The Aea Combined With Differential Evolution Algorithm And Its Application On Constrained Optimization Problems

There exist a large number of optimization problems in the real word, especially in the fields of science and engineering. Because of the different characteristics of the problems, it is ineffective to solve these problems while using mathematic methods. As a population-based global optimization method, and its simplicity and robustness, intelligent evolutionary algorithm showed strong competitiveness in solving optimization problems. The appearance of intelligent evolutionary algorithm provided a new idea on the solution of problems, and it has been a hot topic in the area of evolutionary computation.Alopex-based Evolutionary Algorithm(AEA) is a new optimization algorithm which integrated the swarm intelligence of evolutionary algorithm and the heuristic way of Alopex together. AEA inherits both the merits of the simulated annealing, and the advantages of gradient descent. In order to solve optimization problem, the trial population in AEA is improved by an improved differential evolution operation. Moreover, the improved operations not simply balance the local search and the global search, but increase the population diversity. The possibility of trapping into local optimum was dramatically decreased, and the optimization process is more reasonable. Tested by benchmark functions, the comparison results indicate that the performance of the modified algorithm is significantly improved in both accuracy and stability. Furthermore, the algorithm was applied to the parameter estimation of the models of fermentation dynamics, and satisfactory results were obtained.This paper also proposed a new type of self-adaptive penalty function and constraint handling approach for constraint optimization problems. With the introduction of the parameter ε, the purpose of retaining favorable information in infeasible solution is reached. Controlled by ε, the searching region converges to feasible region gradually, which contributes to the effective search of boundaries of feasible area and infeasible area. By constructing adaptive penalty mechanism, and designing the penalty coefficient self-adaptive adjustment, the irrational penalties are avoided. Testing by benchmark functions, the experimental results indicate that the new handling method is stable and efficient while solving constraint optimization problems.