Data-Driven Predictive Current Control for Active Front Ends with Neural Networks

Active front end (AFE) acts as the interface of energy conversion for renewable energy generation systems and gradually becomes more and more prominent. For controlling AFE, finite-control-set model predictive control (FCS-MPC) has been considered a promising alternative. However, owing to its high dependence on system models, system parameter variations (in particular, the grid-side inductance) and external disturbance will seriously result in degradation of its control performance and even instability. Therefore, in this work, a data-driven predictive control (DDPC) method with a neural network (NN) is proposed and validated for an AFE. Based on the existing NN predictor, the proposed solution not only covers the robustness of state variables against parameter variations, but also takes the input variables into account, which further enhances the system robustness. Control performances of the proposed method are validated and compared with the classical FCS-MPC scheme through both simulation and experimental results.