# Pooling problem, Gurobi example

I tried to model and solve this Pooling problem below. I used Excel and Opensolver and used Couenne solver to get the .nl problem file. I then used Neos Knitro and Scip 8 to (try to) solve the problem.

I got different results than the Gurobi example page. Then I entered Gurobi numbers to my Excel model and noticed the numbers do not match.

• pool 1 input 325 + 258.3 + 500 = 1083.3 but pool 1 output is 92.8 + 990.6 = 1083.4 which is not the same number and I think they should be

• Gurobi optimal objective is 439182.xxx but in my model with the same input and output numbers the objective becomes 541278

"Best objective 4.391825892869e+05, best bound 4.391825893721e+05, gap 0.0000%"

Is Gurobi example something wrong or is my model wrong?

https://www.gurobi.com/resource/standard-pooling-problem/

https://colab.research.google.com/github/Gurobi/modeling-examples/blob/master/pooling/std_pooling_gcl.ipynb

• Welcome to OR.SE. Would you please, elaborate more on the problem you are trying yourself? What you mentioned in the question is totally about Gurobi's example and does not mirror the problem you have. :) Jan 31, 2022 at 7:56
• I am trying to practice milp modeling and solver use. But now it seems that I am wasting my time because the example is something wrong. If I use given optimal oil flow values I get objective 541278 dollars instead of 439182.xxx. I run Scip about 2 hour but it is not going under 200% gap. Neos Knitro gave me solution but it seems it does not satisfy all constraints. This could be infeasible. And I need some practice problems that are feasible. Jan 31, 2022 at 17:19
• Do you ensure that you have coded the same problem in Gurobi and SCIP? Jan 31, 2022 at 18:23
• I think I have. I made it four times from scratch. If I concentrate only Gurobi example, the numbers (pool 1 input & output) does not match. And later the example values does not match my model. Jan 31, 2022 at 18:41
• I found at least one reason. In the example all variable optimal values are rounded maybe too much and if I use less rounded values everything looks better. I got better values with Scip 8. May 10, 2022 at 13:16