Predictor neural network-based model-free predictive control using virtual voltage vector for multiparallel power converters

In recent years, multilevel converter has attracted attention due to its advantages such as higher voltage capability and lower voltage distortion rate. To meet high-power demands, the parallel operation of inverters has become a necessary choice. However, parallel operation of inverters can lead to severe zero-sequence circulating current problems, affecting the quality of system output power. The conventional control methods, moreover, require the knowledge of the exact model of the system and suffer from the problem of poor robustness. In this paper, an innovative control scheme is proposed to address this issue. Specifically, this scheme rapidly identifies and models unknown nonlinearities and uncertainties of the system, combines a feedback mechanism for prediction errors to update neural predictors, and introduces virtual voltage vectors to prevent the occurrence of zero-sequence circulating currents. Furthermore, the main contribution of this paper is that the proposed method can smoothly and quickly capture system dynamics, improve the robustness and reliability of the control system in the presence of parameter uncertainties, achieve suppression of zero-sequence circulating currents, and exhibit good current tracking accuracy. Finally, comprehensive simulation and experimental results are presented to verify the efficacy of the proposed control method for multiparallel power converters.