Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

SB04

4 - Revisiting Approximate Linear Programming Using a Saddle Point Approach Selvaprabu Nadarajah, College of Business, University of Illinois at Chicago, 601 South Morgan Street,, UH 2406, Chicago, IL, 60607, United States, Qihang Lin, Negar Soheili Approximate linear programs (ALPs) are well-known models for approximating high dimensional Markov decision processes (MDPs). Solving ALPs exactly remains challenging, for instance, in applications where (i) the MDP includes nonlinear reward and transition dynamics and/or (ii) rich basis functions are required to obtain a good VFA. We address this tension between ALP theory and solvability by proposing a novel ALP saddle point reformulation and mirror descent solution approach that embeds iterative learning of constraint violation distributions. We present convergence guarantees for our approach and test it on inventory control and energy storage applications. Joint Session OPT/Practice Curated: Optimization under Uncertainty in Energy and the Environment Sponsored: Optimization/Optimization Under Uncertainty Sponsored Session Chair: Sarah G. Nurre, University of Arkansas, Walnut Creek, CA, 94596, United States 1 - Resilient Transmission Hardening Planning in a High Renewable Penetration Erra Ali Bagheri, Oklahoma State University, Stillwater, OK, 74075, United States, Chaoyue Zhao, Feng Qiu, Jianhui Wang Transmission system hardening is a practice to improve system resilience against possible disruptions. In a power system with a very high penetration of renewable energy, the system hardening will be further complicated by the uncertainty of renewable energy. We study the transmission line hardening planning problem in the context of probabilistic power flows injected by the high penetration of renewable energy. We assume that the information of renewable energy is incomplete. A data-driven two-stage stochastic model is formulated by considering the joint worst-case wind output distribution and transmission line contingencies. 2 - An Integrated Two-level Inventory Problem: Applications to Battery Management in Electric Vehicle and Drone Swap Stations Amin Asadi, University of Arkansas, 4207 Bell Engineering Center, 800 W. Dickson Street, Fayetteville, AR, 72701, United States, Sarah G. Nurre We examine the new class of stochastic two-level integrated inventory problems (TOLIPs) with applications in drone and electric vehicle (EV) battery management. In the TOLIP, we model how first-level battery charge with battery charging, discharging, and replacement actions impact the deterioration and necessary actions for second-level battery capacity. In the context of a battery swap station, we formulate the TOLIP using a Markov Decision Process model with an uncertain demand for swaps. We perform computational experiments to derive optimal policies and deduce insights. 3 - Stochastic Multi-Objective Water Allocation with Hedging Rule Ming (Arthur) Yang, The Ohio State University, Columbus, OH, United States, Guzin Bayraksan The problem of water shortage is usually caused by uneven distribution of rainfall, increasing water demand, or other environmental and human factors. In this study, we develop a mathematical model that applies hedging rules under inflow and demand uncertainties to provide an optimal strategy in managing and operating reservoirs. Hedging rules determine different rationing levels between users at certain trigger volumes of reservoirs during droughts. We develop time series models for the uncertain inflows and demands and model the problem as a multistage stochastic mixed integer program with multiple objectives. We present numerical results on a real-world multi-reservoir system. 4 - Industrial Demand Response in Electricity Markets Golbon Zakeri, University of Auckland, Dept of Eng Science, Private Bag 92019, Auckland, New Zealand We will present a single stage optimization problem faced by a large industrial consumer of electricity who is capable responding to price, and also capable of offering interruptible load reserves. We will then extend our model to a multistage setting and report some computational results. n SB02 North Bldg 121B

n SB03 North Bldg 121C Entrepreneurship and Innovation Sponsored: Technology, Innovation Management

& Entrepreneurship Sponsored Session Chair: Sinan Erzurumlu, Babson College, Babson Park, MA, 02457, United States 1 - Lean Startup Goal Conflict: Can Startups Manage Survival and Revenue Growth Simultaneously? Emre Guzelsu, Boston University Questrom School of Business, 595 Commonwealth Ave, Boston, MA, 02215, United States, Nitin Joglekar, Moren Levesque We examine if survival and revenue growth are separately or jointly determined at different stages of a business startup’s life. By examining a startup from the Lean Startup perspective and incorporating dynamic capabilities, and in particular a micro strategy framework, we hypothesize that startups act differently in the early stages versus late stages of development. We test our hypothesis by applying a Hausman simultaneity test to data from the Kauffman Foundation Survey. We find that survival and revenue growth are separately determined in the early stages of a startup’s life, but become more jointly determined in later stages. 2 - Evaluating Telemedicine Adoption in Clinics: Accounting for Socioeconomic, Geographical, Organizational and Technological Antecedents Xiaojin (Jim) Liu, University of Virginia, Darden, 100 Darden Boulevard, Charlottesville, VA, 22903, United States, Susan Goldstein, Kingshuk K. Sinha This study involves a theoretically grounded empirical analysis on how socioeconomic, geographical, organizational and technological antecedents impact the adoption and use of telemedicine in health care delivery. 3 - Shortages of Resources, Routines, Reputation or Regulations: Can Data and Analytics-driven Capabilities Assist Technology Entrepreneurs’ Decisions? Nitin Joglekar, Boston University Questrom School of Business, 595 Commonwealth Ave, Boston, MA, 02215, United States, Moren Levesque We describe emerging studies that contribute to theory and create practical insights for managers in the technology arena, noting the dominance of either time or timing as crucial concepts in technology entrepreneurship. We also highlight the importance of information availability, which directs our attention to the timely availability and usage of data, along with computational technologies. We then argue that data-driven and analytics-driven capabilities can shape every aspect of the tradeoffs associated with the shortages of resources, routines, reputation or regulations and call for novel modes of decision-making for technology entrepreneurs. 4 - Managing the Sources of Uncertainty in Entrepreneurial Decision Making: A Data Analysis Approach to Entrepreneurship and Innovation Sinan Erzurumlu, Babson College, 231 Forest St, Tomasso 123, Babson Park, MA, 02457, United States Abstract not available Theory of Integer and Nonconvex Optimization Sponsored: Optimization/Integer and Discrete Optimization Sponsored Session Chair: Eli Towle, University of Wisconsin-Madison, Madison, WI, 53705, United States 1 - Mining Expression Trees to Improve Factorable Relaxations Taotao He, Purdue University, 403 W. State St, West Lafayette, IN, 47906, United States, Mohit Tawarmalani We introduce new relaxations for composite functions by convexifying the outer- function over a polytope, which models an ordering structure of outer-approximators of inner functions. We devise a fast combinatorial algorithm to separate the hypograph of concave-extendable supermodular outer-functions over the polytope. As a consequence, we obtain large classes of inequalities that tighten prevalent factorable programming. Moreover, these inequalities can be seamlessly embedded into a discretization scheme to approximate nonlinear programs with MIP. Combined with linearization, our techniques provide a framework for deriving cutting planes used in relaxation hierarchies and more. n SB04 North Bldg 122A

31