A Real-Time Edge-AI Multimodal Framework Integrating SCADA Telemetry, Weather Radar, and Social Sensing for Grid Failure Prediction and Adaptive Load Dispatching

• Abbas Ahsun

  Texas University

  Author

Contemporary power grids face escalating threats from extreme weather events and cascading failures, yet existing monitoring systems remain fundamentally reactive and disconnected across heterogeneous data streams. This study addresses the critical gap in grid resilience by designing and validating a real-time Edge-AI multimodal framework that fuses SCADA telemetry, weather radar data, and social sensing signals for predictive failure detection and adaptive load dispatching. Drawing on a quantitative design-based approach, the framework was evaluated using retrospective grid data and simulated edge deployments across 2,847 contingency scenarios. The multimodal fusion model achieved 89.4% prediction accuracy with a mean lead time of 18.3 minutes, substantially outperforming single-source baselines. Feature importance analysis revealed weather radar reflectivity and SCADA voltage deviation as dominant predictors, while social sensing signals enhanced situational awareness during evolving events. The framework demonstrates that edge-deployed intelligence can process heterogeneous data streams within latency constraints suitable for real-time grid operations. These findings establish a replicable architecture for integrating distributed sensing modalities into resilient grid management, offering practitioners a deployable pathway toward proactive load dispatching during extreme weather conditions.

• pdf

Copyright (c) 2026 Abbas Ahsun (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.