Predictive Converter Control Using Real Time Quadratically Constrained Optimization

The paper considers model predictive control of a voltage source converter with inductive-capacitive (LC) filter. The model of the converter describes the nonlinear effect of the switching on the converter state and the MPC problem is therefore nonlinear and non-convex. An in-depth analysis reveals a convex structure of the converter model, and a nonlinear variable transformation is introduced which allows to equivalently state the MPC problem as a convex, quadratically constrained quadratic problem (QCQP). Thus, a problem which has previously been formulated as a non-convex problem and solved approximately, can be solved exactly without approximation error. The QCQP is solved using a novel first-order method suitable for real time implementation.