Model Predictive Control Based Data-Driven Retrofit Controller for Network Systems

In this paper, we propose a data-driven retrofit controller design method based on the robust model predictive control. Retrofit control is a distributed design of decentralized controllers for a large-scale networked system. By using the retrofit control paradigm, the damping performance of the subsystem of interest can be improved by only relying on the local information of the subsystem. In this paper, we propose the extension of the existing method by utilizing the approximate model of the environment through the system identification approach. Then, by using the robust model predictive control framework, we formulate an optimization problem which produces a retrofit controller that complies to the constraints under the presence of the interconnection signal noise. We show that although the retrofit control framework only requires the local information, the approximate model of the environment will improve the damping performance even with a certain degree of inaccuracy. Lastly, the effectiveness and robustness of the proposed method are demonstrated through numerical simulation.