Robust model predictive control with embedded multi-scenario closed-loop prediction. (June 2021)

• Record Type: Journal Article
• Authors: MacKinnon, Lloyd; Li, Hao; Swartz, Christopher L.E.
• Is Part Of: Computers & chemical engineering. Volume 149(2021)
• Publication Date: 2021-06
• Subjects: Robust MPC -- Multi-scenario -- Closed-loop prediction -- Uncertainty; Chemical engineering -- Data processing -- Periodicals; 660.0285
• DOI: 10.1016/j.compchemeng.2021.107283
• Languages: English
• ISSNs: 0098-1354

Abstract

Highlights: Novel multi-scenario robust MPC formulation. Embedded MPC subproblems generate predicted closed-loop response of uncertain plant scenarios. Inner MPC subproblems reformulated as algebraic constraints corresponding to their first-order optimality conditions. Implemented on linear and nonlinear case studies with linear prediction model. Shown to outperform standard MPC under a variety of uncertainty scenarios. Abstract: Robust MPC seeks to mitigate the effects of uncertainty which can lead to suboptimal MPC performance. Previous work on robust MPC includes a stochastic multi-scenario approach to simulate multiple plant realizations and create a control scheme to optimize a performance metric based on the plant scenarios. This paper seeks to combine a scenario-based approach with embedded closed-loop prediction under future MPC control action by directly incorporating MPC subproblems into the overall robust MPC formulation. This allows the current MPC to predict closed-loop MPC responses to a range of uncertain future plant realizations. The resulting multilevel programming problem is solved by reformulating the inner MPC optimization subproblems as algebraic constraints corresponding to their first-order optimality conditions, resulting in a single level mathematical program with complementarity constraints (MPCC). The performance of the robust MPC scheme is evaluated against a standard MPC formulation in linear and nonlinear case studies.