改进遗传算法在过程系统工程中的求解策略

The solution strategy based on modified genetic algorithms for process system engineering

针对过程系统综合问题的多峰、奇异等特性 ,将遗传算法同可行域序贯搜索技术结合起来 ,实现对混合整数非线性规划问题 (MINLP)的有效求解。为克服遗传算法在可行域边界搜索效率较为低下的弊病 ,将惩罚函数同个体的生成函数有机地结合起来 ,利用惩罚函数将跨越可行域的不可行点拉回到可行域内。对过程系统综合中典型的MINLP问题的求解 ,表明该方法在求解过程中能有效地实现全局浏览 ,得到全局最优解或近优解。

Many problems are in the mixed integer non linear program(MINLP) category in the processing system synthesis and they almost are singular, multi peaks and rigid. There is not effective means to get their stable global optimal. In this paper, genetic algorithms(GAs)+feasible domain condensing was proposed for solving MINLP to get a global optimal or near global optimal solution. The penalty function was combined with genetic reproduction to overcome the drawback of genetic algorithms searching inefficiently in feasible domain boundary. Some numerical experiments of MINLP test functions and the optimization problems of system synthesis, which belong to MINLP domain, were illustrated the efficiency of that method. In addition, this method was also applied to solve several optimization problems on process system engineering and the results were useful and efficient.