AI-Driven Anomaly Detection in Eastern Canada's Grid Operations

The operational integrity of modern energy grids depends on real-time visibility and predictive oversight. At Atlantic GridOps, our research focuses on deploying advanced AI frameworks to enhance the digital oversight of Eastern Canada's interconnected energy systems.

This article details our latest pilot project in Nova Scotia, where a modular AI monitoring system was integrated with existing SCADA infrastructure. The primary objective was to move from reactive alarm management to predictive anomaly detection.

Figure 1: Operational monitoring dashboard used in the Nova Scotia pilot.

Modular Analysis Framework

The developed framework is built on a series of independent but interoperable modules:

• Data Ingestion Layer: Aggregates real-time telemetry from sensors, weather feeds, and market data.
• Pattern Recognition Engine: Uses unsupervised learning to establish baseline operational "signatures" for different grid segments.
• Anomaly Scoring System: Flags deviations from established patterns, assigning a severity and confidence score.
• Coordinated Alerting: Integrates alerts into operator dashboards with suggested context and historical parallels.

Over a six-month observation period, the system demonstrated a 94% accuracy rate in identifying potential fault conditions up to 45 minutes before traditional threshold-based alarms were triggered. This lead time allows for coordinated manual intervention or the activation of automated stability protocols.

Challenges in System-Wide Coordination

A significant finding was the challenge of inter-provincial data standardization. While the AI models performed well within a controlled provincial network, scaling oversight across the Eastern Canadian corridor requires harmonized data protocols between Nova Scotia, New Brunswick, and Newfoundland and Labrador.

Future work, slated for 2027, will involve deploying a cross-border test of the coordination module, focusing on load-balancing predictions during extreme weather events. The goal is to create a resilient digital oversight layer that supports both autonomous operations and human decision-making.

This research underscores that the future of grid reliability lies not in more data, but in smarter, more interpretable analysis frameworks that provide clear operational visibility to system engineers.