Data-driven control design for load disturbance rejection by prediction error identification

This paper presents a new direct data-driven control method for the load disturbance problem in a Model Reference Matching framework. It consists in embedding the controller's design under a prediction error approach, where a flexible reference model is also identified in order to guarantee the causality and stability of the ideal controller. Due to the complexity of the proposed approach, a dedicated iterative optimization algorithm is developed to properly solve the problem. Finally, the statistical properties of the obtained estimates are explored through simulation examples, where the enhancement obtained through the proposed methodology is compared to least-squares and instrumental variable solutions.