EnergyEminence: Source-Aware Environmental Calibration and Evaluation in a Physics-Grounded Grid Digital Twin

2026-09-28

AI Summary

This study addresses the challenges of missing environmental observation traceability and the integration of data-driven methods with physical simulations in power grid digital twins by proposing a source-aware calibration framework. Methodologically, a prototype system is constructed by combining graph-temporal forecasting, nonlinear AC cascading simulations, and video synthesis playback techniques. Furthermore, bounded calibration transformations and a wildfire evidence mechanism are innovatively introduced to reveal the issue of environmental shortcut learning obscured by scene-level segmentation. Evaluated on 160 test scenarios generated from the IEEE 118-bus system, the proposed approach achieves zero false negatives, thereby validating the effectiveness of decoupling environmental early warning from electrical inference.

Abstract

Power-grid digital twins must combine data-driven prediction with physically meaningful state evolution while preserving the provenance of environmental observations. This paper presents an early-stage EnergyEminence testbed that couples an IEEE 118-bus-style graph-temporal predictor, nonlinear AC cascade simulation, and operator-dashboard-like temporal replay. In addition, we introduce a shared bounded calibration that converts wildfire-detection confidence and spatial extent into source-comparable wildfire interpretable and explainable evidence. We then evaluate it with visually diverse fire and hard-negative videos. Sixteen synthetic environmental videos are curated to generate 160 source-tracked grid scenarios, and a source-video-disjoint test yields 10 true positives, 8 false positives, 22 true negatives, and no false negatives. The errors occur in stressed, non-cascading scenarios conditioned on an unseen growing-fire source. Our diagnostic then reveals environmental shortcut learning that is obscured by scenario-level random splitting. The paper therefore contributes a data-centric and inspectable evaluation workflow for multimodal grid-resilience models, together with evidence supporting separation of environmental alerting from electrical cascade inference

Problem

• grid digital twin
• environmental calibration
• shortcut learning
• grid resilience
• wildfire detection

Innovation

• Grid Digital Twin
• Source-Aware Calibration
• Shortcut Learning
• Graph-Temporal Prediction
• Data-Centric Evaluation