Andrew Strelzoff, Sanja Cvijic

2026 IEEE Power & Energy Society General Meeting (PESGM) (19-23 July 2026)

Abstract

This paper presents POWERED III (Probabilistic Operations Warranted for Energy Reliability Evaluation and Diagnostics), research effort that develops advanced Condition-Based Maintenance Plus (CBM+) modeling techniques for power transformers. The project leverages power quality (PQ) event data from Powerside Power Monitoring Systems sensors deployed at 25 commercial grid substations over 18 months to predict transformer outages. The developed model achieves approximately 80% accuracy with low false positive rates, providing 3 to 21-day advance warnings of service interruptions. By analyzing correlations between high-frequency impulse events and transformer failures, POWERED III extends traditional CBM+ approaches to incorporate real-time power quality disturbances. The system includes an operator-facing dashboard that visualizes model predictions, underlying sensor data, and historical events. This work demonstrates the practical application of hybrid AI modeling—combining physics-based and machine learning approaches—for predictive maintenance in critical power infrastructure.

For More Information

To learn more contact Andrew Strelzoff or Sanja Cvijic.

Available for download at IEEE Xplore, https://ieeexplore.ieee.org/document/11693530.