Code for Numerical Instability question

GAMS
1

Hello.
My model becomes unstable with the inclusion of just one more
equation. I thought it could be some kind of numerical instability
because the new variable does not feed back into the calculation of
endogenous variables.

I upload the code titled ExtraEquation in the Files section.
The new equation is named as EXTRAEQUATION (lines 449 and 601) that
calculates an vEXTRAVARIABLE

When they are included in the SOLVE command, the error message is as
follows:

MODEL SimpleModel OBJECTIVE vCumCost
TYPE NLP DIRECTION MINIMIZE
SOLVER CONOPT FROM LINE 688

**** SOLVER STATUS 1 NORMAL COMPLETION
**** MODEL STATUS 5 LOCALLY INFEASIBLE
**** OBJECTIVE VALUE 614642434.0384

=====================================================================

When the EXTRAEQUATION is commented out, the result is as follows.

MODEL SimpleModel OBJECTIVE vCumCost
TYPE NLP DIRECTION MINIMIZE
SOLVER CONOPT FROM LINE 688

**** SOLVER STATUS 1 NORMAL COMPLETION
**** MODEL STATUS 2 LOCALLY OPTIMAL
**** OBJECTIVE VALUE 743128421.4200

The EXTRAVARIABLE will be used in the objective function that will be
minimized. I am having problems already.

Thank you.
Subbu

\

2

Subbu,

Conopt declares the model infeasible but if you look closely, it gives
very good hints:

The log reads:

** Infeasible solution. A variable has reached ‘Infinity’.
Largest legal value (Rtmaxv) is 1.00E+10

In the status file which gets party copied into the LST file you find:

** Warning ** The variance of the derivatives in the initial
point is large (= 10. ). A better initial
point, a better scaling, or better bounds on the
variables will probably help the optimization.

vExtraVariable(z4,q80): The variable has reached ‘infinity’

** Infeasible solution. A variable has reached ‘Infinity’.
Largest legal value (Rtmaxv) is 1.00E+10

The allowable range can be changed with option:

Rtmaxv=x.xxe+xx

So after I used a Conopt option file with rtvmax 1e20 the model solved
fine again. You should take the first warning of Conopt very
seriously. Your model is not well scaled and you don’t try to set a
good starting point. Spend some time on this. The GAMS/Conopt manual
has very good section on scaling and starting points. I suggest you
study that.

Hope this helps,
Michael Bussieck - GAMSWorld Coordinator



On Jun 16, 10:20 pm, whoami wrote:

Hello.
My model becomes unstable with the inclusion of just one more
equation. I thought it could be some kind of numerical instability
because the new variable does not feed back into the calculation of
endogenous variables.

I upload the code titled ExtraEquation in the Files section.
The new equation is named as EXTRAEQUATION (lines 449 and 601) that
calculates an vEXTRAVARIABLE

When they are included in the SOLVE command, the error message is as
follows:

MODEL   SimpleModel         OBJECTIVE  vCumCost
 TYPE    NLP                 DIRECTION  MINIMIZE
 SOLVER  CONOPT              FROM LINE  688

**** SOLVER STATUS 1 NORMAL COMPLETION
**** MODEL STATUS 5 LOCALLY INFEASIBLE
**** OBJECTIVE VALUE 614642434.0384

=====================================================================

When the EXTRAEQUATION is commented out, the result is as follows.

 MODEL   SimpleModel         OBJECTIVE  vCumCost
 TYPE    NLP                 DIRECTION  MINIMIZE
 SOLVER  CONOPT              FROM LINE  688

**** SOLVER STATUS 1 NORMAL COMPLETION
**** MODEL STATUS 2 LOCALLY OPTIMAL
**** OBJECTIVE VALUE 743128421.4200

The EXTRAVARIABLE will be used in the objective function that will be
minimized. I am having problems already.

Thank you.
Subbu

\

3

Thank you to Mike and Arne.
Your suggestions and pointers are very helpful.

Subbu


\