融合条件去噪扩散模型与主动迁移学习的电力系统暂态稳定自适应评估方法OA

With the increasing complexity of power system structures,data-driven transient stability assessment(TSA)methods have gained significant attention due to their fast response and flexible modeling capabilities.However,two key challenges hinder their practical application:1)transient instability events occur infrequently,resulting in extremely scarce unstable samples and severely imbalanced training data,which degrades model generalization;2)models are typically trained offline and struggle to adapt to frequent changes in system topology and operating conditions,limiting their online assessment accuracy.To address these issues,this paper proposes an adaptive TSA framework based on conditional denoising diffusion probabilistic models(CDDPM)and active transfer learning.First,to mitigate sample imbalance,a CDDPM is introduced,where system stability indicators are used as conditional priors to guide the sample generation process,thereby enhancing the distribution of unstable samples and improving the model's ability to identify extreme scenarios.Second,an active transfer learning mechanism is developed by integrating transfer learning with active sample selection strategies,enabling rapid adaptation and efficient model updating in new scenarios.Finally,case studies on the IEEE 39-bus and 118-bus systems validate the effectiveness and superiority of the proposed method.