Data-driven load disturbance rejection (2307.01700v1)

Abstract: Data-driven direct methods are still growing in popularity almost three decades after they were introduced. These methods use data collected from the process to identify optimal controller's parameters with little knowledge about the process itself. However, most of those works focus on the problem of reference tracking, whereas many of the problems faced in real-life are of disturbance rejection or attenuation. Also, the vastly majority of those works identify the parameters of linearly parametrized controllers, which amounts to fixing the poles of the controller's transfer function. Although the identification of the controller's poles is not prohibitive, as hinted by some of the papers, there is little effort on presenting a data-driven solution capable of doing so. With all that in mind, this work proposes a data-driven approach which is able to identify the zeros and the poles of a linear controller aiming at disturbance rejection. Two different one-step ahead predictors are proposed, one that is linear on the parameters and another that is non-linear. Also, two different techniques are employed to estimate the controller parameters, the first one minimizes the quadratic norm of the prediction error while the second one minimizes the correlation between the prediction error and an external signal. Simulations show the effectiveness of the proposed methods to estimate the optimal controller parameters of restricted order controllers aiming at disturbance rejection.