Constraint Formulations for Bayesian Optimization of Process Simulations: General Approach and Application to Post-Combustion Carbon Capture

Abstract

Some of the most highly trusted and ubiquitous process simulators have solution methods that are incompatible with algorithms designed for equation-oriented optimization. The natively unconstrained Efficient Global Optimization (EGO) algorithm approximates a black-box simulation with kriging surrogate models to convert the simulation results into a reduced-order model more suitable for optimization. This work evaluates several established constraint-handling approaches for EGO to compare their accuracy, computational efficiency, and reliability using an example simulation of an amine post-combustion carbon capture process. While each approach returned a feasible operating point in the number of iterations provided, none of them effectively converged to a solution, exploring the search space without effectively exploiting promising regions. Using the product of expected improvement and probability of feasibility as next point selection criteria resulted in the best solution value and reliability. Constraining probability of feasibility while solving for the next sample point was the least likely to solve, but the solutions found were most likely to be feasible operating points.