A 2019 CITRIS Seed Award project led by Yu Zhang, a UC Santa Cruz assistant professor of electrical and computer engineering, has sparked the creation of a new artificial intelligence (AI) model that provides smart control for electric microgrids that helps them restore power more quickly than traditional energy restoration methods.
Microgrids are used to distribute electricity to small areas and can be connected to a main power utility source or function independently using alternative energy sources, making them a viable strategy for self-sufficient power.
Employing an AI-based technique called deep reinforcement learning, Zhang’s team modeled practical constraints of real world power systems, including branch flow capabilities, alternative energy sources, and communities’ electricity demands to develop a system that responds to a changing environment. Their model goes a step beyond traditional systems by pioneering the use of constrained policy optimization (CPO) that takes real-time conditions into account and uses machine learning to identify long-term patterns that will affect energy outputs.
A successful algorithm in simulation has shown that the predictive capabilities of CPO significantly outperform traditional methods and respond much faster in a power outage, proving the technology’s potential as community energy needs continue to increase.
“Nowadays, microgrids are really the thing that both people in industry and in academia are focusing on for the future power distribution systems,” Zhang said.