arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2607.05729eess.SYcs.SY
收藏

电力系统在线暂态稳定轨迹预测的因果结构动态图学习

Causal--Structural Dynamic Graph Learning for Online Transient Stability Trajectory Prediction in Power Systems

Texas A&M University(德克萨斯农工大学)

Ibrahim Shahbaz, Omar Al-Refai, Isaac Lagoy, Ahmad Al-Khateeb, Sathvik Sankaranarayanan, Eman Hammad

AI总结:

针对电力系统在线暂态稳定轨迹预测,提出C-DNR框架,融合动态结构图与方向因果图,经端到端边融合机制和GRU预测。在IEEE 39母线系统评估中,相比基线大幅降低预测误差,还提供可解释诊断。

摘要:

电力系统由动态耦合的发电机组成,这促使使用图神经网络(GNN)进行在线暂态稳定性预测。传统GNN框架常受限于基于固定导纳的拓扑结构无法捕捉状态依赖耦合,或受忽略方向影响的数据驱动方法。本文提出因果动态网络表示(C-DNR),一种在时间建模前融合发电机间相互作用的两种互补表示的新框架:从测量中推断出的动态结构图和通过非线性因果发现获得的方向因果图。一种端到端学习的边融合机制为每个发电机对自适应加权这些表示,所得图通过门控循环单元(GRU)传播以预测故障后轨迹。在IEEE 39母线系统上评估,C-DNR与动态结构基线相比将自回归预测误差降低73%。在评估的因果方法中,只有彼得 - 克拉克瞬时条件独立性(PCMCI)实现了持续改进,因其能从误导性振荡相关性中分离方向依赖性。学习到的融合权重还提供了与电气拓扑一致的可解释诊断,对预测过程提供透明的成对洞察。

英文摘要:

Power systems consist of dynamically coupled generators, motivating the use of Graph Neural Networks (GNNs) for online transient stability prediction. Traditional GNN frameworks are often constrained by fixed admittance-based topologies that fail to capture state-dependent coupling, or by data-driven methods that neglect directional influences. This paper proposes Causal Dynamic Network Representation (C-DNR), a novel framework that fuses two complementary representations of inter-generator interactions prior to temporal modeling: a dynamic structural graph inferred from measurements and a directional causal graph obtained via nonlinear causal discovery. An end-to-end learned edge-wise fusion mechanism adaptively weights these representations for each generator pair, and the resulting graph is propagated through a Gated Recurrent Unit (GRU) to predict post-fault trajectories. Evaluated on the IEEE 39-bus system, C-DNR reduces autoregressive prediction error by 73% compared to a dynamic structural baseline. Among the evaluated causal methods, only Peter--Clark Momentary Conditional Independence (PCMCI) achieves consistent improvements, owing to its ability to isolate directional dependencies from misleading oscillatory correlations. The learned fusion weights further provide interpretable diagnostics aligned with the electrical topology, offering transparent, pairwise insight into the prediction process.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑