Optimal data-driven predictive control for transcritical CO2 refrigeration systems with application to supermarket

This study presents an optimal data-driven predictive control (DDPC) framework for transcritical CO2 supermarket refrigeration systems, characterized by high interconnection and nonlinear dynamics. Classical feedforward PID (FFPID) controllers are effective for Multiple Input Multiple Output (MIMO) systems with low coupling but struggle in highly interconnected systems without precise models. To overcome these limitations and the necessity of having a precise model in more advanced model-based controllers, a DDPC approach is proposed that utilizes Singular Value Decomposition (SVD) to eliminate the need for regularization and incorporates the measurable disturbance in the Hankel matrices to achieve higher performance in disturbance rejection. A key focus of this work is on disturbance rejection, which is more challenging than trajectory tracking due to the sudden nature of disturbances, as opposed to the smoother changes driven by ambient temperature. The framework’s effectiveness is demonstrated in two severe scenarios. In the first, a 45 % load increase tests the system’s ability to stabilize evaporator air temperature and gas cooler pressure, outperforming conventional methods. The second scenario, involving disturbances from an on/off evaporator, highlights the algorithm’s robustness in maintaining suction and gas cooler pressures under demanding conditions. The results underscore the offset-free DDPC framework’s capability to enhance energy efficiency, robustness, and practicality, providing a reliable alternative for managing the complex dynamics of transcritical CO2 systems.