Informs Annual Meeting 2017

MB26

INFORMS Houston – 2017

MB24

adaptive bounding of nonlinearities and certification of the self-mapping of the convex set as required by the Brouwer fixed point theorem. The proposed framework is scalable to large cases and covers most of the feasible region in practice. 2 - Value of Flexible Wind Dispatch in Stochastic Unit Commitment We explore the expected benefit on social welfare from integrating in power systems the flexibility to dispatch the wind lower than its available output. Initially, we provide motivation for instances when wind spilling could prove beneficial. We then utilize a stochastic unit commitment approach for a reduced California power system. Two decomposition techniques are used to solve the problem. The first one is based on recent literature and utilizes global cuts and Lagrangian penalties to reach an optimal solution, for a small number of scenarios. The second one is a heuristic combination of progressive hedging with scenario reduction. 3 - Pricing in Multi-interval Real-time Markets Bowen Hua, University of Texas at Austin, 1650 West Sixth Street, Apt D, Austin, TX, 78703, United States, bowenhua90@gmail.com, Ross Baldick A multi-interval real-time market bases itself on a look-ahead economic dispatch problem implemented in a model predictive control fashion. Under the current pricing structure in the multi-interval real-time market of CAISO and NYISO, generators may have incentive to deviate from the ISO’s dispatch. We propose a pricing method that mitigates this incentive problem by explicitly considering historical loss. Numerical tests show that our method improves price signal and reduces out-of-market payments. 4 - Spurious Critical Points in Power Systems State Estimation Richard Y Zhang, UC Berkeley, Berkeley, CA, 94720, United States, ryz@berkeley.edu, Javad Lavaei, Ross Baldick The power systems state estimation problem computes the set of complex voltage phasors given quadratic measurements using nonlinear least squares. Even in the absence of measurement errors, local search algorithms can become “stuck” at local minima, which correspond to nonsensical estimations. In this talk, we observe that local minima cease to be an issue as redundant measurements are added. Posing state estimation as an instance of the quadratic recovery problem, we derive a bound for the distance between the true solution and the nearest spurious local minimum, in order to show that they become increasing rare and far-away from the true solution with the addition of redundant information. 350B Efficiency and Productivity – YR Invited: Data Envelopment Analysis Invited Session Chair: Ole Bent Olesen, University of Southern Denmark, University of Southern Denmark, Odense, 5230, Denmark, ole@sam.sdu.dk 1 - Shape Constrained Kernel-weighted Least Squares Daisuke Yagi, Texas A&M.University, College Station, TX, United States, d.yagi@tamu.edu, Yining Chen, Andrew L. Johnson, Timo Kuosmanen Nonparametric estimation methods avoid functional form misspecification which is caused by parametric assumptions. However, the flexibility of nonparametric methods often cause difficulties in interpreting results in many applications of productivity analysis. However, microeconomic theory provides additional structure for modeling a production or cost function which can be interpreted as shape constraints. We propose nonparametric shape constrained estimators that combine the advantage of avoiding functional misspecification with improving the interpretability of estimation results relative to unconstrained nonparametric methods. 2 - Direction Selection in Stochastic Directional Distance Functions A stochastic directional distance function (SDDF) allows for noise in potentially all input and output variables; however when estimated, the direction selected will affect the functional estimates. This paper addresses the question, how should the direction be selected to improve the estimates of the underlying production behavior. From Monte Carlo simulations we find two main guidelines: picking a direction orthogonal to the region where most of the data lies and a direction that is consistent with the noise in the DGP. Finally we confirm the analysis with an application on a cost function for hospital production in the U.S. Georgios Patsakis, UC Berkeley, Berkeley, CA, 94720, United States, gpatsakis@berkeley.edu, Shmuel S.Oren MB26 Kevin Layer, Texas A&M.University, College Station, TX, United States, kevin.layer@tamu.edu, Andrew L. Johnson, Robin C. Sickles

342F Customer Choices and Pricing Sponsored: Revenue Management & Pricing Sponsored Session Chair: Hongmin Li, Arizona State University, Tempe, AZ, 85287, United States, hongmin.li@asu.edu 1 - Pricing under the Generalized Extreme Value Models with Homogeneous Price Sensitivities Heng Zhang, University of Southern California, Los Angeles, CA, 90007, United States, hengz@usc.edu, Paat Rusmevichientong, Huseyin Topaloglu We study multi-product pricing problems under generalized extreme value models and homogeneous price sensitivity parameter for all products. First, we consider the unconstrained problem, and show that the optimal prices of different products have a constant markup over costs, and give an explicit formula for optimal markups. We next study the pricing problem subject to the convex inventory constraints, the formulation of which includes the deterministic approximation for the dynamic pricing problems over a network of resources. We show the problem can be transformed into a convex problem with purchase probabilities as the decision variables, and solved efficiently. 2 - Consumer Choice and Market Expansion: Modeling, Optimization and Implementation Ruxian Wang, The Johns Hopkins Carey Business School, 100 International Dr, Baltimore, MD, 21202, United States, ruxian.wang@jhu.edu We incorporate market expansion effects into consumer choice models and investigate the revenue management and pricing problems. 3 - Analysis of Discrete Choice Models: A Welfare-based Framework Xiaobo Li, University of Minnesota, 1006 27th Avenue SE,, based model as an intermediate, we show that the RAM and the SCM are equivalent. Furthermore, we show that both models as well as the welfare based model strictly subsume therandom utility model when there are three or more alternatives, while the four are equivalent when there are only two alternatives. Thus, this paper presents a complete picture on the relation between these choice models. We also study substitutability and complementarity properties in choice models. 4 - Product Design under Multinomial Logit Choices: Optimization of Costly Attributes and Prices in an Evolving Product Line Gwangjae Yu, Arizona State University, 300 E.Lemon St, BA 451G, Tempe, AZ, 85281, United States, gjyu@asu.edu We study a product-line design problem in which customers’ choice among multiple products is given by a multinomial logit (MNL) model. A firm determines product attributes and prices in an evolving product line to maximize profit. In particular, given the prices and attributes of products that already exist in a product line, the firm optimizes prices and/or attributes of the new products to be added to the same product line. 350A Control and Optimization Techniques for Power Systems I Invited: Energy Systems Management Invited Session Chair: Richard Y Zhang, UC Berkeley, Berkeley, CA, 94720, United States, ryz@berkeley.edu 1 - Inner Approximations of Power Flow Feasibility Sets Konstantin Turitsyn, Massachusetts Institute of Technology, Cambridge, MA, 02139, United States, turitsyn@mit.edu, Krishnamurthy Dvijotham, Hung Nguyen Inner approximations of power flow feasibility sets have a variety of applications for example operational security. While the outer approximations can be naturally constructed via convex relaxations, the problem of constructing inner approximations is generally hard. This talk will introduce a new strategy of constructing inner approximations for general power flow equations not restricted to any topologies and bus types. The framework is based on the APT.‘E’, Minneapolis, MN, 55414, United States, lixx3195@umn.edu, Guiyun Feng, Zizhuo Wang In this paper, we study the relation between several well known classes of discrete choice models, namely the random utility model (RUM), the representative agent model (RAM) and the semi-parametric choice model (SCM). Using a welfare MB25

173