一种求解非线性约束优化问题的新方法

An approach to solve nonlinear constrained optimization

针对标准遗传算法的缺陷,提出一种基于实数编码技术的新型自适应混沌遗传算法,求解复杂非线性约束优化问题.算法根据实数编码的特点,依据概率分布函数构造杂交算子,结合混沌动力学特性和人工神经网络理论,设计了一种自适应混沌变异算子,使算法有效维持群体多样性,防止和克服进化中的“早熟”现象,同时采用不需要惩罚因子的直接比较惩罚函数方法,对约束条件加以处理.通过算例数值实验,验证了算法在提高解的精度和加快收敛速度方面都有明显改善.

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.