2020/11/14 by Joel A. Paulson, Ali Mesbah, Paulson, Joel A. +1
Engineering · #Advanced Control Systems Optimization #Process Optimization and Integration #Fault Detection and Control Systems
paper · pdf · doi:10.48550/arxiv.2011.07445
Systematic design and verification of advanced control strategies for complex\nsystems under uncertainty largely remains an open problem. Despite the promise\nof blackbox optimization methods for automated controller tuning, they\ngenerally lack formal guarantees on the solution quality, which is especially\nimportant in the control of safety-critical systems. This paper focuses on\nobtaining closed-loop performance guarantees for automated controller tuning,\nwhich can be formulated as a black-box optimization problem under uncertainty.\nWe use recent advances in non-convex scenario theory to provide a\ndistribution-free bound on the probability of the closed-loop performance\nmeasures. To mitigate the computational complexity of the data-driven scenario\noptimization method, we restrict ourselves to a discrete set of candidate\ntuning parameters. We propose to generate these candidates using constrained\nBayesian optimization run multiple times from different random seed points. We\napply the proposed method for tuning an economic nonlinear model predictive\ncontroller for a semibatch reactor modeled by seven highly nonlinear\ndifferential equations.\n