An Adaptive Transient Stability Assessment Method for Power Systems Based on DANN

Data-driven transient stability assessment (TSA) methods have become a focal point in power system stability and security research. However, when the operational modes and topological structures of power systems change frequently, the performance of data-driven models deteriorates, and their generalization capability declines. To address this issue, the paper proposes an adaptive transient stability assessment method for power systems based on Domain-Adversarial Neural Networks (DANN). After a power system operational scenario switches, domain-adversarial training is introduced via a gradient reversal layer to minimize the distribution discrepancy between source and target domain data, enabling the transfer of the source domain assessment model and thereby enhancing its generalization performance. Tests on a modified IEEE 39-bus system with renewable energy sources show that the proposed method exhibits strong adaptability to scenario changes and improves accuracy of TSA after operational scenario transitions.