OPTIMAL DESIGN VIA CHANCE-CONSTRAINED OR TWO-STAGE STOCHASTIC PROGRAMMING

The solution of design problems in chemical engineering is highly desirable given the multitude of sources of uncertainty during the early design phase of a chemical plant. Two-stage stochastic programming and chance-constrained programming are two alternative formulations for the solution of these types of problems. In this contribution, a joint implementation is outlined, which is able to cope with both steady-state and dynamic underlying system models, as a fair basis of comparison of both. Specific details of the implementation to ensure both fast and reliable solution of these optimization problems are given. The comparison is carried out on two case studies, which are both nonlinear and representative of typical design problems in chemical engineering. The results highlight the curse of dimensionality caused by the underlying discretization of the uncertainty space, but also emphasize the different qualities of the results regarding interpretability and robustness, which typically favors chance-constrained programming despite the apparently more complex solution.