Self-healing smart grids have rapid and intelligent control mechanisms to minimize power disruptions during outages. During power distribution network outages, corrective actions involve reconfiguration through switching control and emergency load shedding. However, traditional decision-making models for outage mitigation are not appropriate for smart grids due to their slow response and computational inefficiency.
Researchers at the University at Buffalo have created an artificial intelligence model to assist electrical grids in preventing power outages by automatically redirecting electricity within milliseconds.
This technology is an early demonstration of “self-healing grid” technology, which utilizes AI to autonomously detect and repair issues, such as outages without the need for human intervention when problems arise, such as storm-damaged power lines.
Before the automated system can be implemented and scaled to real-world power grids, further research is required. Nonetheless, researchers believe that this is an exciting development for the nation’s troubled power grid.
“Power grids across the world are being challenged by the growing number of extreme weather events, the likelihood of cyberattacks, and projected increases in demand,” says co-corresponding author Souma Chowdhury, PhD, associate professor in the UB Department of Mechanical and Aerospace Engineering. “Therefore, it is imperative that we develop tools that modernize the system and make it more resilient against future power outages.”
The North American grid comprises a dense and intricate network of transmission and distribution lines, power generation facilities, and transformers that distribute electricity from power sources to consumers.
Through the use of different scenarios in test networks, the team of researchers showed that their solution has the capability to automatically identify alternative pathways for delivering electricity to users before an outage occurs. Once trained, AI offers the benefit of speed: The system can autonomously redirect electrical flow within microseconds, whereas current processes involving traditional engineering methods (or human intervention) to determine alternative paths could take from minutes to hours.
“Our goal is to find the optimal path to send power to the majority of users as quickly as possible,” says co-corresponding author Jie Zhang, PhD, associate professor of mechanical engineering in the Erik Jonsson School of Engineering and Computer Science at UT Dallas.
The research team utilized algorithms that leverage machine learning to analyze the intricate connections between entities in a power distribution network. In this context, graph machine learning focuses on characterizing the network’s topology, the spatial arrangement and interconnections of its components, and the flow of electricity within the system.
Additionally, the team employed reinforcement learning, where a virtual agent operates within a simulated environment of the actual problem, to systematically simulate scenarios and continuously learn from these simulations.
For instance, this approach enables the team to gain insights into issues such as electricity blockages caused by line faults. The system could then reconfigure by utilizing switches and accessing power from nearby sources, like big solar panels or batteries located on a university campus or within a business.
“These are decisions that the model can make almost instantaneously, which in turn has the potential to eliminate or greatly reduce the severity of power outages,” says co-first author Steve Paul, who worked on the project while earning a PhD from UB earlier this year. Paul is now a postdoctoral scholar at the University of Connecticut.
Roshni Anna Jacob, a doctoral student in electrical engineering at UT Dallas, and Yulia Gel, PhD, a professor of mathematical sciences in the School of Natural Sciences and Mathematics at UT Dallas, are additional co-authors of the study.
Having previously concentrated on outage prevention, the researchers are now working on creating similar technology to mend and recover the grid after a power interruption, such as one triggered by a natural disaster.
Journal reference:
- Roshni Anna Jacob, Steve Paul, Souma Chowdhury, Yulia R. Gel & Jie Zhang. Real-time outage management in active distribution networks using reinforcement learning over graphs. Nature Communications, 2024; DOI: 10.1038/s41467-024-49207-y