Presented By: Industrial & Operations Engineering
IOE Lunch & Learn Seminar Series: Elnaz Kabir, U-M IOE
This event is open to all IOE PhD students, faculty, and staff. Lunch will be provided. In order to get an accurate count for food, please RSVP by Thursday, December 5, 2019.
Title:
Predictive and Risk Analytics for Weather-Induced Power Outage Management
Abstract:
A wide variety of weather conditions, from windstorms to prolonged heat events, can have substantial impacts on power systems, posing many risks and inconveniences due to power outages. Being able to accurately estimate the probability distribution of the number of customers without power by using data about the power utility system, environmental and weather conditions has the potential to help utilities restore power more quickly and efficiently. In this research, we develop two frameworks to address these issues. In the first framework, we propose an adaptive two-stage algorithm based on Bayesian model averaging in order to form an ensemble model predicting daily distributions of customer outages. In this algorithm, weights of the base learners depend on the features and they get updated as new data is observed. In the second framework, we focus on the zero inflation issue of power outage data in resolutions smaller than county level. To overcome the challenges caused by zero-inflation, e.g., bias and inaccuracy, we propose a novel approach integrating mixture models with cost-sensitive learning. For both frameworks, we conduct numerical studies using large, real datasets of power outages. We show that our approaches offer more accurate point and probabilistic predictions than traditional approaches, better supporting utility restoration planning.
Bio:
Elnaz Kabir is a PhD Candidate in the Industrial and Operations Engineering Department at the University of Michigan. Her research is grounded in predictive analytics, data-driven decision making, and risk analysis. In her research, Elnaz is interested to use statistical learning theories, and optimization techniques to better understand and solve important problems related to power outages caused by weather events. The results of her studies are used by practitioners of the utility companies in order to make better decisions to reduce the risk of weather-events to the power system.
Title:
Predictive and Risk Analytics for Weather-Induced Power Outage Management
Abstract:
A wide variety of weather conditions, from windstorms to prolonged heat events, can have substantial impacts on power systems, posing many risks and inconveniences due to power outages. Being able to accurately estimate the probability distribution of the number of customers without power by using data about the power utility system, environmental and weather conditions has the potential to help utilities restore power more quickly and efficiently. In this research, we develop two frameworks to address these issues. In the first framework, we propose an adaptive two-stage algorithm based on Bayesian model averaging in order to form an ensemble model predicting daily distributions of customer outages. In this algorithm, weights of the base learners depend on the features and they get updated as new data is observed. In the second framework, we focus on the zero inflation issue of power outage data in resolutions smaller than county level. To overcome the challenges caused by zero-inflation, e.g., bias and inaccuracy, we propose a novel approach integrating mixture models with cost-sensitive learning. For both frameworks, we conduct numerical studies using large, real datasets of power outages. We show that our approaches offer more accurate point and probabilistic predictions than traditional approaches, better supporting utility restoration planning.
Bio:
Elnaz Kabir is a PhD Candidate in the Industrial and Operations Engineering Department at the University of Michigan. Her research is grounded in predictive analytics, data-driven decision making, and risk analysis. In her research, Elnaz is interested to use statistical learning theories, and optimization techniques to better understand and solve important problems related to power outages caused by weather events. The results of her studies are used by practitioners of the utility companies in order to make better decisions to reduce the risk of weather-events to the power system.
Related Links
Explore Similar Events
-
Loading Similar Events...
