Informs Annual Meeting 2017

MB77

INFORMS Houston – 2017

2 - Preference Elicitation for Participatory Budgeting Gerdus Benade, Carnegie Mellon University, 6235 Fifth Avenue, Apt B17, Pittsburgh, PA, 15232, United States, gerdusbenade@gmail.com, Swaprava Nath, Ariel Procacci a, Nisarg Shah Participatory budgeting is a practical variant of the knapsack problem which enables the allocation of public funds by collecting and aggregating individual preferences; it has already had a sizable real-world impact. Making the most of this new paradigm requires a rethinking of some of the basics of computational social choice, including the very way in which individuals express their preferences. We analytically compare four preference elicitation methods through the lens of implicit utilitarian voting, and find that threshold approval votes are qualitatively superior. This conclusion is supported by experiments using data from real participatory budgeting elections. 3 - The Replenishment Schedule to Minimize Peak Storage Problem Dorit Simona Hochbaum, University of California-Berkeley, Dept of IEOR, 4135 Etcheverry Hall MC 177, Berkeley, CA, 94720- 1777, United States, hochbaum@ieor.berkeley.edu, Xu Rao The replenishment storage problem is to determine the integer timing of orders of items, each of given size, and each order is depleted within a given number of days (cycle time), so that the peak storage capacity is minimized. The problem is known to be NP-hard. We demonstrate here that RSP is weakly NP-hard for constant joint cycle length (the least common multiple of all individual cycles) and strongly NP-hard for non-constant joint cycle length. For constant joint cycle discrete-RSP, we present a pseudo-polynomial optimization algorithm and the first FPTAS for identical cycles. 4 - Ranking Students using Noisy Self-generated Questions Nam Ho-Nguyen, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, hnh@andrew.cmu.edu, Gerdus Benade, Wolfgang Gatterbauer, R. Ravi We consider the question of evaluating students based on student-generated questions. Unlike teacher-generated questions, student-generated questions are more unreliable and can misleading, i.e., unable to distinguish between strong and weak students. We explore several novel methods for this problem which are scalable, robust to misleading questions and missing responses, and capable of identifying ill-posed questions. We perform a comprehensive computational study to compare our methods against traditional parametric frameworks. 372F Joint session OPT/Practice: Recent Progress in Optimization Software II Sponsored: Optimization, Computational Optimization and Software Sponsored Session Chair: Hans Mittelmann, Arizona State University, Tempe, AZ, 85287-1804, United States, mittelmann@asu.edu 1 - The SAS MILP Solver: Current Status and Future Developments Philipp M. Christophel, SAS.Institute Inc., Maria-Sibylla-Merian-Str. 36, Mainz, 55122, Germany, Philipp.Christophel@sas.com, Imre Polik, Menal Guzelsoy, Amar Kumar Narisetty We give an overview of the current status of the SAS mixed integer linear programming (MILP) solver. The focus will be on describing recent implementation efforts for the MILP solver as well as future development directions. 2 - Recent Progress in the Xpress Solvers Michael Perregaard, Xpress Developer, FICO, Birmingham, United Kingdom, MichaelPerregaard@fico.com We will present some of the recent advances in the FICO Xpress solvers. The emphasis will be on how we can bring in different technologies, such as interior point solvers, in order to improve MILP performance. 3 - Matlab Optimization Update Mary Fenelon, MathWorks, 3 Apple Hill Drive, Natick, MA, 01760, United States, Mary.Fenelon@mathworks.com, Steve Grikschat The MATLAB Optimization Toolbox has solvers for linear, quadratic, mixed- integer linear and nonlinear optimization problems. Recent improvements to these solvers will be presented along with their application to building prescriptive analytics applications with MATLAB. 4 - Recent Improvements in SCIP 4 Robert Lion Gottwald, Zuse Institute Berlin, Rubensstr. 102, Berlin, 12157, Germany, robert.gottwald@zib.de SCIP is a branch-and-cut based solver for mixed-integer-nonlinear programming (MINLP) problems. Its extensible design makes it particularly well suited for research and it is currently one of the fastest solvers available in source code. We MB77

will present improvements in the latest release of SCIP and give a prospect of new features such as presolving methods for the sparsification of the constraint matrix and enhancements in the cutting plane separation.

MB78

381A Deterministic and Stochastic Optimization for Electric Power Systems Sponsored: Energy, Natural Res & the Environment Electricity Sponsored Session Chair: Andy Sun, Georgia Institute of Technology, Atlanta, GA, 30312, United States, andy.sun@isye.gatech.edu 1 - Market-based Transmission Expansion Planning using an Area and Unit Decomposition Approach Philipp Hartel, Fraunhofer IWES, Konigstor 59, Kassel, 34119, Germany, philipp.haertel@iwes.fraunhofer.de Motivated by future offshore grid connections in the North Sea region, a market- based TEP model with a particular focus on long-term operational flexibility in multiple market areas of the energy system in Europe is presented. While accounting for interactions of power, heat, and transport sectors in the future, aggregated formulations of hydro-thermal units are used to capture short- to long-term power market dynamics. To solve the problem with a planning horizon of a full consecutive year, a solution strategy involving an area and unit decomposition for the resulting large-scale convex linearly-constrained non- smooth problem is pursued using a doubly stabilized proximal bundle algorithm. 2 - Valid Inequalities for Hydro Genco Self-scheduling Optimization Minseok Ryu, University of Michigan, msryu@umich.edu, Antonio J. Conejo, Ruiwei Jiang We focus on a self-scheduling optimization problem for a hydro generation company that manages a set of interconnected hydro reservoirs. We consider a class of key physical and operating characteristics of hydro reservoirs and generators in practical settings, including the prohibited operating zones and the nonlinear and nonconvex performance curves. We employ a mixed-integer linear programming (MILP) approach to approximate these characteristics. The MILP approach facilitates the use of many efficient computational tools, e.g., optimization solvers and valid inequalities. Finally, numerical experiments are conducted based on a real-world case study. 3 - Leveraging Predictive Analytics to Control and Coordinate Operations Scheduling and Maintenance Murat Yildirim, Georgia Institute of Technology, Atlanta, GA, United States, murat@gatech.edu, Andy Sun, Nagi Gebraeel Penetration of the renewable sources, distributed generation, and market dynamics increase the congestion and volatility in modern power grids, leading to frequent on-off cycles and stringent production requirements on power plants that significantly affect their lifetime. This presentation proposes a unified analytics framework that integrates i) a stochastic degradation model to provide accurate sensor-driven predictions of the remaining life distribution for power plants operating under varying loading conditions, with ii) an adaptive optimization model that finetunes the loading on the power plants by determining the optimal maintenance and operational decisions. 4 - Multistage Stochastic Unit Commitment with SDDIP Andy Sun, Georgia Institute of Technology, 755 Ferst Drive, Atlanta, GA, 30312, United States, andy.sun@isye.gatech.edu Unit commitment is one of the key operational problems in power systems. We propose a new type of decomposition algorithm based on Stochastic Dual Dynamic Integer Programming (SDDiP) to solve a dynamic programming formulation of a multistage stochastic unit commitment (MSUC) problem. We propose a variety of computational enhancements to adapt SDDiP to MSUC, and conduct extensive computational experiments to demonstrate that the proposed method is able to handle elaborate stochastic processes and can solve MSUCs with a huge number of scenarios that are impossible to handle byexisting methods. 5 - Techniques to Assess Performance of Electricity System Operation in the Presence of Ex Post Pricing Tuong Nguyen, The New Zealand Electricity Authority, Level 7 - 2 Hunter Street, Wellington, 5024, New Zealand, manhtuong@yahoo.com This paper examines an approach for the analysis of the system operator’s real- time dispatch performance. In the New Zealand electricity system, generators are economically dispatched every five minutes in real time based on a real-time dispatch schedule. Computation of this schedule takes account of offers to supply, network topology, and demand forecasts. Market settlement prices are determined for each 30-minute trading period the following day using the final pricing schedule. The paper discusses the major causes of the differences between the real-time dispatch schedule and the final pricing schedule, and describes various options to measure system operation performance.

192