The management of power grids during fault events and other dynamic conditions is expected to improve with an artificial intelligence (AI)-based scheduling system that considers static as well as dynamic operating parameters.

The neural network developed at U.S. Argonne National Laboratory addresses both static and dynamic features to enhance decision making and power system resiliency. The technology could help operators anticipate when they can turn on and off generation resources while ensuring that all the resources that are online can withstand certain disruptions. The method is intended to provide a quicker, safer route to power restoration after an outage.

The approach was tested by integrating the deep neural network with a diesel generator-wind turbine microgrid scheduling model as a mixed-integer linear program to perform frequency-constrained energy management where detailed wind turbine generator characteristics are considered. The effectiveness of the constraint encoding technique was verified by detailed three-phase network simulation, and the dispatch and control commands generated during the exercise were shown to ensure both islanding success and adequate frequency response.

Future work on the neural network approach, which is also applicable to bulk-power systems, will seek solutions to secure dynamic voltage constraints at critical loads as well as frequency.