A Novel Method for Transient Stability Assessment of Power Systems Based on DenseNet and Transfer Learning

With the large-scale and rapid growth of source-grid-load-storage and the increasingly close connection between AC and DC power grids, power system transient stability analysis will face more unknown operating scenarios, posing great challenges to the requirements of online transient stability assessment. When the operation mode and topological structure of the power system change, the prediction effect of the model trained only based on historical data will decline sharply. To address the above issues, this paper proposes a power system transient stability assessment method based on the densely connected convolutional network (DenseNet) and transfer learning. DenseNet is used for feature extraction and classification of power system states, and transfer learning is employed to apply existing models to new domains, thereby saving time and resources for model training. Verification is carried out on the IEEE-39 bus test case, and the results show that the proposed method in this paper has better performance and higher prediction accuracy compared with traditional transfer learning methods.