Researchers Farshad Amani, Faezeh Ardali, and Amin Kargarian from the University of Texas at Austin have developed a new approach to aid power grid restoration efforts following natural disasters. Their work, published in the journal IEEE Transactions on Power Systems, focuses on improving the efficiency and effectiveness of crew dispatch during post-storm restoration processes.
Natural disasters like hurricanes and floods often cause significant damage to power grid equipment, leading to repeated replanning of restoration efforts as new information becomes available. The researchers have created a deep reinforcement learning (DRL) dispatcher that acts as a real-time decision engine for assigning repair crews to tasks. This system models the restoration process as a sequential, information-revealing process, taking into account various factors such as component status, travel and repair times, crew availability, and the value of restoring power to different areas.
The DRL dispatcher uses a policy that learns from these factors to make informed decisions. It also includes a feasibility mask to prevent unsafe or inoperable actions, such as violating power flow limits, switching rules, and crew-time constraints. To provide realistic runtime inputs without relying on heavy computational solvers, the researchers use lightweight surrogates for various factors like wind and flood intensities, fragility-based failure, spatial clustering of damage, access impairments, and progressive ticket arrivals.
In simulated hurricane and flood events, the learned policy updates crew decisions in real time as new field reports arrive. The lightweight nature of the runtime logic allows it to improve online performance metrics such as energy-not-supplied, critical-load restoration time, and travel distance compared to traditional methods like mixed-integer programs and standard heuristics. The proposed approach was tested on the IEEE 13- and 123-bus feeders with mixed hurricane/flood scenarios, demonstrating its potential for practical applications in the energy sector.
This research highlights the potential of advanced machine learning techniques to enhance the resilience and efficiency of power grid restoration efforts, ultimately leading to faster recovery times and improved service reliability for customers.
This article is based on research available at arXiv.