A transfer learning based CNN for dynamic security assessment with small datasets and unknown data

Dynamic security assessment (DSA) methods based on deep learning (DL) have shown promising results, leading to stable power system operation and rotor angle stability of synchronous generators. However, DL-based DSA faces challenges related to model robustness against topology changes, unknown errors, and small data sets. This paper addresses these two challenges with the help of transfer learning (TL) and conditional tabular generative adversarial networks (CTGAN). First, this paper introduces CTGAN-based data augmentation to deal with expensive data collection and labeling. This data augmentation makes the proposed model applicable to small data sets. Then, with the help of augmented database, a DSA model based on convolutional neural network (CNN )is trained. Finally, when the model encounters unknown faults and topologies, it is updated using fine-tuning and only with a small data set. The results of implementing the proposed model on IEEE 39 bus systems show its superior performance.