1-Day Course on AI in Power System Reliability Monitoring
Date:
This 1-day course, hosted by Tallinn University, focuses on the application of AI in monitoring the reliability of power systems. It is part of the Smart and Green Energy Systems and Business Models program, aiming to educate participants on the integration of AI technologies to enhance the efficiency and reliability of modern power grids.
Abstract
This course addresses the challenges and opportunities of applying AI to power system reliability amid renewable energy integration and electrification. It covers neural networks (including convolutional and graph-based), techniques to handle data imbalances and noise, and applications for distribution and transmission systems. Key topics include dynamic stability assessment, state estimation, and AI’s trustworthiness, reliability, and generalization for future-ready systems. The course equips researchers with insights and tools to leverage AI effectively in addressing power system challenges. Topics covered:
Security assessment and data in control rooms
- Balancing datasets
- Optimisation-based sampling
- Model requirements in control rooms
Learning models for secure system operation
- Interpretable models
- Convolutions Neurel Networks and Graph Neural Networks
- Learning with domain knowledge
- Cost-sensitive learning
- End-to-end learning for secure operation
Graph Neural Networks (GNN)
- Introduction to graphs with Pytorch Geometric and Networkx
- Introduction to Graph Neural Networks (GNN)
Exercise: Solving Optimal Power Flow (DCOPF) problem with GNN
- Creating power system graphs
- Training DCOPF model
- Improving constraint satisfaction