Transfer Learning-Based Power System Online Dynamic Security Assessment: Using One Model to Assess Many Unlearned Faults

This letter proposes a novel data-driven method for pre-fault dynamic security assessment (DSA) of power systems. To address the large number of potential faults, the proposed method aims to use one trained model to work for multiple faults. Firstly, a hybrid learning based DSA model is initially trained by one fault database. Then, based on transfer learning technique, the model is transferred to an unknown different but related fault by iteratively minimize the marginal and conditional distribution differences between the trained data and unknown data. Thus, the extensibility of the DSA model is greatly enhanced and the need for training a large number of models is eliminated. Test results have demonstrated the high accuracy of the proposed method.