Constrained Modulated Model Predictive Control for a Three-Phase Three-Level Voltage Source Inverter

In the last decade, model predictive control (MPC) has been widely studied in power converters, such as voltage source inverters (VSIs). Unfortunately, MPC often presents a high computational burden that limits their applicability, especially when driving multilevel (MLIs) because of higher number switching combinations than two-level inverters. As result, some strategies have developed to reduce complexity MPC. One most relevant is use artificial neural networks (ANNs) approximate behavior an However, ANNs require be evaluated at bounded inputs. Otherwise, response cannot guaranteed good approximation controller they learned from. Furthermore, LC-filtered VSI, inductor current can present peaks due cross-coupling effect between and capacitor. These cause physical damages loss performance ANN-emulated This paper new constrained modulated (M 2 PC), better suited for ANN emulation, overcome these issues. The proposed strategy evaluates cost function once per region, allowing easy intuitive constraint inclusion. Additionally, overmodulation stage used handle negative duty cycles enhance disturbance rejection. Finally, proposal validated through simulations, using MATLAB-Simulink, taking into account different load conditions. Simulations show M PC keeps below its desired limit while having (low harmonic distortions fast dynamics) even inverter operates near boundaries.