Informs Annual Meeting 2017

MB80

INFORMS Houston – 2017

MB78B

the ultimate task at hand. For instance, instead of merely predicting electricity demand in a standalone setting, one may want to use these predictions within planning and control tasks such as power generator allocation in the presence of energy storage. We propose an end-to-end approach for learning probabilistic machine learning models in stochastic programming that directly capture the objective of the task for which they will be used. 3 - Collaborative Demand-response Planner for Smart Buildings with Storage Juan A. Gomez, Polytechnique Montréal, 3190 Boul. Edouard Montpetit 104, Montreal, QC, QC H3T.1K2, Canada, juan.gomez@polymtl.ca, Miguel F.Anjos We present a collaborative scheme for the end-users in a smart building with multiple housing units. This approach determines a day-ahead operational plan that provides demand-response services by taking into account the amount of energy consumed per household, the use of storage and solar panels, and the amount of shifted load. We use a biobjective optimization model to trade off total user satisfaction versus total cost of energy consumption. Experimental results and a sensitivity analysis validate the performance of the proposed approach and help to clarify its strengths, its limits, and the requirements for ensuring the desired outcome. 381C Interdependent System Modeling – The Nexus of Power, Natural Gas, and Water Systems Sponsored: Energy, Natural Res & the Environment, Energy Sponsored Session Chair: Andrew Lu Liu, Purdue University, West Lafayette, IN, 47907, United States, andrewliu@purdue.edu 1 - Quantifying the Effect of Natural Gas Price Uncertainty on Economic Dispatch Cost Uncertainty Dan Hu, Iowa State University, 422 Stonehaven Dr. Unit 17, Ames, IA, 50010, United States, danhu@iastate.edu, Sarah M. Ryan In competitive electricity markets, vulnerability in gas supply to electricity generators creates a risk of high electricity prices. We propose an hourly economic dispatch model that accounts for natural gas availability and cost from both contracts and the spot market. With probabilistic inputs estimated from historical data, we use Monte Carlo simulation to generate the resulting distributions of electricity dispatch cost both with and without gas price uncertainty. The effect of gas price uncertainty is assessed in terms of distance between the distributions. 2 - An Economic Equilibrium Model for the Integrated Electricity Market Lihui Bai, University of Louisville, Dept. of Industrial Engineering, J.B. Speed School Of Engineering, Louisville, KY, 40292, United States, lihui.bai@louisville.edu, Andrew Lu Liu, Qipeng Zheng We consider an equilibrium model for an integrated electricity market system consisting of electricity consumers, power generators, grid owners, coal producers, natural gas producers, marketers and pipeline owners. In the equilibrium model, each individual player optimizes its own subsystem while market-clearance conditions are satisfied wherever players interact with each other. Numerical results will be reported including validation with the EIA’s published data. 3 - Joint Optimization of Power and Water Systems through Distributed Algorithms Andrew Lu Liu, Purdue University, 315 North Grant Street, School of Industrial Engineering, West Lafayette, IN, 47907, United States, andrewliu@purdue.edu There has been an increasing interdependence between the power and water sector. Meeting growing electricity demand would require more fresh water for power plants’ cooling, causing tension on water systems. While co-optimization of all the inter-linked systems may reach the most efficient solutions, it is impossible to do so with regulatory and computational constraints. We propose a provably-convergent distributed algorithm so that through each sector’s individual optimization and information sharing, the joint-systems-wide optimal solution can be reached. MB80

380B Humanitarian Logistics Contributed Session Chair: Yi Chu, Industrial Engineering, Beijing, China, chuy14@mails.tsinghua.edu.cn 1 - Research on Diabetes Screening Strategies in China using Agent Based Simulation Model Bowen Pang, Tsinghua University, Tsinghua Univ. Zijing 14#, Beijing, 100084, China, pzkaixin@foxmail.com, Xiaolei Xie, Yan Li, José A. Pagán Diabetes is one of the major chronic diseases in China with a prevalence rate of more than 10%. Screening is an effective way to recognize diabetes at early stage in order to control and reduce relative costs. However, considering the large population with high variety, screening without sound strategy often incurs considerable amount of costs. Therefore, it is significant and challenging to make policies on diabetes screening. An agent based model (ABM) is developed and validated to simulate the status quo in order to provide policy implications for nation-wide screening strategy, which is currently lacking. 2 - Determinants of Older Patient Choice of Medical Service Provider in China Yi Chu, PhD Student, Industrial Engineering, Beijing, China, chuy14@mails.tsinghua.edu.cn, Xie Xiaolei, Zuojun Shen It is crucial to understand factors influencing health service choice to manage patient flow on a national level. We aim at finding the determinations of medical service provider choice between older patients in China. We use conditional logit model to analyze nationwide data from China Health and Retirement Longitudinal Study and local yearbooks in 2013. Outpatient and inpatient choice patterns in different regions are also compared. Our research shows that policy makers should design a flexible policy and pay attention to important factors, such as regional effect. 3 - Impact of Affordable Care Act (ACA) on Patient Choices and Care Sarah Berkin, California State University, Hayward, CA, United States, sberkin@horizon.csueastbay.edu, Philip Cole-Regis, Kimberly Reyes, Balaraman Rajan, Surendra Sarnikar We analyze the impact of the Affordable Care Act (ACA) on healthcare coverage in the United States. Rollout of the plans, protections, and mandates began in 2010, progressing until full implementation in 2015. There has since been a notable decrease in uninsured citizens. To assess the quality of healthcare, we focus on: number of, type, and cost of emergency room visits including wait times and appointment availability. We also look at the impact on preventive care through different indicators. 381B Operational Modeling for Energy Storage Sponsored: Energy, Natural Res & the Environment Electricity Sponsored Session Chair: Ramteen Sioshansi, The Ohio State University, 240 Baker Systems Engineering Building, 1971 Neil Avenue, Columbus, OH, 43210-1271, United States, sioshansi.1@osu.edu Co-Chair: Antonio J. Conejo, The Ohio State University, 286 Baker Systems Engineering Building, 1971 Neil Avenue, Columbus, OH, 43210-1271, United States, conejonavarro.1@osu.edu 1 - Reducing Degradation in Batteries used for Frequency Regulation via Nonlinear Control Johanna Mathieu, University of Michigan, 1301 Beal Avenue, EECS.Building, Ann Arbor, MI, 48109, United States, jlmath@umich.edu, Joshua Adam Taylor Batteries can be used to provide frequency regulation to electric power systems; however, a battery’s energy storage capacity degrades as it consumes and produces power. We extend an existing nonlinear battery model, pose the battery control/management problem as a tracking control problem, and develop several nonlinear controllers. In simulation, we show how the controllers enable us to easily tradeoff tracking accuracy and battery life, and how explicitly considering nonlinear battery dynamics in the controllers improves performance. 2 - Task-based End-to-end Model Learning in Stochastic Optimization Priya L. Donti, Carnegie Mellon University, Pittsburgh, PA, 15213, United States, pdonti@cs.cmu.edu, Brandon Amos, J. Zico Kolter While prediction algorithms commonly operate within some larger process, the criteria by which we train these algorithms often differ from the criteria on which we actually evaluate them: the performance of the full “closed-loop” system on MB79

193