Informs Annual Meeting 2017

MA78B

INFORMS Houston – 2017

MA78B

3 - Stronger Relaxations for Bilinear Terms using Linear Constraints Benjamin Mueller, Zuse Institute Berlin, Takustr 7, Berlin, 14195, Germany, benjamin.mueller@zib.de, Ambros Gleixner, Felipe Serrano One of the most important techniques for relaxing nonconvex quadratic constraints are the well-known McCormick inequalities, which describe the convex envelope for a single bilinear term over a rectangular domain. However, it might not be the best choice when taking additional (e.g., linear) constraints into account. Following this line of thought, we use linear programming to obtain inequalities from the problem formulation that can be used to improve the McCormick relaxation of a bilinear term. We use the constraint integer programming framework SCIP to show the impact of the tighter relaxations on instances of the MINLPLib2. 4 - Latest Benchmarks of Optimization Software Hans Mittelmann, mittelmann@asu.edu We will report on selected of our benchmarks including on discrete problems and for both open source and commercial software. 381A Decomposition Approaches for Power Systems Planning under Uncertainty Sponsored: Energy, Natural Res & the Environment Electricity Sponsored Session Chair: Lim Gino, ginolim@central.uh.edu Co-Chair: Aida Khayatian, University of Houston, Houston, TX, 77057- 3197, United States, aida.khayatian@gmail.com 1 - Policymaking for Microgrid Expansion Planning in Electric Power Systems under Uncertainty Aida Khayatian, University of Houston, 1801 Bering Dr, Apt 736, Houston, TX, 77057-3197, United States, aida.khayatian@gmail.com, Masoud Barati, Gino J. Lim This paper integrates microgrid expansion planning with generation and transmission planning to study the potential advantages of the grid-connected microgrid. This planning problem faces several challenges such as the competitiveness of power investor companies, environmental restrictions, rural electrification, and uncertainty. To address these challenges via integrating microgrid, this study proposes long-term planning policies. These policies are established based on evaluating the performance of the microgrid in terms of the planning criterions. Computational results from two IEEE 6 and 118-bus test systems are presented to analyze the effectiveness of the proposed policies. 2 - Managing Stored Energy in Microgrids via Multistage Stochastic Programming Arnab Bhattacharya, University of Pittsburgh, 6236 5th Avenue, MA78 Energy storage systems can mitigate adverse effects of renewable sources in a microgrid where procurement and storage decisions are made under uncertain demand, renewable supply and prices. A multistage stochastic linear programming model is formulated to minimize the expected total costs in a microgrid, subject to power flow, storage capacity and supply-demand balance constraints. To improve computational tractability, a new regularized stochastic dual-dynamic programming (SDDP) algorithm is employed to obtain high-quality solutions within a reasonable time. A computational study highlights significant cost reductions and computational benefits. 3 - Power Management in Micro Grids with Renewable Source and Storage Natarajan Gautam, Texas A&M.University, Dept of Industrial and Systems Engineering, Room 4012 Emerging Technologies Building, College Station, TX, 77843-3131, United States, gautam@tamu.edu, Soongeol Kwon We consider a micro-grid scenario where a significant amount of energy is generated in-house from renewable sources. To combat uncertainty in demand, supply and smart pricing, we consider managing power consumption by the usage of storage devices, energy market and demand response. We formulate and solve a stochastic optimization problem. Apt 102A, Pittsburgh, PA, 15232, United States, cfcarnabiitkgp@gmail.com, Jeffrey P. Kharoufeh

380B Dynamic Programming/Control Contributed Session Chair: Xin Wang, University of Wisconsin-Madison, Middleton, WI, United States, xin.wang@wisc.edu 1 - Dynamic Expediting of an Urgent Order with Uncertain Progress Luca Bertazzi, University of Brescia, Contrada Santa Chiara, 50, Brescia, 25122, Italy, luca.bertazzi@unibs.it, Riccardo Mogre A supplier manages an urgent order with uncertain progress for which her client has set a deadline. The supplier observes in real time the order progress and chooses dynamically the effort level to expedite the order. The problem is to identify expediting policies to minimize the expected cost. We formulate a discrete stochastic dynamic programming problem. By conducting a worst-case analysis, we show that decreasing the level of flexibility may lead to a large increase in the cost. Then, we model the case in which the supplier takes into account the negative effects of late order completion. Finally, we compare the two policies using a computational study based on a car seat assembly case. 2 - Accurately Approximating the Value Function in Stochastic Dynamic Programming with Neural Networks Kyle Perline, Cornell University, Ithaca, NY, 14850, United States, krp73@cornell.edu, Christine A.Shoemaker In Approximate Dynamic Programming the value function can be approximated with neural networks. We develop a Unimodal Approximation Optimization (UAO) algorithm to tune the neural network architecture more efficiently than existing methods to create a more accurate value function approximation. In general, UAO can be applied to computationally expensive discrete domain optimization problems with noisy objective functions that have a unimodal, but not necessarily convex, mean. We demonstrate on hydropower control problems that this approach yields more accurate control solutions. 3 - How a Cities Open Data Portal Helps Drive the Local Economy This study explores how the flow of information in smart cities improves the local economy. An open data portal allows municipalities to publish data for public and private analysis that drives decision making. The results of this study also provide insights into why providing data to citizens creates economic opportunity. 4 - Dynamic Pricing with Demand Learning and Reference Price Effect Nenggui Zhao, University of Science and Technology of China, Jinzhai, He Fei, China, zhaoneng@mail.ustc.edu.cn, Cao Ping, Xiaoming Yan In this paper, we consider a retailer selling a durable product over a finite horizon, with the objective of maximizing the expected total discounted revenue. In this problem, the customers’ arrival rate is unknown and will be learned in a Bayesian way. Moreover, arriving customers’ purchase behavior is affected by reference price. First, we analyze the structural properties of the optimal revenue function and the optimal pricing policy. Second, we find that the problem can be simplified in the case of sufficient inventory. Third, we investigate the value of market size and the effect of reference price. Finally, we conduct several numerical examples to justify our theoretical results. 5 - Dynamic Pricing in a Bike Sharing System Mikhail Gordon, University of Pittsburgh, 241 Mervis Hall, Pittsburgh, PA, 15260, United States, m.gordon@pitt.edu A bike sharing system is subject to rebalancing daily to keep each station operating at its potential. The use of dynamic pricing allows for incentives to be given to consumers for their starting and ending locations to be altered to lessen the amount of rebalancing necessary for a system. 6 - Optimal Pricing Strategy for Electric Vehicle Sharing Fleet Xin Wang, University of Wisconsin-Madison, 8916 Red Beryl Drive, Middleton, WI, 53562-4278, United States, xin.wang@wisc.edu, Yuguang Wu We investigate a dynamic pricing decision problem for an electric vehicle (EV) sharing system who faces time varying stochastic customer demand and electricity price. Demand in each time period reacts to the announced price (like Uber Surge Pricing), while the EV fleet operations and charging decisions are optimized to gain revenue. An approximate dynamic programming based algorithm is proposed to derive the optimal decision policy. Abdulrahman Habib, Student, University of North Texas, 1404 Sombre Vista Dr, Denton, TX, 76205, United States, abdulrahmanhabib@my.unt.edu, Victor Prybutok

161