Bonmin Solver

GAMS
1

Hello everybody,

I’m new to GAMS and this group and I hope to find some help here :slight_smile:

For my model, which I have to use in my bachelor thesis, I have to use a minlp solver. So the only one, which is included in the basic GAMS version is Bonmin.
When I run my model with the B-Hyb algorithm, it solves very fast, but I get a solver message: OA on non-convex constraint is very experimental.
When I change the used algorithm to B-BB it takes veeeerry long… So do you have any suggestions for the option file?

Using the B-Hyb algorithm there is another mystery and I don’t really don’t know whats wrong with it…
In my model I use two sets:

Sets
t time_steps / t1t24 /,
u units / u1u18 /
;

When I change t to 35 the model solves very fast, so does it at t = 37. But when I use t=36 it takes “hours” and returns no feasible solution. So does it with 47 (fast), 48 (slow), 49 (fast), 71 (fast), 72 (slow) and 73 (fast). So it can be said, when there is factor 12, it has problems solving the model. Any ideas?

Thank you all!




\

To view this discussion on the web visit https://groups.google.com/d/msg/gamsworld/-/bE5Kb6OeQMoJ.
To post to this group, send email to gamsworld@googlegroups.com.
To unsubscribe from this group, send email to gamsworld+unsubscribe@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/gamsworld?hl=en.

2

Reply-to: gamsworld@googlegroups.com

Hi,

On 11/14/2012 08:13 PM, MrUnknown wrote:

Hello everybody,

I’m new to GAMS and this group and I hope to find some help here :slight_smile:

For my model, which I have to use in my bachelor thesis, I have to use a
minlp solver. So the only one, which is included in the basic GAMS version
is Bonmin.

There is also Couenne.
If you have an academic license, then there is also SCIP.

When I run my model with the B-Hyb algorithm, it solves very fast, but I
get a solver message: OA on non-convex constraint is very experimental.
When I change the used algorithm to B-BB it takes veeeerry long… So do
you have any suggestions for the option file?

Using the B-Hyb algorithm there is another mystery and I don’t really don’t
know whats wrong with it…
In my model I use two sets:

Sets
t time_steps / t1t24 /,
u units / u1u18 /
;

When I change t to 35 the model solves very fast, so does it at t = 37. But
when I use t=36 it takes “hours” and returns no feasible solution. So does
it with 47 (fast), 48 (slow), 49 (fast), 71 (fast), 72 (slow) and 73
(fast). So it can be said, when there is factor 12, it has problems solving
the model. Any ideas?

As the warning says, you have a nonconvex model.
Bonmin is made for convex models.
B-BB is an option that is most robust for nonconvex models.
For other algorithmic options, it can be random when Bonmin stops and what result it gives, since the generate outer-approximation may not be an outer-approximation in this case.

If you want to use Bonmin’s B-Hyb option, it would be best to reformulate your model so that it is convex.
Otherwise, either stick with the B-BB option (and check the section “Getting good solutions to nonconvex problems” in the GAMS/Bonmin manual) or use a solver for general MINLPs.

Stefan

Thank you all!

–
To post to this group, send email to gamsworld@googlegroups.com.
To unsubscribe from this group, send email to gamsworld+unsubscribe@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/gamsworld?hl=en.

\

3

Reply-to: gamsworld@googlegroups.com

Hi Stefan,

thanks for your answer.
I gave up with Bonmin B-BB, hopeless.

Formulated the problem for a MIP Solver (CBC) and now it works fine :slight_smile:
Solves in seconds…

Greetings


Am Montag, 19. November 2012 12:16:28 UTC+1 schrieb Stefan Vigerske:

Hi,

On 11/14/2012 08:13 PM, MrUnknown wrote:

Hello everybody,

I’m new to GAMS and this group and I hope to find some help here :slight_smile:

For my model, which I have to use in my bachelor thesis, I have to use a
minlp solver. So the only one, which is included in the basic GAMS version
is Bonmin.

There is also Couenne.
If you have an academic license, then there is also SCIP.

When I run my model with the B-Hyb algorithm, it solves very fast, but I
get a solver message: OA on non-convex constraint is very experimental.
When I change the used algorithm to B-BB it takes veeeerry long… So do
you have any suggestions for the option file?

Using the B-Hyb algorithm there is another mystery and I don’t really don’t
know whats wrong with it…
In my model I use two sets:

Sets
t time_steps / t1t24 /,
u units / u1u18 /
;

When I change t to 35 the model solves very fast, so does it at t = 37. But
when I use t=36 it takes “hours” and returns no feasible solution. So does
it with 47 (fast), 48 (slow), 49 (fast), 71 (fast), 72 (slow) and 73
(fast). So it can be said, when there is factor 12, it has problems solving
the model. Any ideas?

As the warning says, you have a nonconvex model.
Bonmin is made for convex models.
B-BB is an option that is most robust for nonconvex models.
For other algorithmic options, it can be random when Bonmin stops and
what result it gives, since the generate outer-approximation may not be
an outer-approximation in this case.

If you want to use Bonmin’s B-Hyb option, it would be best to
reformulate your model so that it is convex.
Otherwise, either stick with the B-BB option (and check the section
“Getting good solutions to nonconvex problems” in the GAMS/Bonmin
manual) or use a solver for general MINLPs.

Stefan

Thank you all!

–
To view this discussion on the web visit https://groups.google.com/d/msg/gamsworld/-/8Ihp0EC3LxIJ.
To post to this group, send email to gamsworld@googlegroups.com.
To unsubscribe from this group, send email to gamsworld+unsubscribe@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/gamsworld?hl=en.