A Tracing Approach for a Subsynchronous Oscillation Source in a Power System with Grid-Connected PMSG

This paper proposes a method to identify and locate the source unit which induces subsynchronous oscillation (SSO) in a power system with grid-connected permanent magnet synchronous generator (PMSG). The method is carried out based on the open-loop modal coupling theory and deep transfer learning (DTL) algorithm. The location model contains two parts: training model and transferring model, the training model uses the simulation system to obtain the training data and uses Convolutional Neural Network (CNN) to extract the features and then establish the relationship model of oscillation source and measurements. The transfer model is obtained by adding regular parameters to the training model and fine-tuning the final trained model, which is used for the oscillation source localization of the actual system. In order to demonstrate and validate the proposed approach, this paper designs two different simulation systems, and the results indicate that the proposed method has the advantages of high location accuracy and convenient online application, besides, this method avoids the establishment of theoretical models, especially, the induced mechanism has become more complicated.