Performance Comparison of Finite Control Set Model Predictive Control Approaches Applied to Three-Phase Two-Level Inverters

This paper presents a comparison among different finite control set model predictive control approaches for voltage control in a three-phase, two-level voltage source inverter. The trade-off between computational cost and closed-loop performance, measured with mean squared error (MSE), total harmonic distortion (THD), and power losses related to switching frequency, is used to evaluate the controllers and the impact of increasing their prediction and control horizons. A cost function that uses additional terms considering the accumulated averaged difference between the absolute values of references and outputs has proven to give the best trade-off between reference tracking and power losses in all the considered cases, at the cost of always being the slowest option. On the other hand, a different cost function that considers a simpler formulation for the accumulated errors was almost 30% faster than the previous one when unit horizons were considered, with reasonably similar performance. The use of two-step horizons has proven to be a good choice, because it brought the biggest improvement in MSE and THD, while requiring a relatively low computational effort, regardless of the technique chosen. The increase in the horizon length to values greater than two samples did not bring sufficient improvement to justify the resulting computational burden.