Multi-objective optimization of dividing wall columns under model and process parametric uncertainties

与传统的蒸馏系统相比，分隔壁塔具有显著的节能潜力，同时具有高的产品纯度。如果柱设计最佳地适应相应的分离任务，则可以实现这种潜力。在申请人的初步工作中，已经为此目的开发了用于基于模型的模拟和优化的新方法。在这个项目中，目的是考虑到这些优化结果的不确定性，并制定战略，以最好的方式对这种不确定性作出反应。一方面，这涉及操作的不确定性-例如反应物成分的波动-另一方面，涉及物理模型参数的不确定性，例如热力学平衡的描述。本项目的目的是在塔的设计和操作中选择自由度时开发基于模型的策略，以便这些不确定性对上述节能潜力和产品纯度的影响尽可能小。科学的挑战是区分设计和运营管理的自由度：前者在开始时是固定的，此后无法改变;后者在一定的窗口内可以灵活调整。这导致了一类新的多目标优化问题。基于申请人的初步工作，现在有可能解决这些问题，从而获得与应用相关的见解和方法。

Dividing wall columns have the potential of significant energy savings compared to conventional distillation trains, with high product purities at the same time. This potential can be realized if the column design is optimally adapted to the respective separation task. In preliminary work by the applicants, novel approaches for model-based simulation and optimization have been developed for this purpose. In this project, the aim is to take uncertainties into account in these optimization results and to develop strategies for reacting to such uncertainties in the best possible way. On the one hand, this involves operational uncertainties - such as fluctuating reactant compositions - and on the other hand, uncertainties in physical model parameters, such as for the description of thermodynamic equilibria. The aim of this project is to develop model-based strategies in the choice of degrees of freedom in the design and operation of the columns, so that these uncertainties have as little impact as possible on the energy-saving potential and product purities addressed above. The scientific challenge is to distinguish between the degrees of freedom in design and operational management: The former are fixed once at the beginning and cannot be changed thereafter; the latter are flexibly adjustable during operation within a certain window. This leads to a novel class of multi-objective optimization problems. Based on the preliminary work of the applicants, there is now the possibility to solve these and thus to arrive at insights and methods relevant for the application.