Hybrid model generation for superstructure optimization with Generalized Disjunctive Programming

We propose a novel iterative procedure to generate hybrid models (HMs) within an optimization framework solve design problems. HMs are based on first principle and surrogate (SMs) they may represent potential plant units embedded superstructure. initial SMs with simple algebraic regression refine them by adding Gaussian Radial Basis Functions in three steps: SM refinement, domain exploration, and, after solving the optimal problem, further exploitation, until convergence criterion is fulfilled. The superstructure problem formulated Generalized Disjunctive Programming solved Logic-based Outer Approximation algorithm. addressed methanol synthesis propylene Compared rigorous model-based design, proposed gave same configuration, objective function decision variables maximum relative differences of 1 7 %, respectively. A sensitivity analysis shows that strategy reduced CPU time 33 %.