基于OpenMP的并行GA加速求解换热网络设计

Parallel Genetic Algorithm for Design of HEN Based on OpenMP System

为了提高化工过程系统中大规模优化问题的求解效率,提出了一个基于OpenMP系统的并行遗传算法。该算法实现了CPU主线程和GPU线程的同步并行化,达到了加速求解优化问题的目的。该算法在基本遗传算法的基础上引入了一系列调节和控制策略,用于改善算法的收敛性,提高算法获得最优解的概率。通过对算法中各项操作的并行性分析,设计了CPU-GPU异构系统下的并行遗传算法,并最终在OpenMP系统下得以实现。以2个不同规模的换热网络优化问题为例,验证算法的准确性和有效性。优化结果表明:基于OpenMP的并行遗传算法不但可以得到比文献中更优的换热网络设计方案,而且与串行的遗传算法相比具有明显的加速效果。而且加速比随着换热网络优化问题规模的增大而增大这一特征将有利于化工过程系统中各类优化问题的快速准确求解。

In order to improve the solving efficiency of the large-scale optimization problem in chemical process industries, we proposed a parallel genetic algorithm based on OpenMP system, which realizes the acceleration on solution of the optimization problem by synchronous parallelization of the CPU and GPU threads. In the proposed algorithm, a series of adjustment and control strategies were introduced to the basic GA to increase the possibility of finding the global optimum. A parallel GA on CPU-GPU heterogeneous system was designed via an analysis of the operators, which was finally implemented on OpenMP system. We took two HEN optimization problems with different scales as examples to verify the effectiveness of the parallel GA. Results indicated that the proposed GA on OpenMP system can not only obtain a better solution than those in literature, but also present a significant speedup ratio when comparing with the one in sequential. Moreover, the speedup ratio of the parallel GA increases with an increase in the scale of the HEN problems. Thus, our work will facilitate an accurate and quick solution of the optimization problems in practical process industries.