An approach to solve nonlinear constrained optimization

Yuan Xiaohui,Yuan Yanbin,Zhang Yongchuan,Wang Cheng

DOI: https://doi.org/10.3321/j.issn:1671-4512.2006.04.021

2006-01-01

Abstract:The shortcomings of the standard genetic algorithm was discussed. On the basis of real-value encoding technology, the genetic algorithm for self-adaptive chaotic was presented to solve nonlinear constrained optimization with the complexity. In accordance with the characteristics of real-value encoding and probability distribution function, a crossover operator was constructed. In combination of the chaotic dynamic characteristics with the artificial neural network theory, a self-adaptive chaotic mutation operator was designed. The population diversity of the algorithm was kept by this operator to prevent and overcome the premature phenomena in the evolutionary process. Constrained conditions were dealt with by direct comparison penalty function method without penalty factor. It was from the numerical experiments seen that this algorithm showed its good solution precision and convergence speed.