High winds causing power outages have become a common headache for communities, leading to inconveniences and potential dangers. UC Santa Cruz Assistant Professor Yu Zhang and his team are tackling this issue head-on with an artificial intelligence (AI) solution that enhances the efficiency, reliability, and resilience of power systems, specifically focusing on microgrids for rapid power restoration during outages.

In a recent publication in the prestigious IEEE Transactions on Control of Network Systems, Zhang's lab details their innovative AI model, demonstrating its superiority over traditional power restoration methods. The paper, led by Ph.D. student Shourya Bose, sheds light on the potential of microgrids as a key player in future power distribution systems.

Unlike conventional setups where communities rely solely on a central utility company, microgrids provide a decentralized solution. These local grids, often incorporating renewable energy sources and backup generators, can operate independently or connect to the main utility, ensuring uninterrupted power supply even during disasters or outages.

Zhang's team employs deep reinforcement learning, a form of AI that rewards the algorithm for successfully adapting to changing conditions. Their approach, termed constrained policy optimization (CPO), factors in real-time conditions, machine learning, and long-term patterns affecting renewable energy output. This method outperforms traditional model predictive control (MPC) by anticipating fluctuations in renewable sources and responding swiftly during power outages.

The success of Zhang's AI-powered microgrid control was recently validated in a global competition, L2RPN Delft 2023, where their team clinched the top spot. This victory, co-sponsored by France's electricity transmission system operator, indicates a growing recognition of the potential for AI and renewable energy techniques in large-scale grid operations.

Moving forward, the researchers aim to test their algorithm on real microgrids in their lab, with a long-term goal of implementing it in the UC Santa Cruz campus's energy system. By bringing power generation closer to the demand side, they aim to create smaller, stronger, and more resilient grids, addressing outage issues and garnering interest from industry collaborators. The journey from successful simulations to practical implementation marks a significant stride toward a smarter, more adaptive future for power systems.

Based on: https://www.sciencedaily.com
Source: University of California - Santa Cruz