Controller Tuning by Bayesian Optimization An Application to a Heat Pump

Abstract

In this paper, we consider the problem of controller tuning for an operating unit in a building energy system. As an illustrative plant example we focus on a heat pump. Since the plant is in use, the tuning method is supposed to not intervene with its operation. Moreover, the tuning procedure is supposed to be online, model-free, based only on historical data and needs to provide safety guarantees of the plant in operation. In this regard, we formulate the problem as a black-box optimization and adopt safe Bayesian optimization approaches for controller parameter tuning. These approaches are relatively new to the control community and not intensively studied in control applications. Meanwhile, the underlying systems are often expensive and performing relevant experiments is time consuming. Therefore, a crucial step prior to implementation in reality is validating the methods in simulation to verify their applicability. Toward this end, we derive a physical-based model for the heat pump and identify the unknown parameters using gray-box identification methods. Given the simulation model, we tune the controller parameters in simulation for optimal performance while considering safety constraints of the system.

Fast Posterior Estimation of Cardiac Electrophysiological Model Parameters via Bayesian Active Learning

“…In Bayesian optimization, a GP surrogate is built to approximate the objective function of the optimization, using a small number of sampling to query the expensive objective function where the samples are selected based on an acquisition function. In many areas such as nuclear physics (Ekström et al, 2019 ), material science (Ueno et al, 2016 ), and many more (Khosravi et al, 2019 ; Vargas-Hernández et al, 2019 ; Duris et al, 2020 ), Bayesian optimization is applied to estimate complex model parameters. However, all these techniques are focused on deterministic optimization to find a single optimal parameter value that best fits the simulation output to measurement data without considering the associated uncertainty.…”

Performance-based Trajectory Optimization for Path Following Control Using Bayesian Optimization

“…To reduce the number of times this experimental "oracle" is invoked, we employ Bayesian optimization (BO) [16], [17], which is an effective method for controller tuning [13], [18], [19] and optimization of industrial processes [20]. The constrained Bayesian optimization samples and learns both the objective function and the constraints online and finds the global optimum iteratively.…”

VABO: Violation-Aware Bayesian Optimization for Closed-Loop Control Performance Optimization with Unmodeled Constraints

“…BO is a sample-efficient derivative-free global optimization method [2,3] that utilizes probabilistic machine learning to intelligently search through high-dimensional parameter spaces. In recent work, BO has demonstrated potential in controller gain tuning [4][5][6][7], MPC tuning [8][9][10] and in various real-world control applications, such as wind energy systems [11,12], engines [13] and space cooling [1].…”