Parallel particle swarm optimization algorithm with tabu table for operation parameter optimization of fractionator

DOI：10.7511/dllgxb201902012

 作者 单位 王雅琳,孙家舟,薛永飞,尚丹丹,袁小锋

多元分馏塔操作优化需要反复求解其机理模型中的高维耦合非线性方程组，计算十分耗时．针对此昂贵优化问题，提出了一种带禁忌表的并行粒子群优化(parallel particle swarm optimization，PPSO)算法．以分馏过程机理知识初始化操作参数的禁忌表，再根据已测试候选参数的优化性能动态更新该禁忌表，实现对每次迭代产生的候选解初选，减少禁忌解的计算；接着，以MapReduce的任务分发思想构建算法的并行处理框架，设计任务预分配策略减少通信成本，提高计算效率；最后在子系统中通过对无法求解粒子与不合格粒子的剔除，筛选出原约束优化问题的等价解，进一步减小计算开销．以实际操作优化问题验证了所提算法的有效性，结果表明该算法能够更快地找到分馏系统操作参数的最优设定值．

The operation optimization for a fractionator in multicomponent system requires solving the mechanism model repeatedly, which contains a large number of high dimensional coupled nonlinear equations. This is very time-consuming. To solve the problem, a parallel particle swarm optimization (PPSO) algorithm with tabu table is proposed. Firstly, a tabu table of the operating parameters is initialized according to the mechanism knowledge of fractionation, and then it is dynamically updated according to the optimization performance of the tested candidate parameters. For every iteration, the tabu table is used to pass the solutions in the tabu table. To further improve the computational efficiency and reduce the communication cost, the parallel processing framework and the task pre-allocation strategy are adopted in the proposed algorithm, whose idea is similar to MapReduce. Finally, in the subsystem, the original constrained optimization problem is solved by rejecting the particle that can not be solved or does not meet the production demand. The effectiveness of the proposed algorithm is validated by an actual operation optimization problem, which can provide a set of optimum operating parameters for a fractionation system quickly.