Model Predictive Control of Step-up Matrix Converters

The paper proposes a predictive control scheme for a reversed operation of Matrix Converter (MxC) where the input is at the low-voltage side and acts as a current source feeding over MxC low-impedance high-voltage output side. A novel Model Predictive Control (MPC) is developed to control the step-up MxC input current and output voltage. The MPC is based on a search algorithm that selects the switching states to optimize a cost function depending on converter topology and control objectives. In addition to terms for quality of current tracking for the inner control loop, the objective function has terms for voltage tracking for the outer control loop. Since the applied load is not known in advance, a current observer is implemented to eliminate the need for additional current load sensors. The observer gains are chosen based on analytic analysis and verified using experimental results. The effectiveness of the proposed MPC control approach is validated using lab experiments on a 15 kVA custom-built prototype where testing is conducted in a step-up configuration. Additionally, the system delays due to the rapid control prototyping are characterized and a compensation of the delay in the control algorithm results in a low total harmonic distortion of output voltages.