2016 INFORMS Annual Meeting Program

19 / 561

INFORMS Nashville – 2016
19
SA07
SA05
101E-MCC
Power System Resilient Design and Optimization
Sponsored: Energy, Natural Res & the Environment,
Energy I Electricity
Sponsored Session
Chair: Seyedamirabbas Mousavian, Clarkson University,
8 Clarkson Avenue, Potsdam, NY, 13699-5790,
1 - Self-healing Attack-resilient PMU Network For Power
System Operation
Chen Chen, Argonne National Laboratory, Lemont, IL, United
States, , Hui Lin, Jianhui Wang, Junjian Qi,
Dong Jin
We propose a self-healing PMU network by exploiting the features of
programmable configuration in a software-defined networking (SDN) to achieve
resiliency against cyber-attacks. After a cyber-attack, by changing the
configuration of the network switches, the disconnected yet uncompromised
PMUs will be reconnected to the network to “self-heal” and thus restore the
observability of the power system. Specifically, we formulate an integer linear
programming (ILP) model to minimize the overhead of the self-healing process,
while considering the constraints of power system observability, hardware
resources, and network topology.
2 - Minimum Risk-maximum Availability Response To Electric
Vehicle-initiated Smart Grid Attacks
Seyedamirabbas Mousavian, Clarkson University,
, Melilke Erol-Kantarci, Thomas Ortmeyer
Malware pose a significant threat to the power grid and the connected electric
vehicle infrastructure. Penetration and propagation of cyber attacks including
worms and viruses vary depending on the nature of the connected systems.
Electric vehicles (EVs) being the mobile portion of the smart grid may easily
spread worms and viruses in a large geographic area. We propose a probabilistic
model for the worm propagation in EV to Electric Vehicle Supply Equipment
(EVSE) networks, formulate threat levels and then, we propose a Mixed Integer
Linear Programming (MILP) model as a protection scheme that relies on isolating
infected nodes.
3 - Storage And Generation Expansion Problem Considering
Primary Response
Hrvoje Pandzic, University of Zagreb, Zagreb, Croatia,
, Yury Dvorkin, Miguel Carrion
A sustainable and efficient generation and storage expansion program needs to
consider both the capacity needs and short-term operational requirements of a
power system. A generation expansion formulation considering frequency
regulation using both generators and storage units will be presented.
4 - Optimal Resilient Grid Design Distribution And
Transmission Systems
Russell Bent, Los Alamos National Laboratory,
Emre Yamangil, Harsha Nagarajan, Pascal Van Hentenryck
Modern society is critically dependent on the services provided by engineered
infrastructure networks, particularly distribution and transmission grids. When
natural disasters (e.g. Hurricane Sandy) occur, the ability of these networks to
provide service is often degraded. However, well-placed upgrades to these grids
can greatly improve post-event network performance. Hence, we pose the
optimal electrical grid resilient design problem as a two-stage, stochastic mixed-
integer program with damage scenarios and propose decomposition-based
algorithms to solve and analyze medium-sized networks.
SA06
102A-MCC
Joint Session DM/Optimization: Discrete
Optimization and Machine Learning
Sponsored: Data Mining
Sponsored Session
Chair: Berk Ustun, Massachusetts Institute of Technology, Cambridge,
MA, United States,
1 - A Multi-group Discrete Support Vector Machine – Theory And
Computation
Eva Lee, Georgia Tech,
We describe a general-purpose machine learning framework, DAMIP, for
discovering gene signatures that can predict vaccine immunity and efficacy.
DAMIP is a multi-group ‘concurrent’ classifier that offers unique features not
present in other models: a nonlinear data transformation to manage the curse of
dimensionality and noise; a reserved-judgment region that handles fuzzy entities;
and constraints on the allowed percentage of misclassifications.Computational
results for biological and medicine problems will be discussed.
2 - Optimized Risk Scores
Berk Ustun, MIT, Massachusetts Institute of Technology, Sloan
School of Management, Cambridge, MA, 02142, United States,
Cynthia Rudin
Risk scores are simple models that let users quickly assess risk by adding,
subtracting, and multiplying a few small numbers. These models are widely used
in healthcare and criminology, but difficult to create because they need to be risk-
calibrated, use small integer coefficients, and obey operational constraints. We
present a new approach to learn risk scores from data by solving a discrete
optimization problem. We formulate the risk score problem as a MINLP, and
present a cutting-plane algorithm to efficiently recover the optimal solution by
solving a MIP. We demonstrate the benefits of our approach by creating risk
scores for real world problems.
3 - Supersparse Integer Regression Model For Nonparametric Failure
Time Analysis
Keivan Sadeghzadeh, MIT Sloan School of Management,
Cambridge, MA, United States, , Cynthia Rudin
Analysis of failure time data has an inevitable role in predicting events
occurrence. We develop an integer-based predictive model that is accurate and
also interpretable, in order to determine effective features and predict potential
failures. The strategy is to select appropriate covariates for censored large-scale
and high-dimensional failure time data in a regression model. Our approach is to
design robust algorithm to find the optimal integer solution for supersparse linear
model. This optimal solution is reached by using machine learning techniques
over a high-dimensional closed quadric hypersurface.
4 - Nested Clustering On A Graph
Gokce Kahvecioglu, Northwestern University, 2145 Sheridan Road
Room C210, Evanston, IL, 60208, United States,
David P. Morton
We study a clustering problem defined on an undirected graph with weight
function defined on the edges, which denotes the importance of the connection
between vertices. We remove a set of edges in order to maximize the number of
clusters in the residual graph while minimizing the weight of deleting edges.
Solving this graph clustering problem parametrically identifies the solutions that
lie on the concave envelope of efficient frontier and the breakpoints on this
envelope have a nested structure. We propose to solve this parametric model in
polynomial time by solving a sequence of parametric maximum flow problems,
which yields the family of nested clusters on the efficient frontier.
SA07
102B-MCC
Undergraduate OR Prize – I
Invited: Undergraduate Operations Research Prize
Invited Session
Chair: Pavithra Harsha, IBM Research, 1101 Kitchawan Road,
Room 34-225, Yorktown Heights, NY, 10598, United States,
1 - Car Sharing Fleet Location Design With Mixed Vehicle Types For
CO2 Emission Reduction
Joy Chang, University of Michigan, Ann Arbor, MI, United States,
Siqian Shen, Ming Xu
As carsharing companies integrate electric vehicles to reduce CO2 emissions, we
optimize a mixed-integer linear program to study a carsharing fleet location
design problem with mixed vehicle types. We create a minimum cost flow
formulation on a spatial-temporal network to model round-trip and one-way
vehicle flows. We test different one-way trip demands, and provide a model that
ensures the first-come first-serve principle while satisfying CO2 restrictions. Via
testing instances from the Boston Zipcar fleet, we provide insights on carsharing
fleet location and vehicle-type composition.
2 - Collaborative Decision Making For Air Traffic Management
Hale Erkan, Bilkent University, Ankara, Turkey,
, Nesim K. Erkip, Ozge Safak
We propose a model which can be utilized within a collaborative decision making
(CDM) framework for rescheduling of flights. The proposed mathematical
program is expected to be utilized by major stakeholders, airlines and air
navigation service providers. After providing the constraints, we list possible
equity and efficiency performance measures that will make-up the objective
function to be used by a stakeholder. We suggest guidelines to utilize the model
for any stakeholder within CDM. Finally, a case study is prepared using publicly
available data to demonstrate possible benefits.