Trustworthiness Layer for Foundation Models in Power Systems: Application to N-k Contingency Screening

TL;DR

This paper introduces a model-agnostic trustworthiness layer for foundation models in power systems, providing statistically valid prediction intervals to improve reliability in N-k contingency screening.

Contribution

It presents a novel calibration layer using conformal prediction methods that enhances foundation models with reliable uncertainty estimates in power system security assessments.

Findings

Over 90% of critical violations captured across N-k levels.

Achieves up to 5 times fewer false alarms than DC Power Flow.

Negligible computational overhead for the added trustworthiness layer.

Abstract

We propose a model-agnostic trustworthiness layer that equips any foundation model (FM) for power systems with statistically valid prediction intervals. The layer offers two calibration approaches: (i) stratified conformal prediction (SCP), which partitions residuals by contingency severity and grid element, and (ii) kernel-weighted conformal prediction (KCP), which localizes the calibration to each test scenario via scenario representations, yielding tighter, approximately conditional bounds. Using GridFM as a guiding example, we demonstrate the framework on N-k contingency screening for IEEE 24- and 118-bus systems. The trustworthiness layer ensures that over 90% of all critical violations are captured across N-k levels, minimizing missed detections while maintaining up to 5 times fewer false alarms than DC Power Flow. With negligible computational overhead over the underlying FM, this approach enables reliable large-scale security assessment beyond routine N-1 screening.