Informs Annual Meeting 2017

TC80

INFORMS Houston – 2017

3 - When to Walk Away – Assessing Net Pay Thresholds in an Exploration Well Patrick Burdett, Chevron, Houston, TX, United States, PBurdett@chevron.com With drilling costs exceeding a million dollars per day in deep water wells, it is essential to understand when, in real time, to walk away from an exploration well that is sub-economic. A critical indicator of the economic viability of a field is the net pay thickness seen in the exploration well. This presentation address how a minimum economic pay thickness can be calculated beforehand that will give a real-time signpost on economic viability, and how that threshold can be adjusted on the fly for other information obtained from the well. 4 - A New Approach to Analyzing Data Acquisition Programs Before and After Data Come in Jincong He, Chevron, Heyj@chevron.com Data acquisition programs, such as surveillance and pilot, play an important role in reservoir management and are crucial for minimizing subsurface risks and improving decision quality. However, these projects often involve significant cost in terms of both capital investment and production downtime. In this talk I will present some recently developed methodologies to maximize “the bang for the buck”, i.e., to optimize the program design before data come in, and rapidly integrate and interpret the data after they come in. 5 - Decisions using Monte Carlo: More Samples May not be Better Mark A.Powell, Attwater Consulting, Webster, TX, United States, mark.powell@attwaterconsulting.com Monte Carlo is one of the most ubiquitous analytics tools used in risk and decision making. The literature on Monte Carlo errors is rich, and the general consensus is that more samples reduce the errors. Quantification of the absolute error in any particular Monte Carlo however remains elusive. A new method to quantitatively bound Monte Carlo errors for any simulation using any number of samples will be presented, along with a decision strategy to take advantage of these error bounds. We will find that many decisions do not require tons of expensive Monte Carlo samples using this strategy. 381C Resilience in the Energy Sector Sponsored: Energy, Natural Res & the Environment, Energy Sponsored Session Chair: Spada Matteo, Paul Scherrer Institute, Villigen PSI, 5232, Switzerland, matteo.spada@psi.ch Co-Chair: Feng Qiu, Argonne National Laboratory, Lemont, IL, 60439, United States, fqiu@anl.gov 1 - Resilience of the European Natural Gas Network Peter Lustenberger, Future Resilient Systems (FRS), Singapore- ETH Centre (SEC), Singapore, Singapore, lustenberger@frs.ethz.ch, Patrick Gasser, Wansub Kim, Matteo Spada, Peter Burgherr, Stefan Hirschberg, Bozidar Stojadinovic Energy systems are regularly subjected to major unforeseen disruptions that are triggered by natural and human-made causes, which affect the economy and society. This calls for resilience assessment, including consideration of disruption and recovery processes of the system under investigation. This study presents, on the one hand, an assessment of probabilistic disruption processes caused by both natural hazards and technical failures and, on the other hand, potential recovery processes, applicable to the European natural gas network and its components. 2 - Smart and Resilient Energy There is a growing effort in the electricity generation and distribution industry to use smart technologies to promote efficiency and minimize the frequency and duration of outages. Smart energy technologies collect, analyze and utilize data in real time, and include SCADA systems, unmanned aerial vehicle surveillance, remote vibration monitoring, etc. Many smart technologies contribute to energy sector resilience. However, the reliance on highly-connected smart technologies leaves the energy sector vulnerable to unanticipated or targeted disruptions. This presentation serves to discuss the contribution, both positive and negative, of smart energy technologies on system resilience. 3 - Optimal Post-disruption Resilience-oriented Restoration for Interdependent Infrastructure Network Yasser Adel Almoghathawi, University of Oklahoma, 1506 E Lindsey St., Apt. P, Norman, OK, 73071, United States, moghathawi@ou.edu, Kash Barker We propose a multi-objective optimization model using mixed-integer programming for restoring the interdependent infrastructure networks after a disruptive event. The model considers the availability of limited time and TC80 Matthew Wood, United States Army Engineer R&D Center, 686 Virginia Road, Concord, MA, 01742, United States, Matthew.D.Wood@usace.army.mil, Igor Linkov

resources with two disruption scenarios, complete and partial, for the disrupted network components, nodes or links. It aims to (i) minimize the total cost of restoration, and (ii) maximize the combined resilience of interdependent infrastructure networks considering the physical interdependencies between them. The proposed model is illustrated through a set of interdependent infrastructure networks after hypothetical earthquakes in Shelby County, TN, United States. 4 - Quantifying and Visualizing the Economic Impact of Commercial Districts Due to an Electric Power Outage Jose Emmanuel Ramirez-Marquez, Stevens Institute of Technology, Castle Point on Hudson, School of Systems & Enterprises, Hoboken, NJ, 07030, United States, jmarquez@stevens.edu, Andrea Garcia tapia, Kash Barker The level of dependence of communities on electric power, from social behavior to economic productivity, is significantly high, revealing how vulnerable communities are to a power outage. Assessing the economic impact of electricity outages after a disruption in the service is a challenging task, especially when costumer’s interruption costs vary from type of user. A geospatial visualization is presented to observe and compare the areas that are more vulnerable in terms of economic impact based on the areas covered by each distribution substation. 382A Dynamic Robust Optimization and Learning Sponsored: Optimization, Optimization Under Uncertainty Sponsored Session Chair: Yehua Wei, Boston College, Chestnut Hill, MA, 02467-3809, United States, yehua.wei@bc.edu Co-Chair: Dan Andrei Iancu, Stanford University, Stanford, CA, 94107, TC81 Do Young Yoon, 37 Angell Ct, Apt 319, Stanford, CA, 94305, United States, doyoung@stanford.edu, Dan Andrei Iancu, Nikolaos Trichakis We study a dynamic decision problem under limited information, in which a decision maker chooses when to monitor a system - which allows gaining more information about current and future states of the system - as well as whether to stop the monitoring process and take a particular action. With robust optimization framework, we solve for a monitoring policy that maximizes the worst-case stopping reward. Our main results show that, under mild conditions, a static monitoring policy is equivalent to a fully dynamic policy along the worst-case path. We then provide several methods for obtaining an optimal static monitoring policy and present a robust optimal algorithm for the dynamic decision problem. 2 - Discrete Convexity and Dynamic Robust Optimization Dan Andrei Iancu, Stanford University, 655 Knight Way, Stanford, CA, 94107, United States, daniancu@stanford.edu, Yehua Wei We discuss necessary and sufficient conditions for the optimality of specific classes of policies in dynamic robust optimization. We then focus on the specific case of affine policies, and show how our conditions can be used to recover and generalize several existing results in the literature. Our treatment draws interesting connections with the theory of discrete convexity (L-natural / M- natural convexity and multimodularity) and global concave envelopes, which may be of independent interest. 3 - A Constraint Generation Algorithm for Two-stage Robust Network Flow Problem Yehua Wei, Boston College, 140 Commonwealth Avenue, Chestnut Hill, MA, 02467-3809, United States, yehua.wei@bc.edu, David Simchi-Levi, He Wang We propose a new method for solving two-stage robust network flow problems with polyhedral uncertainty set. By exploiting the network structure of the problem, we suggest a reformulation where its number of constraints is exponential with the number of nodes in the network, but independent with the number vertices in the uncertainty set. We then develop an effective constraint generation algorithm for solving the two-stage robust optimization problem, and apply our method to several application domains. 4 - Out-of-Sample Analysis of Robust Empirical Optimization Andrew Lim, National University of Singapore, 15 Kent Ridge Drive, Mochtar Riady Building, Singapore, 119245, Singapore, andrewlim@nus.edu.sg We study the out-of-sample performance of distributionally robust data driven optimization. The key result is that the reduction in the variance of the out-of- sample reward that results from “robustifying” empirical optimization is an order of magnitude greater than the impact on the mean. Implications for calibration of robust optimization models is also discussed. United States, daniancu@stanford.edu 1 - Monitoring with Limited Information

366