2016 INFORMS Annual Meeting Program

SC10

INFORMS Nashville – 2016

SC10 103C-MCC Advances in Energy Systems Modeling Sponsored: Energy, Natural Res & the Environment, Energy II Other Sponsored Session Chair: Rudolf Gerardus Egging, Norwegian University of Science & Technology, Trondheim, Trondheim, 00000, Norway, ruud.egging@iot.ntnu.no 1 - Plug And Abandonment Of Offshore Oil And Gas Fields. Steffen J. Bakker, Norwegian University of Science and Technology, steffen.bakker@iot.ntnu.no At the end of a wells life-cycle, the well has to be permanently abandoned. This process is called plug and abandonment (P&A). Decisions in the P&A process depend heavily on uncertain factors such as oil and gas prices, rig rates or well states. Moreover, these decisions have to be taken at different levels. In this presentation we discuss a classification of the P&A process into an operational, tactical and strategic level. For each of these levels we present a corresponding model, where we make use of real options theory and the frameworks of integer and stochastic programming. 2 - Trading Off Demand Side Flexibility Vs. Supply Side Flexibility And Storage In The Electricity System Hector Maranon-Ledesma, Norwegian University of Science and Technology, hector.maranon-ledesma@iot.ntnu.no Demand Side Management (DSM) permits a reduction in peak load consumption, a more adaptable demand of electricity, allowing shuting down high emission power plants, and the intermittent renewable resources to be better exploited by the use of flexibility mechanisms. The EMPIRE electricity sector model is a long-term investment stochastic model. This model has been improved by including DSM at the operational level. The contributions of this work are including DSM in a large scale electric system and highlighting the importance that DSM might acquire in the future European power system. 3 - The Effect Of Drivers Elasticity On The Optimal Pricing And Management Of Electric Vehicles Charging Chiara Bordin, Norwegian University of Science and Technology, Trondheim, Norway, chiara.bordin@iot.ntnu.no, Stig Ødegaard Ottesen, Asgeir Tomasgard, Siri Bruskeland Ager-Hanssen, Siri Olimb Myhre The increasing demand for Electric Vehicles (EV) charging puts pressure on the power grids as in some situations the power consumption can exceed the grid capacity. We propose a mathematical model for the indirect control of EV charging that finds an optimal set of price signals to be sent to the drivers according to their flexibility. The objective is to satisfy the demand when there is a capacity lack by minimizing the curtailment of loads and prioritizing the loads shifting. The key contribution is the use of the elasticity concept to forecast the drivers reactions to the price signals. Sensitivity analyses are presented to investigate the elasticity effect on prices and loads management. 4 - Risk Aversion In Imperfect Natural Gas Markets. Rudolf Gerardus Egging, Norwegian University of Science & Technology, ruud.egging@iot.ntnu.no We consider risk aversion by natural gas supply companies considering investment in conventional and shale gas resources in a stochastic multi-period mixed complementarity problem. Uncertainty considered includes political risk and resource sizes. We consider shale gas investment in Poland and Ukraine in a realistic market setting in Europe. We discuss investment decisions and profits for varying levels of risk aversion.

networks, as well as related asymptotic results on phase transitions in random graphs. 2 - Robust Network Clusters With Small-world Property

Jongeun Kim, University of Florida, Gainesville, FL, United States, kje0510@ufl.edu, Alexander Veremyev, Vladimir Boginski, Oleg A Prokopyev

Networks are popular and effective tools for analyzing real-world systems, such as telecommunication, transportation, and social networks. Network robustness is one of the important issues, because some undesired failures may affect connectivity and functionality of a network. The ideal robust cluster in a network is a clique and clique-relaxation research have been developed in recent decades. In this talk we will address small-world clusters that are robust but also have certain natural properties. 3 - Detecting Essential Proteins Using A Novel Star Centrality Metric Mustafa Can Camur, North Dakota State University, Fargo, ND, United States, mcancamur@gmail.com In this talk, we propose a new centrality metric (referred to as star centrality), which aims to incorporate information from the closed neighborhood of the node, rather than strictly from the node itself. More specifically, we turn our focus to degree centrality and show that in the complex protein-protein interaction networks it is a naive metric that can lead to misclassifying importance in the network. We portray the success of the new metric using protein-protein interaction networks, and investigating the significant difference in the importance of individual nodes we observe when transitioning from node degree centrality to star degree centrality. 4 - Estimating The Maximum IUC Using SDP Relaxations Eugene Lykhovyd, Texas A&M University, lykhovyd@tamu.edu, Sergiy Butenko If you have a simple, undirected graph, the Independent Union of Cliques (IUC) problem is to find the maximum subset of vertices, in which every connected component is a clique. It is known that this problem can be formulated on 3- uniform hypergraphs as the maximum weak independent set. We propose the estimates for IUC problem based on different SDP relaxations, extending the Lov\’asz estimate for the maximum stable set. The comparison of different approaches is also presented. SC12 104B-MCC Convex Relaxations for Nonconvex Polynomial Optimization Sponsored: Optimization, Integer and Discrete Optimization Sponsored Session Chair: Daniel Bienstock, Columbia University, 116th and Broadway, New York, NY, 10027, United States, dano@columbia.edu 1 - LP And SOCP-based Algebraic Relaxations For Polynomial Optimization Amir Ali Ahmadi, Princeton University, a_a_a@princeton.edu We present ongoing work on solving polynomial optimization problems using linear and convex relaxations based on a number of ideas, including separation from the set of rank-1 psd matrices, and, in particular, the method of approximate representation of continuous variables as weighted sums of binary variables. We will discuss theory and computational practice. Joint work (Gonzalo Munoz, Chen Chen and Daniel Bienstock). 2 - Online First-order Framework For Robust Convex Optimization Fatma Kilinc-Karzan, Carnegie Mellon University, fkilinc@andrew.cmu.edu, Nam Ho-Nguyen We present a flexible iterative framework to approximately solve robust convex optimization problems. Our results are based on weighted regret online convex optimization and online saddle point problems. A key distinguishing feature of our approach from prior literature is that it requires access to only cheap first- order oracles for each constraint individually and does simple online updates in each iteration while maintaining the same convergence rate. For strongly convex functions, we also establish a new improved iteration complexity. As a result, our approach becomes much more scalable and hence preferable in large-scale applications from machine learning and statistics domains. 3 - New And Old Results On Polynomial Optimization Daniel Bienstock, Columbia University, dano@columbia.edu We present ongoing work on solving polynomial optimization problems using linear and convex relaxations based on a number of ideas, including separation from the set of rank-1 psd matrices, and, in particular, the method of approximate representation of continuous variables as weighted sums of binary variables. We will discuss theory and computational practice, and attempt to relate our work to earlier results by Renegar and Barvinok. Joint work (Gonzalo Munoz, Chen Chen and Daniel Bienstock).

SC11 104A-MCC

Dense Clusters in Network Optimization Sponsored: Optimization, Network Optimization Sponsored Session

Chair: Vladimir Stozhkov, University of Florida, 2330 SW Williston Rd, Apt 2826, Gainesville, FL, 32608, United States, vstozhkov@ufl.edu 1 - Relative Clique Relaxations In Complex Networks Vladimir Boginski, University of Central Florida, Vladimir.Boginski@ucf.edu Real-world complex networks exhibit clustered structure: certain groups of nodes (vertices) form “cohesive” or “highly connected” clusters (can also be referred to as “communities”), which can be rigorously characterized using graph-theoretic concepts. In this presentation, we will focus on so-called relative clique relaxation models, which are obtained by relaxing certain metrics that attain their maximum possible values on a clique: edge density, minimum vertex degree, and vertex connectivity. We will discuss optimization problems of identifying such clusters in

72