QA_CHECK Errors

[Tanzeel]

I couldn't understand the "FLO_SHAREs" defective sum even in VEDA Template, the numbers are exactly to 1 for dual cars (gasoline and CNG). Moreover, sometimes, having the same errors in Log file, I get successful run and often it crashes. 

[Antti-L]

Well, the tolerance for reporting these inconsistencies is quite small (9.0E-5), and so the sums may appear exactly 1 in Excel.

If the runs are "crashing", I don't think that is related to these small share violations, because they are relaxed by TIMES.

[Tanzeel]

Well, the tolerance for reporting these inconsistencies is quite small (9.0E-5), and so the sums may appear exactly 1 in Excel.

If the runs are "crashing", I don't think that is related to these small share violations, because they are relaxed by TIMES.

Good....What is the effective way to validate the results? Hence, it's somehow loose question but not sure with different GAMS source Code (V331 to V391) that provide the different results of the same model. Please refer any useful documentation in this regard. I try to debug errors by seeing LST and GDX files but yet not smart enough.

[Antti-L]

Over the past years, there have been changes in the interplay between VEDA-FE and TIMES, and therefore you cannot use TIMES V331 with VEDA versions from recent years, or TIMES V391 with very old VEDA versions. One can always recommend to use the latest versions of both (unless you can identify some problem in either, which you should then report).

Concerning TIMES on a more general perspective, all versions have been tested before releasing them, and any discrepancies in the results between successive versions should be caused by either small numerical differences (caused e.g. by slightly different order of the computations leading to different rounding errors), or by bug fixes. I think that the most recent version, TIMES v4.0.1 is very robust. To my knowledge, no bugs have been reported in the past six months. Both the mathematical formulation and the implementation in GAMS have been thoroughly reviewed over the past ten years.

Concerning the validation of a specific TIMES model, I think you can thus be quite confident in the underlying mathematics, and concentrate on the quality of model data and fine-tuning the behaviour of the model, with both statistical data and by comparing the results to those from other models.