Lightweight Digital Twin of Power Electronics

This paper presents a method for constructing a lightweight digital twin of power electronic devices based on data-driven approaches. Initially, a data-driven model of the power electronic device is developed using data collected from the device. The predictive capabilities of this data-driven model are then utilized to digitally enhance model predictive control (MPC), thereby creating an MPC strategy based on the data-driven model. Finally, the offline model is trained and updated online, enabling the construction of the lightweight digital twin of the power electronic device. The effectiveness of the data-driven model predictive control is validated through experiments, with frequency stability adjustment of virtual synchronous generator (VSG) used as a case study.