A physics-informed neural network (PINN)-based grid optimizer that performs real-time capacity assessment, temperature monitoring, and dynamic stress analysis of power distribution components by fusing sensor, SCADA, and weather data.

Patentradars sammanfattning

The disclosed system combines a grid optimizer with a physics-informed neural network (PINN) to enable real-time capacity assessment and performance prediction of power distribution grid components including transformers and powerlines. The PINN ingests power component data, historical load records, SCADA outputs, and current/forecasted weather variables, using physical heat-transfer and electromagnetic equations embedded directly in the network architecture to improve accuracy beyond purely data-driven approaches. Training data consists of approximately 80% real-world operational data and 20% simulated extreme-scenario data (e.g., heatwaves, demand surges), and the validated model continuously refines itself via feedback from live predictions. Outputs include real-time maximum capacity estimates, hot-spot temperature predictions, and loss-of-life tracking, supporting proactive maintenance, EV charger management, and renewable energy integration.

AI-genererad och redaktionellt processad.