| 引用本文: | 牛哲文,武超辉,冀 岳,等.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[J].电力系统保护与控制,2026,54(08):104-115. |
| NIU Zhewen,WU Chaohui,JI Yue,et al.Adaptive power system transient stability assessment based on conditional denoising diffusion models and active transfer learning[J].Power System Protection and Control,2026,54(08):104-115 |
|
| 摘要: |
| 随着电力系统结构日益复杂，数据驱动的暂态稳定评估(transient stability assessment, TSA)方法因其响应速
度快、建模灵活等优势，受到广泛关注。然而，该类方法在实际应用中面临两个关键问题：1) 暂态失稳事件发生
概率低，导致失稳样本极度稀缺，训练数据呈现严重不均衡，影响模型泛化能力；2) 模型通常在离线状态下训练，
难以适应实际运行中系统结构与工况的频繁变化，制约其在线评估精度。为此，提出一种融合条件扩散模型与主
动迁移学习的暂态稳定自适应评估方法。首先，针对样本分布不均衡问题，引入条件去噪扩散概率模型(conditional
denoising diffusion probabilistic model, CDDPM)，以系统稳定性指标为条件先验，引导样本生成过程，从而增强失
稳样本分布，提升模型对极端工况的识别能力。其次，构建主动迁移机制，联合迁移学习与主动样本选择策略，
实现评估模型在新场景下的快速适应与高效更新。最后，在IEEE39节点与118 节点系统中验证了所提方法的有
效性和优越性。 |
| 关键词: 暂态稳定评估 样本增强 主动学习 迁移学习 自适应评估 |
| DOI：10.19783/j.cnki.pspc.251127 |
| 分类号: |
| 基金项目:国家自然科学基金项目 (52507132)；煤电清洁智能控制教育部重点实验室开放基金项目 (CICCE202413) |
|
| Adaptive power system transient stability assessment based on conditional denoising diffusion models and active transfer learning |
|
NIU Zhewen, WU Chaohui, JI Yue, HAN Xiaoqing, QU Ying
|
|
1. Shanxi Key Laboratory of Power System Operation and Control (Taiyuan University of Technology), Taiyuan 030024, China; 2. State Grid Shanxi Electric Power Research Institute, Taiyuan 030001, China
|
| Abstract: |
| Aiming at the problems of scarce transient instability samples, unbalanced data and insufficient model generalization ability, an adaptive assessment method integrating conditional denoising diffusion model and active transfer learning is proposed. High-quality unstable samples are generated guided by stability indicators, and an active transfer mechanism is constructed to realize rapid model update in new scenarios. The effectiveness and superiority are verified in IEEE 39-bus and 118-bus systems. |
| Key words: transient stability assessment data augmentation active learning transfer learning adaptive assessment |
