Neural-Predictor-based Data-Driven Predictive Control for 100kVA, 4.16/0.48 kV Solid-State Transformer for Smart Distribution Grids

A data-driven predictive controller (DDPC) is proposed for the solid-state transformer (SST) to manage the power flows in a distribution network. The proposed DDPC is derived to be used as a potential substitute to vector control which the design requires knowledge of the grid-connected system. The strategy uses input and output data from the system, the ultra-local model based on a neural network for modeling the dynamic of the current. The neural network enhances the robustness of the controller for nonlinear systems and external disturbances. To improve the current accuracy of the controller, a discrete space vector is used to discretize the operating region of the converter into an equivalent virtual multi-level converter. Simulation results demonstrate the effectiveness of the proposed control in managing the power flows in a medium-voltage distribution network.