Leveraging Advanced Modeling Capabilities in Pyomo for National Security

Event Sponsor: 
Mathematics and Computer Science Division Seminar
Start Date: 
May 22 2017 - 10:30am
Building/Room: 
Building 240/Room 1404-1405
Location: 
Argonne National Laboratory
Speaker(s): 
William Hart
Host: 
Barry Smith

Pyomo is a powerful open-source modeling language (ML) that is written in the widely used Python programming language. Complex, real-world optimization problems require the use of MLs to simplify the process of describing and analyzing optimization problems. Specifically, a ML is applied to develop an optimization model, or a high-level description of an optimization problem. MLs also automate the tedious processes of writing input files, executing an optimization solver, and extracting the solution.
 
Sandia has developed Pyomo to support the analysis of complex national security applications. We will review representative applications that illustrate key features of Pyomo, including municipal water security, electric power planning and management and cybersecurity. A key feature of Pyomo is its object-oriented design, which can express structured models in a hierarchical manner.  This design enables Pyomo to model and analyze a wide range of optimization problems, including linear and integer programs, stochastic programs, nonlinear programs, generalized disjunctive programs, mathematical programs with equilibrium conditions, bilevel programs and dynamic optimization problems. Pyomo supports explicit control of the application of modeling transformations, which convert models into canonical forms that are supported by optimization solvers. We will describe how these capabilities provide a generic, re-usable framework for analyzing complex real-world applications.
 
Speaker Biography:
William Hart is the Purdue Partnerships Manager at Sandia National Laboratories, who focuses on developing research partnerships with Purdue University.  At Sandia, Bill has led research teams developing methods related to stochastic programming, parallel optimization, mathematical modeling, theoretical computer science, and graph algorithms.  Bill’s research focuses on optimization techniques, including: parallel branch-and-bound, heuristic global optimization, and derivative-free local search. He has also applied optimization to many real-world applications, including computational biology, engineering design, logistics planning, and sensor placement in municipal water systems. He has developed a variety of optimization software libraries, including the widely used DAKOTA and Pyomo software packages. In 2016, Bill led a team that was awarded an R&D100 Award for their work on Pyomo.  Bill is a member of the COIN-OR Senior Leadership Board, and he is a former member of the board of the INFORMS Computing Society.