Research On Single Objective Optimization Employing Evolutionary Computation

Many optimization problems exist widely in scientific research and engineeringapplication. With the increasing of dimension, the number of local optima andconstrained condition, the classical optimization algorithms are difficult to obtainqualified solution. However the evolutionary algorithms perform well.To solve single-objective unconstrained single-objective optimization problemand single-objective multiple constrained optimization problems, the evolutionaryalgorithms are employed. The qualities of optimization problem solutions areimproved by modifying the exploring and exploiting ability of evolutionaryalgorithms. The research results are profound and systematic. The Main resultsinclude the followings:(1) In order to improve the ability of neighborhood search of differentialevolutionary (DE) algorithm, we propose a new variant of DE with linearneighborhood search, called LiNDE, for global optimization problems (GOPs).LiNDE employs a linear combination of triple vectors taken randomly fromevolutionary population. The main characteristics of LiNDE are less parameters andpowerful neighborhood search ability. Experimental studies are carried out on abenchmark set, and the results show that LiNDE significantly improved theperformance of DE. A ccording to experimental results of LiNDE, we can learn thatLiNDE is better than SANSDE, SaDE and NSDE. LiNDE has innovation.(2) In order to exploit and preserve the diversity of immune optimizationalgorithm when solving high dimensional global optimization problems, a novelevolutionary optimization algorithm based lifespan (LIO) model is proposed. LIOincorporates a lifespan model, local and global search procedure to improve theoverall performance in solving global optimization instance. The experimental resultsof the LIO are significantly better than that of the conventional clonal selectionalgorithm (CSA) in terms of the performance evaluation criterion proposed.According to experimental results of LIO, we can learn that LIO is better thanCSA/FC, CSA/RC and CSA/SC. LIO has innovation.(3) A new hybrid Good Point set algorithm is applied to solve the constrainedproblems. Good Points Set, can make the local search achieve the same sound results just as the state-of-the-art methods do, such as orthogonal method. But the precisionof the algorithm is not confined by the dimension of the space. An integratedmechanism is used to enrich the exploration and exploitation abilities of the approachproposed. Experiment results on a set of benchmark problems show the efficiency ofthe algorithm. According to experimental results of COAGPN, we can learn thatCOAGPN is better than SAFFand SMES.