How Artificial Intelligence and Advanced Optimization Help Improve Outage Management

By Zhaoyu Wang

Enormous pressure is looming over utility companies as they try to improve responses to severe weather events and power outages. Natural disasters have been causing major disruptions around the world and the trend is increasing. In 2016, the average outage duration for customers in the United States ranged from 27 minutes in Nebraska to 6 hours in West Virginia, to 20 hours in South Carolina due to Hurricane Matthew. More recently in 2017, nearly 280,000 Texas customers were left without power after Hurricane Harvey. In 2018, approximately 1.9 million customers experienced power outages during Hurricane Florence, while 1.7 million customers lost power after Hurricane Michael in the Southeast United States. In 2019, severe storms caused multiple outages in Michigan and Wisconsin, where more than 500,000 customers were left without electricity. Power outage can cause significant social-economic impacts and is potentially life threatening. Utilities use outage management systems (OMS) to manage the grid during outages and restore power to customers. An OMS identifies and predicts potential grid outages and manages restoration activities. Moreover, OMS can assist in prioritizing restoration efforts and managing repair resources after outages. OMS can potentially reduce the outage duration by up to 25%. Recent development in Artificial intelligence (AI) and advanced optimization provides an opportunity to greatly improve existing OMS. This article will introduce how these new techniques help disaster preparation, outage prediction, damage assessment and service restoration.

Outage management and power restoration cannot be efficient without proper preparation, which includes labor force preparation, equipment management, and pre-staging resources. The pre-disaster preparation is to ensure there is an appropriate amount of resources in the right locations. Lack of resources was one major reason of the slow response after Hurricane Maria in Puerto Rico. AI and machine learning can leverage the data collected from previous outages and extreme weather events to make more informed decisions. For example, AI can improve weather forecasting and predict potential outages using fragility models of grid components and historic outage data. Advanced optimization techniques can be used to efficiently prepare and coordinate the available resources before the event. In a recent study, we showed how to use stochastic programming for optimal allocation of resources before natural disasters. The key is to consider different types of labor/equipment resources, capacity constraints, and damage uncertainties. An interesting future direction is to investigate resource coordination between multiple utilities and different agencies.

During and after extreme events, one of the main concerns for utilities is lack of observability and situational awareness. Currently, many utilities rely on customer phone calls and field inspections to locate and assess damages. In recent extreme events, drones were utilized for scanning grids to identify damages. However, AI has the capability to consolidate the information from different sources, such as smart meters, social media, customer phone calls, and protection devices, for faster and more accurate damage assessment. A viable solution is the use of AI-based assessment to prioritize and down-select inspection areas, which will significantly improve the efficiency of damage inspection. Furthermore, this is important for utilities who have limited inspection crews while experiencing large-scale outages after natural disasters. In addition to damage assessment, another example is to use AI for estimating repair and restoration time. This information is critical to ensure efficient dispatch of repair crews. Moreover, it can provide valuable information to customers who are desperate for electricity and could help them better prepare their lives. Recently, we have proposed a deep learning method for estimating the repair and restoration time using historic outage data and weather forecasts.

After damage assessment, the next step is repair and restoration. This includes dispatching crews to repair damaged components and conducting recovery operation using smart grid technologies such as distributed energy resources, microgrids, and automatic switches. The difficulty is to build co-optimization models that coordinate the two processes which are interdependent yet have different timescales. In a recent work, we have shown that such a co-optimization model will significantly improve the restoration compared to solving the two processes independently. However, the large-scale mixed-integer co-optimization model is extremely difficult to solve. We have leveraged advanced optimization techniques such as progressive hedging and metaheuristics to deal with the computational challenge.

Outage management includes multiple interdependent phases. AI and advanced optimization can improve all these phases, from outage prediction, pre-event preparation, damage assessment, to repair and service restoration. However, implementing these methods is still challenging. We must make the decision-making process of AI and optimization easier-to-understand, clearly demonstrate their effectiveness, and communicate the financial and social benefits to the policy makers and engineers.

 For a downloadable copy of September 2019 eNewsletter which includes this article, please visit the IEEE Smart Grid Resource Center.

Zhaoyu Wang

Zhaoyu Wang (S’13-M’15) is the Harpole-Pentair Assistant Professor with Iowa State University. He received the B.S. and M.S. degrees in electrical engineering from Shanghai Jiaotong University in 2009 and 2012, respectively, and the M.S. and Ph.D. degrees in electrical and computer engineering from Georgia Institute of Technology in 2012 and 2015, respectively. He was a Research Aid at Argonne National Laboratory in 2013 and an Electrical Engineer Intern at Corning Inc. in 2014. His research interests include power distribution systems, microgrids, renewable integration, power system resilience, and data-driven system modeling. He is the Principal Investigator for a multitude of projects focused on these topics and funded by the National Science Foundation, the Department of Energy, National Laboratories, PSERC, and Iowa Energy Center. Dr. Wang is the Secretary of IEEE Power and Energy Society (PES) Award Subcommittee, Co-Vice Chair of PES Distribution System Operation and Planning Subcommittee, and Vice Chair of PES Task Force on Advances in Natural Disaster Mitigation Methods. He is an editor of IEEE Transactions on Power Systems, IEEE Transactions on Smart Grid and IEEE PES Letters, and an associate editor of IET Smart Grid.


IEEE Smart Grid Newsletter Editors

 

Rostami

 

 

 

 

 

Dr. Mehrdad Rostami
Managing Editor

IEEE Smart Grid Publications Editorial Board
Frances Bell
Pardis Khayyer
Mehmet Cintuglu
Geev Mokryani
Jose Medina
Merhrdad Boloorchi
Vigna Kumaran
Hossam Gaber
Click here for more info

Dr. Panayiotis (Panos) Moutis
Editor-in-Chief
IEEE Smart Grid eNewsletter

Past Issues

To view archived articles, and issues, which deliver rich insight into the forces shaping the future of the smart grid, please visit the IEEE Smart Grid Resource Center.

IEEE Smart Grid Newsletter Compendium

The IEEE Smart Grid Newsletter Compendium "Smart Grid: The Next Decade" is the first of its kind promotional compilation featuring 32 "best of the best" insightful articles from recent issues of the IEEE Smart Grid Newsletter and will be the go-to resource for industry professionals for years to come. Click here to read "Smart Grid: The Next Decade"