Data-driven generalized minimum variance regulatory control

The present work proposes a design method for a data-driven generalized minimum variance (GMV) regulatory control. The new design method derives a GMV control law directly from plant operating data generated by stochastic disturbances. Thus, it does not require a plant model and an extra plant test for identifying the plant model or tuning control parameters. A novel cost function for solving data-driven GMV control parameters is introduced. The proposed cost function can be minimized by using the input-output data without using the plant model. The data-driven GMV control parameters which is obtained by using the proposed cost function correspond to the true values which minimize the cost function of the original GMV control. The efficiency of the proposed method is demonstrated through simulations.