进化算法与确定性算法在优化控制问题中的收敛性对比

对比了进化算法(基因算法)与确定性算法(共轭梯度法)在优化控制问题中的优化效率．两种方法都与分散武优化策略-Nash对策进行了结合，并成功地应用于优化控制问题。计算模型采用绕NACA0012翼型的位流流场．区域分裂技术的引用使得全局流场被分裂为多个带有重叠区的子流场，使用4种不同的方法进行当地流场解的耦合，这些算法可以通过当地的流场解求得全局流场解。数值计算结果的对比表明．进化算法可以得到与共轭梯度法相同的计算结果．并且进化算法的不依赖梯度信息的特性使其在复杂问题及非线性问题中具有广泛的应用前景。

Comparison of convergence between evolutionary algorithms and deterministic algorithms for optimal control problem

The comparison for optimization efficiency between evolutionary algorithms (Genetic Algorithms, GAs) and deterministic algorithms (Conjugate Gradient, CG) is presented. Both two different methods are combined with Nash strategy-decentralized optimization strategy in Game Theory-and implemented into an optimal control problem using a technique DDM (Domain Decomposition Method). The problem consists in simulating the perfect potential flow field around a NACA0012 airfoil with the technique DDM, the global calculation domain is then split into sub-domains with overlaps, the accord of local solutions on interfaces is obtained using four different algorithms which permit the resolution of global problem via local sub-problems on sub-domains and their interfaces. Comparable numerical results are obtained by different algorithms and show that the property of independence of gradient makes GAs based algorithms serious and robust research tools for great dimension problems or non-linear problems.