人工免疫网络多Agent的分馏装置建模

分馏过程是石油化工流程模拟中的重要模块之一。在分馏装置的模拟计算中,塔板Murphree效率直接影响模型质量,且随装置的原料、工况不同而有所改变,因此探索Murphree效率的准确确定方案是分馏建模的关键。本文提出用人工免疫网络多Agent(Maopt-aiNet)确定最佳的Murphree效率。不同于其他智能算法,Maopt-aiNet结合人工免疫网络的核心思想与多Agent技术,搜索算子包括:邻域克隆选择、邻域竞争、自信度激励、自信度邻域学习和邻域协作算子,算法有机结合全局与局部搜索能力,实验表明其对高维系统搜索能力较强。在分馏装置建模应用中,用Maopt-aiNet确定的Murphree效率,能够使分馏塔的塔板温度分布及塔顶、塔釜主要产品的产出与实际吻合得较好,表明用Maopt-aiNet建立的分馏塔模型能较好地描述实际分馏塔的生产过程,可以用来指导分馏装置的操作优化。

Simulation of distillation unit based on artificial immune network multi-agent algorithm

In petrochemical field, the process simulation for distillation is an important task.The key parameter in the distillation process simulation is the tray efficiency, which can not be obtained easily.Thus the determination of appropriate tray efficiency is an important issue.In this study, artificial immune network multi-agent optimization strategy(Maopt-aiNet), which combines immune mechanics and multi-agent technology, is used to determine the Murphree efficiency.The main search operators of Maopt-aiNet include neighborhood clonal selection, neighborhood competition, self-confidence motivation, self-confidence neighborhood learning, and neighborhood collaborative operators.Based on the process and analysis data of the distillation unit, Maopt-aiNet is applied to determine the Murphree efficiency for each stage and to minimize the square summation of model’s analog relative error of the stage temperature.The experimental results show that with the tray efficiency determined by Maopt-aiNet the model fits the actual distillation unit fairly well.The method can be used to guide the operation of the distillation process.