融合DBSCAN与GAN集成学习的电力系统安全评估与边界生成方法OA

In the context of hybrid AC/DC power grids,the increasingly complex operating conditions and growing un-certainties of a system have imposed limitations on the traditional thermal security region(THSR)analysis method,making it difficult to accurately handle operation scenarios where multiple transmission lines simultaneously approach their stability limits.In this paper,a fast security boundary generation method integrating density based spatial cluster-ing of applications with noise(DBSCAN)and ensemble learning is proposed.First,it classifies the operating data ac-cording to line load status using DBSCAN clustering and constructs a set of security boundaries corresponding to differ-ent load levels.Second,a boundary selection model is built with the improved AdaBoost.M2 algorithm as the core,which outputs potential over-limit boundaries with high confidence and thus avoids the misjudgment risk arising from single-classification results.Finally,generative adversarial network(GAN)is utilized to integrate multi-boundary fea-tures and confidence levels,generating a high-dimensional security boundary that reflects the overall security margin.The results of a case study verify that the proposed method can realize the online rapid generation of high-dimensional security boundaries and improve the credibility and interpretability of the model,providing support for the advanced op-eration functions of power grids such as security assessment,monitoring and early-warning.