Optimization

[vc_row][vc_column width=”2/3″][vc_row_inner][vc_column_inner width=”1/6″][/vc_column_inner][vc_column_inner width=”1/6″][vc_column_text]speaker-focacci[/vc_column_text][/vc_column_inner][vc_column_inner width=”2/3″][vc_column_text]

Filippo Focacci

Co-founder & CEO
DecisionBrain

From Optimization Models To Advanced Decision Support : A Manufacturing Planning And Scheduling Case Study.

In this presentation, we describe a real world Manufacturing Planning and Scheduling solution implemented for a market leader in the domain of disposable product (e.g. plastic Cutlery, Cups, Containers, etc.). We will focus on the different dimensions needed to implement a successful application: business analysis, advanced optimization models, user interface design, and implementation and development best practices. We will start by analyzing the business problem addressed, focusing on the main business challenges and the benefits of the application. A large part of our discussion will then focus on the details of the optimization models. While the core optimization models (capacitated lot sizing models and scheduling models) have been widely studied, we will discuss the many real-world extensions that have been necessary to generate solutions able to drive business value to our customer. These extensions include multi-level production (raw material, intermediates and finished goods), transportation times, warehouse capacity limitations, maturity and shelf life constraints, multi-attribute setups, secondary resources for production and setup, parallel machines, etc. We will describe heuristics and tips that enabled us to effectively solve large scale problems with good quality of solution and good performance. While many of these are not needed when solving well structured academic instances, they are indeed critical to effectively solve real world cases. We will also describe a decomposition approach that enabled us to smoothly integrate long term planning and detailed scheduling. Indeed, our application covers both the Supply Chain level and the Plant Scheduling level where these two levels are tightly integrated. Finally, we will provide insights on how to build a cost structure that correctly represents the key economical trade-offs between inventory, operational efficiency and service level. A third part of the presentation will focus on the requirements for an effective user interface (UI) for decision support. Used by planners without any knowledge of optimization technology, the UI is a critical component for the success of the project. It must quickly provide insights on the quality of the plans and schedules and enable planners to interact with the optimization system. We will discuss capabilities for interactive planning, what-if analysis and support to “solution explanation”. The last part of the presentation will focus on development and implementation best practices. We will discuss the advantages of using a sophisticated software platform and how to build specific solutions by assembling standard software components. We will compare this approach against a “packaged solution” approach and discuss pros and cons. We will also stress the importance of using an iterative and agile project methodology to minimize project costs and risks. We will conclude with the benefits of the system and the changes that the new application generated in the role of the planning team.

Bio

Before founding DecisionBrain, Filippo Focacci worked for ILOG and IBM for 20 years where he held several leadership positions in Consulting, R&D, Product Management and Product Marketing in the area of Supply Chain, Logistics and Optimization. He received a Ph.D. in Operations Research (OR) from the University of Ferrara (Italy) and has over 20 years experience applying OR techniques in industrial applications in several optimization domains. He has published several Supply Chain and Optimization articles for international conferences and journals. He has been granted a patent for Optimization Models.[/vc_column_text][/vc_column_inner][/vc_row_inner][vc_row_inner][vc_column_inner width=”1/6″][/vc_column_inner][vc_column_inner width=”1/6″][vc_column_text]no-photo-icon[/vc_column_text][/vc_column_inner][vc_column_inner width=”2/3″][vc_column_text]

G. N. Srinivasa Prasanna

Professor
International Institute of Information Technology

Optimization Under Uncertainty For Demand Response In Smart Grid

In this presentation, we are going to discuss an application of OR techniques, in smart grid. Optimizations in smart grids are challenging due to large size, uncertainty in the demand and supply, especially due to penetration of distributed renewable energy sources and consumers becoming prosumers. Demand Response (DR) programs are getting focused attention due to sustainability and green environment considerations. These programs can be more interactive and automatic by satisfying preferences of both the consumers and utility companies. Fine grained level of monitoring and controlling till the appliance level is possible with the advances in sensing systems, smart appliances and smart meters, which give rise to new optimization problems. We highlight the identified gaps from literature and our approach to solve them. We propose modeling methodologies to handle large scale demand response problems of smart grid with uncertainty in price, demand and solution methods for such models. Our model is of Mixed Integer Linear Programming (MILP) to handle constraints such as user comfort, emission limits, and fine grained appliance level constraints. Another feature in our model is the incorporation of uncertainty, in an intuitive fashion, using substitutive, complementary and general constraints as an extension of robust optimization framework. Convexification and hierarchical model with aggregation are used to make the problems computationally tractable, with targeted near real-time response. The convexified model is solved by nested optimization using state-of-art ILP solver both at consumer and utility levels in decoupled rolling horizon approach. All-or-None heuristic is developed as a solution method for quick run times. Fair demand response scheme is proposed with a view to increase the participation of consumers in DR programs. We observed benefits of up to 20% of energy savings during the large scale optimization with heterogeneous consumers and practical implementation of our model at a campus level network with 100s of smart plugs. These MILP results are compared with our heuristic that is scalable up to millions of consumers. While the heuristic offers simple run times of few seconds, the results indicate 5-10% of extra energy bill reduction by using optimization methods, compared to heuristics, on area wide grids. We analyze the stability and accuracy of the system for millions of users with sensitivity analysis. Utility is able to cut down the costs up to 7% and reduce the % of difference in allocated and actual usage at an area level to 10%. The analysis of results indicates that our model is mutually beneficial to the consumers by minimizing consumer’s energy bill while keeping their equipment in desirable operating regions and to utility companies by flattening the demand curve over a day either by shifting or reducing the load and automating the demand response programs. Significant takeaways from the presentation are the ways to model demand response programs in smart grid, solution methods for such models, and results and interpretation of these solutions.

[/vc_column_text][/vc_column_inner][/vc_row_inner][vc_row_inner][vc_column_inner width=”1/6″][/vc_column_inner][vc_column_inner width=”1/6″][vc_column_text]speaker-li[/vc_column_text][/vc_column_inner][vc_column_inner width=”2/3″][vc_column_text]

Haitaio Li

Associate Professor of Logistics & Operations Management and Research Fellow
Center for Transportation Studies, University of Missouri- St. Louis

Optimizing The Labor Strategy Of A Professional Service Firm

A professional service firm often faces challenges in matching its available workforce with the demands for service delivery. While most of the existing research and practice address such challenges at the operational level, in this talk, we coin the labor strategy optimization (LSO) problem to optimize a service firm’s labor plan that supports its target revenues at the strategic level. The LSO helps a firm optimize the capacity, capability, location and flexibility of its workforce, which may serve as inputs to the more detailed resource assignment and scheduling decision-support. We will present a new modeling framework that captures the complexity and richness of a general service transformation process. It offers a data-driven approach to model the inefficiency in service transformation and risks involved in offshoring operations. We will also show case the implementation of a prototype version of our LSO methodology at the former HP Consulting & Integration organization.

Bio

Haitao Li is an Associate Professor of Logistics and Operations Management, and a Research Fellow of Center for Transportation Studies, at University of Missouri – St. Louis (UMSL). He holds his Ph.D. in Operations Management from the University of Mississippi (2005), Master of Arts in Economics also from University of Mississippi (2002), and Bachelor of Engineering in Foreign Trade in Industry with minor in Aeronautical Engineering from Beihang University, China (2000).
Dr. Li’s research interests include optimization modeling, simulation, and algorithm design in the application domains of scheduling, workforce optimization, and supply chain configuration. He has worked as a Statistical Analyst at the Naval Personnel Research, Study and Technology (NPRST) in Millington, TN, and was a Visiting Scholar at the Hewlett-Packard Laboratory (HPL) in Palo Alto, CA. His past research projects include developing capacity and capability planning (CCP) models for strategic workforce optimization, new models for tactical resource planning (RP) and project portfolio optimization at HP; manpower optimization and scheduling of DDX battleship for the U.S. Navy; approximate dynamic programming (ADP) algorithms for solving high-dimensional stochastic resource-constrained project scheduling and its applications in unmanned aerial vehicle (UAV) scheduling for the U.S. Army; dynamic models and solution approaches for resource distribution and scheduling of large-scale construction projects at J.E. Dunn.
Dr. Li has published in reputable scholarly journals including European Journal of Operational Research, Computers and OR, Interfaces, Omega, Military Operations Research, Journal of Scheduling and Annals of Operations Research among others. He was a recipient of the Young Investigator Award from the US Army Research Office (ARO). With one U.S. Patent pending and several invention disclosures, he was named 2015 UMSL Inventor of the Year.[/vc_column_text][/vc_column_inner][/vc_row_inner][vc_row_inner][vc_column_inner width=”1/6″][/vc_column_inner][vc_column_inner width=”1/6″][vc_column_text]speaker-melkote[/vc_column_text][/vc_column_inner][vc_column_inner width=”2/3″][vc_column_text]

Sanjay Melkote

Managing Consultant & Senior Data Scientist
IBM Global Business Services

Optimizing The Workforce Of An Electric Utility

We describe the solution of a fundamental problem faced by all electric utilities: when should company employees perform service requests, and when should contractors be used? We solve this problem at a major North American electric utility using the following approach. First, a forecasting model based on exponential smoothing is developed to predict the daily demand for service requests. Then, historical service request data are scrutinized to identify job types which are performed significantly more cheaply by contractors or company employees. An integer programming model is then developed to optimize the workforce to meet predicted demand for service requests. The models are used to (1) determine the ideal mix between employees and contractors and (2) stress test the workforce’s ability to respond to fluctuations in demand. The model results indicate the utility can save $11 million per year by increasing its use of contractors for certain job types. The forecasting and optimization models will be used for tactical planning and to maximize the efficiency of the utility’s workforce for 2016 and beyond.

Bio

Sanjay Melkote is a Managing Consultant and Senior Data Scientist at IBM’s North America Analytics Center in Dublin, Ohio. He has also held a number of positions in government and academia. He has applied linear and integer programming to solving a number of transportation, logistics, and energy-related problems in government and industry. He holds a Ph.D. in Industrial Engineering and Management Sciences from Northwestern University.[/vc_column_text][/vc_column_inner][/vc_row_inner][vc_row_inner][vc_column_inner width=”1/6″][/vc_column_inner][vc_column_inner width=”1/6″][vc_column_text]speaker-tarel[/vc_column_text][/vc_column_inner][vc_column_inner width=”2/3″][vc_column_text]

Guillaume Jean Tarel

Vice President
Artelys Canada, Inc.

100% Renewable Electricity Generation In France? Key Lessons

Given the sharp fall in renewable energy costs, studies are needed to assess whether it will soon be feasible and economically sound to produce 100% of our power through renewable sources. We have recently delivered such a study to the French Environment and Energy Management Agency, which answers positively in the case of France. We used advanced analytics to optimize the electricity mix, taking into account the demand-supply equilibrium on short time-scales, using a regionalized model and several climate/technology/social acceptance scenarios. We shall deliver five key lessons during the presentation:
– #1 – Several 100% renewable mixes can balance supply and demand on an hourly basis, including in unfavourable weather conditions
– #2 – Overall electricity costs would vary between $110 to $150/MWh
– #3 – Demand flexibility and storage solutions are required
– #4 – Complementarity between technologies is key
– #5 – The transmission network must be reinforced to pool the regional potentials
The study can be obtained at http://mixenr.ademe.fr/en

Bio

Guillaume Tarel started his career in the energy field in Switzerland where he was involved as a consultant in a prospective energy modeling activity for the Swiss Federal Office of Energy (Energy strategy 2050). He is now based in Montréal, Canada, where he manages the Canadian subsidiary of the company Artelys. Artelys develops, sells and supports modeling and optimization software and solvers. In addition Artelys provides expert advisory to energy-intensive companies, governments, energy producers, regulation authorities and transmission operators. Guillaume has a BA in engineering from the Ecole Centrale de Paris engineering school and a MSc and PhD from the Swiss Federal Institute of technology in Lausanne (EPFL).[/vc_column_text][/vc_column_inner][/vc_row_inner][/vc_column][vc_column width=”1/3″][vc_row_inner][vc_column_inner width=”1/4″][vc_column_text]microphone[/vc_column_text][/vc_column_inner][vc_column_inner width=”3/4″][vc_column_text]

Focused Tracks

Speakers organized by Track

small-arrow-bullet-gray Plenary Speakers
small-arrow-bullet-gray 2016 Franz Edelman Award Competition
small-arrow-bullet-gray Analytics Leadership & Soft Skills
small-arrow-bullet-gray
Analytics Process
small-arrow-bullet-gray
Data Mining
small-arrow-bullet-gray Decision & Risk Analysis
small-arrow-bullet-gray Fraud Detection & Cyber Security
small-arrow-bullet-gray Health Care & Life Sciences
small-arrow-bullet-gray
INFORMS Prizes & Special Sessions
small-arrow-bullet-gray Internet of Things
small-arrow-bullet-gray Marketing Analytics
small-arrow-bullet-gray Optimization
small-arrow-bullet-gray
Predictive Analytics/Forecasting
small-arrow-bullet-gray Revenue Management & Pricing
small-arrow-bullet-gray Sports & Entertainment
small-arrow-bullet-gray Supply Chain
small-arrow-bullet-gray
Supply Chain Analytics
small-arrow-bullet-gray Technology Tutorials
small-arrow-bullet-gray
Technology Workshops – Sunday
small-arrow-bullet-gray
Transportation[/vc_column_text][/vc_column_inner][/vc_row_inner][/vc_column][/vc_row]