I was trying out one of the 2017 advanced demos on the TIMES-Macro-MSA implementation. (starting from the link I found in another post on the forum: https://forum.kanors-emr.org/showthread.php?tid=299&highlight=macro)
I copied all the files from the demo, synced and did the following:
1) Ran with the CSA switch and the SMA_parameters scenario file activated
2) Ran with the MSA switch and the SMA_parameters scenario file deactivated
However, I did not manage to reproduce the results, I am getting errors in the MSA run.
Two things I noticed:
1) The contents of the MSADDF.DD file look weird to me: some parameter value equal 'UNDF'. It seems only the last period is interpreted correctly. (see text below)
2) I get a similar looking MSADDF.DD file when comparing in VEDA 1 and VEDA 2 (including UNDF values) but the PARAMETER TM_DDF(R,ALLYEAR,C) is empty in VEDA 2. (and UNDF in VEDA 1) I suppose this could be due to the demo model version that is not compatible with VEDA 2? Or is the TIMES-Macro-MSA feature not yet operational in VEDA 2?
I just tested that DemoS_Adv (I found a VEDA2 version of it in my archives), with the CSA switch (MSA calibration), and got no such problem. The convergence of the algorithm was also quite good. The run completed normally and MSADDF.DD looked fine to me. So, the reasons to the issues you see remain undisclosed to me.
13-12-2024, 02:54 PM (This post was last modified: 13-12-2024, 02:56 PM by Antti-L.)
The listing file (*.LST) could disclose some useful information about your issue, if the problem persists. One can include such files to posts here as attachments (zipped).
(13-12-2024, 02:54 PM)Antti-L Wrote: The listing file (*.LST) could disclose some useful information about your issue, if the problem persists. One can include such files to posts here as attachments (zipped).
Ok, the listing file shows that Conopt finds the calibration model infeasible in the very first round.
** Infeasible solution. The Preprocessor has determined that the model is infeasible.
Such does not happen in my run, and so your model is in some way behaving badly. I can see that the initial LP run nonetheless solves to optimality:
**** MODEL STATUS 1 Optimal
**** OBJECTIVE VALUE 499835.4619
However, that objective value is vastly different from what I am getting:
**** MODEL STATUS 1 Optimal
**** OBJECTIVE VALUE 25302357.9221
This latter objective value is neither identical to the old documented value for this case, but it is quite close to it. Anyway, your objective value for the initial LP is only 2% of the old documented value, which indicates some major problem in your model. Maybe it has indeed not been converted to VEDA2 correctly. Perhaps Amit could tell where the correct VEDA2 version is available (I cannot recall where I found a converted version, it was back in 2020).
Hi! Thank you for your support so far, I am getting closer to figuring this out!
Small update and a new question:
I retried the procedure on another version of one of the demo models, as I no longer trust the version I was using.
I used the demo model DemoS_007 and included the SMA_parameters scenario file (adapted for the new period definition).
The MDADDF.DD file now looks good after the calibration step! (all values are there, and nothing is UNDF)
So it seems like the demo version I was using was indeed somehow faulty.
But, now a new issue arises in the MSA step... I can see the following in the *msa.run file:
*** Error 282 in D:\VEDA2\GAMS_SrcTIMES.v4.6.8\preppm.msa Unable to open include file Hint: Does the filename need to be quoted?
I attached the *csa.lst and *msa.lst files here again.
It shows a similar message for the MSA step.
**** LINE 96 INPUT D:\VEDA2\Gams_Wrk_TIMESxMacro\DemoS_007-MSA\demos_007-msa.RUN **** 282 Unable to open include file **** Hint: Does the filename need to be quoted?
I guess so. VEDA2 tends to create a separate working folder for all cases, and then the MSADDF.DD created by the CSA (calibration) run may not be found when running in the MSA folder.
Try copying it from the CSA folder into the MSA run folder, and see if that helps.
(16-12-2024, 05:57 PM)Antti-L Wrote: >Do you know what could cause this error?
I guess so. VEDA2 tends to create a separate working folder for all cases, and then the MSADDF.DD created by the CSA (calibration) run may not be found when running in the MSA folder.
Try copying it from the CSA folder into the MSA run folder, and see if that helps.
