectiveoptimization problems. Multi-objective evolutionary algorithms inherit the advanntages ofevolutionary algorithms: their random search is parallel; they have strong capability for globaloptimization; and they can solve highly complex nonlinear problems. Multi-objective evolutionaryalgorithms are used to solve the optimization problems with multiple conflicting goals. In recent years,various multi-objective evolutionary algorithms are proposed, and many of the new evolutionaryparadigms are constantly introduced, multi-objective evolutionary algorithms have received wideattention and become a research hotspot.In the begaining of this thesis the concept related to multi-objective optimization, the basicframework of multi-objective evolutionary algorithm, the development and application ofmulti-objective evolutionary algorithm are introduced. And two classic multi-objective evolutionaryalgorithms, the evaluation indicators for algorithms are produced in details. All of above lay thefoundation for the further improvement and application of the algorithms.There are some problems with the classical algorithm: the convergence and distribution indicatorsof NSGA2are inadequate; and the computing time of SPEA2is too long.To solve above problems and the problem that the classical multi-objective evolutionary algorithmscan not converge or the solutions set can’t across the Pareto equilibrium surface, when the complexityof Pareto solution set is increased, a real-coded quantum clone multi-objective evolutionary algorithm(RQC-MOEA) is proposed. In RQC-MOEA, probability amplitude, quantum rotation gate, andcloning operator is introduced, use chaotic coding to initilize the probability amplitude;complementary single-gene Gauss mutation is used to improve the diversity of algorithm; a dynamiccrowding strategy is designed to regulate the distribution of the solution set. Experiments show thatthe running time, distribution, and convergence of RQC-MOEA is balanced. And for the test problems,when the complexity of Pareto solution set is increased, RQC-MOEA shows an obvious advantage.Then, a multi-objective evolutionary algorithm based on binary indicators (IB-MOEA) is proposed.When the target dimension of the optimization problem is increased, IB-MOEA can also converge. Afitness fuction based on binary indicators and an environmental selection operator is designed inIB-MOEA. IB-MOEA is simple, high efficiency and has good distribution and convergence. In dealing with high-dimensional target DTLZ problems, IB-MOEA shows an obvious advantage. Theapplication of IB-MOEA on IIR digital filters designing futher validates the effectiveness ofIB-MOEA.Finally, research the multi-objective optimization methord on air to groud missile tajectory, explorehow to transform the continuous optimization problem into a parameter optimization problem, thetajectory multi-objective optimization model is established, and then RQC-MOEA is used to optimizemissile tajectory. Compared with nonlinear programming method, the correctness of the optimizationmodel and the effectiveness of RQC-MOEA are proved.