2016 INFORMS Annual Meeting Program

MA09

INFORMS Nashville – 2016

MA09 103B-MCC Energy System Design and Optimization Invited: Energy Systems Management Invited Session Chair: Kai Pan, University of Florida, 303 Weil Hall, Gainesville, FL, 32611, United States, kpan@ufl.edu 1 - An Asynchronous Dual Decomposition Algorithm For Stochastic Unit Commitment Ignacio Aravena, PhD Student, Université catholique de Louvain, Voie du Roman Pays 34 bte L1.03.01, Louvain-la-Neuve, 1348, Belgium, ignacio.aravena@uclouvain.be, Anthony Papavasiliou We present an asynchronous dual decomposition algorithm for solving transmission constrained stochastic unit commitment (UC) under multi-area renewable production uncertainty with a sub-hourly resolution. Dual iterations rely upon asynchronous subgradient methods, while primal candidates are recovered from dual subproblem solutions. The algorithm is implemented on a high performance computing cluster and its performance is compared to a deterministic UC model with exogenous reserve targets. The superior performance of stochastic UC in terms of expected cost, reliability, and run time is demonstrated on an industrial scale test case of the Central Western European region. 2 - Valid Inequalities For Hydro Genco Self-scheduling Optimization Minseok Ryu, University of Michigan, Ann Arbor, MI, United States, msryu@umich.edu, Antonio J. Conejo, Ruiwei Jiang We study on a self-scheduling optimization problem for a hydro generation company (GENCO) that manages a set of interconnected hydro reservoirs. We consider a class of key physical and operating characteristics of hydro reservoirs and generators in practical settings, including the minimum up/down time, the prohibited operating zones, and the nonlinear performance curves. We employ a mixed-integer linear programming (MILP) approach to formulate or approximate these characteristics. The MILP approach facilitates the use of many efficient computational tools, e.g., optimization solvers and valid inequalities. Finally, numerical experiments are conducted based on a real-world case study. 3 - A Bi-level Decision Dependent Stochastic Programming Model For Generation Investment Planning Yiduo Zhan, University of Central Florida, yzhan@knights.ucf.edu Qipeng Zheng A multistage bilevel decision dependent stochastic model is presented to tackle the generation investment planning problem. This model addresses both exogenous and endogenous uncertainties. The upper-level focuses on a long-term generation planning problem. The lower-level represents an electricity pricing problem that addresses the market clearing consideration with local transmission network. A linear reformulation solution approach is developed for nonlinear terms. The optimization model is implemented to CPLEX with C++. Real-world scenarios are tested. 4 - Optimal Bidding Strategy For Electricity Market Participants Considering Wind And Price Uncertainties Kai Pan, University of Florida, 411 Weil Hall, Gainesville, FL, 32611, United States, kpan@ufl.edu, Yongpei Guan, Jean-Paul Watson An optimal bidding strategy is derived for independent power producers (IPPs) by attending both day-ahead (DA) and real-time (RT) markets as a price taker. The IPP submits an offer of generation amounts to the DA market, for which a multistage adaptive optimization setting is explored for submitting RT market offers for each hour as a recourse by utilizing the more accurate forecasting of renewable output and RT price as the forecast range shrinks. This proposed strategy is theoretically justified of its significant advantages over existing alternative ones. The numerical studies show the promising future of adapting the proposed strategy and verify the effectiveness of the proposed cutting planes.

MA10 103C-MCC SpORts: Bracketology

Sponsored: SpORts Sponsored Session

Chair: Laura Albert McLay, University of Wisconsin-Madison, 1513 University Avenue, Madison, WI, 53706, United States, laura@engr.wisc.edu 1 - A Modified Logistic Regression Markov Chain Model For Ranking College Basketball And Football Teams And Forecasting Game Outcomes Laura Albert McLay, University of Wisconsin-Madison, laura@engr.wisc.edu Selecting the teams for the College Football Playoff for NCAA Division IA men’s football is a controversial process performed by the selection committee. We present a method for forecasting the four team playoff weeks before the selection committee makes this decision. Our method uses a modified logistic regression/Markov chain model for rating the teams, predicting the outcomes of the unplayed games, and simulating the unplayed games in the remainder of the season to forecast the teams that will be selected for the four team playoff. You can check out the methodology and results at http://bracketology.engr.wisc.edu/ 2 - Sampling From The 9,223,372,036,854,775,808 Possible Brackets In The Ncaa Men’s Basketball Tournament Using The Power Model This paper proposes the Power Model to estimate the winning seed distribution out of 9,223,372,036,854,775,808 possible brackets for the NCAA basketball tournament. The Power Model incorporates both the possibility of upsets and the better performance of stronger seeds by quantifying the relative strength of each pair of teams as a power function of their seed numbers. The Power Model is assessed based on the aggregate performance of one million brackets, which are generated for the five most recent tournaments (2012-2016) and scored using the ESPN scoring system. 3 - Predicting The Other Bracket Analysis Of The Selection Process For The National Invitation Tournament Stephen Hill, University of North Carolina - Wilmington, hills@uncw.edu In this work the selection process for college basketball’s National Invitation Tournament (NIT) is examined. Using historical selection data, models are constructed from variables that are shown to be strong predictors of NIT tournament selection. Model quality is also assessed. MA11 104A-MCC Network Optimization Models and Applications I Sponsored: Optimization, Network Optimization Sponsored Session Chair: Jorge A Sefair, Arizona State Univerity, 699 S. Mill Ave., BYENG 330, Tempe, AZ, 85281, United States, jorge.sefair@asu.edu 1 - Optimizing The Recovery Of Disrupted Multi-echelon Assembly Supply Chain Networks Huy Q Nguyen, Rensselaer Polytechnic Institute, 110 8th Street, We consider the problem of recovering multi-echelon assembly supply chain networks from large-scale disruptive events. Each supplier within this network assembles a component from a series of sub-components received from other suppliers. We show that scheduling rules applied locally at each supplier can optimize key recovery metrics including minimizing the maximum tardiness of any order of the final product of the supply chain network and minimizing the time to recover from the event. Our approaches are applied to a data set modeling an industry partner’s supply chain. Arash Khatibi, University of Illinois, Urbana, IL, 61802, United States, khatibi2@illinois.edu, Douglas M King, Maryam Kazerooni, Sheldon H Jacobson 5119 Center for Industrial Innovation, Troy, NY, 12180, United States, nguyeh7@rpi.edu, Thomas Sharkey, John E. Mitchell, William (Al) Wallace

124