By Amy Pielow 

Power systems operations is a topic of extensive research and the focus of many sessions at this year's annual meeting. The Contributed cluster features several sessions on the topic, including Tuesday morning's presentations on management of power systems.

Geoffrey Pritchard of the University of Auckland spoke first on modeling uncertainty from inflows for use in hydrothermal scheduling. The model must consider seasonal variations, serial dependence, and heavy right tails of week-to-week water inflows. Autoregressive general linear models with different structures for the errors were among those fit. Outputs can be used as scenarios for scheduling optimization models as shown with a test problem. The resulting energy prices from an SDDP optimization varied significantly depending on the source of scenarios.

The second presentation on storage technology choices in electricity systems was delivered by PhD student Santhosh Suresh, University of Michigan. His work focused on the interaction between battery storage losses, transmission losses, and renewable utilization. Local storage mitigates losses from transmission, but centralized storage may benefit from increased efficiency; thus, the trade-off between options leads to different decisions given the realities of the system. In general, the researchers found that it is better to pool with increased efficiency of transmission or storage.

National Renewable Energy Laboratory research scientist Marissa Hummon presented work on accelerating unit commitment and dispatch models. Presently, solving their rolling 48-hour optimization for an entire year may take several weeks of computation. Instead of solving the year sequentially, the researchers parallelized optimization of shorter time horizons then compared the results of the single optimization to the simultaneous computations. Because unit commitment involves long-term decisions, there may be optimality loss with weekly or monthly simulations. Despite this, the researchers found overlapping the shorter simulations by only two days was sufficient to reasonably emulate the annual simulation results.

By Justin Fung 

As a 2012 McKinsey report predicting a future worldwide shortage of analysts currently makes the rounds on the Internet, INFORMS is doing its part to support and recognize the achievements of the next generation of operations researchers and management scientists. The four finalists of the Undergraduate OR Prize, representing the U.S., Puerto Rico, and Turkey, were invited on Tuesday afternoon to present their work in a session chaired by Vincent Slaugh of Carnegie Mellon University.

"It was good to meet people ...  it's both a networking opportunity for us and a chance of seeing (other) education and cultures," said Basak Kepir of Bilkent University after her group's presentation. Winners of the Undergraduate OR Prize, Kepir and her group -  accompanied by their advisor  - traveled from Ankara, Turkey to present their flight scheduling optimization decision support system that was eventually adopted by AnadoluJet.

Rebecca Swaszek made the trip from Lehigh University to discuss her research on the utilization of decision trees in healthcare applications, designing a dynamic program for medication management. "It's overwhelming," she said, in regards to the number of proceedings available to first time attendees. Whether a shortage of analysts comes to fruition, the 2013 INFORMS Undergraduate OR finalists appear enthusiastic and well-positioned to follow the lead of their senior counterparts. Kepir also stated, "We've met a lot of professors ... it's been a great influence."

Pascal Van Hentenryck, Optimization Research Group Leader, National ICT Australia and professor at University of Melbourne, gave an IFORs Distinguished Lecture in disaster management and computational approaches. He started his talk with the impact of natural disasters like Hurricane Irene, Hurricane Sandy, San Diego Blackout, Tohoku Tsunami, and Black Saturday. He mentioned that OR/MS tools are important for preparation and response to the disasters, and continued with a recent implication of OR/MS tools for disaster management. He noted that good understanding of the aftermath disasters is important for disaster prediction. He continued his talk with the characteristics of disaster management and explained decisions are often used in the disaster management. He pointed out the challenges of the problems of disaster management such as complex, joint combinatorial optimization problems, fast response time, stochasticity, and nonstandard objectives and multiobjectives. He also indicated the computational challenges which are complex infrastructures in uncertain environments such that off-the-shelf packages cannot solve the disaster management problems in even small instances. Van Hentenryck then described three case studies.

The first case study is scheduling and planning evacuations in flood management. He showed that GIS, flood simulation, planning, and scheduling evacuations must be dealt with. He focused on scheduling evacuations in practice. The decisions to be made are when, where, and how to evacuate the residents. The objective can be selected as maximizing the number of evacuated people. Secondly, he studied relief distribution in disaster management. The decisions should be defined as where to store the relief before the disaster and delivery plans for the recovery phase. The objective in this case can be minimizing unsatisfied demand, latest delivery time, or storage costs in worst case scenarios. The last case study is based on power flow optimization. The objective is finding how much power to generate and the phase angles to maximize the service lead. He explained three-step approaches with hybrid optimization to solve this problem. He concluded his talk by identifying the essential research topics for OR/MS studies in disaster management.

Solving real-world problems is at the heart of operations research and practitioners who successfully implement mathematical theory in innovative and practical ways should be recognized. The Daniel H. Wagner Prize is awarded every year to researchers employing operations research methodology to practical problems. Applicants are judged based on both the analytical content and quality of exposition of their research. The 2013 winners, announced during Tuesday afternoon's keynote address, are researchers working on a newsvendor problem for the Yedioth Ahronoth weekly newsletter. Allen Butler, chair of the 2013 award committee, introduced Yale T. Herer, who presented the work.

The classic newsvendor problem served as the motivation for the work. Production decisions must be made before demand is realized, so choosing the optimal level of papers to print must consider stochastic sales levels. In reality, up to 25% of papers produced are returned unsold, so better production decisions can lead to greatly reduced costs of both production and salvage. The proposed model implements mid-week redistribution of unsold papers based on demand in the model's first subperiod. Mathematically, the delayed two-phase distribution becomes a cost-minimizing stochastic planning problem with recourse and two subperiods.

La'isha magazine (owned by Yedioth Ahronoth media) was used as a pilot case study for 10 weeks and 50 retailers to test the model. After collecting data, the researchers suggested producing fewer copies of the magazine than before. Despite lower quantities available, similar sales numbers were realized, leading to decreased costs. Additionally, fewer stock-outs occurred. The pilot's success led to model implementation for more than 15 magazines and 400 retailers with cost savings up to $250,000 per year, depending on the day of redistribution. Future work will focus on applying the model to weekend newspapers and combining the model with data generated from RFID tags on the magazines to test if RFID tracking is worthwhile.

By Justin Fung 

In an age when "interdisciplinary" has become a popular buzzword, Professor Dimitris Bertsimas of MIT is furthering the extant interdisciplinary nature of operations research. An academic with industry experience and an industry man with an education, Professor Bertsimas has spearheaded two innovations in healthcare analytics that are expanding the definition of what it means to be "interdisciplinary."

To that end, during Tuesday's plenary, intrigued attendees looked on as Bertsimas gave an on-screen demo of Lifestyle Analytics (LiA), a web-based application that uses OR analytics to design food and exercise plans tailored to the user's preferences to control blood glucose - an ideal tool for persons at risk of diabetes, and just as accessible as any social-networking mobile app.

Professor Bertsimas has not limited the scope of healthcare analytic tools to patients, however. He outlined an OR/health initiative aimed at healthcare professionals in which a combination of statistical modeling and optimization is employed in predicting clinical trial outcomes and designing treatment regiments for gastric and gastroesophageal cancers, with important, exciting inference for cancer of all types.

Perhaps spurred by his personal investment in the projects (Bertsimas dedicated the presentation to his parents, both of whom were victims of cancer), he does not stop the bridge-building at the research stage. Utilizing business acumen and advances in the social technology revolution, Bertsimas puts the research in the hands of the end-user, pushing the OR envelope and "adding value" to interdisciplinary problems.