2016 INFORMS Annual Meeting Program

TC06

INFORMS Nashville – 2016

TC04 101D-MCC Power System Operations Under Increasing Uncertainty Sponsored: Energy, Natural Res & the Environment, Energy I Electricity Sponsored Session

2 - Reliable Fuel Supply Chain Design Bo Zeng, University of Pittsburgh, bzeng@pitt.edu, Anna Danandeh, Brent Caldwell To ensure reliable operations of a power plant, an optimization based fuel supply chain model is developed and implemented. TC06 102A-MCC Data-Intensive Computational Methods for Large-scale Infrastructure Systems Sponsored: Data Mining Sponsored Session Chair: Adrian Albert, C3IoT, 1300 Seaport Boulevard, Suite 500, Redwood City, CA, 94063, United States, adrian.t.albert@gmail.com 1 - Sparse Data Analytics For Modern Engineering Systems Borhan Sanandaji, Risk Management Systems (RMS), Hall, Newark, CA, 24061, United States, sanandaji@eecs.berkeley.edu Forecasting plays a vital role in reliable operation of modern engineering systems such as smart grids and transportation systems. These systems are often large- scale and generate a huge amount of data. It is, therefore, quite important to come up with forecasting schemes that can deal with such high-dimensionality. In this work, we propose a Sparse Spatio-Temporal Forecasting (SSTF) scheme which exploits the intrinsic low-dimensionality and structure of the generated data. We applied SSTF to predict wind speed, residential electric load, and solar irradiance in different scenarios to prove its significance as compared to other benchmark models. 2 - A Learning Based Method For Real Time Prediction Of Cascading Failures Yue Zhao, Stony Brook University, Stony Brook, NY, United States, yue.zhao.2@stonybrook.edu, Jianshu Chen Real time prediction of imminent cascading failures in a dynamically evolving power grid is studied. As the cascade look-ahead window increases, the number of future cascade scenarios grows exponentially. A novel learning based method is developed to compute the marginal failure probability of each line due to cascades at times deep into the future. The proposed method enjoys the unique advantage that a labeled data set can be generated in an arbitrarily large amount at very low cost. Numerical results demonstrate that the off-line trained predictive model provides very fast online and accurate prediction of cascading failures. 3 - New Approaches In Data Analysis For Infrastructural Networks: Combinatorial Hodge Theory Chase Dowling, University of Washington, Seattle, WA, United States, Cdowling@uw.edu, Lillian Ratliff, Baosen Zhang Recent advances in Hodge theory have shed light on a deep relationship between graph theory and calculus. One important theorem in calculus—the Helmholtz decomposition—splits a vector field into conservative and solenoidal components. The combinatorial Hodge decomposition extends this technique to graphs, and gives conservation law respecting flows on edges. Power, gas, and traffic networks all respect some form of conservation law, and their optimal utilization has proven difficult owing to nonlinearities in flow. We explore a novel application of the Hodge decomposition in traffic and power networks with the aim of developing control strategies in face of these nonlinearities. 4 - Energy Profile Prediction: Implications For Electric Vehicle Demand Response Caroline Camille Le Floch, University of California, Berkeley, Berkeley, CA, United States, caroline.le-floch@berkeley.edu, Scott Moura This work shows a predictive framework that uses demographic data to predict energy profiles and acceptance of smart grid tariffs. Our analysis is based on the Australian Smart Grid Data, including electricity use interval readings, customer demographics, peak event offers and acceptances. First, we use clustering methods to define a representative dictionary of hourly load shapes, and assign individual energy profiles as his/her most frequently used shapes. Second, we present the performance of several estimators to predict energy profiles and peak event responses from demographic data. Third, we discuss implications for designing smart grid programs for Electric Vehicles owners.

Chair: Antonio J. Conejo, Prof., The Ohio State University, 1971 Neil Avenue, 286 Baker Systems Engineering, Columbus, OH, 43210, United States, conejonavarro.1@osu.edu Co-Chair: Ramteen Sioshansi, Ohio State University, 1971 Neil Avenue, Columbus, OH, 43210, United States, sioshansi.1@osu.edu 1 - Ramp Capability Modeling For Reliable And Efficient Integration Of Renewable Energy Congcong Wang, MISO, Carmel, IN, United States, cwang@misoenergy.org, Dhiman Chatterjee With increasing penetration of renewable energy, net load variations and uncertainties impose challenges to maintain real-time power balance. This presentation highlights MISO’s recent development of Ramp Capability Product to manage system ramping needs. It starts with an examination of recent market evolutions that drive both operational and economic needs of resource flexibility and then presents the design of Ramp Capability Product that systematically pre- position resources with flexibility to meet future net load at a specified level of confidence. More importantly, explicit price signals are developed to reflect the Farzaneh Abbaspourtorbati, EPFL, Lausanne, Switzerland, Farzaneh.Abbaspourtorbati@swissgrid.ch, Antonio J. Conejo, Jianhui Wang, Rachid Cherkaoui This paper analyzes the impacts of flexible demands on day-ahead market outcomes in a system with significant wind power production. We use a two- stage stochastic market-clearing model, where the first stage represents the day-ahead market and the second stage the real-time operation. On one hand, flexibility of demands is beneficial to the system as a whole since such flexibility reduces the operation cost, but on the other hand, shifts in demands from peak periods to off-peak periods may influence prices in such a way that demands may not be willing to provide flexibility. Specifically, we investigate the impacts of different degree of demand flexibility on day-ahead prices. 3 - Aggregating (almost) Symmetric Generators In Unit Commitment Ben Knueven, University of Tennessee, Knoxville, TN, United States, bknueven@vols.utk.edu, Jim Ostrowski, Jean-Paul Watson, Jianhui Wang We consider a method to precisely aggregate symmetric ramping unconstrained generators in unit commitment formulations. We apply the same methods to nearly symmetric generators to create symmetric relaxations of the unit commitment problem, and empirically test the strength of the relaxation. We demonstrate massive computational improvements over the standard formulation for the CAISO set of generators. Extensions to accelerate stochastic unit commitment are also examined. Reliable Power System Design and Operations Sponsored: Energy, Natural Res & the Environment, Energy I Electricity Sponsored Session Chair: Bo Zeng, University of Pittsburgh, Benedum Hall 1009, Pittsburgh, PA, 15261, United States, bzeng@pitt.edu 1 - Tighter Modeling And Enhanced Solutions For Power System Operations Under Uncertain Environment Lei Wu, University of Clarkson, lwu@clarkson.edu In emerging power systems, as the generation side gets more distributed and the demand side becomes more active, it is of critical importance to evaluate the impacts of individual assets on the reliable and economic operation of power systems. This presentation will highlight several key issues in the operation of power systems with significant penetration of renewable energy and DR assets, and discuss advanced modeling and optimization techniques, robust security- constrained unit commitment (SCUC) models in particular, for enhancing the reliability and economics of power system operations under uncertain environment. underlying cost causation and provide economic incentives. 2 - Is Being Flexible Advantageous For Demands? TC05 101E-MCC

303