Seminars

Fall 2019

Oct. 3: Hise Gibson, COL, US Army, and phd, harvard business school

Hosted by: Colorado School of Mines
When: October 3
Where: Brown Building W475

Spring 2019

Leading the Way to Better Strategic Decisions: Jennifer Meyer, Strategic Decisions Group

Title: Leading the Way to Better Strategic Decisions By Jennifer Meyer, Strategic Decisions Group

Hosted by: the Operations Research Group at Jeppesen, A Boeing Company
When: Thursday 31 January; 5:30pm Networking/Snacks; 6:00pm Presentation
Where: 55 Inverness Drive East, Englewood, CO 80112

Abstract: We’ve all seen the consequences of poor strategic decisions. A missed opportunity. A botched attempt at eliminating risk. A course charted on flights of hubris rather than grounded in understanding. These poor decisions cost individuals and organizations millions and billions of dollars every year. And yet, what can we do?

In fact, we can do a lot. We can apply a proven framework to judge the quality of a decision – at the time the decision is made. Then, we can facilitate dialogue to bring leaders to a good decision with confidence and clarity. Such is the calling of a decision professional.

In this talk we’ll explore the decision quality (DQ) framework and how it improves strategic decisions. We’ll discuss how a decision professional brings insight into situations that are complex both analytically and organizationally. And we’ll highlight how you and your organization can learn more in the annual conference of the Society of Decision Professionals, this year being held right here in Denver, March 5 to 8. The 2019 conference theme is “Data and Decisions.” (Early bird pricing and student discounts are available only through January 18! Read more and register at http://daag.io)

Bio: Dr. Jennifer Meyer has spent more than 20 years with Strategic Decisions Group, working as a decision professional and executive educator to help clients in industries ranging from oil and gas to technology. She is co-author of the book “Decision Quality: Value Creation from Better Business Decisions” and is a Fellow in the Society of Decision Professionals. She holds a PhD in Operations Research from Stanford University.

Note: This chapter meeting is being hosted by the Operations Research Group at Jeppesen, A Boeing Company. Before Jennifer’s talk, members of the OR group will provide a brief overview of the work that they do as part of Best Practices and Integration Services at Jeppesen.

Due to security restrictions at the Jeppesen facility, attendees must RSVP no later than January 22. 
To RSVP, please send your full name to Brian Lambert william.lambert@jeppesen.com.

Renewable Energy Optimization (REO): an Heuristic Approach to Planning Renewable Energy Projects at a Site: andy walker, phd, pe, national renewable energy lab

Thursday, September 16, University of Colorado Denver, Room CU1110, 1250 14th Street, Denver

Speakers: Dr (COL) Armacost, Dr Hewitt, Dr Laguna

Title: OR in the Rocky Mountains

Abstract: How does Operations Research contribute sustaining value and interest to our academic and professional community here in the Rocky Mountains? Our distinguished panelists, who represent a cross-section of OR programs in our region and have followed diverse and exhilarating career paths, share their experience and perspective on successes,
challenges, and opportunities in learning, teaching, and applying OR.

Fall 2018

OptTek – Tools for Optimizing Complex Systems: James kelly, phd, opttek systems inc.

Rocky Mountain Chapter of INFORMS Meeting

Title: OptTek- Tools for Optimizing Complex Systems

Speaker: James Kelly, Ph.D

Date: Thursday, September 27, 2018

Time: 5:30 p.m. Hors d’oeuvres and networking
6:00 p.m. Presentation and discussion

Location: OptTek Systems, Inc., 2241 17th Street, Boulder, CO 80302
(Parking lot is off 17th Street directly adjacent to building (on south side). The area around our office is residential, so 3-hr street parking at no-charge is also available.)

Abstract: OptTek Systems is a recognized world leader in complex systems optimization. We utilize an advanced set of problem-solving and analytical techniques to help clients in the commercial and governmental sectors resolve a vast array of mission-critical challenges. These techniques include simulation modeling, statistical analysis, metaheuristics optimization, evolutionary algorithms, and tabu search and scatter search, to name a few. Everything we do is grounded in helping organizations make the most informed decisions and achieve the best possible outcomes. We will provide an overview of our core products and discuss ongoing enhancements and use cases for these products.

Spring 2018

Impulse Buying and Store Shelf Space Allocation: prof. tulay flamand, division of economics and business, colorado school of mines

Title: Impulse Buying and Store Shelf Space Allocation

Speaker: Professor Tulay Flamand, Division of Economics and Business, Colorado School of Mines

Date: Thursday, March 15, 2018

Time: 5:30 p.m. Hors d’oeuvres and networking
6:00 p.m. Presentation and discussion

Location: Brown Building West Wing 280 (enter on the northwest side of the building and the room will be the first one on your left)

Abstract: Impulse (unplanned) buying is a common shopping behavior that is responsible for over 50% of the revenue in some retail settings. Here, we investigate how to optimize the  store-wide shelf space allocation of retailers in a way that guides in-store traffic and induces impulse buying. A data analysis is conducted for a grocery store in Beirut, which prompts the development of a predictive regression model for in-store traffic. The customer traffic model is embedded within an optimization model in order to prescribe shelf space allocation solutions. Due to the non-linearity and computational challenges that the proposed model poses, we propose a variable neighborhood search algorithm. We compare the allocation obtained by our approach against the current shelf space allocation of the grocery store to provide managerial insights. Further, we investigate alternative allocation scenarios and sensitivity analysis with respect to key model parameters.

Spring 2017

Determinants of Feedback Effectiveness in Production Planning: Dr. Peter Letmathe, RWTH Aachen university, aachen, germany

Rocky Mountain Chapter of INFORMS Meeting

Title: Determinants of Feedback Effectiveness in Production Planning*

Speaker: Dr. Peter Letmathe, Faculty of Business and Economics, RWTH Aachen University, Aachen, Germany

Date: Thursday, Mar 23, 2017

Time: 5:30 p.m. Hors d’oeuvres and networking
6:00 p.m. Presentation and discussion

Location: KOBL 308 at Leeds School of Business (995 Regent Dr., Boulder, CO 80309). Paid parking is available right in front of the building.

Abstract: The focus is the interplay of cognitive capabilities (mathematical understanding and heuristic problem solving) and learning from feedback. Furthermore, the authors analyze the role of individual factors in designing appropriate feedback systems for complex decision-making situations. Based on a learning model the purpose of this paper is to present an experimental study analyzing the feedback effectiveness in a repeated complex production planning task. Referring to individual characteristics in terms of educational background and problem solving capabilities of the decision maker the authors compare different forms of feedback systems.

geometric modeling techniques: steffen borgwardt

Speaker: Steffen Borgwardt

Date and time: 6pm, Thursday, February 2, 2017

Location: Room 2504, Commons Building I, 1201 Larimer Street, Denver

Abstract: With geometric modeling techniques, one can represent the feasible solutions of problems in operations research as objects in high-dimensional space. The properties of these objects reveal information about the underlying problems and lead to algorithms. We model an application in land exchange as a clustering problem where the clusters have to adhere to prescribed cluster sizes. In this approach, we connect least-squares assignments, cell complexes, and the studies of polyhedra. Further, we report on how these results were implemented in practice. The devised methods lead to various tools for more general questions on big data.

uncertainty quantification: roger ghanem, university of southern california

Speaker: Roger Ghanem, University of Southern California

Date and time: Friday (1/13) at 3pm in Chauvenet 143

Topic: Uncertainty Quantification

fall 2016

A decomposition method for convex optimization problems, the Bienstock-Zuckerberg (BZ) algorithm revisited: renaud chicoisne

Date and Time: Thursday, October 13, 9-10am, University of Colorado, Denver, Math Department Academic building, fourth floor, room 4113

Speaker: Renaud Chicoisne

Title: A decomposition method for convex optimization problems, the Bienstock-Zuckerberg (BZ) algorithm revisited.

Abstract: In this talk, we will briefly introduce the BZ algorithm as it originally appeared for open pit mining scheduling problems. We then discuss its equivalence with a specialized  column generation scheme and how its framework can be generalized to solve convex optimization problems. We illustrate this generalization with a resource constrained nonlinear objective routing problem.

Beyond the Black Box in Derivative-Free and Simulation-Based Optimization: stefan wild

Date and Time: Friday, September 30, 3-4pm, CH143

Title: Beyond the Black Box in Derivative-Free and Simulation-Based Optimization

Speaker: Stefan Wild

Abstract: The advent of computational science has unveiled large classes of nonlinear optimization problems where derivatives of the objective and/or constraints are unavailable. Often, these problems are posed as black-box optimization problems, but rarely is this by necessity. We report on our experience extracting additional structure on problems consisting of both black-box and algebraic or otherwise known components. We provide diverse examples of such problems that are being solved at Argonne National Laboratory. In each case, we use quadratic surrogates to locally model both the black-box and algebraic components and obtain new, globally convergent grey-box optimization methods.

Short Bio: Stefan Wild is a Computational Mathematician in the Mathematics and Computer Science Division at Argonne National Laboratory and a Fellow in the Computation Institute at the University of Chicago. Prior to his current appointment, he was an Argonne Director’s Postdoctoral Fellow and a DOE Computational Science Graduate Fellow at Cornell University. He obtained his Ph.D. in operations research from Cornell University and B.S. and M.S. degrees in applied mathematics from the University of Colorado-Boulder. His primary research focus is on algorithms and software for challenging numerical optimization problems.

Image-based Machine Learning: Peigang Zhang, PhD, Uber

Title: Image-based Machine Learning

Speaker: Peigang Zhang, PhD, Uber

Wednesday, Sep 14, 2016

5:30 p.m. Hors d’oeuvres and networking

6:00 p.m. Presentation and discussion

Location: Room DCB 100 in Daniels College Building, (if entering DCB from the North, DCB 100 is the first room on the left). DCB is located at 2101 S University Blvd, Denver CO. PARKING: There is limited on-street metered parking on University (free after 6pm). Also, there is a parking garage at the corner of High and Evans that has hourly parking.

Abstract: Automatic information extraction from imagery (satellite and street) is one of the most important techniques to improve map services. It usually includes object detection and object classification steps. Deep learning is used in both steps and recent advances in this area lead to dramatic improvement in automation results. This talk will focus on information extraction from imagery, deep learning, and techniques to process large-scale petabyte data.

Bio: Peigang Zhang is a senior software engineer at Uber. Before joining Uber, he worked at Microsoft Bing Maps division for 7 years. He received his bachelors degree from Peking university and Ph.D in computer science from the University of North Carolina, Charlotte. 

spring 2016

The Role of an Enterprise Analytics Team at a Large Retail Company: michael fuhr, cap, sports authority

LOCATION: University of Colorado Denver, Auraria Campus, 1201 Larimer
Street (Academic Building I), Room 4125

TIME: 5:30 p.m. Hors d?oeuvres and networking, 6:00 p.m. Presentation

TITLE: The Role of an Enterprise Analytics Team at a Large Retail Company

SPEAKER: Mr. Michael Fuhr, CAP, Sports Authority

ABSTRACT: The Sports Authority Enterprise Analytics team is a centralized source of analytics expertise for the entire company. Its members come from a variety of backgrounds and have a diverse set of skills. In this talk, we’ll discuss the role of an Enterprise Analytics team in a large retail company and will present a few examples of the kind of work we do. We’ll also discuss how we’ve used the INFORMS Certified Analytics Professional (CAP) program for training and career progression and to provide a consistent analytics experience to internal customers across the domains of retail.

BIO: Michael Fuhr spent most of his career in Information Technology before moving into analytics. He enlisted in the Air Force in 1990 and spent six years as a computer programmer and Unix system administrator working closely with meteorologists and intelligence analysts. He then worked mostly in the cable TV and telecommunications industries as a system administrator, security consultant, and database developer. He started at Sports Authority as a contractor in 2011 and joined the company full time in 2012.

fall 2015

Innovative Aisle Configurations for Unit-Load Warehouses: russell d. meller, vp for R&D, fortna inc.

Title: Innovative Aisle Configurations for Unit-Load Warehouses

Speaker: Russell D. Meller, Vice President for Research & Development at Fortna Inc.

Abstract: Unit-load warehouses are used to store items—typically pallets—that can be stowed or retrieved in a single trip. In the traditional, ubiquitous design, storage racks are  arranged to create parallel picking aisles, which force workers to travel rectilinear distances to picking locations. We consider the problem of arranging aisles in new ways to reduce the cost of travel for a single-command cycle within these warehouses. Our models produce alternative designs with piecewise diagonal cross aisles, and with picking aisles that are not parallel. One of the designs promises to reduce the expected distance that workers travel by more than 20 percent for warehouses of reasonable size. We report on the expected performance of these designs under various warehouse configurations and operating policies, as well as relate experiences from the implementation of these designs.

Location and Time: MZ326, 8-9:15am

spring 2015

Adjustable Robust Optimization of Process Scheduling Under Uncertainty: chrysanthos e. gounaris, dept. of chemical engineering, carnegie mellon university

Title: Adjustable Robust Optimization of Process Scheduling Under Uncertainty

Speaker: Chrysanthos E. Gounaris, Department of Chemical Engineering, Carnegie Mellon University

Rocky Mountain INFORMS Chapter Meeting
Tuesday, May 19th

5:30 p.m. Snacks and networking
6:00 p.m. Chapter Business
6:10 p.m. Presentation and discussion
7:00 p.m. Optional dinner afterward

Location: Academic Building I, 1201 Larimer Street, Room 4125

Note: This is the brand new building on the west corner of Speer and Larimer. There is 4 hour visitor parking available outside the building ($2/hr) (enter from Auraria Pkwy). Participants can also park in the usual parking garages/lots.

RSVP to Adam Clark at Adamclark.USA@gmail.com <Adamclark.USA@gmail.com> by May18th

Abstract: We develop an Adjustable Robust Optimization (ARO) framework to address uncertainty in the parameters of Process Scheduling models. Unlike the traditional RO approach, which results in a static, here-and-now solution, ARO results in a solution policy that is a function of parameter realizations. We discuss the derivation of the ARO counterpart in this context, and we propose decision-dependent uncertainty sets to enforce that the policy depends only on observable realizations. Our results show that the ARO approach provides robust solutions that are considerably less conservative than those obtained with the static approach. In addition, we show that ARO can provide feasible solutions to instances with zero-wait task restrictions for which the traditional approach inherently cannot.

Bio: Professor Chrysanthos Gounaris received a Dipl. in Chemical Engineering (2002) and an M.Sc. in Process Control (2003) from the National Technical University of Athens. He then attended Princeton University, where he earned an M.A. (2005) and a Ph.D. in Chemical Engineering (2008). His doctoral thesis, pursued under the supervision of Professor Chris A. Floudas, is entitled “Advances in Global Optimization and the Rational Design of Shape-Selective Separations” and explores the use of nonlinear modeling and global optimization techniques in the study of porous materials. After graduation, Professor Gounaris joined McKinsey & Co. as an Associate, where he provided consultation to petrochemical, pharmaceutical and consumer packaged-goods companies on a variety of projects of operational and strategic nature (2008-2010). He returned to academia to pursue post-doctoral research at Princeton University (2010-2013), after which he joined the Department of Chemical Engineering at Carnegie Mellon University as an Assistant Professor.

Forbidding solutions in combinatorial optimization: shabbir ahmed, industrial & systems engineering, georgia tech

Title: Forbidding solutions in combinatorial optimization

Speaker: Shabbir Ahmed, H. Milton Stewart School of Industrial &
Systems Engineering, Georgia Institute of Technology

Time: 6pm, Thursday February 5th

Location: Room 350, Leeds School of Business, CU Boulder, 995 Regent
Dr, Boulder, CO 80309

Abstract:

Various important applications give rise to combinatorial optimization problems with difficult side constraints that can be easily checked for satisfiability but cannot be easily formulated using standard mathematical programming constructs. We refer to such constraints as “black box” constraints. In this talk we present effective formulations to cut-off or forbid a given list of solutions that violate the black box constraints from the set of all solutions. Such formulations pave the way for Benders-like branch and cut approaches for combinatorial optimization with black box constraints.

This talk is based on joint on the paper “Forbidden vertices” (to appear in Mathematics of Operations Research, 2015) with Gustavo Angulo, Santanu S. Dey, and Volker Kaibel. A preprint is available at http://www.optimization-online.org/DB_HTML/2013/09/4041.html Speaker bio: Shabbir Ahmed is the College of Engineering Dean’s Professor in the H.

Milton Stewart School of Industrial & Systems Engineering at the Georgia Institute of Technology. He received his PhD from the University of Illinois at Urbana-Champaign in 2000. His research interests are in optimization, specifically stochastic and integer programming. Dr. Ahmed served as the Chair of the Stochastic Programming Society and as a Vice-chair of the INFORMS Optimization Society. He serves on the editorial board of various journals include Mathematical Programming and Operations Research. Dr. Ahmed’s honors include the Stewart Fellowship from Georgia Tech, the National Science Foundation CAREER award, two IBM Faculty Awards, the Coca-Cola Junior Professorship from ISyE, and the INFORMS
Dantzig Dissertation award. 

spring 2013

Joint Optimization of Virtual Capacities and Bid-Prices for Revenue Management: Prof. vulcano, new york university, stern school of business

Time and location: 2:00 pm on April 19th in Koelbel 340, CU Boulder campus

Speaker: Professor Vulcano, New York University, Stern School of Business

Title: Joint Optimization of Virtual Capacities and Bid-Prices for Revenue Management

Abstract:

We consider a network revenue management problem where the physical capacity at the perish time of the asset is uncertain while the firm processes requests and/or not all the accepted requests show-up at the service delivery time. For both cases, the controller sets a virtual capacity and a bid-price for each of the resources at the beginning of the finite horizon, and collects revenues by accepting or rejecting requests for products using a standard bid-price policy; i.e., a reservation is accepted as long as there is enough virtual capacity available and the collected revenue exceeds the sum of the bid-prices of the resources defining the product. At the end of the horizon, the effective capacities and demands are realized. If there is not enough room and part of the accepted requests cannot be accommodated, the controller incurs a penalty cost for each bumped reservation. The firm’s objective is to maximize the total cumulative adjusted revenue (sales revenue minus penalty cost) obtained by the end of the horizon.

We present a continuous capacity formulation for this problem which allows for the partial acceptance of requests. The model admits a simple calculation of the sample path gradient of the adjusted revenue function. This gradient is then used to construct a stochastic steepest ascent algorithm. We show that the algorithm converges (w.p.1) to a stationary point of the expected adjusted revenue function under some mild conditions. Then, through an exhaustive numerical study, we show that our controls are computed within an order of magnitude faster computational times than other recent proposals and deliver revenues that are comparable or higher than the ones obtained from those benchmarks.

Joint work with Alex Weil, University of Buenos Aires and University di Tella, Argentina.

fall 2011

Scheduling an open-pit mine for extraction: challenges in the optimization of very large integer programming problems

Date: Friday, November 18, 1pm

Location: Engineering Hall 211, School of Mines campus

Title: Scheduling an open-pit mine for extraction: challenges in the optimization of very large integer programming problems.

Speakers: Daniel Espinoza, Marcos Goycoolea, Eduardo Moreno and Gonzalo Munoz

Abstract: For the purpose of production scheduling, open pit mines are discretized into three-dimensional arrays known as block models. Production scheduling consists in deciding which blocks in the model should be extracted, when they should be extracted and how each extracted block should be processed. Blocks which are on top should be extracted first, and capacity constraints limit the production each time period. Since the 1960s it has been known that this problem can be modeled with integer programming. However, the large size of real instances (3-10 million blocks, 15-100 time periods) has made these models impractical for use in real planning applications, thus leading to the use of numerous heuristic methods.

In this talk I will discuss recent advances in Linear and Integer Programming that could potentially change the way industry schedules their open pit mining operations. Specifically, I will discuss some new linear programming decomposition methods for solving precedence-constrained multiple-knapsack problems, and some simple heuristics for obtaining solutions from these relaxations. Finally, I will show how these techniques apply to solving real planning problems.

Approximate Dynamic Programming for Network Revenue Management : dan zhang, phd

Date: Tuesday, November 8, 2011, 4pm

Location: Engineering Hall 211, School of Mines campus

Speaker: Dan Zhang, PhD

Title: Approximate Dynamic Programming for Network Revenue Management

Abstract: Revenue management (RM) entails tactical controls of product availability or pricing in order to maximize seller revenues. A central component of many practical RM systems is the so-called network RM problem, where availabilities of products with different revenue contributions and resource consumption requirements are managed dynamically to  maximize expected total revenue from a network of resources. A canonical example lies in the airline industry where products are itineraries on an airline network with different fares, and resources are seats on scheduled flights. The problem has also been applied to many other industries, including hotels, car rentals, manufacturing, retailing, etc. The problem can often be formulated as large-scale stochastic dynamic programs, the solution of which is difficult due to the well-known curse of dimensionality. This talk describes some recent advances in approximate dynamic programming that deal with the network RM problem. 

spring 2011

Air Traffic Flow Management: thomas vossen, cu boulder leeds school of business

Date: Thursday, January 20, 2011

Location: University of Colorado Denver, Room TBA
CU-Denver Building (1250 14th Street between Larimer and Lawrence)
Directions: http://www-math.ucdenver.edu/contact/map2office.shtml
Walking distance from RTD bus (Market Street and Union Stations)
and Light Rail (Union Station and Theatre District/Convention Center)
Get directions at http://www.rtd-denver.com/, Trip Planner

Time: 5:30 refreshments; 6pm start of presentation

Speaker: Thomas Vossen, CU Boulder Leeds School of Business

Title: Air Traffic Flow Management

Abstract: While nearly everyone has experienced flight delays or cancelations, few know how the Federal Aviation Administration and the airlines go about managing disruptions to the air transportation system. In this talk, I will provide an overview of the processes and procedures used in air traffic flow management, and discuss how Operations Research can be used to help resolve congestion.

fall 2010

Practical Guidelines for Solving Difficult Linear and Mixed Integer Programs: alexandra newman

Date: Thursday, November 18, 2010

Location: Colorado School of Mines, Hill Hall 202

Time: 5:30 refreshments; 6pm start of presentation

Speaker: Alexandra Newman

Title: Practical Guidelines for Solving Difficult Linear and Mixed Integer
Programs

Abstract: Even with state-of-the-art hardware and software, large linear and mixed integer programs can require hours, or even days, of run time and are not guaranteed to yield an optimal (or near-optimal, or any!) solution. We present suggestions for appropriate use of commercially available optimization software and guidelines for efficient formulation and that can vastly improve performance. These suggestions and guidelines concern: (i) algorithmic tuning, (ii) variable elimination, (iii) variable definition, (iv) cut generation, and (v) data manipulation. We draw on examples from energy and mining, inter alia.

Strategies for employing renewable energy on the grid (Application of Benders Decomposition to an Extended Unit Commitment Problem with Renewable Energy and Storage): jennifer van dinter, colorado school of mines

Thursday, November 11, Colorado School of Mines, Brown Building 206, 2pm

Speaker: Jennifer Van Dinter, Colorado School of Mines PhD Candidate, Mineral and
Energy Economics

Title: Strategies for employing renewable energy on the grid (Application of Benders
Decomposition to an Extended Unit Commitment Problem with Renewable Energy and Storage)

Abstract: We present a mixed integer linear programming unit commitment model which determines generator schedules, associated production and storage quantities, and spinning and non-spinning reserve requirements to respond to demand. Our model minimizes operational costs and quantifies the CO2 and NOx emissions from the optimal schedule. A complex constraint set balances the load, imposes requirements on the way in which plants and storage devices
operate, and tracks reserve requirements. We capture cost functions and emissions curves with piecewise-linear constructs. We propose a hybrid solution method including a sliding time window heuristic and Benders Decomposition to solve our version of the unit commitment model more quickly than standard optimization algorithms. 

Optimizing the Acquisition and Operation of Distributed Energy Systems: kris pruitt, colorado school of mines

Thursday, October 28, Colorado School of Mines, Brown Building 206, 2pm

Speaker: Kris Pruitt, Colorado School of Mines PhD Candidate, Mineral and
Energy Economics and U.S. Air Force

Title: Optimizing the Acquisition and Operation of Distributed Energy Systems

Abstract: We present a mixed-integer programming model for designing and operating a distributed energy generation system to supply electricity to a large, commercial building. The model determines the optimal on-site generation capacity to acquire, along with operating levels over time, to minimize total cost subject to system performance characteristics and the building’s demand.

Hydro-Thermal Scheduling with Electricity Demand Uncertainty and CO2 Emission Constraints: prof. steffen rebennack

Friday, September 24, Colorado School of Mines, Chauvenet Seminar Room (143), 3pm

Speaker: Prof. Steffen Rebennack

Title: Hydro-Thermal Scheduling with Electricity Demand Uncertainty and CO2 Emission
Constraints

Abstract: Hydro-thermal scheduling addresses the problem of determining optimal decisions related to the dispatch of power plants (e.g., hydro and thermal) under the presence of uncertainty. The classical theory in this field focuses on stochastic hydro inflows. This problem can be solved for real instances in a reasonable amount of time with so-called stochastic dual dynamic programming (SDDP) algorithms (Pereira and Pinto, 1991). We present several extensions of the problem.

Stochastic fuel prices play an increasingly important role when the installed capacity of thermal plants increases relative to that of the hydro-thermal plants. We present a scenario tree approach for modeling the fuel cost uncertainty and adopt the classical SDDP algorithm to treat this additional uncertainty. The proposed method can also easily be applied to address electricity demand uncertainty. The results for case studies in Panama and Costa Rica are discussed.

After the Kyoto protocol in 1997, the power industry has been faced with the challenges of reducing its CO2 emissions. Therefore, we present a model for CO2 emission quotas in the context of SDDP. The emissions are modeled via reservoirs, admitting a time decomposition and an efficient solution algorithm. This new model is applied to a Guatemalan power problem.

OR in the rocky mountains: dr. (Col) armacost, dr. hewitt, dr. laguna

Thursday, September 16, University of Colorado Denver, Room CU1110, 1250 14th Street, Denver

Speakers: Dr (COL) Armacost, Dr Hewitt, Dr Laguna

Title: OR in the Rocky Mountains

Abstract: How does Operations Research contribute sustaining value and interest to our academic and professional community here in the Rocky Mountains? Our distinguished panelists, who represent a cross-section of OR programs in our region and have followed diverse and exhilarating career paths, share their experience and perspective on successes,
challenges, and opportunities in learning, teaching, and applying OR.

spring 2010

Renewable Energy Optimization (REO): an Heuristic Approach to Planning Renewable Energy Projects at a Site: andy walker, phd, pe, national renewable energy lab

Tuesday, April 6, 5pm

Speaker: Andy Walker PhD PE, Principal Engineer, National Renewable Energy Laboratory

Title: Renewable Energy Optimization (REO): an Heuristic Approach to
Planning Renewable Energy Projects at a Site

Abstract: Many organizations that operate multiple facilities have  implemented renewable energy projects, but selection of the sites has been anecdotal, often driven by a local champion, and may not represent the optimal solution. This presentation will describe a simple method developed to identify and prioritize cost-effective renewable energy
projects in a system of real property that covers areas with different renewable energy resources; utility rates; incentive regimes; and construction costs. The objective of the optimization is to minimize life cycle cost and the variables are the sizes of each renewable energy technology: photovoltaics; wind; solar water heating; solar ventilation air preheating; concentrating solar heat and power, biomass heat and power; ground source heat pumps; landfill gas; and daylighting. An integration algorithm accounts for the interactions between the technologies. Constraints on the optimization may include: a percent-energy-from-renewables goal; land area constraint; initial cost constraint; and many others that change the nature of the problem being solved and the solution. Examples presented include several Federal and industrial facilities around the country.

Optimization of feeder additions with respect to reliability and cost: hilary e. brown

Tuesday, March 2, 5pm.
————————

Speaker: Hilary E. Brown

Title: Optimization of feeder additions with respect to reliability and cost

Abstract: The Smart Grid is gaining visibility as the preferred program for the modernization of the United States electric power grid. It is expected that distributed generation sources will be able to participate in the local electric system and, thus, help to improve the reliability of the electric distribution system. In order to improve the reliability, it is proposed that new
connections be made between feeders in the distribution system, changing the topology from radial to networked. The use of a heuristic technique and a multi-objective genetic  algorithm to optimize the connections between feeders with respect to reliability and cost will be discussed. Results will be given for a small test system.

How to be an Effective OR Analyst: dr. bill tarantino

HILL HALL 204, JANUARY 21 (SPECIAL DATE, TIME AND LOCATION)

Speaker: Dr. Bill Tarantino (winner of the Edelman prize for OR practice)

Title: How to be an Effective OR Analyst

Absract: Many seasoned OR practitioners have their definition of success as well as rules of thumb to be a consultant. Many of these rules focus on profitability — their profitability. I will discuss my rules of thumb for how to be an effective consultant. My focus is how to influence decision makers with insightful analysis. In my discussion, I assume each analyst is qualified to be an operations researcher and discuss how he or she can become an effective operations research consultant, a distinction that I believe highlights where our profession is heading.

fall 2009

The Network Diversion Problem: Complexity and Stronger Integer Programming Formulations : chris cullenbine

Tuesday, December 1, 5pm.
————————

Speaker: Chris Cullenbine, Operations Research/Division of Economics and Business, CSM

Title: The Network Diversion Problem: Complexity and Stronger Integer Programming Formulations

Abstract: The network-diversion problem (NDP) seeks a minimum-weight, minimal s-t cut in a graph that contains a pre-specified edge. The problem arises in intelligence-gathering and
war-fighting scenarios. We develop a polynomial-time algorithm for undirected, s-t planar NPDs. We identify other NDP variants that may be solvable in polynomial time as well as those that are known to be intractable. The latter case requires integer linear programming to find optimal solutions and we improve upon an existing formulation by (a) replacing a submodel for a single-commodity flow-preserving path with two-commodity constructs, and (b) adding constraints that link a flow-preserving path with each vertex’s location with respect to the optimal cut. Duality gaps decrease by as much as 51% for the new formulation compared to the original, and we solve test problems with as many as 10,000 vertices and 39,800 edges in less that one hour using CPLEX 12.1 on a Sun Fire x4150 workstation. In one hour, the original formulation will not solve for problems containing more than 2,502 vertices and 9,900 edges. 

NURBs-Enabled Design Space Optimization: cameron turner

Tuesday, November 3, 5pm.

Speaker: Cameron Turner, Engineering Division, CSM

Title: NURBs-Enabled Design Space Optimization

Abstract: Engineering relies upon science and mathematics to predict the performance of a design within the design space. Within this space, the inherent tradeoffs between design variables and performance indices occurs. Understanding the ramifications of these tradeoffs is vital to the process of engineering design. Yet the design space is hyper-dimensional, complex and nonlinear and it is difficult to comprehend or visualize. Even systems composed of simple components can exhibit highly complex relationships between their design variables and performance indices. Understanding and manipulating these design variable-performance index relationships is a fundamental challenge in the area of engineering design. Engineers can employ metamodels that build approximate or surrogate models of other models, to generate useful representations of the design space. Metamodels may be constructed from a wide variety of mathematical basis functions; those derived from Non-Uniform Rational B-splines (NURBs), offer unique advantages compared to other metamodeling approaches. Through adaptive design space surveys, data can be collected and used to produce a NURBs-based metamodel by adapting NURBs fitting algorithms originally developed for computer graphics representations. This formulation leads to a design space representation that exhibits properties which support design space visualization, adaptive sequential sampling algorithms, and the detection of robust design space regions. Significantly, NURBs-based metamodels enable multi-start optimization algorithms to locate the global metamodel optimum in finite time. The same features that have made NURBs the de facto standard for computer graphics applications also support these unique metamodeling capabilities. Through these models, engineers obtain a window into the hyperdimensional design space, allowing the designer to explore the design space for undiscovered design variable combinations with superior performance capabilities and open a host of possibilities for further research. 

Optimizing the design of networks for detecting trends: betsy weatherhead, cu boulder

Tuesday, October 6, 5pm.

Speaker: Betsy Weatherhead, University of Colorado at Boulder

Title: Optimizing the design of networks for detecting trends

Abstract: Climate change has presented a number of important questions, including how fast should we expect the climate to change? Where are the changes likely to be largest? How can we use data to determine the drivers of change? Fundamental to many of these questions is identifying how the Earth is currently changing. Past efforts at long-term monitoring have
often been problematic with local influences corrupting data and instrument changes making trends difficult. In recent years, a number of efforts have started to carefully consider the question, “How do we monitor in order to detect trends?” Various groups highlight expensive high-quality monitoring devices as a solution; others promote measuring in many remote locations, while still others propose the value of explanatory measurements. This presentation will outline the problem of maximizing a monitoring network’s ability to detect trends
with the fundamental constraint of financial resources. Metrics of successful monitoring will be proposed. This is a currently unsolved problem, so colleagues of all levels of interest are welcomed to participate in what will likely be an interesting discussion.