Veda2.0 Released!


Breaking down objective...
#1
Hi there,

I want to break down objective variable by components and periods. I thought it is easy question, but cannot find where I'm wrong.

Here I'm attaching an example (uploads/51/VedaExport_MultTabs_68509.zip) with every year milestone years for simplicity, and costs components exported from VEDA-BE: Cost_Act, Cost_Flo, Cost_Fom, Cost_Inv, Cost_Salv.

As I understand,
Cost_Act + Cost_Flo should give OBJVAR component,
Cost_Fom, Cost_Inv, and Cost_Salv should be OBJFIX, OBJINV, and OBJSAL respectively,
in the objective equation:

EQ_OBJ..  VAR_OBJ(REG,OBJINV,CUR) + VAR_OBJ(REG,OBJFIX,CUR) - VAR_OBJ(REG,OBJSAL,CUR) + VAR_OBJ(REG,OBJVAR,CUR) - OBJz =E=  0 ; (LHS = 0)

But discounting and summing doesn't provide value of OBJz.

Did I missed something?

Thanks,
Oleg


Attached Files
.zip   VedaExport_MultTabs_68509.zip (Size: 12.01 KB / Downloads: 4)
Reply
#2

It is difficult to comment on this reconstruction without more detailed information.

But I can see strange drops in the activity and fixed costs after 2007. I suspect that you have modeled PRC_RESID capacities in some peculiar way, which is causing the reported fixed costs drop suddenly after the Base Year in the annual cost reporting, but are accounted differently in the objective function.  However, note that the fixed costs for the RESIDs are only a constant term in the objective.
For the RESID capacities, the reporting of the investment and fixed cost is only rudimentary, because these costs cannot be "correctly" reported for the residual capacities anyway, and the RESID capacities are not even supported by the Standard version of TIMES (but have been added into the VDA extension of TIMES). Without seeing the model I cannot be sure what is causing the difference in your calculation, but can only suspect that the issue is related to the reporting of RESID capacity costs. In addition, note that the salvage costs should not be subtracted from the sum of the discounted annual investment costs (because the annual costs are reported only for the years within the model horizon).
You should be able to easily identify the differences by main objective component, by comparing your reconstructed figures to the reported objective component values (Reg_wobj, which sum up to the Objz value).
Reply
#3
Thanks Antti,

The drop of fixed costs is a result of technologies which don't have demand, but exist in the base year (I cut one industry from a larger model). There are no INVCOST for RESID capacities here, but FIXOM and VAROM are there. As I've found using Reg_wobj, the divergence goes from FIXOM. I understand that fixed costs are constant for RESID capacities, and don't affect solution. But do I understand correctly that fixed costs are calculated in different ways for objective and for reporting in VEDA-BE? Are the any rational reason for that? Thanks,
Reply
#4

Thanks for verifying the difference by objective component. It seems that I guessed well, and the difference was due to the fixed costs for the RESID capacities.

Yes, you have understood correctly: The investment and fixed costs for the RESID capacities are accounted in the objective function differently from how they are reported. Although this difference does not affect the solution in any way, I admit that it may be considered undesirable, if you want to reconstruct the objective function from the reported annual cost results. There is no rational reason for the difference; the reason is essentially a historical one. Because the RESID capacities are not supported by the Standard TIMES code, the investment and fixed costs are accounted in the objective function in the standard straightforward manner, which is inconsistent with the user-defined RESID capacity evolution. But because these costs represent just a constant term in the objective, this was originally not considered any problem. On the other hand, the reported annual costs do take into account the RESID capacity evolution.

The accounting in the objective function could, nonetheless, be revised by intoducing in the objective function a special handling for the RESIDs, but doing so might also cause a lot of confusion, because for most models the value of the objective function would then change, perhaps even considerably, and users might get worried about what has been changed.

I am sorry about this discrepancy, even if it has otherwise no impact on the model solution.  I will discuss the matter with Amit, to see if we can do something about it.

Reply
#5
Thanks for the clarification.

Best,
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)