Low-Computational Model Predictive Control of T-Type Three-Level Inverter with Fixed Switching Frequency

Conventional Model Predictive Control (C-MPC) typically suffers from an unfixed switching frequency and requires significant computational power from the controller, which restricts its widespread industrial application. To overcome this limitation, this paper proposes a low-computational control strategy for achieving a fixed switching frequency in aT-type three-level inverter. In this strategy, the 27 vectors are initially grouped together, and the candidate vector group for optimized rolling is determined by evaluating the position of the reference voltage vector in space. Subsequently, the vectors are categorized into P-type vectors and N-type vectors based on their impact on the capacitance voltage of the DC side. By comparing the magnitude of the capacitance voltage and simultaneously balancing the neutral-point voltage of the DC side, the candidate vectors are further narrowed down. Importantly, the objective function no longer needs to consider neutral-point voltage balancing, thus simplifying the computational process. Lastly, the candidate vectors are evaluated, ranked according to the objective function, and assigned action times. Through simulation and experimentation, it is demonstrated that this approach enables the T -type inverter to achieve high performance while minimizing computational load.