부산대학교 도서관

An approach for the small-signal stability assessment (SSSA) of power systems using a Convolutional Neural Network (CNN) model with transfer learning is presented in this paper. The concept of permutation feature importance (PFI) is included in model development to identify and drop the most irrelevant features in a given dataset, which minimizes the input information required by the model to achieve a certain performance and reduces the set of measurement locations for the related application. Then, a transfer learning approach using weight initialization and feature extraction is applied to leverage the knowledge of a pretrained model when a new independent dataset (obtained from the integration of converter-interfaced generation) is considered. Simulation results demonstrate that the transfer learning-based CNN model is able to exploit previous knowledge and provide a superior performance, as compared to the traditional rebuilt-from-scratch model.