Analysis of Transfer Learning for Output Impedance Estimation of Power Electronics Converters

The large integration of electronic power converters (EPCs) in modern power systems has led to an increased interest in the stability analysis of the interaction between EPCs and the grid. The output impedance of an EPC is a key parameter for assessing the stability of the system and the impedance-based stability criteria are widely used for this purpose. However, the conventional methods for impedance characterization require a large amount of data and time-consuming repetitive processes. Recently, data-driven techniques that utilize prior knowledge through transfer learning have shown that impedance models can be developed with a significantly reduced amount of data. This paper aims to explore the generalization capabilities of transfer learning approach for estimating the output impedance in ac/dc or dc/dc converters using transfer learning regardless of their topology and control parameters. The proposed framework is tested on a three-phase grid-tied converter, a dc-dc boost converter, and a single-phase inverter. The results show that the proposed approach can significantly reduce the amount of data needed to develop the model on the new converter, while maintaining a good level of accuracy.