4 présentations

• 
  14h45 - 15h07

  Diagnostic tools for nonlinear optimization problems

  • Robert Parker, prés., Los Alamos National Laboratory
  • Andrew Lee, OLI Systems

  Nonlinear optimization is a powerful tool for engineers and operations researchers, but implementing a nonlinear algebraic model of a complicated process is error-prone. To help diagnose common modeling issues, we have developed a diagnostics toolbox for algebraic models written in Pyomo. Specifically, we have implemented the Dulmage-Mendelsohn decomposition to diagnose structural singularities, bound propagation to identify potential function evaluation errors, and tools to identify certificates (small subsets of constraints or variables) of rank-deficiency or ill-conditioning. In this talk, we discuss the methods implemented, present our user-friendly API provided by the IDAES process modeling environment, and provide an example of debugging a nuanced error in a chemical reactor optimization problem.

• 
  15h07 - 15h29

  InfiniteOpt.jl: Accelerating the Solution of Infinite-Dimensional Optimization Problems

  • Joshua Pulsipher, prés., University of Waterloo

  Infinite-dimensional problems are a general and challenging problem class with applications in the optimal control, optimization under uncertainty, PDE-constrained optimization, and more. This talk will discuss recent updates to the Julia package, InfiniteOpt.jl, which enable a large collection of modeling logic via disjunctive programming and accelerated solution via specialized parallel workflows that are GPU compatible. For posing logical constraints, we will show how InfiniteOpt.jl integrates with DisjunctiveProgramming.jl to easily model infinite-dimensional logic (e.g., dynamic switching constraints). We will also show how InfiniteOpt.jl integrates with ExaModels.jl to solve infinite-dimensional optimization problems significantly faster than existing state-of-the-art methods.

• 
  15h29 - 15h51

  PAMSO.jl: Parametric Autotuning Algorithm for Multi-Time Scale Optimization in JuMP

  • Li Can, prés., Purdue University
  • Ramanujam Asha, Purdue University

  Optimization models with decision variables in multiple time scales are widely used across various fields such as integrated planning and scheduling. To address scalability challenges in these models, we present the Parametric Autotuning Multi-Time Scale Optimization (PAMSO) algorithm. PAMSO tunes parameters in a low-fidelity model to assist in solving a higher-fidelity multi-time scale optimization model. These parameters represent the mismatch between the two models. PAMSO defines a black-box function with tunable parameters as inputs and multi-scale cost as output, optimized using Derivative-Free Optimization methods. This scalable algorithm allows optimal parameters from one problem to be transferred to similar problems. Case studies demonstrate its effectiveness on an MINLP model for integrated design and scheduling in a resource task network with around 67,000 variables and an MILP model for integrated planning and scheduling of electrified chemical plants and renewable resources with around 26 million variables.

• 
  15h51 - 16h13

  Implementing non-standard automatic QUBO reformulation in JuMP

  • Pedro Maciel Xavier, prés., Purdue University
  • David E. Bernal Neira, Purdue University

  Quadratic Unconstrained Binary Optimization (QUBO) models are the most used framework for solving optimization problems using novel computing paradigms such as Quantum Computers. To make these devices more accessible, automatic reformulation software was developed to translate general MINLP models into the QUBO formalism. When representing constraints, these tools typically introduce penalty terms that often degrade the final model’s conditioning. To avoid that, non-standard reformulation schemes that do not rely on penalization can be employed to encode certain types of constraints. Implementing such techniques in an automatic reformulation system is not trivial and will require additional mechanisms that are beyond what algebraic modeling languages usually provide. In this presentation, we would like to discuss the details and common caveats of implementing these methods within the Julia Mathematical Programming (JuMP) ecosystem. This is done as an extension to the QUBO.jl environment.