INFORMS 2021 Program Book

INFORMS Anaheim 2021

ME03

2 - Latest Benchmark Results Hans Mittelmann Selected results from our optimization software benchmarks will be reported.

convergence features and pricing under Sample Average Approximation as well as Benders decomposition and Lagrangian-based relaxations applied to an energy and reserve scheduling problem. Characterized by non-convex unit commitment decisions, this stochastic mixed-integer programming problem entails a large number of scenarios and technical parameters inducing a high computational burden. Decomposition and approximation techniques increase solution efficiency and convex relaxations enable the derivation of marginal cost of energy and reserve, with each method depicting varied solution quality and convergence. 2 - Hydropower Scheduling by Use of SDDP Treatment of State-dependent Constraints Arild Helseth, SINTEF Energy Research, Trondheim, Norway We study a specific and challenging type of state-dependency within the framework of stochastic dual dynamic programming (SDDP) applied to hydropower scheduling. For environmental purposes, the maximum allowed discharge from hydropower stations may be limited depending on the state variables (reservoir level and inflow), reducing the hydropower flexibility. From a mathematical point of view, such state-dependent limits introduce a discontinuity that is not straightforward to embed in the SDDP algorithm. We report on the results from numerically testing a set of possible solution strategies. 3 - Multi-Stage Modeling with Recourse for Solving Stochastic Complementarity Problems with an Application In Energy Pattanun Chanpiwat, Graduate Student, University of Maryland, College Park, MD, United States, Steven A. Gabriel The intermittency of the variable renewable energy in power generation could be mitigated using energy storage systems. This project concerns utilizing battery storage for the implementation of renewable energy technologies in the electricity market from multiple perspectives of uncertainty. We have developed multi-stage modeling with recourse decisions for solving stochastic complementarity problems with an application in energy. The model is based on a Nash-Cournot formulation of imperfect competition among power producers. This energy system optimization modeling, based on game theory and energy market equilibria, is expected to provide great insights for energy market planners. 4 - Contextual Merit-order Dispatch under Uncertain Supply Miguel Ángel Muñoz Diaz, PhD Student, University of Malaga, Malaga, Spain, Juan Miguel Morales, Salvador Pineda We consider a forward (e.g., day-ahead) electricity market with uncertain supply, where production quantities are dispatched following a merit order based on marginal production costs. Unlike the traditional practice of using the forecast value of the supply to clear the market, in this talk, we introduce a procedure to find the estimate of the supply (generally different from its expected value) that leads to the most cost-efficient dispatch taking into account the subsequent real- time operation of the power system. This procedure utilizes the forecast supply as the context and exploits a novel bilevel framework for decision-making under uncertainty with contextual information. ME03 CC Ballroom C / Virtual Theater 3 Hybrid Matching Markets Sponsored: Auctions and Market Design Sponsored Session Chair: Thayer Morrill, North Carolina State University, Raleigh, NC, 27695, United States, United States 1 - Matching and Money Ravi Jagadeesan, Stanford University, Stanford, CA, United States, Alexander Teytelboym We study the implications of budget constraints for matching with contracts. We assume preferences satisfy net substitutability: i.e, if a price of a good increases, then buyers (resp. sellers) who minimize the cost of obtaining a given level of utility will buy (resp. sell) more (resp. less) of other goods. Net substitutability coincides with gross substitutability for quasilinear preferences, but is strictly weaker otherwise. If agents have sufficient incomes for hard budget constraints not to bind, stable outcomes exist and coincide with competitive equilibrium outcomes. Otherwise, competitive equilibria can fail to exist, but stable outcomes exist and coincide with quasiequilibrium outcomes. Stable outcomes are weakly Pareto-efficient, but do not form a lattice or satisfy a Lone Wolf Theorem. Our results suggest a new scope for matching with budget constraints. 2 - Ranking Objects That Rank Back Thayer Morrill, North Carolina State University, Raleigh, NC, 27695, United States We consider ranking alternatives that are outcomes of a competitive process. Examples include students ranking colleges, doctors ranking residency programs, and academics ranking journals. We introduce a new approach based on desire. An object is desired if an agent prefers it to her outcome. We characterize the class of desirable rankings and argue that these rankings are superior to rankings based on revealed preference.

Monday, 4:30PM 6:00PM

ME01 CC Ballroom A / Virtual Theater 1 Hybrid QSR Best Refereed Paper Competition Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Fugee Tsung, HKUST, Kowloon, Hong Kong Chair: Kamran Paynabar, ISyE Georgia Tech, Atlanta, GA, 30332-0205, United States 1 - An Approach for Group and Individual Fairness in Federated Learning Xubo Yue, University of Michigan, Ann Arbor, Ann Arbor, MI, 48105-2179, United States, Raed Al Kontar, Maher Nouiehed 2 - Partitioned Active Learning for Heterogeneous Systems Cheolhei Lee, Virginia Tech, Blacksburg, VA, 24060, United States, Kaiwen Wang, Jianguo Wu, Wenjun Cai, Xiaowei Yue 3 - GPS: Gaussian Process Subspace Regression for Model Reduction Ruda Zhang, The Statistical and Applied Mathematical Sciences Institute, Cary, NC, 27519, United States, Simon Mak, David Dunson Subspace-valued functions arise in a wide range of problems, including parametric reduced order modeling (PROM). In PROM, each parameter point can be associated with a subspace, which is used for Petrov-Galerkin projections of large system matrices. Previous efforts to approximate such functions use interpolations on manifolds, which can be inaccurate and slow. To tackle this, we propose a novel Bayesian nonparametric model for subspace prediction: the Gaussian Process Subspace regression (GPS) model. This method is extrinsic and intrinsic at the same time: with multivariate Gaussian distributions on the Euclidean space, it induces a joint probability model on the Grassmann manifold, the set of fixed-dimensional subspaces. The GPS adopts a simple yet general correlation structure, and a principled approach for model selection. Its predictive distribution admits an analytical form, which allows for efficient subspace prediction over the parameter space. For PROM, the GPS provides a probabilistic prediction at a new parameter point that retains the accuracy of local reduced models, at a computational complexity that does not depend on system dimension, and thus is suitable for online computation. We give four numerical examples to compare our method to subspace interpolation, as well as two methods that interpolate local reduced models. Overall, GPS is the most data efficient, more computationally efficient than subspace interpolation, and gives smooth predictions with uncertainty quantification. 4 - Data-driven Pathwise Sampling Approaches for Online Anomaly Detection Dongmin Li, University of Florida, Gainesville, FL, United States, Miao Bai, Xiaochen Xian We propose a data-driven strategy for quick anomaly detection with Moving Vehicle-based sensors. We integrate statistical process control and mathematicaloptimization to monitor the system and adaptively sample from suspicious locations based on real-time data. We provide theoretical investigations and present its performance in a numerical study on wildfire detection. ME02 CC Ballroom B / Virtual Theater 2 Hybrid Decision analytics in Electricity Markets Sponsored: OPT/Optimization Under Uncertainty Sponsored Session Chair: Bruno Fanzeres, Co-Chair: Nuran Cihangir, Pontifical Catholic University of Rio de Janeiro 1 - Approximation and Decomposition Techniques for Stochastic Energy and Reserve Scheduling: Algorithmic Efficiency and Pricing Nuran Cihangir Martin, Pontifical Catholic University of Rio de Janeiro, Rio de Janeiro, Brazil, Bruno Fanzeres dos Santos This work evaluates the computational performance, solution quality,

87