INFORMS Nashville – 2016
364
4 - Representing Uncertainty With Convex Model Databases
Anushka Chandrababu, Research Scholar, International Institute
of Information Technology, Bangalore, 26/C, Electronic City,
Bangalore, India,
anushka.babu@iiitb.org, Srinivasa Prasanna
We present Convex Model Databases for storing and querying uncertain data sets.
The database stores convex models which could be a collection of polytopes,
ellipsoids etc. We will discuss methods to represent them, generate them from
uncertain or big-data applications, associated relational algebra and results from
new optimized queries based on I-structures, a generalization of database indices.
We will also show case real world applications of this database representing
uncertain or big data.
5 - A Recommender Systems Approach For Predicting Utility Of
Various Mobile Services
Abhay Kumar Bhadani, Indian Institute of Technology-Delhi,
Ground Floor office, New Delhi, 110016, India,
abhaybhadani@gmail.com, Ravi Shankar, Vijay Rao
Modeling human behavior and predicting their preferences is of interest to many
organizations. The ability of predict one’s preference in mobile ser- vices arena
pertains high business value for the service providers. This paper attempts to
model user’s preference using collaborative filtering based recommender system.
WA04
101D-MCC
Optimization in converter-based power systems
Sponsored: Energy, Natural Res & the Environment, Energy I
Electricity
Sponsored Session
Chair: Joshua Adam Taylor, University of Toronto,
10 King’s College Road, SF 1021C, Toronto, ON, M5S 3G4, Canada,
taylor.a.josh@gmail.com1 - Designing Microgrid Dynamics
Baosen Zhang, University of Washington,
zhangbao@uw.eduPower electronics interfaces allows us to design dynamical behaviors in
microgrids. As an important example, the moment of inertia and damping
coefficients can be designed to optimize system performance. In this talk, we will
consider the optimal design problem of choosing these coefficients, subject to the
physical constraints of the power electronics. We show how to cast this problem
as an eigenvalue problem, which can then be approximated via a convex
program. We also investigate the role of network topology.
2 - Frequency Control In Microgrids: Distributed Implementation And
Intrusion Detection
Lin-Yu Lu, University of Illinois at Urbana Champaign,
Urbana, IL, United States,
lyl@illinois.edu, Hao Zhu
High penetration of distributed generation in microgrids has raised the frequency
stability issue. Secondary control can resolve it by dispatching active power
resources, which may be vulnerable to malicious attacks on the communication
infrastructure. A distributed secondary control for distributed energy resources is
developed in this paper along with the cyber-security considerations. The control
scheme can achieve the goal of power sharing while restoring frequency in a
distributed fashion. Attack detection and localization strategies are developed
using local measurements and neighbor information.
3 - Optimizing The Interplay Between The Micro And Macro Grids:
From Challenges To Perspectives
Luckny Zephyr, Postdoctoral Associate, Cornell University, Ithaca,
NY, 14853, United States,
lz395@cornell.edu, C. Lindsay Anderson
Efficient management of power networks is a difficult task. This is exacerbated by
the integration of renewables. Finding good operating policies depends upon two
complementary tasks (i) finding an acceptable representation for the underlying
stochastic process, and (ii) given an approximation, finding an optimal operating
policy for the power network. The interface between these two steps is a
challenge. We assert that a significant proportion of the flexible loads are located
in the distribution system, we then want to develop a comprehensive stochastic
co-optimization scheme for the interplay between the generation and
transmission system, and the distribution-system-as-micro grid.
4 - Optimizing Power Electronic Converters Using
Geometric Programming
Andrija Stupar, University of Toronto,
andrija@stupar.comWe formulate the design of power electronic converters as a geometric program
(GP). The GP formulation allows the use of existing convex optimization solvers
and techniques which greatly speeds up the optimization process. Some
components are naturally modeled as posynomials, and those that are not can be
accurately approximated via empirical fitting. The GP formulation allows a quick
generation of Pareto curves and surfaces over a number of operating points and
component combinations for a converter. This is illustrated using an example
where loss-volume Pareto-optimal solutions for different multi-level converter
topologies are compared.
WA05
101E-MCC
Forest Management II
Sponsored: Energy, Natural Res & the Environment II Forestry
Sponsored Session
Chair: Peter Rauch, BOKU-Univ of Natural Resources & Life Sciences,
Institute of Production and Logistics, Wien, 1180, Austria,
Peter.rauch@boku.ac.at1 - Quantifying The Conflict between Competing Forest Ecosystem
Services under Alternative Climate Scenarios
Nicholas Kullman, University of Washington,
Nick.Kullman@gmail.com, Sandor Toth
The potential impact of climate change on the production of forest ecosystem
services is well-documented. Much less understood is how the tradeoffs among
these services would change. Would more conflicts arise? How would one
measure such a conflict at the first place? We introduce a new method to quantify
conflict as the ratio between the volume of n-dimensional objective space under
the production possibilities frontier of a set of competing ecosystem services vs.
the space defined by the ideal solution. We illustrate the method via use of real
examples from management.
2 - A Replanning Model For Maximizing Woodland Caribou Habitat
Alongside Timber Production
David L Martell, University of Toronto,
david.martell@utoronto.ca,
Andrew B. Martin, Jonathan Leo William Ruppert, Eldon Gunn
Woodland caribou in the boreal forest region of Canada tend to prefer older jack
pine forest stands but such habitat needs can conflict with industrial fibre needs.
We present a forest harvest scheduling model that meets timber harvest targets
while maximizing a proxy for woodland caribou habitat, the configuration of
preferred habitat on the landscape. We used our model to carry out a case study
of the Trout Lake forest in northern Ontario, Canada, and found that our model
creates about 10% more caribou habitat than an earlier heuristic procedure and
30% more than the current plan for the forest.
3 - Modeling The Spatial Interactions Of Timber Harvesting And Sitka
Deer Habitat On The Tongass National Forest
Yu Wei, Colorado State University,
yu.wei@colostate.edu,
Michael Bevers, Curt Flather, Greg Hayward, Ben Case,
Mary Friberg, Thomas Hanley
We developed and implemented a spatially explicit timber harvest scheduling
model to optimize the joint production of timber and deer habitat capability on
management units of the Tongass NF. We found model solutions to be sensitive to
variation in sea level snow depths, with notable effects on timber harvest
schedules, deer habitat capacity, and the amount and location of old-growth
remaining at the end of a forest planning horizon. We also discovered that a
spatially scattered harvesting pattern helped create diversified forest compositions;
which consequently could improve the overall deer forage production.
4 - How Does Climate Change Impact On Wood Supply Security?
Peter Rauch, BOKU-Univ of Natural Resources & Life Sciences,
Peter.rauch@boku.ac.atIn order to assess climate change risks and their mid and long-term impacts on
the bio-based industry a System Dynamics model of the Austrian wood supply
was developed that includes a stochastic simulation of the main risk agents. The
model examines future annual cut and evaluates wood supply security
considering climate change impacts. Simulation results provide insights on
probabilistic future wood supply security and reveal a contra-intuitive system
effect for the climate change scenario.
WA06
102A-MCC
Text Mining I
Sponsored: Data Mining
Sponsored Session
Chair: Majeed Simaan, RPI, 231 Congress Street, Troy, NY, 12180,
United States,
simaam@rpi.edu1 - Text Mining Based Prediction Model For Incident Occurrences In
Steel Plant
Sobhan Sarkar, Research Scholar, IIT, IIT kharagpur, kharagpur,
721302, India,
sobhan.sarkar@gmail.com,Vishal Lakha,
Irshad Ansari, Jhareswar Maiti
The aim of this study is to provide the predictive solution using text mining and
classification algorithms. Data on accident occurrences for a period of four years
from a steel industry was collected. The outputs of text mining have been fed into
four binary classification algorithms (SVM, k-NN, Random Forest, Maximum
Entropy) which were tested further for evaluation of the best fit model to predict
WC04