Heßler Martin, Kamps Oliver
Center for Nonlinear Science, University of Münster, Münster, Germany.
Institute of Theoretical Physics, University of Münster, Münster, Germany.
Nat Commun. 2025 Jul 7;16(1):6246. doi: 10.1038/s41467-025-60877-0.
Critical transitions necessitate anticipation to prevent adverse outcomes. While many studies focus on bifurcation-induced tipping, noise-induced tipping is also possible. We propose to use the open-source (non-Markovian) Bayesian Langevin estimation to quantify deterministic and stochastic dynamics simultaneously. By analysing bus voltage frequency time series from the Western Interconnection blackout on 10th August 1996, complemented by conceptual network models of its key events, we reveal the interplay of changing local restoring rates and noise levels. Furthermore, a comparison of these findings to the blackout's timeline supports our frequency Langevin model driven by correlated noise. A state change is indicated two minutes before the official triggering event, potentially by establishing a tree-to-line fault. This study highlights the importance of distinguishing destabilising factors for anticipating critical transitions and provides a tool for understanding such events across various disciplines.
关键转变需要提前预测以防止不良后果。虽然许多研究关注分岔引发的突变，但噪声引发的突变也是可能的。我们建议使用开源（非马尔可夫）贝叶斯朗之万估计来同时量化确定性和随机动力学。通过分析1996年8月10日西部互联电网停电事件中的母线电压频率时间序列，并辅以关键事件的概念网络模型，我们揭示了局部恢复率变化与噪声水平之间的相互作用。此外，将这些发现与停电时间线进行比较，支持了我们由相关噪声驱动的频率朗之万模型。在官方触发事件前两分钟就显示出状态变化，这可能是通过建立树线故障实现的。这项研究强调了区分不稳定因素对于预测关键转变的重要性，并提供了一种跨学科理解此类事件的工具。
