Informs Annual Meeting 2017

TE70

INFORMS Houston – 2017

4 - A Competitive Multiperiod Supply Chain Network Model with Freight Carriers and Green Technology Investment Option Sara Saberi, Worcester Polytechnic Institute (WPI), Washburn Rm 217, Foisie School of Business, Worcester, MA, 01609, United States, ssaberi@wpi.edu, Jose Cruz, Anna B. Nagurney This paper presents a multiperiod supply chain with freight carriers model. Manufacturers, retailers, and carriers maximize the net present value of their investments in technology. Future production, inventory, transaction, and transportation costs savings are used to help fund investments. The environmental impact of the costs are all integrated. The tradeoff between the initial technology investment and its ecological footprint effect are considered. We provide variational inequality formulations of the equilibrium conditions and propose the modified projection method. Numerical examples are examined with an analysis of the effects of technology investments on supply chain network. 5 - Polling, Psychology and Propaganda in Forecasting Electoral Outcomes Dimitris Vayenas, Metapoll, London, United Kingdom, dimitris@metapoll.co.uk Having predicted the outcomes of recent elections the session addresses the underlying theories that are calling for a re-evaluation of political campaigning. By utilising formal models developed to quantify transparency in many-to-many communications, accurate forecasting is attainable based on the motives of the population as addressed by the campaigns; in the case of Brexit with a 0.1% deviation one month before the elections. The shortcomings of polling methodologies and how analytics can help mitigate the impact of unreliable responses are also addressed. Models of human socialty and motivators are employed in the mapping of quantifiable sentiments and reactions to propaganda. 6 - Optimization of Feedstock Blending for Biochemical Conversion Mohammad Sadekuzzaman Roni, Idaho National Laboratory, 21 T. Wallace Circle, Starkville, MS, 39759, United States, msr131@gmail.com Identifying a least cost blend is challenging, especially when decisions about feedstock blending must be based on a number of criteria. This study developed an optimization framework to identify optimal blending of feedstock to meet biochemical conversion specifications. Results reveal the tradeoff among blend components in the least cost blend based on biomass resource availability, quality requirement and logistics cost. 371E Data Mining Contributed Session Chair: Jimi Oke, Massachusetts Institute of Technology, Cambridge, MA, United States, oke@mit.edu 1 - Mathematical Programming Approaches and Data Representations for Multiple Instance Learning Emel Seyma Kucukasci, Istanbul Commerce University, Kucukyali E5 Kavsagi Inonu Cad No 4, Istanbul, 34840, Turkey, emelseyma@gmail.com Emel Seyma Kucukasci, Boaziçi University, Istanbul, Turkey, emelseyma@gmail.com, Mustafa Gokce Baydogan, Z. Caner Ta kın We propose a multiple instance learning framework including new mathematical models of multiple instance classification and enhanced data representations. We efficiently solve MIL problem without imposing strict assumptions on object descriptions. Our approach embeds instance relationships via inputting various data representations and determines class memberships of the objects. We compare our learning procedure with state-of-the-art MIL methods on a wide range of machine learning datasets to highlight the classification success on different application domains. 2 - Forecasting Gathering Events through Continuous Destination Prediction on Big Trajectory Data Amin Vahedian Khezerlou, University of Iowa, S283 Pappajohn Business Building, Iowa City, IA, 52242, United States, amin-vahediankhezerlou@uiowa.edu, Xun Zhou Gathering Event Forecasting (GEF) in urban areas is the problem of forecasting unexpectedly high number of arrivals in an urban region. In this work, we propose a trajectory-based approach to forecast the gathering events by predicting the destinations of incomplete trip trajectories at a given time. To the best of our knowledge, this method is the first to adopt this approach to GEF. This method includes a novel destination prediction model, which is superior to the models in the literature in terms of accuracy and computational cost (both running time and memory). We show through experiments that our method successfully forecasts real events. TE70

3 - Discovering Urban Typologies for Future Mobility Scenarios in Prototype Cities Jimi Oke, Postdoctoral Associate, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Building 1-178, Cambridge, MA, 02139, United States, oke@mit.edu, Yafei Han, Sean Hua, Carlos L. Azevedo, P. Christopher Zegras, Moshe E. Ben-Akiva Ongoing trends in the development of vehicle fuels and technologies along with potential environmental and policy regulations will produce a complex space of futures of mobility outcomes in urban areas. We gather the most recent land-use, economic, environmental and mobility data from over 331 cities. We extract the key factors and then cluster the cities to find the prevailing urban typologies. Further, we estimate a latent class choice model for mobility responses based on 43000 individual observations from a global sample, thus confirming our typology classes. In a novel effort, we will then generate new representative prototype cities in order to model various urban mobility scenarios. 371F Modeling and Analysis of Power System Expansion Planning Sponsored: Energy, Natural Res & the Environment Electricity Sponsored Session Chair: Enzo E Sauma, PhD, Pontificia Universidad Catolica de Chile, Pontificia Universidad Catolica de Chile, Santiago, Chile, esauma@ing.puc.cl 1 - Quantifying the Effect of Head-sensitive Hydropower Approximations on Investments and Operations Francisco Munoz, Universidad Adolfo Ibáñez, Santiago, Chile, fdmunoz@uai.cl, Gonzalo Ramírez Planning for new generation infrastructure in hydrothermal power systems requires consideration of a series of nonlinearities that are often ignored in capacity-planning models. In this article we use three different capacity-planning models, with different degrees of complexity and accuracy, to quantify the impact of simplifying the head dependency of hydropower generation on investments in thermal units and system operations. We find that simplified investment models can bias the optimal generation portfolios significantly and increase system costs compared to more detailed nonlinear models. 2 - Joint Expansion Planning of Distribution Networks, EV Charging Stations and Wind Power Generation under Uncertainty Javier Contreras, Univ de Castilla-La Mancha, E.T.S.de Ingenieros Industriales, Campus Universitario s/n, Ciudad Real, 13071, Spain, Javier.Contreras@uclm.es, Pilar Meneses de Quevedo, Gregorio Muñoz-Delgado This paper describes the consideration of uncertainty in distribution system expansion planning (DSEP) including electric vehicles (EV) and the possibility of joint investment in EV charging stations (EVCS) and wind power generators. The charging demand necessary for EVs transportation is performed using a vehicle decision model based on travel patterns. The problem is defined as a stochastic- programming-based model driven by the minimization of the annualized investment, maintenance, production, losses and non-supplied energy costs. A 24-bus system case study is provided illustrating the results of the proposed approach. 3 - Transmission Expansion Planning: A Distributionally Robust Optimization Approach David Pozo, Skolkovo Institute of Science and Technology, Mosco, Russian Federation, davidpozocamara@gmail.com, Alexandre Street, Alexandre Velloso Investment decisions on power systems should be robust among all (infinite) possible future scenarios. This talk addresses the transmission expansion planning (TEP) problem where decisions have to be optimized given that the future is uncertain, there are several scales of uncertain parameters, and there is no knowledge of the true probabilities. A Distributionally Robust Optimization framework is proposed. Numerical test results are reported in comparison with the conventional approaches such as two-stage stochastic optimization and adaptive robust optimization. TE71

436