Real-time optimization meets Bayesian optimization and derivative-free optimization: A tale of modifier adaptation

• Published in: Computers & chemical engineering 2021-04, Vol.147, p.107249, Article 107249
• Main Authors: Chanona, E. A. del Rio, Petsagkourakis, P., Bradford, E., Graciano, J. E. Alves, Chachuat, B.
• Format: Article
• Language: English
• Subjects: Acquisition function; Bayesian optimization; Gaussian process regression; Model-free RTO; Modifier adaptation; Real-time optimization; Trust region
• ISSN: 0098-1354; 1873-4375
• DOI: 10.1016/j.compchemeng.2021.107249

•Modifier adaptation with embedded GPs to capture plant-model mismatch in cost and constraints.•Combination of trust-region ideas and acquisition functions to balance exploration vs exploitation.•Enhanced reliability of modifier adaptation with prior (nominal) models compared to model-free RTO.•Effectiveness demonstrated on a batch-to-batch photobioreactor optimization with a dozen inputs. This paper investigates a new class of modifier-adaptation schemes to overcome plant-model mismatch in real-time optimization of uncertain processes. The main contribution lies in the integration of concepts from the fields of Bayesian optimization and derivative-free optimization. The proposed schemes embed a physical model and rely on trust-region ideas to minimize risk during the exploration, while employing Gaussian process regression to capture the plant-model mismatch in a non-parametric way and drive the exploration by means of acquisition functions. The benefits of using an acquisition function, knowing the process noise level, or specifying a nominal process model are analyzed on numerical case studies, including a semi-batch photobioreactor optimization problem with a dozen decision variables.