Informs Annual Meeting 2017

TDO3C

INFORMS Houston – 2017

TD03A Grand Ballroom A Energy & Uncertainty: Optimization & Control Sponsored: Applied Probability Sponsored Session Chair: Adam Wierman, California Institute of Technology, Pasadena, CA, 91125, United States, adamw@caltech.edu 1 - Distributed Optimization Decomposition for Joint Economic Dispatch and Frequency Regulation Enrique Mallada, Johns Hopkins University, Baltimore, MD, 21201, United States, mallada@jhu.edu This work studies the joint problem of co-optimizing economic dispatch and frequency regulation resources in power systems. We provide sufficient conditions under which the joint problem can be decomposed without loss of optimality into slow and fast timescale problems. These problems have appealing interpretations as the economic dispatch and frequency regulation problems respectively. Moreover, the fast timescale problem can be implemented using distributed load and/or generation control. We apply this optimal decomposition to propose an efficient market mechanism for economic dispatch that coordinates with frequency regulation. 2 - Opportunities for Price Manipulation by Aggregators in Electricity Markets Navid Azizan Ruhi, California Institute of Technology, 1200 E. California Blvd., MC 305-16, Pasadena, CA, 91125, United States, nazizanr@caltech.edu, Krishnamurthy Dvijotham, Niangjun Chen, Adam Wierman Aggregators are playing an increasingly crucial role in the integration of renewable energy in power systems. However, the intermittent nature of renewables makes market interactions difficult to monitor and regulate, raising concerns about potential market manipulation by aggregators. We study this issue by quantifying the profit an aggregator can obtain through strategic curtailment of generation. We show that, while the problem of maximizing the benefit from curtailment is hard in general, efficient algorithms exist when the topology of the network is acyclic. Further, we highlight that significant increases in profit are possible via strategic curtailment in practical settings. 3 - Cascading Failures in Power Grids: A Large Deviations Approach Alessandro Zocca, Caltech, Pasadena, CA, United States, zocca.ale@gmail.com, Tommaso Nesti, Bert Zwart Power grids are increasingly affected by uncertainty due to the intermittent nature of renewable generation. In this talk I will present a stochastic model of power grids under uncertainty, aiming to get insight in the interplay between renewable energy sources and grid reliability. Using the DC approximation and large deviation techniques, we investigate line overload probabilities in a small- noise regime and identify the most likely way for overloads to occur. We also investigate how such overloads and failures lead to cascades. In particular, such cascading failures do not propagate in a nearest neighbor fashion, unlike epidemic models, and we illustrate this analytically in our model. 4 - Interaction Between Retail and Wholesale Electricity Markets: Modeling and Stability Analysis Thiagrajan Ramachandran, Pacific Northwest National Laboratory, Richland, WA, 99354, United States, thiagarajan.ramachandran@pnnl.gov, Krishnamurthy Dvijotham, Karanjit Kalsi As power systems cope with increasing amounts of renewable generation, there is an increasing need to engage flexible demand-side resources into power systems operations. Retail markets can facilitate this by providing a platform to aggregate demand-side flexibility and bid the aggregate flexibility into wholesale markets. However, flexible loads have physical dynamics that couples their capacity over time which could have unintended and undesirable consequences. We develop a simple model to capture evolution of load flexibility over time and analyze stability of the market dynamics under this model. We prove that oscillations cannot occur and prices converge to a equilibrium levels quickly.

1 - Information Revelation and Consumer Privacy Alessandro Bonatti, MIT, Cambridge, MA, United States, bonatti@mit.edu, Rossella Argenziano, Gonzalo Cisternas We develop a dynamic model in which a forward-looking consumer interacts with heterogeneous sellers. In each period, the consumer makes both a purchase decision that is imperfectly observed by future firms. We characterize the amount of information the consumer’s equilibrium actions reveal about her true preferences. Because consumers are wary of revealing information through purchases, sellers lower their prices to stimulate demand. These results have implications for privacy regulation that restricts the flows of information. 2 - Learning from Reviews Many online platforms present summaries of reviews by previous users. Even though such reviews could be useful, previous users leaving reviews are typically a selected sample of those who have purchased the good in question, and may consequently have a biased assessment. We construct a model of dynamic Bayesian learning and profit-maximizing behavior of online platforms to investigate whether such review systems can successfully aggregate past information and the incentives of the online platform to choose the relevant features of the review system. 3 - Origins and Control of Misinformation Gad Allon, gadallon@wharton.upenn.edu, Kimon Drakopoulos, Vahideh Manshadi During the last several years we have witnessed a dramatic change in the way information is being distributed to citizens and consumers: From a market controlled by several key players we moved to a highly fragmented market with many players. The paper is motivated by the the observation that while on the one hand, we see more of these independent news organizations, we also see an increased level of polarization even when it come to agreeing about facts. In order to study this phenomenon, we aim to study the role social networks, and these independent news organization play in discovering and transmitting the real “truth” when the information is discovered sequentially. 4 - Small-loss Online Learning with Partial Information Thodoris Lykouris, Cornell University, 107 Hoy Rd, Ithaca, NY, United States, teddlyk@cs.cornell.edu, Karthik Sridharan, Eva Tardos In many applications (online advertising, internet routing, clinical decisions), decisions are based on past experience. Online learning algorithms trade off exploiting knowledge about options chosen in the past and exploring new options to gain useful information for the future. In this talk, we focus on two challenges. First, we discuss the effect of side-information: learners collecting information about options relevant to the ones selected. Second, we consider evolving settings, in which players learn in repeated games where some parameters (e.g. set of opponents) change over time. We provide a black-box framework to make full information learning algorithms robust in both settings. TD03C Grand Ballroom C Food Supply Chain Analytics Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Retsef Levi, MIT, Cambridge, MA, 02142, United States, retsef@mit.edu 1 - Network Based Risk Analytics in Global Food Supply Chains Tauhid R. Zaman, Massachusetts Institute of Technology, MA, United States, zlisto@mit.edu, Amine Anoun, Retsef Levi We develop a network-based approach to identify high-risk food importing consignees. We illustrate the power of the approach by applying it to consignees of shrimp, a product subject to high food safety and adulteration risks. Using detailed shipment data, as well as F.D.A. site inspection and shipment sampling records, we construct a set of novel network features which predict the risk of individual shrimp consignees failing site inspection. We find that increased supply chain network complexity correlates with increased risk. Our results suggest that network-based risk analytics could significantly improve the effectiveness of regulatory activities related to food supply chains. 2 - Novel Methods for Identifying Shell Companies Nicholas J. Renegar, Massachusetts Institute of Technology, Cambridge, MA, United States, renegar@mit.edu Motivated by transshipping networks in the import/export industry, we explored novel methods of identifying shell companies. Using website information and business databases we were able to create an automated method of predicting shell companies. This information will be useful in identifying illegal shipping activity, among many widespread applications. Ali Makhdoumi, Massachusetts Institute of Technology, 77 Massachusetts Ave. , 32D-640, Cambridge, MA, 02139, United States, makhdoum@mit.edu

TD03B Grand Ballroom B

Information Markets and Learning Sponsored: Revenue Management & Pricing Sponsored Session

Chair: Ali Makhdoumi, Massachusetts Institute of Technology, 77 Massachusetts Ave. , 32D-640, Cambridge, MA, 02139, United States, makhdoum@mit.edu Co-Chair: Kimon Drakopoulos, Massachusetts Institute of Technology, Los Angeles, CA, 90066, United States, kimondr@mit.edu

375