Veda2.0 Released!


Regional Attributes
#1
In a model that contains several regions, for the attribute Reg_obj the sum over all regions equals ObjZ - so far so good.
According to the Documentation part V p.61 the attribute Reg_ACost provides the 'Regional total annualized costs by period'.
However, for one region, the sum over all periods does not equal the one given in Reg_obj, when each periodic value is multiplied by the length of the periods as given in SysSettings.
Therefore, we thought this difference is somehow related to the discounting and manually calculated the discounting factor for each year via
Code:
(1+discount_rate)^(Year-Baseyear)
which deviates from what the attribute Time_NPV gives us.


Could somebody maybe clarify the relation between Reg_obj and  Reg_ACost and explain the sense of and how Time_NPV is being calculated to us?
Reply
#2
Reg_obj gives the total present value of all costs by region, in the discounting base year (G_DYEAR).
Reg_ACost gives you the sum of the annualized costs by region and cost type.
Time_NPV gives the present value of the time (in years) in each period, in the discounting base year (G_DYEAR).
The annualized costs are by default reported for each Milestone year, and therefore you cannot reconstruct Reg_obj from Reg_ACost. And why should you, because you already have Reg_obj reported?
However, if you use the following switches, you should, indeed, be able to reconstruct Reg_obj from Reg_ACost with high accuracy, by multiplying the sum of Reg_ACost with Time_NPV for each period, and summing together.

$SET OBLONG YES
$SET ANNCOST LEV

You are right that (1+discount_rate)^(Year-Baseyear) deviates from what the attribute Time_NPV gives, because Time_NPV gives the present value of each whole period, i.e. sum over all the years in each period.
Reply
#3
Thank you, Antti! TimeNPV is now completely clear to me, indeed really handy to have!

I did this whole comparison as a plausibilty check: The sum of the discounted annual costs for any region should equal the discounted costs of this region, right?!
It just ran a scenario with
    $SET OBLONG YES
    $SET ANNCOST LEV
and computed the scalar product of Reg_ACost and TimeNPV for each region as you suggested.
Still, l find huge regional differences (ranging between -70% and +212%), even though the sum over all regions matches with high accuracy, as you said.

Could this possible be related to this midyear discounting option?
Reply
#4
You should exclude IRE from the Reg_ACost components when doing that exercise.  
See if that helps.

IRE stands for the implied trade costs, and is thus not a true cost, and its sum is zero over regions.
Reply
#5
Ah yes, I see, now these values match at very good accuracy: The worst regional deviation is now <0.2%, which is a numerical effect I think.
Thank you again! Smile
Reply
#6
Dear Fabian,

Ok, thanks for confirming.
However, while <0.2% is already small, it is unexpectedly large.  For example, in the DEMO model the relative difference is in the order of 1e-10.

Although I am pretty sure your remaining differences are "cosmetic" ones, could you do me a favour, and post the relevant Reg_Wobj, Reg_ACost and Time_NPV data here for the worst region (like shown in the picture below)?
In addition, could you also post the listing file (*.LST, zipped) for the same run?  
This additional information might help me detect why you have such differences, and possibly suggest an improvement in the reporting routine.



Attached Files Thumbnail(s)
   
Reply
#7
Dear Antti,
of course!  I've attached you one xls file in your desired format and the other one, in which I initially found that the largest deviation is <0.2% (for region FR).
Additionally, you can find the .lst file in the zip archive.
Best!
Fabian


Attached Files
.zip   package.zip (Size: 69.71 KB / Downloads: 7)
Reply
#8
Thanks.  
On the basis of the data and the listing file, I was able to determine that the difference is probably a nuclear issue.

I found out that the decommissioning costs of existing plants are not reported consistently with the objective function: In the reporting routines these costs are reported on full cost basis, while the objective function adjusts these costs according to the lifetime falling under the model horizon, in the same way as for investment costs and when calculating the salvage values. And the regions where you had differences appear to be nuclear regions. Do you think I am right, and confirm having decommissioning costs modelled for existing nuclear facilities?

I am not sure which way should actually be preferred, but I do understand the convention adopted in the objective function, that the costs are accounted only according to the (discounted) proportion of the process lifetime falling within the model horizon (with the salvage value possibly further adjusted by NCAP_FDR).

Anyway, the difference is, indeed, cosmetic in the sense that the capital costs (including decomm.) for the existing plants are basically sunk costs that only appear as a constant term in the objective function, and will not cause any distortion when comparing the costs between scenarios. I think I would suggest adjusting the reported costs to be consistent with the convention in the objective function. Would you agree with that, or would you prefer seeing the full costs in both?
Reply
#9
(27-10-2017, 11:05 PM)Antti-L Wrote: I found out that the decommissioning costs of existing plants are not reported consistently with the objective function: In the reporting routines these costs are reported on full cost basis, while the objective function adjusts these costs according to the lifetime falling under the model horizon, in the same way as for investment costs and when calculating the salvage values. And the regions where you had differences appear to be nuclear regions. Do you think I am right, and confirm having decommissioning costs modelled for existing nuclear facilities?
Yes, you are totally right: All affected regions are indeed nuclear regions and I do have decommissioning costs modelled for nuclear plants.

(27-10-2017, 11:05 PM)Antti-L Wrote: I am not sure which way should actually be preferred, but I do understand the convention adopted in the objective function, that the costs are accounted only according to the (discounted) proportion of the process lifetime falling within the model horizon (with the salvage value possibly further adjusted by NCAP_FDR).

Anyway, the difference is, indeed, cosmetic in the sense that the capital costs (including decomm.) for the existing plants are basically sunk costs that only appear as a constant term in the objective function, and will not cause any distortion when comparing the costs between scenarios. I think I would suggest adjusting the reported costs to be consistent with the convention in the objective function. Would you agree with that, or would you prefer seeing the full costs in both?
In any case I agree completely on being consistent. That is (a) either put full costs into the objective function and also report them accordingly or (b) keep the convention that only the mentioned proportion within the horizon is being kept in the objective function but then also only report this proportion.
The more I think about it, the more I am convinced that your suggestion is sensible. The end user would probably not be interested in any costs beyond his/her chosen horizon I guess (since it is usually decades away from now (or base year) and either way highly uncertain) ...
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)