Informs Annual Meeting 2017

POSTER SESSION

INFORMS Houston – 2017

MB83

2 - Managing Volunteer Convergence at Disaster Relief Centers Hussain Abualkhair, North Carolina A&T.State University, 614 Eagle Road, Apt 2C, Greensboro, NC, 27407, United States, hfabualk@aggies.ncat.edu Disaster Relief center managers face uncertainty in both the timing and quantity of donations, as well as the number of spontaneous volunteers. Volunteer convergence is one of the biggest challenges for relief center managers. To better understand this phenomenon and find effective management strategies, we develop an agent-based simulation model consisting of donors providing relief items, beneficiaries in need of relief items and random arrivals and departures of volunteers. We investigate volunteer assignment policies that reduce donor beneficiary waiting time under given changes in volunteer capacity, available inventory, and beneficiary’s arrival rates. 3 - Budgeting Dredging Projects to Improve Resilience of the Inland Waterway Network Khatereh Ahadi, PhD Candidate, University of Arkansas, 4207 Bell Engineering Center, Fayetteville, AR, 72701, United States, kahadi@uark.edu We consider the problem of selecting a budget-limited subset of maintenance actions to maximize the expected tonnage of commodities that can be transported through the inland waterway system. Uncertainty in our model is the budget required at each district for emergency dredging arises from unpredictable conditions. Thus, our main decision is allocating budget to each district to cover the cost of reactive dredging. We model this problem as a stochastic programming model, develop solution approaches, and analyze computational results. 4 - Robust Synthetic Control Muhammad J. Amjad, Massachusetts Institute of Technology, 235 Albany Street, Suite 3121A, Cambridge, MA, 02139, United States, mamjad@mit.edu, Devavrat Shah, Dennis Shen Abadie et al introduced the idea of using ``synthetic control’’ for comparative case studies, where the control unit is a convex combination of unaffected (donor) units. We present ``robust’’ synthetic control where we propose a generic latent variable model which subsumes the classical method. Algorithmically, we first ``de-noise’’ the observations and then learn the linear synthetic control allowing us to do away with needing preprocessing “expert” knowledge or covariate information to identify a good subset of donor units. Additionally, our algorithm is asymptotically consistent in MSE and robust to missing and/or noisy observations. 5 - Predicting Power Price Profile in the Electricity Market by Solving a Bi-level Optimization Problem In this project we address a practical bi-level optimization model in the electricity market. By considering market data for a particular operating horizon, our goal is to find an hourly price profile incentivizing global production levels that closely match the market demand. To solve the problem, a two-stage heuristic algorithm is developed. An initial price profile is first found and a sequence of steps then iteratively perturbs the price profile closer to an optimal price profile. The designed heuristic algorithm can also be solved using disruptive hardware such as a quantum annealer. Our numerical results on a real data set illustrate a close match between supply and demand for 95% of the hours. 6 - Interdicting Layered Physical and Information Flow Networks Nail Orkun Baycik, Rensselaer Polytechnic Institute, 110 8th Street Center for Industrial Innovation, Suite 5015, Troy, NY, 12180, United States, baycin@rpi.edu We study the interdiction of layered networks that involve a physical flow and an information flow network. The objective of the defender is to maximize the physical flow and the objective of the attacker is to minimize this maximum amount. There exist interdependencies between the networks which leads to a network interdiction problem with a discrete inner problem. By using duality, we reformulate the problem and obtain a single-level model. We apply this technique to the application of combating illegal drug trafficking and protecting cyber infrastructures, and present computational results. 7 - Optimal Operation of Large Scale Flexible Hydrogen Production in Constrained Transmission Grids with Stochastic Wind Power Espen Flo Bødal, PhD Candidate, Norwegian University of Science and Technology, O.S.Bragstads Plass 2E, Trondheim, 7030, Norway, espen.bodal@ntnu.no The focus of this work is on developing and assessing the value of using a stochastic model when planning the operation of the flexible hydrogen production in a region with large wind power potential and constrained transmission grids, and how it affects storage strategies. Specific situations when the stochastic model is superior to a deterministic model based on the expected value is identified and analyzed in detail. The effects from stochastic wind power production on the optimal size of storage is also investigated by conducting an analysis of the dual variables of the storage constraints. Maliheh Aramon, Applied Researcher, 1QBit, Unit 403, 1461 Harwood Street, Vancouver, BC, V6G 1X7, Canada, maliheh.aramon@1qbit.com

382C Stochastic Mixed-Integer Programming and Applications Sponsored: Optimization, Optimization Under Uncertainty Sponsored Session Chair: Semih Atakan, University of Southern California, Los Angeles, CA, 90089-0193, United States, atakan@usc.edu 1 - Error Bounds for Inexact Cutting Plane Techniques for Stochastic Mixed-integer Programming Problems Ward Romeijnders, University of Groningen, Nettelbosje 2, Groningen, 9747 AE, Netherlands, w.romeijnders@rug.nl We derive error bounds for a class of inexact cutting plane techniques for stochastic mixed-integer programming problems. These error bounds converge to zero if the total variations of the probability density functions of the random variables in the model converge to zero. Moreover, we show several numerical experiments, illustrating the actual error of using these inexact methods. 2 - Solving Stochastic Mixed-Integer Programs using a Generalized Value Function We introduce the generalized value function of a mixed-integer program, which is simultaneously parameterized by its objective and right-hand side. We describe its fundamental properties, which we exploit through three algorithms to calculate it. We then show how this generalized value function can be used to reformulate two classes of very difficult mixed-integer optimization problems: two-stage stochastic mixed-integer programming and multi-follower bilevel mixed-integer programming. For both of these problem classes, the generalized value function approach allows the solution of instances that are significantly larger than those solved in the literature. 3 - Subgradient Methods for Stochastic Mixed-Integer Programs Cong Han Lim, University of Wisconsin-Madison, Madison, WI, United States, clim9@wisc.edu, Jeff T.Linderoth, James Luedtke, Stephen J.Wright We present our work on improving the subgradient method for solving the Langrangian dual of a stochastic mixed-integer program (SMIP). Our techniques include subsampling and asynchronicity, and leads to methods that have similar guarantees as the full subgradient method. The algorithms run effectively on both large-scale distributed platforms and multi-core servers. We present computational results on some classic SMIP problems. 4 - PH-BAB Applications in Power Systems Planning Semih Atakan, University of Southern California, 3715 McClintock Ave, GER.240, Los Angeles, CA, 90089-0193, United States, atakan@usc.edu, Suvrajeet Sen Progressive Hedging (PH) has been used as a heuristic optimization method for many SMIP problems. In this talk, we briefly describe our framework (PH-BAB) for using PH to compute optimal solutions to stochastic mixed-integer convex programs (SMICPs). We, then, present an algorithm within this framework for specially-structured SMICPs. The performance of the algorithm will be demonstrated on academic benchmark instances and stochastic Unit Commitment problems. Monday Poster Monday Poster Session Poster Session 1 - Weighted Networks: Edge-splitting Procedure in Centrality Computations Ivan Belik, Dr., Norwegian School of Economics, Helleveien 30, Bergen, 5045, Norway, ivan.belik@nhh.no The analysis of social network’s centralities has a high-level significance for many real-world applications. The variety of game and graph theoretical approaches has a paramount purpose to formalize a relative importance of nodes in social networks. In the given research, we present a formalized symbiotic graph theoretical and game theoretical algorithm for the centrality computation in the domain of weighted networks. The given algorithm calculates network centralities for weighted graphs based on the proposed edge-splitting procedure. The approach is tested and illustrated based on different types of network topologies. Onur Tavaslioglu, University of Pittsburgh, 7171 Buffalo Speedway, Apt 1934, Houston, TX, 77025, United States, ont1@pitt.edu, Oleg A.Prokopyev, Andrew J. Schaefer Monday, 12:30 - 2:30PM

195