State estimation improves with tabular foundation models in power grids
State Estimation as In-context Tabular Prediction
Machine Learning
Summary
Estimating the electrical state of power networks is key to keeping the system stable and efficient. This paper shows how arranging past measurements and estimates like a table helps a special type of pretrained model make accurate predictions for new measurements. Unlike traditional methods that struggle with missing or faulty data, their approach is more reliable in such conditions. The models don’t need retraining with real power system data, only example contexts, making them easier to use.
What this means in practice
- •For power system operators: Improve grid state monitoring accuracy under limited or faulty measurement data using pretrained tabular models without retraining.
- •For smart grid technology developers: Embed frozen tabular foundation models to enhance state estimation in power networks subject to sensor noise and measurement bias.$Commercial implications: Enables more resilient real-time grid monitoring products for utility companies by using pretrained models requiring no additional training on customer data.
Authors
Ayushi Jolotia, Parikshit Pareek
Abstract
This letter poses state estimation (SE) as an in-context tabular prediction (ICTP) problem. The key idea is a tabular arrangement in which historical measurement vectors and corresponding state estimates serve as the context, and a new measurement vector forms the query for which we predict a state vector. We propose to use tabular foundation models (TFMs), pretrained on synthetic data, whose weights are never updated, so power system data is needed only as context. With frozen TFM weights we estimate the state of 14-, 30-, 33-, 70-, and 118-bus transmission and distribution networks, and do so where weighted least squares (WLS) fails or degrades, under low observability and with bad data. ICTP errors are 11 to 57$\times$ below the per-bus average of the context at full metering, and 1.3 to 4.0$\times$ below it with only 5% of measurements available. With measurements at six buses biased by 8$σ$, the ICTP angle error rises 1.8 to 5.0$\times$ over its bias-free value against 2.5 to 12.7$\times$ for WLS with bad-data detection. Without retraining, ICTP has lower voltage error than DSS$^2$ on both of its feeders, and lower than a retrained TAPINE on two of three reconfiguration sets.