| --- |
| title: Eigenvalue early warning for power grid stability — the physical basis AI transfer-learning models learn to detect |
| emoji: 📚 |
| colorFrom: gray |
| colorTo: gray |
| sdk: static |
| pinned: false |
| license: mit |
| short_description: Eigenvalue early warning for power grid stability — the phys |
| --- |
| |
| Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference |
|
|
| --- |
| **TaiScience CSD-EWS** is an educational simulator for *critical slowing down* (CSD) as an early-warning signal in power grid stability. It visualizes how system eigenvalues (poles) approach the stability boundary, using a Jacobian eigenvalue proxy and damping-ratio thresholds. |
|
|
| This work is part of the TokaMind transfer-learning framework, which adapts a tokamak-plasma foundation model to power-grid dynamics. While Microsoft's GridSFM addresses static AC-OPF feasibility, this system targets *dynamic* CSD precursors — complementary directions for grid security analysis, relevant to cascading-failure events such as the 2025 Iberian blackout and the 2016 South Australia blackout. |
|
|
| ## What It Shows |
|
|
| - **Complex s-plane map** — conjugate pole pair (s = σ ± jω) with real-time position relative to the stability boundary (Re = 0) |
| - **Damping ratio** — ζ = −σ / √(σ² + ω²), with a 5% threshold separating adequate damping from warning and danger zones |
| - **Time-domain response** — illustrative y(t) = e^(σt)·cos(ωt) showing how a near-critical mode decays or diverges |
|
|
| ## Core Concept |
|
|
| Critical slowing down is a generic precursor to instability across dynamical systems: as a control parameter pushes a system toward a bifurcation, its dominant eigenvalue's real part approaches zero, recovery from perturbations slows, and damping collapses. In power systems, detecting this slowing *before* a mode crosses into the right-half plane provides an early-warning window for operators. |
|
|
| The full CSD detection pipeline (time-domain Jacobian proxy + spectral LFPR engine) is released as open source: |
| https://github.com/chihsingwu/CSD-dual-engine |
|
|
| ## Use Case |
|
|
| Teaching and intuition-building for power-system stability monitoring, transient stability assessment (TSA), and PMU-based early-warning research. This page is a manual teaching demo — not a live PMU or Toeplitz identification system. The waveform shows e^(σt)·cos(ωt) only, without modal amplitude or phase. |
|
|
| ## Related Work |
|
|
| - TaiScience Research Group (Fu Jen Catholic University): https://taiscience.org |
| - StromaPath platform: https://stromapath.com |
| - TokaMind PMU transfer-learning preprint: https://arxiv.org/abs/2605.11033 |
| - CSD dual-engine (open source): https://github.com/chihsingwu/CSD-dual-engine |
|
|
| ## Keywords |
|
|
| `critical slowing down` · `power grid early warning system` · `Jacobian eigenvalue proxy` · `damping ratio monitoring` · `transient stability assessment` · `PMU` · `TokaMind transfer learning` · `s-plane pole map` |
|
|
