Getting dual variables in multi-scenario model

回答済み

Paolo Bignardi

2020年08月29日 19:00

• 
• 
• 

Hello,

I'm trying to solve a stochastic programming model by using benders decomposition. I built a model with multiple scenarios for better speed and save on ram usage, though I need to extract the dual variables (Pi attribute) for each of the different scenarios. Is there any way to do this? On this aspect the documentation is a bit vague, if someone has any experience with this or knows a way forward I'd be very greatful.

Thank you in advance

Paolo

 

1

• 正式なコメント

  

  Simranjit Kaur

  • Gurobi Staff

  2024年12月06日 00:42

  This post is more than three years old. Some information may not be up to date. For current information, please check the Gurobi Documentation or Knowledge Base. If you need more help, please create a new post in the community forum. Or why not try our AI Gurobot?.
• 

  Jaromił Najman

  • Gurobi Staff

  2020年09月03日 09:06

  Hi Paolo,

  A multi scenario model is a MIP, meaning that the shadow prices are not well-defined in this case. Please have a look at our knowledge base article on Pi values for MIPs.

  Is your original model which you formulated as multiple scenarios an LP? If yes, then the best way to get shadow prices with a proper meaning is to solve this LP.

  Best regards,
  Jaromił

  0
• 

  Paolo Bignardi

  2020年09月10日 06:55

  • 
  • 
  • 

  Thank you for the reply. Yes, I am working with LP. The thing is, I have many LP problems to solve and they are all quite similar (and quite big as well), so it seems a bit wasteful to solve them individually. If I could extract dual variables from each of the multi scenario solution I could save quite a bit of time in the solution. 

  Moreover, I am just curious to know how multi-scenario works under the hood. Does it just set the rhs's, lower bounds and upper bound and then uses warm start to speed up the solution or is there more to it (maybe parallel execution or something similar)?

  Thank you very much for the insights.

  Best regards,

  Paolo

  0
• 

  Jaromił Najman

  • Gurobi Staff

  2020年09月10日 09:03

  Hi Paolo,

  The multi scenario feature is designed for mixed-integer problems. Thus, we always recommend to not apply it for purely continuous LPs. It is in most cases faster to solve the LPs independently, e.g., by parallelizing the solution process. You can then access all information, e.g., the duals of the LPs.

  Applying the multi scenario feature to LPs may result in a very complex problem as binary variables are added to model the different scenarios with different right hand sides etc. resulting in a MIP. This MIP is then solved using Gurobi's whole arsenal for MIPs, i.e., all heuristics, parallelism etc. 

  Best regards,
  Jaromił

   

  0

投稿コメントは受け付けていません。