Quadratic Regularization of Data-Enabled Predictive Control: Theory and Application to Power Converter Experiments

TL;DR

This paper investigates the impact of quadratic regularization on data-enabled predictive control (DeePC), providing theoretical insights and demonstrating its effectiveness in power converter applications through simulations and experiments.

Contribution

It offers a theoretical understanding of quadratic regularization in DeePC and introduces a framework for power converter control, validated by high-fidelity tests.

Findings

Quadratic regularization improves robustness against disturbances.

Closed-form solutions enable faster computations when constraints are inactive.

The proposed framework effectively controls power converters in simulations and experiments.

Abstract

Data-driven control that circumvents the process of system identification by providing optimal control inputs directly from system data has attracted renewed attention in recent years. In this paper, we focus on understanding the effects of the regularization on the data-enabled predictive control (DeePC) algorithm. We provide theoretical motivation and interpretation for including a quadratic regularization term. Our analysis shows that the quadratic regularization term leads to robust and optimal solutions with regards to disturbances affecting the data. Moreover, when the input/output constraints are inactive, the quadratic regularization leads to a closed-form solution of the DeePC algorithm and thus enables fast calculations. On this basis, we propose a framework for data-driven synchronization and power regulations of power converters, which is tested by high-fidelity simulations and experiments.