A two-step model-free predictive control strategy based on extended SMO and swift voltage vector localisation for power converters

The finite control set model predictive control strategy simplifies the controller design by directly optimising a finite number of switching states. It has the advantages of fast dynamic response, easy handling of constraints, and flexible multivariate control, which makes it suitable for fast dynamic systems such as power electronics and motor drives. However, it also imposes several technical challenges, such as poor robustness and high computational burden. To address these drawbacks, this work is concerned with a new two-step model-free predictive control method based on extended sliding mode observer (SMO) and voltage vector swift localisation for three-level neutral point clamped inverters, which aims to suppress the switching frequency (SF) while solving the aforementioned challenges. Specifically, a swift voltage vector localisation method based on two-step sector division is proposed to reduce the controller's computational burden. Furthermore, on this basis, the new reference voltage vector is obtained by observing the ultra-local model with an extended SMO, and the output voltage of the controller is kept constant over two cycles, thus improving the robustness while reducing the SF. Finally, the simulation and experimental results show that the proposed scheme has the advantages of high robustness, low SF, and small computational burden, both under parameter matching and mismatching conditions.