Setting up high and low scenario files

[james]

Hi

Query regarding setting up high and low scenarios using a * multiplier to scale investment costs for a particular technology. So the high might scale investment costs by 10% up (*1.1) while the low might scale them down by 10% (*0.9). I expected this functionality to be dependent on what is set in the Case Manager, i.e. if the high scenario is selected only it should scale up the investment cost in the SubRES by 10%. However, it seems that the scaling is driven by what scenarios are present in the scenario window of the Navigator. So for instance if both the low and high scenario are in that window, the cost coming out of the SubRES file is first scaled down and then scaled up from that lower value. That is you don't get the 10% up scale relative to the base value in the SubRES that you might expect. Essentially the calculation chain for the high case appears to be:

SubRes -> low -> high -> to .dd

when you want:

SubRes -> high -> to .dd

As you can't change the relative order of the scenario files in the navigator and this happens if only the high case is selected in the case manager (low is not ticked), I wanted to check if this was intended?

Thanks

James

[AKanudia]

Thanks for raising this issue James; I am sure other users must have encountered this issue as well.

The rule at this point is that if there are multiple candidates, the UPD tag will pick the (alphabetically) last scenario that is alphabetically inferior to the scenario where this tag exists. For example, if a particular parameter exists in scenarios S1, and S3, then a UPD table in a scenario S2 will use S1 value as the seed. So, the simplest but cumbersome workaround would be to name your scenarios in a sequence and import them in reverse order, one at a time. The risk here is that if you ever import one of these scenarios out of sequence, then the values will get messed up.

A better alternative is to use the new "fast fill" method to get the seed values you want in the scenario file and make INS tables based on them.

[cyang]

Two questions about this.  I'm also trying to do some sensitivity analysis and want to understand this behavior more.

1) Is there a dd file or somewhere we can actually see what the specific value that will be used in the model run will be?  My example isn't using a multiplier for the UPD value but rather just another value to REPLACE the value.  For example I might have vehicle service demand specified in the base year file and another value specified in a scenario.  I can see both of these values in the dd files one in the base.dd file and the other in the scenario.dd file.  Is there some way to see/verify which value is chosen?

2) What if I have multiple scenario files with UPD tables and they are conflicting with one another?  Let's say one scenario file specifies the price of an electric vehicle at one value and another scenario file specified the price at another value.  If they are both selected in the Case Manager, which one will be used in the model run?

[AKanudia]

Short answer - the last one that GAMS encounters from all the DD files.

First thing that can help you here is the Browse facility of VEDA-FE. Have you used it to investigate the issue at hand? Like I have pointed out several times, it is absolutely critical to visualize your model with Browse/Process info/RES features of VEDA and not through Excel files. This is even more important in situations when people other than you make changes or have made changes.

To find out exactly what values go into a particular run, you can look at the GDX file that is created from these steps.

The problem will become even more interesting if you inherit values for different years from different scenarios. GAMS will see a *merged* timeseries in this case. If you want VEDA to ignore values for *any* year that come from previous (case manager-wise) scenarios, then you can put a 0 in Year2 index of the parameter. The easiest way to do this is to specify the year as <year>-0. "2015-0", for example.

[cyang]

Amit,

Thanks for the helpful hints.  Yes, I do use the Browse facility and more commonly the Commodity and PRocess Master to view the RES and the imported data parameters.  But this doesn't tell me which value will be used in a model run if there are conflicting scenarios chosen.  But as a general rule, yes, it is a helpful thing to use.

It looks like your answer that "the last one that GAMS encounters from all the DD files" means that a scenario that is later in the case manager will be the value chosen, in the case of conflicting scenarios, and presumably the order of dd files listed in the .RUN file.  Correct?

I'm trying to follow the instructions in the link you provided to look at the GDX file, by adding

$SET INTEXT_ONLY YES

$SET PREP_ANS YES

above the $BATINCLUDE calls

But when I do this I get an error when I try to run the file (using VTRUN) and the model run terminates.  When I remove these two lines, the model run is able to continue.

Any tips for solving this issue would be appreciated.  

Chris

[AKanudia]

Yes, correct. Order of scenarios on the case manager is very important for this reason. Of course, this is exactly the order in which the DD files are listed in the RUN file.

Sorry, I did not realize that I had pointed you to the manual instructions for creating the data-only GDX. A feature was added for this: check "Create data-only GDX" on the general tab of user options. After this, choosing "create DD only" on the case manager will generate the data-only GDX file (in addition to creating the DD files).

[cyang]

Thanks Amit for the clarification.  This is helpful as I am doing some automated sensitivity runs.

And looking at the GDX file is quite useful just to verify the values that I think are being used are actually being used.