Learning by Examples: Power Grid Steady State Analysis with a Tabular Foundation Model

Olayiwola Arowolo; Oluwadara Adedeji; Peter Palensky; Jochen L. Cremer

cs.LGOct 14, 2026

Learning by Examples: Power Grid Steady State Analysis with a Tabular Foundation Model

Authors: Olayiwola Arowolo, Oluwadara Adedeji, Peter Palensky, Jochen L. Cremer

Organizations: Delft University of Technology, Delft, the Netherlands · University College Dublin, Belfield Dublin, Ireland · Austrian Institute of Technology, Vienna, Austria

Abstract

Foundation models have shown strong generalization across scientific domains, but their use in power systems remains largely task-specific. We investigate whether a pretrained tabular foundation model (TFM), TabPFN, can serve as a general-purpose neural solver for power grid steady-state analysis. We evaluate the TFM on three tasks: power flow (PF) prediction, N-k contingency screening (NCS), and voltage stability margin (VSM) prediction. We performed experiments on 118- and 2000-bus systems, comparing the TFM against task-specific neural baselines. Without retraining or fine-tuning, the TFM matches or outperforms the best task-specific model in several settings across the three tasks. For 118-bus PF prediction, a variant of the TFM achieves errors of $1\times 10^{-3}$ pu for voltage magnitudes and $0.30^\circ$ for voltage angles, outperforming all baselines. On unseen N-1 contingency and close-to-infeasible PF cases, the TFM is competitive with the DCPF approximation. For millions of unseen N-2 contingencies, the TFM achieves AUCs of 0.92 and 0.95 for branch and voltage violation screening, respectively. The TFM also attains accurate VSM prediction performance for the 2000-bus system with limited labelled data. The TFM attains strong results across the tested tasks despite pretraining without any power-system-specific dataset and requiring no fine-tuning step. These results offer a glimpse into the potential of pretrained TFMs as a neural solver for a range of power grid analysis tasks.

Figures & tables

	Variable
Bus type	Known	Unknown
PQ bus	$P_{i},\,Q_{i}$	$V_{i},\,\theta_{i}$
PV bus	$P_{i},\,V_{i}$	$Q_{i},\,\theta_{i}$
Slack bus	$V_{i},\,\theta_{i}$	$P_{i},\,Q_{i}$

TABLE I: Known and unknown variables per bus type.

Fig. 1: The whole table of labelled context rows and unlabelled query rows is passed to the pretrained TFM in one step. The TFM returns labels for all query rows in a single forward pass.

Fig. 2: In-distribution accuracy on the full-topology test sets: RMSE of voltage magnitude

V

, voltage angle

\theta

, active power

P

and reactive power

Q

(DCPF does not output

Q

). Bars show the mean and error bars the standard deviation over three seeds in log scale.

Fig. 3: Generalization to unseen line outages on the 118-bus (top) and 2000-bus (bottom) systems. Zero-shot: outage topologies absent from training (and from TabPFN’s context); mixed:

N

-1 samples included. Mean and standard deviation over three seeds in log scale; DCPF does not output

Q

Fig. 4: Generalization to close-to-infeasible (C2I) operating points on the 118-bus (top) and 2000-bus (bottom) systems. Zero-shot: C2I samples absent from training (and from TabPFN’s context); mixed: C2I samples included. Mean and standard deviation over three seeds (log scale); DCPF does not output

Q

Fig. 5: Mean (top) and maximum (bottom) power-balance residual of the learned models across all test conditions: We report the mean and standard deviation over three seeds (log scale).

Fig. 6: N-k contingency classification on the 118-bus grid for

N

-1 contingencies (top) and unseen

N

-2 contingencies (bottom): average recall, macro F1 and AUC for the combined, branch and voltage violation prediction. We report mean and standard deviation over three seeds. DCPF cannot be used to assess voltage violations.

Fig. 7: Average recall of contingency screening obtained by thresholding the PF regression outputs (hatched) compared with the directly trained classifiers (solid). The dashed line shows the performance of the direct TabPFN+ classifier. We report mean and standard deviation over three seeds.

Fig. 8: Prediction error of the loadability factor, $\hat{\sigma}-\sigma$ , against the true margin $\sigma-1$ in the in-distribution experiment on the 118-bus (top) and 2000-bus (bottom) grids. The solid line is the median error, and the shaded bands contain 50% and 90% of the test points. The dashed lines show the predicted 90% interval of TabPFN.

	118-bus		2000-bus
True margin $\sigma-1$	$\leq 0.05$	$>0.05$	$\leq 0.05$	$>0.05$
$\|\hat{\sigma}-\sigma\|$	$\leq 0.01$	$\leq 0.05$	$\leq 0.01$	$\leq 0.05$
	( $n=978$ )	( $n=1022$ )	( $n=797$ )	( $n=123$ )
GNS	57.7 $\pm$ 1.2	90.4 $\pm$ 1.3	69.6 $\pm$ 0.8	94.6 $\pm$ 2.0
TabPFN	33.6 $\pm$ 0.4	57.1 $\pm$ 0.4	83.8 $\pm$ 1.0	96.5 $\pm$ 1.2
PI-EVGNN	30.2 $\pm$ 0.6	49.0 $\pm$ 2.3	60.1 $\pm$ 4.5	76.4 $\pm$ 11.0

TABLE II: Percentage of VSM predictions within specified tolerance for in-distribution generalization experiment:

|\hat{\sigma}-\sigma|\leq 0.01

for true margins

\sigma-1\leq 0.05

and

|\hat{\sigma}-\sigma|\leq 0.05

for

\sigma-1>0.05.

We report mean

\pm

standard deviation over three seeds;

n

is the number of test samples. Best value per column in bold.

	118-bus		2000-bus
True margin $\sigma-1$	$\leq 0.05$	$>0.05$	$\leq 0.05$	$>0.05$
$\|\hat{\sigma}-\sigma\|$	$\leq 0.01$	$\leq 0.05$	$\leq 0.01$	$\leq 0.05$
	( $n=978$ )	( $n=1022$ )	( $n=797$ )	( $n=123$ )
GNS	4.7 $\pm$ 0.7	40.8 $\pm$ 7.6	5.0 $\pm$ 0.1	69.1 $\pm$ 4.5
TabPFN	0.0 $\pm$ 0.0	12.8 $\pm$ 0.2	3.2 $\pm$ 0.4	75.9 $\pm$ 0.5
PI-EVGNN	0.0 $\pm$ 0.0	20.7 $\pm$ 3.1	5.2 $\pm$ 0.6	75.6 $\pm$ 4.5

TABLE III: Percentage of VSM predictions within specified tolerance for out-of-distribution generalization experiment:

|\hat{\sigma}-\sigma|\leq 0.01

for true margins

\sigma-1\leq 0.05

and

|\hat{\sigma}-\sigma|\leq 0.05

for

\sigma-1>0.05

We report mean

\pm

standard deviation over three seeds;

n

is the number of test samples. Best value per column in bold.

	PF 118-bus		PF 2000-bus		N-2 screening
	( $n=5{,}600$ )		( $n=2{,}800$ )		( $n=3{,}133{,}892$ )
Model	$tr$	$inf$	$tr$	$inf$	$tr$	$inf$
DCPF	–	10	–	10	–	637
LeapNet	164	11	140	11	101	38
PowerFlowNet	829	8	4,369	10	512	176
Typed-GNN	7,291	10	9,944	12	4,392	621

TABLE IV: Average computational cost in seconds: total training time (

tr.

) and total inference time on the test set (

inf.

);

n

is the number of test samples. TabPFN requires no training; its context is processed at inference and included in

inf.

Power Systems

Power systems research studies the planning, operation, control, and protection of electrical grids and their integration with energy resources.

See all 231 papers

GENCO - A Unified Neural Solver Embedded in a Development Framework for Steady-State Grid Analysis
Aug 10, 2026Alban Puech, Matteo Mazzonelli, Tamara R. Govindasamy +19Neural Surrogate Modeling Optimal Power Flow
IBM Research · Reliability and Risk Engineering Laboratory, ETH Zurich, Switzerland · Hydro-Qu´ebec Research Institute, Canada +3
Revisiting data-driven dynamic security assessment with a tabular foundation model
Jul 17, 2026Olayiwola Arowolo, Maosheng Yang, Jochen CremerTabular Foundation Models Few-Shot Learning
Electrical and Sustainable Energy, TU Delft
GridSFM: A Foundation Model for Solving AC Optimal Power Flow
Sep 24, 2026Luke Bhan, Weiwei Yang, Margaret Capetz +1Neural Surrogate Modeling Optimal Power Flow
University of California, San Diego · Microsoft Research · University of Washington

Aug 10, 2026J/K move · Enter open · S save

GENCO - A Unified Neural Solver Embedded in a Development Framework for Steady-State Grid Analysis

Alban Puech, Matteo Mazzonelli, Tamara R. Govindasamy, Mangaliso Mngomezulu+18

IBM Research · Reliability and Risk Engineering Laboratory, ETH Zurich, Switzerland · Hydro-Qu´ebec Research Institute, Canada+3

Neural Surrogate Modeling Optimal Power Flow Power Systems

Open paper AskPDF arXiv

Foundation models are transforming business workflows and boosting productivity, yet they remain largely absent from engineering domains such as power system analysis, where strict physical consistency must be enforced. We present GENCO (GEometric Neural Corrective Optimizer), a unified neural solver for steady-state transmission grid analysis that handles power flow (PF), optimal power flow (OPF), and state estimation (SE) within a single architecture and shared grid representation. To support advances in neural power system solvers, we introduce the open-source GridFM Development Framework, which standardizes synthetic data generation and training in a low-code environment. We also release large-scale datasets with millions of PF and OPF scenarios across diverse grid topologies to support reproducible benchmarking. We evaluate GENCO on the PFDelta and OPFData benchmarks against state-of-the-art neural solvers and classical solvers, including Newton-Raphson and IPOPT, as well as on real-world Hydro-Québec SCADA data. For large-scale PF, GENCO recovers the full AC operating state, including voltage magnitudes and reactive power that DC-PF cannot provide, while matching DC-PF-level active power-balance residuals. It achieves up to 30x speedups over Newton-Raphson at only 2x the runtime of DC-PF. For OPF, it achieves up to 85x speedups over IPOPT while improving feasibility, optimality, and runtime over DC-OPF. For SE, GENCO is more robust than classical weighted least squares to noisy measurements and grid parameter errors, and always returns a high-quality estimate even when weighted least squares fails to converge. Together, the unified architecture and development framework provide a new approach to large-scale steady-state grid analysis, lowering the barrier to entry for power system engineers and marking a step toward Grid Foundation Models.

Abstract

Power Systems

Explore similar work

GENCO - A Unified Neural Solver Embedded in a Development Framework for Steady-State Grid Analysis

Revisiting data-driven dynamic security assessment with a tabular foundation model

GridSFM: A Foundation Model for Solving AC Optimal Power Flow