When solve Stackelberg problems with MPEC type, shows 1 error that "571 rows 599 columns 12094 non-zeroes 13372 nl-code 9947 nl-non-zeros "

[lkc...@163.com]

Help:

When I using GAMS to solve an Economic phenomenon based on Stacklberg Modle, once I operate the system ,some error occurs, the information shows in the below picture:

  Below is the whole program:

set i (*)

    j (*);

$load i j

parameters

    t(i,j) unit transport cost

    c(j) unit product cost

    sl(i) supply limit

    da(j) demand constant a

    db(j) demand constant b

    dl(j) demand limit;

$load t c sl da db dl

$ gdxin

display i,j,t,c,sl,da,db,dl;

set leader(i) /6/;

positive variables

    ps(i) supply price

    pd(j) demand price

    x(i,j) traffic volume between supply i to demand j;

variable benefitleader leader's benefit;

equations

    profit_leader leader's profit

    profit_follwer(i,j) follower's profit

    sb(i) supply balance

    deb(j) demand balance;

alias (i,ii);

x.lo(i,j)=0.01;

deb(j).. sum(i,x(i,j))=g=dl(j);

sb(i).. sl(i)=g=sum(j,x(i,j));

profit_follwer(i,j)$(not leader(i)).. t(i,j)+c(j)+ps(i)-pd(j)-exp(da(j))*sum(ii,x(ii,j))**db(j)=g=0;

profit_leader.. benefitleader =e=sum((i,j)$leader(i),exp(da(j))*sum(ii,x(i,j))**db(j)*x(i,j)-c(j)*x(i,j)-t(i,j)*x(i,j));

model sitanboge /sb.ps,deb.pd,profit_follwer.x,profit_leader/;

solve sitanboge using mpec max benefitleader;

option NLP=conopt;

display x.l,ps.l,pd.l;

Parameter report1(i,*) supply summary report

           report2(j,*) demand summary report ;

 report1(i,"supply") = sum(j,x.l(i,j)); report1(i,"capacity") = sl(i); report1(i,"prices") = ps.l(i);

 report2(j,"demandl") = dl(j);report2(j,"demand")=sum(i,x.l(i,j));  report2(j,"priced") = pd.l(j);

Display  report1,report2, x.l;

Sincerely ask for a solution to solve this problem !!!! 

Many Many Thanks !!!!!!

[Renger van Nieuwkoop]

Hi ??

Without the possibility of running your code (as you didn't send the input files) or at least information from the listing file with the exact error message, I must have superpowers to give you an answer (but perhaps others are better in guessing ).

Cheers

Renger 

Sent from my Samsung Note 4

--
You received this message because you are subscribed to the Google Groups "gamsworld" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gamsworld+...@googlegroups.com.
To post to this group, send email to gams...@googlegroups.com.
Visit this group at https://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/d/optout.

[lkc...@163.com]

Hi Renger;

Thanks for your reply!!

Attachment is my input file, looking forwarding your reply !

[Renger van Nieuwkoop]

But then you should also send the code for reading the xlsx. Not that it is much work for me, but if the error is there, I won’t be able to reproduce your error.

Renger

[lk]

Hi Renger:

Sorry. It's my fault, here are the code for reading the xlsx.

$call GDXXRW.EXE sitanboge.xlsx dset=i rng=t!a2 rdim=1 dset=j rng=t!b1 cdim=1 par=sl rng=sl!a2 rdim=1 par=t rng=t! rdim=1 cdim=1 par=da rng=da!a2 rdim=1 par=db rng=db!a2 rdim=1 par=c rng=c!a2 rdim=1 par=dl rng=dl!a2 rdim=1

$gdxin sitanboge.gdx

  

Thanks !

[Renger van Nieuwkoop]

Hi

I had to change the following line to:

set leader(i) /6/; -> set leader(i) /s6/;

After that, I get the following error:

**** MCP pair sb.ps has unmatched equation

     sb(S6)

 

This means that there is one equation missing or one variable too much. (if you fix one of the prices, e.g., PS.FX(“s6”) = 1 the model starts solving but is infeasible.

Try to check your model formulation.

Additional material can be found in the GAMS model library (nash.gms has a Stackelberg) and Foundations of Network Optimization and Games, Chapter 10, by Friesz and Bernstein

Cheers

[lk]

Hi Renger:

Thanks so much, I have solved this problem. I deleted "$not leader(i)" in the "profit_follower" equation. But now aonther prolem is "infeasible solution: reduced gradient less than tolerance " ,I don't know where is going wrong...

please HELP!!

Again thank you very much !!!!! 

[Renger van Nieuwkoop]

THat would take too much of my time.

Sorry

[lk]

and it says infeassible solution..

[lk]

Hi Renger:

It's ok, thanks anyway.

LK