基于禁忌搜索的全局记忆增强型优化方法

针对过程系统优化问题中存在多解、而目前通用高性能实时优化算法均为局部算法的情况,提出了基于禁忌搜索的全局记忆增强型优化方法(global mnemonic enhancement optimization,GMEO)。该方法证明了在一定条件下局部最优解是过程系统参数的连续可微函数,并采用禁忌搜索方法以背景计算的方式获得优化计算经验点上的多解信息。基于这些多解信息,GMEO可为优化计算提供全体局部最优解的逼近值。以这些逼近值为初值,可使局部优化算法达到全局优化的效果,并能有效提高过程系统运行的稳健性。基于原油混合问题的数值仿真试验验证了GMEO方法的有效性。在不过多损失实时性的前提下,优化计算的目标函数值得以显著改善。

Global mnemonic enhancement optimization based on taboo searchs

There are many multi-solution optimization problems in process system engineering;however,almost all practical solvers are local.This paper presents a global mnemonic enhancement optimization(GMEO) approach to endow solvers to find global solutions.This analysis shows that each local optimum is a first-order differentiable function of the system parameters,so a taboo search can be used in the background to explore the local solutions.Thus,the method is able to approximate all the local solutions of the current optimization problem.The local solvers provide global optimization by using these approximate results as initial points.This increases the chance of successful convergence,making the system more robust.Numerical tests on a pooling problem validate the method.The objective function is significantly improved at an acceptable loss of real-time performance.