INFORMS 2021 Program Book

INFORMS Anaheim 2021

SA42

outside of the platform we use historical data on thousands of fruit pickups, gathered through GPS trackers installed on middlemen trucks in our field site in Sumatra. To account for uncertainty in data collection, we rely on a distributionally robust optimization approach based on the Wasserstein metric. We show that such a platform can reduce transportation costs, while also easing the digitization of agricultural data and improving the transparency of smallholder supply chains. 3 - Internet of Things-enabled Information, Dual Sourcing and Supplier Competition Tao Lu, PhD, University of Connecticut, CT, United States, Brian Tomlin Internet of Things (IoT) technologies have been increasingly used to monitor production and transportation processes, thereby predicting a potential supply disruption. In this study, we examine the impact of IoT-enabled supply information on a buyer sourcing from two competing suppliers: an unreliable supplier subject to disruption risks and a reliable one. We show that the buyer may or may not benefit from the IoT-enabled information. While the IoT information enables the buyer to place an emergency order with the reliable supplier if needed, it may soften the competition between suppliers. Our model extensions further discuss the cases when the IoT information is not accurate and when the unreliable supplier can resolve a detected supply issue (with a certain probability). SC40 CC Room 211B In Person: Advanced Game-theoretic Models in Energy Market and Policy Design General Session Chair: Nathan Boyd, College Park, MD, United States 1 - Modeling a Co2 Tax at the Margin Combined with a Subsidy Scheme for the Dutch Industry Marit Van Hout, Netherlands Environmental Assessment Agency, Den Haag, Netherlands, Bert W. Daniels, Robert Koelemeijer During this presentation we will illustrate the use of state-of-the-art modeling at the Netherlands Environmental Assessment Agency (PBL) applied annually for the Dutch Climate and Energy Outlook. More in particular, the focus is on the modeling of the Dutch industry and the interplay between the Dutch CO2 levy and EU ETS prices on the one hand, and subsidy schemes on the other. The model applied is a hybrid simulation and optimization (LP) model where interactions with other relevant segments and markets (such as refineries and electricity) are included in an iterative fashion by exchanging modeling results. 2 - Optimal Pricing in the Long-run in Electricity Markets with Non-convex Costs Conleigh Byers, ETH Zurich, Brunnwiesenstrasse 40, Zurich, 8049, Switzerland, Gabriela Hug Non-convex costs are critical in electricity markets, as startup costs and minimum operating levels yield a non-convex optimal value function over demand levels. We evaluate the performance of different non-convex pricing frameworks by determining long-run adapted resource mixes associated with each pricing framework while fully preserving the non-convex operations. We frame optimal pricing in terms of social surplus achieved and transfer of consumer to producer surplus in adapted long-run market equilibria. We find that approximate convex hull pricing achieves the highest social and consumer surplus, while other methods tend to over-compensate inframarginal units. We find that marginal prices with fixed integer variables can achieve high consumer surplus, but near- optimal solutions significantly affect performance. 3 - Privacy Impact on Generalized Nash Equilibrium in Peer-to-peer Electricity Market Ilia Shilov, Inria, Paris, France, Hélène Le Cadre, Ana Busic We consider a peer-to-peer electricity market, where agents hold private information. The problem is modeled as a noncooperative communication game, in the form of a Generalized Nash Equilibrium Problem, where the agents determine their randomized reports, while anticipating the form of the peer-to- peer market equilibrium. We characterize the equilibrium of the game, prove the uniqueness of the Variational Equilibria and provide an expression for the privacy price. Numerical illustrations are presented on the 14-bus IEEE network. 4 - Game Theoretic Modeling for Improved Management of Water And Wastewater Resources Using Equilibrium Programming and Feedback Mechanisms Nathan Boyd, University of Maryland, College Park, MD, United States, Steven A. Gabriel, George Rest, Tom Dumm, Doug Murphy This work investigates the modeling of novel market-based management approaches for improved cooperation among independent water resources users using equilibrium programming techniques (e.g., MCPs, MPECs). Such cooperation is not naturally incentivized because the actions beneficial to upstream users can often negatively impact downstream users. These asymmetrical benefits can lead to non-cooperative behavior in three key areas: 1)

water withdrawal rights, 2) water quality responsibilities, and 3) risks associated with flooding. The approach is applied to a case study in the Duck River Watershed to promote economic development and ecological preservation.

SC41 CC Room 212A In Person: Recent Advancement of Optimization Methodology in Energy Systems General Session Chair: Guyi Chen, Evanston, IL, United States 1 - Stochastic Planning of Joint Power and Gas System Against Extreme Weather and Climate Events Wenjing Su, PhD Candidate, The Pennsylvania State University, State College, PA, United States Physical infrastructure systems including power grids and natural gas systems become increasingly tightly coupled. The energy infrastructure systems have shown great susceptibility to more frequent and severe extreme weather and climate events. The low-probability and high-impact natural disasters as well as the interdependence between power and gas system need to be taken into consideration in the energy system planning process to ensure its reliability. In this study, the vulnerability of joint electricity and natural gas system under spatially correlated failures induced by extreme weather and climate events and uncorrelated failures are compared. A two-stage stochastic optimization model is proposed to enhance the resilience of joint power and gas system against spatially correlated failures. 2 - Private Risk and Social Resilience Han Shu, Cornell University, Ithaca, NY, United States, Jaob Mays, Michael Craig, Lynne Kiesling, Joshua Macey, Blake Shaffer Energy-only electricity markets rely on the decentralized investment decisions of market participants to provide an efficient level of reliability. During an exceptionally cold winter storm in February 2021, ERCOT experienced shortfalls on an unprecedented scale, with nearly half of the generation fleet experiencing outages. The depth of the resulting blackouts invites questions as to the ability of systems relying on decentralized planning to appropriately prepare for and withstand rare events. Based on two mild assumptions, risk-aversion among investors and incomplete risk trading, we explain why decentralized markets may lead to underinvestment in resilience to rare events. We describe the nature of the incomplete risk trading that arises in the context of electricity markets and discuss potential market, market-like, and non-market remedies. 3 - Parameter Estimation in an Energy Simulation Model using Statistical Learning Guyi Chen, Northwestern University, Evanston, IL, United States, David Morton, Oscar Dowson We consider a simulation model of a highly detailed concentrating solar power system, which requires a large number of input parameters. Because of the novelty of the system, many of the parameters have a high degree of uncertainty. To be practically useful, we must learn these input parameters from limited data. Bayesian optimization is a promising tool to learn parameter values of a possibly expensive oracle. We couple Bayesian optimization method, a local search algorithm, and a scheme to search the space of parameters with a focus on exploration. We present empirical and computational results to show the effectiveness of the framework. SC42 CC Room 212B In Person: Artificial Intelligence II Contributed Session Chair: Soomin Lee, Yahoo! Research, Sunnyvale, CA, 94087, United States 1 - Image-based Characterization of Laser Scribing Quality using Transfer Learning Mohammad N. Bisheh, Kansas State University, Manhattan, KS, United States, Shing Chang, Xinya Wang, Shuting Lei, Jianfeng Ma Due to the processing speed and high-quality requirement in modern industrial applications, it is important to measure and monitor quality characteristics in real time during ultrafast laser scribing process. This research presents a study on image-based characterization of laser scribing quality using a novel transfer deep convolutional neural network (TDCNN) model for several quality characteristics such as debris, scribe width, and straightness of a scribe line using only a few images. Appropriate image processing techniques are provided to measure scribe width and line straightness as well as total scribe and debris area using classified images with 96 percent accuracy.

27