Infeasible in nonc-convex model

Answered

Salvador Pineda Morente

April 11, 2025 10:47

• 
• 

I am solving an OPF problem using polar coordinates (with cosine and sine terms), and Gurobi reports that the model is infeasible. However, I know that a feasible solution exists, and when I warm-start the problem with that solution, Gurobi successfully finds the optimal solution.

This made me wonder: in the context of a non-convex problem, what exactly does 'infeasible' mean? Could it be that the solver is reporting local infeasibility rather than global infeasibility?

Related to

• Nonconvex
• Numerics

-1

• 

  Maliheh Aramon

  • Gurobi Staff

  April 11, 2025 23:13

  Hi, 

  Gurobi is a global solver for both convex and non-convex problems. If the model is solved to infeasibility without a warm-start solution, the Gurobi has concluded that there is no feasible solution globally. 
  
  The discrepancy between the results with and without a warm-start solution can likely be due to numerical issues in the model. You can try solving the model by disabling procedures such as DualReductions=0, Aggregate=0, bumping the NumeircFocus parameter, and relaxing the value of the tolerance parameters, including FeasibilityTol and OptimalityTol.
  
  You can also consider calling the model.computeIIS() method to diagnose infeasibility through computing an IIS. More information can be found in the article "How do I determine why my model is infeasible?". 
  
  

  Best regards,

  Maliheh

  0

Please sign in to leave a comment.