Inter-Domain Dual-Distribution Adaptive Transient Stability Assessment

Transient stability assessment models based on deep learning usually require a large number of labeled samples for offline training. Once the power grid operating condition changes greatly, the pre-trained model deteriorates in performance and even becomes ineffective, resulting in a certain blank window during the online assessment. In order to solve this problem, we examine an inter-domain dual-distribution adaptive transient stability assessment framework. Firstly, a marginal probability distribution alignment mechanism is designed to align the marginal probability distributions of the source and target domains. This establishes a basis for similarity in probability distributions between the two domains. Secondly, a conditional probability distribution alignment mechanism is proposed. Using Bayes’ theorem, it further aligns the conditional probability distributions of the same categories in the source and target domains to improve the adaptability of the assessment framework to the power grid. The case studies conducted on the IEEE 39-bus system demonstrate the outstanding generalization and reliability of the proposed method, particularly in operating scenarios with limited availability of new samples.