2020 PES GM 8/6 Panel Session: Application of Machine Learning in Emergency Control for Resilient System Operationns

Power system emergency and restorative control are generally designed off-line, based on either worst-case scenarios or complex mathematical models and sophisticated algorithms. They are difficult for on-line or real-time applications, due to increasing uncertainties and variations. Machine learning has demonstrated the great capability in load and renewable energy forecasting, cyber intrusion and attack detection, situational awareness enhancement. Although there are limited research yet, machine learning also has great potential in increasing power grid resilience.
This panel will present recent developments of machine learning in outage prediction, load shedding, load restoration, and resilient control. It will discuss main challenges in the training data quality, the learning model design with parameter uncertainties, and the transfer capability to provide useful information for operators. The panel will also discuss the opportunities of enhancing the cyber-physical system resilience with AI and learning technologies.