Electrical grids face mounting pressure from renewable energy integration, climate-driven weather extremes, and the cascading failure risks they create. Existing resilience assessment tools typically classify equipment as either functional or failed, providing little insight into degradation patterns that precede failure or explain system behavior during stress events.
A research team has developed the Agentic Resilience Assessment System (A-RAS), which implements a structured Resilience Assessment Cycle to monitor grid infrastructure more granularly. The system combines multiple analytical layers: anomaly detection identifies early operational shifts in SCADA data from diverse equipment types; diagnostic engines powered by large language models interpret these shifts and assign operational state labels; and resilience agents compute dual-metric outputs tracking both service performance degradation and asset health decline.
The innovation lies in treating equipment condition as a spectrum rather than binary. A-RAS assigns intermediate health states, enabling operators to correlate actual grid behavior during adverse events to specific equipment degradation patterns. This addresses a core limitation of traditional metrics that struggle to explain observed failures during weather stress.
The system employs a dual-scoring mechanism combining quantitative anomaly severity (derived from time-series analysis) with qualitative assessment from AI semantic analysis, improving robustness over single-method approaches. An orchestration agent manages the entire workflow—from data ingestion through extreme weather detection to final resilience quantification—enabling fully automated assessment.
Initial validation on an operational wind power plant demonstrated the system's ability to trace resilience trajectories back to specific residual health deficits and contributing equipment. While this proof-of-concept focuses on a single site and weather hazard class, the modular design provides a foundation for expansion across substations, transmission corridors, distribution networks, and additional climate scenarios.
Future deployment will require standardization of data inputs, integration with existing SCADA architectures, and validation across diverse asset types and geographic regions. The structured assessment framework promises to enhance operator situational awareness and improve preventive maintenance targeting during periods of heightened climate stress.
