Abstract:Current data-driven predictive control (DDPC) methods heavily rely on data collected in open-loop operation with elaborate design of inputs. However, due to safety or economic concerns, systems may have to be under feedback control, where only closed-loop data are available. In this context, it remains challenging to implement DDPC using closed-loop data. In this paper, we propose a new DDPC method using closed-loop data by means of instrumental variables (IVs). By drawing from closed-loop subspace identification, the use of two forms of IVs is suggested to address the closed-loop issues caused by feedback control and the correlation between inputs and noise. Furthermore, a new DDPC formulation with a novel IV-inspired regularizer is proposed, where a balance between control cost minimization and weighted least-squares data fitting can be made for improvement of control performance. Numerical examples and application to a simulated industrial furnace showcase the improved performance of the proposed DDPC based on closed-loop data.

What problem does this paper attempt to address?

The paper primarily focuses on addressing the challenges of applying Data-Driven Predictive Control (DDPC) under closed-loop data conditions. Traditional DDPC methods typically require carefully designed inputs under open-loop operation to collect data that ensures persistent excitation. However, in actual industrial scenarios, systems are often under closed-loop control for safety or economic reasons, thus only closed-loop data can be obtained. In such cases, directly implementing DDPC using closed-loop data becomes highly challenging. The paper proposes a new DDPC method based on closed-loop data by introducing Instrumental Variables (IVs) to address the issue of implementing DDPC under closed-loop conditions. Specifically, the authors draw on the methods of Closed-Loop Subspace Identification (SID) and propose two forms of IVs: one is the future feedback reference signal, and the other is an IV designed based on the Left Coprime Factorization (LCF) of the controller. Additionally, the paper presents a new regularized DDPC formulation that includes a novel IV-inspired regularization term, which can balance minimizing control costs and weighted least squares data fitting, thereby improving control performance. Through numerical examples and an application case of a simulated industrial furnace system, the proposed closed-loop data-based DDPC method is shown to have better control performance compared to traditional DDPC algorithms. These results demonstrate the effectiveness and superiority of implementing DDPC using closed-loop data.

Data-Driven Predictive Control Using Closed-Loop Data: An Instrumental Variable Approach

Closed-loop Data-Enabled Predictive Control and its equivalence with Closed-loop Subspace Predictive Control

On the equivalence of direct and indirect data-driven predictive control approaches

Causality-Informed Data-Driven Predictive Control

Adaptive Predictive Control: A Data-Driven Closed-Loop Subspace Identification Approach

A Data-Driven Closed-loop Subspace-aided Approach to Predictive Controller Design

Open-/Closed-loop Active Learning for Data-driven Predictive Control

A Linear Parameter-Varying Approach to Data Predictive Control

Online Reduced-Order Data-Enabled Predictive Control

Robust and efficient data-driven predictive control

Combining Data-Driven and Event-Driven for Online Learning Predictive Control in Power Converters

From Data to Predictive Control: A Framework for Stochastic Linear Systems with Output Measurements

Uncertainty-aware data-driven predictive control in a stochastic setting

Generalized Data-driven Predictive Control

Dynamic-Linearization-Based Predictive Control of a Voltage-Source Inverter

Data-Driven Modeling and Predictive Control for Boiler–Turbine Unit

Innovation-triggered Learning with Application to Data-driven Predictive Control

Efficient Recursive Data-enabled Predictive Control (Extended Version)

Deep DeePC: Data-enabled predictive control with low or no online optimization using deep learning

Frequency-Domain Data-Driven Predictive Control

Constrained Robust Model Predictive Control Embedded with a New Data-Driven Technique