Quadratic Regression Model-Based Indirect Model Predictive Control of AC Drives

Abstract

Model predictive control is a promising technique for electric drives as it enables optimization for multiple parameters and offers reliable operation with nonlinear systems. In this article, a novel approach is presented that aims to harness the advantages of both finite and continuous set model predictive methods in converter-fed ac drive control. The proposed method requires the calculation of only seven predicted states. These states are then assigned cost function values. Using a quadratic regression model, the cost function is mapped to the entire modulation region. After solving a constrained optimization problem on this cost function mapping, the optimal voltage vector is obtained, which is then applied via pulsewidth modulation. The presented method can also be applied to multilevel converter structures without the need to calculate predictions for additional voltage vectors. Therefore, the proposed method does not increase in complexity with the utilized converter topology. Furthermore, the method offers a fixed switching frequency operation and an exact noniterative solution to the optimization problem due to the formulation of the regression model. As a case study, simulation and experimental results verify the operation of the predictive torque control for permanent magnet synchronous machines with the proposed method.