A Deep Neural Network Based Predictive Control Strategy for High Frequency Multilevel Converters

This paper investigates a new deep neural network (DNN) based predictive control strategy for high frequency multilevel converters. Compared to the conventional model predictive control (MPC) strategy, the new strategy significantly reduces computational burden, and improves calculation accuracy of mathematical models for power converters. A 5-level flying capacitor converter is used as a case study to compare these two strategies. It is found that the DNN based control strategy offers effective control of the load current and capacitor voltages with a much reduced computational requirement. The new concept is validated by both numerical analysis and hardware-in-loop experimental results.