基于均匀布点的全局优化方法

A Global Optimization Method Based on Uniform Design

把实验设计中的均匀设计思想引入优化设计 ,提出了一种将确定性和随机性相结合的优化方法。该方法根据均匀设计原理在优化模型的设计变量空间内均匀分布一系列点 ,然后 ,将可行域内的上述系列布点作为优化计算的系列初始点 ,并选用常规优化算法 ,分别开始进行优化计算 ,得到优化模型的一系列局部最优点。最后 ,比较所有局部最优点的最优值 ,即认为在一定程度上获得了该优化问题的全局最优解。该算法可求取非线性多峰函数的全局最优解。编制了计算程序 ,给出了计算实例 ,计算结果表明该设计方法是可行的。

In this paper, some ideas of uniform design in the experimental design were introduced into optimization design and a new method of design combining deterministic method with random method was discussed. The global optimal solutions of nonlinear multi-peak function can be found by this method. A series of uniformly distributed points are generated by the principle of the uniform design in variable design space. These points are regarded as series start points of the optimization model. The conventional optimize algorithms are selected to compute and a series of local minimum values can be gained.After comparing with each other, the best value of all local minimum values can be found out, the value is considered as the global minimum in some degree. According to the method, a program was compiled and an example of design was implemented. The design results of the example testified that the method is feasible.