Posts: 53
Threads: 21
Joined: Sep 2010
The stochastic
programming is limited to 50 stages and 64 States of the World by the predefined
sets J and ALLSOW. Is there a simple way I can extend the number of stages or
SOW? For example can I change the predefined sets in a specific TIMES model file?
I have also a question related to the
implementation of SPINES. I am wondering if the capacity-related parameters are
modeled with a none-anticipation constraint ( option 2) or not (option 1). See
description below. I do not think the model result will differ with the different approaches but perhaps the solving time for option 1 is faster than the solving
time for option 2.
Option 1:
It calculate the capacity related variables
separately, independen of SOW, and find the non-capacity related variables for each state of the
world.
Option 2:
It calculate all variables for all SOW, and
force the capacity related variables to be equal for all SOW.
Please let me know
if something is not clear or understandable!
Pernille
Posts: 392
Threads: 18
Joined: May 2010
13-03-2012, 06:17 PM
(This post was last modified: 13-03-2012, 06:28 PM by Antti-L.)
Your question about the predefined sets is quite valid, but remember that in MARKAL the number of stages was limited to 2 and the number or SOWs was limited to 9.
Although I have not tested it, I think you can easily extend the domains of the sets J and ALLSOW by defining them in the RUN file, with your own ranges. If I am not mistaken, the redefinition should work well if you put them after the call for initsys.mod, but before loading the DD files. For example, to set both sets up to 200 elements, you could try adding the following lines:
SET J / 1*200 /;
SET ALLSOW / 1*200 /;
I admit that I should have tested that myself before giving this advice. Let me know if it does not work for you (in that case you should get some compiler errors).
Concerning the other question, I am afraid that I don't quite understand what you are aiming at. In the experimental Spines option all the capacity-related variables have only a single SOW index in all stages.