Veda2.0 Released!


Empty Group in FLO_SUM/FLO_FUNC/FLO_SHAR
#16
The setting $SET RETIRE YES alone is not sufficient for enabling early retirements.
You must activate retirements on a technology basis. For that purpose, you can use either PRC_RCAP(r,p) or RCAP_BND(r,y,p,lim).
Reply
#17
Thanks Antti for your help. Sorry for not thinking about this before writing my message.
nevertheless, I should admit that it wasn't easy for me to get familiar with the design of RCAP_BND. Anyways, I set RCAP_BND~LO (2005) = 0 with 15 as I/E rule (which forces the technologies to retire). The counterparty here is that the objective function becomes higher (which is normal since I am adding more constraints). 

I am now running the model between 2010 and 2020 using different pdefs. the major result is that even when allowing dummy imports, the only pdef for which the model is feasible has the following milestone years: 2010, 2012, 2016, 2020. In this configuration, the model uses almost 70 PJ of a postive dummy variable to satisfy a constraint on the minimum conventional power capacity dring the peak, among other dummies. This occurs only in 2016 (please find attached a screenshot of dummy imports for this study case).

When specifying the following milestone years: 2010, 2011, 2015, 2020, the model becomes infeasible (even with dummy imports). The start year of most new capacities is 2011. I don't know in what manner this could impact the feasibility of the model. 

Do you have any advice for me at this situation as I appear to be all out of ideas ? 
Best regards, 

Mahmoud


Attached Files Thumbnail(s)
   
Reply
#18
I thought your demolition rate constraints would just need allowing early retirements, not forcing any retirements. To allow early retirements for a technology, you need to have either PRC_RCAP(r,p) or any RCAP_BND(r,y,p,lim) defined for it.  Putting just some interpolation option would be sufficient, e.g. RCAP_BND(r,'0',p,'LO')=1.

Anyway, allowing early retirements would only make the model constraints more relaxed, and so the objective function (total discounted costs) value should then decrease, and not become higher.  It should make your demolition rate constraints feasible, and eliminate those earlier infeasibilities.  But if your model is still infeasible, I don't see how that could be related to allowing those early retirements. Are you nonetheless saying that allowing retirements makes the model infeasible, or increases the objective value?

In general, if you have an infeasible model, it indicates that you have made some modeling errors, which should be fixed. If infeasiblities occur even when using dummy imports, you can get a report of the model infeasibilities from Cplex, showing the full names of the related constraint and variable names by removing the setting names no from Cplex.opt. Then one should just correct the model accordingly. I am afraid that it is very difficult for anyone else to analyse what is going on in your model without any good knowledge of the model, or even having access to it.
Reply
#19
Big Grin 
Update: 

I switched to RCAP_BND(r,0 , p , "LO") = 1 to allow early retirement 

I finally found the main problem in my model: I had some dummy variables (LowAF_powerplant_type) without any CAPEX/OPEX and for which the following constraint is imposed: 
VAR_CAP(r, 2015, LowAF_power_plant_type) = VAR_CAP(r, 2015, power_plant_type) with 5 as I/E rule. The I/E is the problem since (I don't know why) the model installs very big capacity of the process LowAF_powerplant_type early in 2005 and has to retire it as by 2015. I changed the I/E rule to 3 on this UC and the model is back to optimality [Image: biggrin.png]

It is notable that that the computation time is way lower now. 

I also eliminated a major part of dummy imports (I still have very little flows of dummies into space heating but no problem with this since those flows represent less than 0.1% of the commodity production in each region / year). 



I would like to thank you Antti for your priceless help. 
If by any chance Evangelos Panos sees this thread: thank you for this useful presentation on how to play with cplex parameters https://fr.slideshare.net/IEA-ETSAP/impr...lexbarrier


Attached Files Thumbnail(s)
   
Reply
#20
Music 
Nice to hear that you were able to make some good progress!  Shy
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)