基于元学习与改进Transformer的N-k小样本暂态稳定事故筛选方法OA

The number of N-k contingency combinations is enormous,and existing data-driven transient stability contingency screening methods require large amounts of training data derived from representative operating scenarios,resulting in high computational costs and difficulty in meeting practical N-k screening requirements in power systems.To address this issue,a few-shot N-k transient stability screening method based on meta-learning and an improved Transformer is proposed.The method can infer the transient stability of unknown high-order N-k contingencies using a limited number of low-order N-1 fault samples.First,fault feature matrices are constructed using pre-and post-fault electrical quantities,forming N-k meta-learning tasks where low-order faults serve as the support set and high-order faults as the query set.Then,considering the complex coupling characteristics between low-order and high-order contingencies,a Transformer-based meta-learning algorithm for combined failures(T-MLCF)is proposed,integrating a relation network and a contrastive network.The relation network uses an improved Transformer to learn nonlinear similarity functions between low-and high-order contingencies,while the contrastive network extracts combinatorial knowledge of their cooperative effects.Based on this framework,generalization to unseen N-k contingencies can be achieved under few-shot conditions.Finally,case studies on the IEEE 39-bus system demonstrate that the proposed T-MLCF achieves excellent performance in few-shot learning and generalization,while maintaining robustness under varying sample sizes.