Learning memory forces and bounding frequency errors in fractional Kuramoto grids

Missing history creates two different demands on a forecaster for a fractional grid: predicting frequency accurately and reconstructing memory well enough to bound the prediction error. We connect these demands through the hereditary force that remains after an integral observation correction. A compact decoder retains the directions most relevant to that force, and an exact decomposition separates omitted directions from coefficient error. A response-trained linear predictor and neural refinement learn the coefficients for frequency prediction. A separate force reconstruction supplies the reference for a nonlinear energy bound for the specified finite relaxation model, accounting for numerical propagation defects and the difference between the reference and point forecasts. For the outage transients studied here, refinement improves on the linear predictor across the time steps and memory parameters tested. Replacing the noisy origin states reveals why this gain can coexist with greater memory reconstruction error: the learned correction partly compensates for observation error. Separating prediction from reconstruction enables frequency bounds in cases where the point predictor alone cannot support them. A critical radius compares calibrated uncertainty with the angle margin available along the reference trajectory, explaining both issuance and abstention. Under jointly exchangeable calibration and test events, the calibrated guarantee bounds the marginal probability of issuing an invalid envelope. Together, these results connect uncertainty about history outside the observation window to frequency forecasts and explicit conditions for bounding their errors.

Authors

Publication Details

Journal
arXiv (Cornell University)
Published
2026-11-12
DOI
https://doi.org/10.48550/arxiv.2611.16220
Primary Topic
Power System Optimization and Stability
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Learning memory forces and bounding frequency errors in fractional Kuramoto grids

이환서, Hyunju Kim
arXiv (Cornell University)
Power System Optimization and Stability
preprint

Learning memory forces and bounding frequency errors in fractional Kuramoto grids

이환서, Hyunju Kim
preprint en

Abstract

Missing history creates two different demands on a forecaster for a fractional grid: predicting frequency accurately and reconstructing memory well enough to bound the prediction error. We connect these demands through the hereditary force that remains after an integral observation correction. A compact decoder retains the directions most relevant to that force, and an exact decomposition separates omitted directions from coefficient error. A response-trained linear predictor and neural refinement learn the coefficients for frequency prediction. A separate force reconstruction supplies the reference for a nonlinear energy bound for the specified finite relaxation model, accounting for numerical propagation defects and the difference between the reference and point forecasts. For the outage transients studied here, refinement improves on the linear predictor across the time steps and memory parameters tested. Replacing the noisy origin states reveals why this gain can coexist with greater memory reconstruction error: the learned correction partly compensates for observation error. Separating prediction from reconstruction enables frequency bounds in cases where the point predictor alone cannot support them. A critical radius compares calibrated uncertainty with the angle margin available along the reference trajectory, explaining both issuance and abstention. Under jointly exchangeable calibration and test events, the calibrated guarantee bounds the marginal probability of issuing an invalid envelope. Together, these results connect uncertainty about history outside the observation window to frequency forecasts and explicit conditions for bounding their errors.

arXiv (Cornell University)
Power System Optimization and Stability
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon