Risk-Averse PID Tuning Based on Scenario Programming and Parallel Bayesian Optimization

Abstract

The pervasiveness of PID control in process industries stipulates the critical need for efficient autotuning techniques. Recently, the use of Bayesian optimization (BO) has been popularized to seek optimal PID parameters and automate the tuning procedure. To evaluate the overall risk-averse performance of PID controllers, scenario programming that considers a wide range of uncertain scenarios provides a systematic method, but induces extensive simulations and expensive computations. Parallel computing offers a viable method to address this issue, and thus we propose a novel parallel BO algorithm for the risk-averse tuning, which enjoys a higher efficiency in both surrogate modeling and surrogate optimization. For the latter, a multiacquisition-function strategy with diversity promotion is developed to generate widely scattered query points to parallelize experiments efficiently. For the former, a data-efficient stability-aware Gaussian process modeling strategy is designed, obviating the need for building an additional classifier as required by existing methods. Numerical examples and application to a real-world industrial bio-oil processing unit demonstrate that the proposed parallel BO algorithm considerably improves the efficiency of simulation-aided PID tuning and yields practically viable controller parameters under the risk-averse tuning framework.