Informs Annual Meeting 2017

TD22

INFORMS Houston – 2017

3 - Optimizing Wind Farm Siting to Reduce Variability Impacts of Wind Power Amelia Musselman, Georgia Institute of Technology, 1040 Huff

3 - Maximal Expected Coverage for Network Restoration and Emergency Response After Disrupted Events Suzan Afacan, University of Wisconsin-Madison, 27D University Houses, Madison, WI, 53705, United States, iloglu@wisc.edu, Laura Albert Infrastructure recovery is important for delivering time-sensitive services and commodities after a disaster while also repairing network damage. To examine this issue, we present an extension of the maximal covering problem in the case of extreme events. In the model, we coordinate two types of service providers: (1) recovery crews who repair disrupted roads and (2) emergency responders who deliver services and commodities by relocating emergency service responder between potential locations with restricted access. The objective is to maximize the covered demand with backup coverage over the time horizon. The model is illustrated with a computational example. 4 - An Optimization-based Restoration Approach for Enhanced Interdependent Infrastructure Network Resilience Yasser Adel Almoghathawi, University of Oklahoma, 1506 E Lindsey St., Apt P, Norman, OK, 73071, United States, moghathawi@ou.edu, Kash Barker Infrastructure networks such as power, water, and gas are increasingly becoming more interdependent to be functional, potentially making them highly vulnerable to any disruptive event, which makes their restoration more challenging. We propose a resilience-driven solution approach for the restoration problem of interdependent infrastructure networks following a disruptive event. The proposed approach is based on an optimization model considering fixed and variable recovery times for disrupted network components. The proposed approach is illustrated through generated interdependent networks with different sizes and topologies considering different disruption sizes and scenarios. 342D Innovations in Pricing Sponsored: Revenue Management & Pricing Sponsored Session Chair: Adam Elmachtoub, Columbia University, New York, NY, 10027, United States, adam@ieor.columbia.edu 1 - The Value of Personalization in Limited Information Markets Michael Hamilton, Columbia University, 500 West 120th Street, New York, NY, 10027, United States, mh3461@columbia.edu, Adam Elmachtoub, Vishal Gupta We study a one product pricing problem, and quantify the revenue difference between optimal single price strategies and fully personalized ones based on simple summary statistics about the market. Namely using only the mean, range and volatility of the market, we characterize the best and worst case “gaps” between an optimal one price strategy and fully personalized one. As an application of our work we give a simple, distribution-free and near optimal algorithm for market segmentation. 2 - A Joint Framework for Market Segmentation and Price Curve Estimation Ryan McNellis, Columbia University, New York, NY, 10027, United States, rtm2130@columbia.edu, Adam Elmachtoub We propose a new methodology for incorporating feature information into price- setting problems. Relevant applications include the recommendation of personalized discounts and customizing bids for advertising exchanges. Our method uses recursive partitioning to segment the feature space into a collection of regions. A probability model involving only the decision variable (i.e., price) is then fit locally in each region. A key strength of our method is its interpretability, as the resulting market segmentation is easily visualized and understood. We explore new training algorithms to ensure scalability in the face of high- dimensional data. 3 - Primal-dual Algorithms for Revenue Management Problems with Unknown Demand Model He Wang, Georgia Institute of Technology, Atlanta, GA, 30332- 0205, United States, he.wang@isye.gatech.edu In this talk, we consider a general class of revenue management problems with finite inventory and finite selling horizon. In addition, we assume the customer demand model is unknown prior to the selling season. We propose a primal-dual algorithm that simultaneously estimates the customer demand model and the bid prices associated with inventory constraints. TD22

Road NW, Apt 1218, Atlanta, GA, 30318, United States, amusselman@gatech.edu, Valerie Thomas, Dima Nazzal

We present a multi-objective optimization model to strategically locate wind farms to reduce wind power variability. In order to model the impact of wind variability without explicitly modeling the full electricity system we develop a metric, which we call demand deficit, to measure the load that the remainder of the system would need to account for under wind fluctuations. By selecting which sites to construct from a set of potential wind farm locations our model minimizes both demand deficit and the maximum change in demand deficit over a given period. We solve our model for a case study of the Southwest Power Pool and develop an efficient frontier to compare the trade-offs between the two objectives.

TD20

342B Information Systems Contributed Session Chair: Ozay Ozaydin, Dogus University, Istanbul, Turkey, ozay.ozaydin@gmail.com

1 - Budget Management in International Humanitarian Organizations Milad Keshvari Fard, ESSEC Business School, 3 Avenue Bernard Hirsch, Cergy, 95021, France, milad.keshvarifard@essec.edu, Felix Papier Humanitarian Organizations (HO) have to allocate their limited budget to countries and missions at the beginning of each year. Since the budget comes from donations, it is uncertain and may be partly earmarked for specific uses. Considering an HO’s objective to improve the social welfare of the beneficiaries and to meet the target fill rates, budget allocation is a challenging problem. We model budget allocation as a two-stage decision process: In the first stage, the HO decides the budget target for each country and in the second stage, the HO determines the optimal allocation of the non-earmarked budget. 342C Resilient Infrastructure and Community Networks Invited: InvitedNatural Hazard Planning Invited Session Chair: Kash Barker, University of Oklahoma, University of Oklahoma, Norman, OK, 73019, United States, kashbarker@ou.edu 1 - Building Resilience through Redundancy in a Multi-echelon Supply Chain TD21 Due to the complexity of supply chains (SC) they must increase their resilience to meet demand. A multi-echelon supply chain is analyzed using Discrete Event Simulation (DES) for measuring the effects of disruptions downstream the supply chain - modeled in a probabilistic manner with different impact levels. An experimental design is conducted on multiple input parameters to illustrate the model application and behavior. To lessen the effects of the disruptions, this model considers redundancy as a supply chain capability and aid managers in identifying SC nodes vulnerable to disruptions. SC can increase resilience and a rapid recovery to optimal operation levels. 2 - Analytically Comparing Disaster Resilience across Multiple Dimensions Christopher Zobel, Virginia Tech, 1007 Pamplin Hall, Dept of Business Info Tech, Blacksburg, VA, 24061-0235, United States, czobel@vt.edu It is important to be able to compare the resilience of a community to different types of disaster event in order to assess both its vulnerabilities and the opportunities available to address them. Since resilient behavior can be complex and multi-dimensional, however, one must be able to characterize the different ways in which that behavior is actually exhibited in practice. This presentation explores an approach for visualizing and analyzing the relationships between each of the individual resilience dimensions. Jose Emmanuel Ramirez-Marquez, Stevens Institute of Technology, Castle Point on Hudson, School of Systems & Enterprises, Hoboken, NJ, 07030, United States, jmarquez@stevens.edu, Araceli Zavala

383