Model size and number of input years

[Pernille.S]

We are currently developing a regional TIMES from a short term (2020) to a long term (2070) model. We have seen that the size of the model decrease if we add an additional input year. Does anyone know or have experience with how the number of input & output years affects the size of the model?

Thank you!

 Pernille

 

[Antti-L]

I am not sure what you mean by input and output years.  I don't think such terminology is used in the TIMES documentation.

Surely, if you extend the model horizon from 2020 to 2070 by adding more Milestone Years into the model, the model size (number of equations, variables and non-zeros) should certainly increase.  But if you just add more Data Years for the input data, that should have little (if any) impact on model size.

However, if you are nonetheless seeing that the model size decreases even if you are adding more milestone years, I would find that strange and would be interested to know why that would happen. Maybe you could in that case send me the corresponding GAMS listing files for some investigation?

[Pernille.S]

Perhaps this is only related to Answer-TIMES, see the figure below. Input years = data time periods, Output years = result time periods. 

The model is slightly smaller when we include 2040 as a data time period compared to when we leave out 2040.  The number of milestone years was constant when we tested this.

In general, will the model size decrease if the number of data time periods increase (within the same model horizon)?

Pernille

[Antti-L]

Ok,  Input years = Data Years and Output years = Milestone Years.

In ANSWER-TIMES, if you only add some new Data Year(s), but don't change the Result years (=Milestone Years), the new Data Years should not have an impact on the model size, unless you also actually add some data for the new Data Year(s). But if you add, for example, some zero bounds for a new Data Year, the model size could be reduced, because the new data point having the zero bound might make the process unavailable in a model period where it was earlier considered available. Consequently, a slightly reduced model size might, indeed, be resulting in such cases, like you reported.

However, if you leave the new Data Year column totally blank in the ANSWER database, then there should be no impact at all on the model size, and the solution should be identical.

Hence, I would be curious to know whether you are seeing the model size reducing even if you did not specify any data for the new Data Year? Unless I am missing something, that should not occur.