Sustainable Energy Speaker Series – Dr. Dimitris Floros

• In-person event
• 4020, Nicol, Carleton University
• 1125 Colonel By Drive, Ottawa, ON, K1S 5B6

• Contact
• Ahmed Abdulla, AhmedAbdulla3@cunet.carleton.ca

Abstract

The increased likelihood of extreme weather events and the expansion of variable renewable energy (VRE) sources underscore the need for improved uncertainty characterization in electric power systems. Traditional real-time operation decisions for power generators and storage are predicated on single-point forecasts of electricity demand and energy supply for the planning horizon. Cognizant of the error in such forecasts, electricity system operators schedule assets so that the system has sufficient operating reserves to account for real-time deviations in demand and renewable energy production, which may lead to elevated reserve targets and system costs under higher uncertainty. A promising alternative that reduces system costs and enhances reliability is the stochastic unit commitment (SUC) methodology. Rather than using a single-point forecast, SUC utilizes an ensemble of plausible forecasts to inform operational decisions by pre-positioning all power-generating and storage assets to satisfy demand under each scenario. The applicability of SUC, however, is contingent upon the quality of the probabilistic forecast ensembles and the ability to solve the problem within the required time constraints. We develop a methodology to (i) generate high-quality, ensemble forecasts of electric demand and VRE generation, (ii) optimize large, long-duration energy storage assets via a risk-adjusted SUC model, and (iii) create comprehensive and realistic power-grid transmission networks that represent any US balancing authority. We demonstrate our approach in the Duke Energy system—the largest vertically integrated utility in the United States—with a total of 2.14 GW of pumped-hydro capacity. Our approach achieves near-optimal operational costs and mitigates system risk due to the uncertainty in short-term forecasting. Our findings offer actionable insights for refining and expediting studies needed for policies and market decisions: (i) policies for reducing interconnection queues with different transmission contracts (ERIS vs NRIS), (ii) policies for signing new contracts on new large loads (for demand-response, backup generation, and storage assets), and (iii) generation and transmission capacity-expansion planning.

Speaker Biography

Dr. Dimitris Floros is a postdoctoral researcher at the Nicholas School of the Environment at Duke University. His latest research focuses on integrating uncertainty and risk considerations into short-term power system operations as part of the GRACE project. His research interests and experience encompass theoretical and computational analysis of network data, a field at the intersection of data analytics, network science, and high-performance computing. Dr. Floros applies network analysis techniques to generate high-quality probabilistic forecasts of electricity load and variable renewable energy (VRE) generation and to enhance the planning and operation of electric power systems under uncertainty.

Food will be provided at this seminar.