How to avoid winner-take-all

[Antti-L]

I am sorry but I think it would be too time-consuming to try and investigate your issue with those three Excel files.

> However investment costs for the new technology TCANDSL1 for cases ITC1 and ITC05 were higher then base case but both had same value.

If I understand you correctly, the unexpected issue ("unexpected result was in investment costs") was that the investment costs for the new technology TCANDSL1 were higher in cases ITC1 and ITC05 than in the base case.  Is that correct?

If you would like me to investigate that issue, could you therefore please provide the reproducible cases:

  1) The *.DD and *.RUN files and the *.LST file for the base case.
  2) The *.DD and *.RUN files and the *.LST file for the ITC1 case.

With these files at hand, I would be able to quickly reproduce the issue and investigate it.

[Antti-L]

Ok, as there was no reply, I tried to look at it with the Excel files now anyway.
At first I thought it would be difficult to reproduce the issue, because there was no information about what the Base case should include.
However, it turned out that it was a very simple bug causing the unexpected issue: The Logit market sharing feature had been thus far tested only with a single market, and your example immediately disclosed that with multiple markets a bug was causing spurious interference between markets. This bug will be fixed in the next version.

Thank you nonetheless for trying this experimental feature and for reporting the issue, which was caused by a careless oversight in the implementation.

[Antti-L]

The bug discovered above has been fixed in TIMES v4.7.6, released in November 2023.

[vincedh]

Dear Antti,

I have also been testing the ECB extension on the demo 12 model.

This demo model has been slightly modified to further refine the mobility demand in REG1, first regroupping PUB (public transportation) and CAR demands, but also segmenting this demand into long distance/short distance and rural/dense demands.

I first implemented FLO_MARK for these new end-uses demands to control the contribution of PUB and CAR transportation to this demand. Everything seemed to work well.

Then, I created a new regular scenario file "ECB" where I relaxed the FLO_MARK(UP) to 1 and activated the Logit Market share feature for these demands. At first, I just added an heterogeneity parameter, letting the default value for other parameters. I know that Levelized Costs from PUB and CAR might be very different (a lot smaller for PUB than for CAR), so I would need to implement very small (close to 0) heterogeneity parameter to see a contribution from CAR. However, from 2010 onwards, PUB still takes the whole market, even for heterogeneity = 0.001 or 0.

Do you understand what happens here ?
Here is a link to download the .DD, .LST and .RUN files for the case where heterogeneity equals 0 : https://we.tl/t-1r1osatViU

Thanks a lot !

Vincedh

[Antti-L]

You have missed some requirements for levelized cost calculations:

  •  Levelized costs are by definition calculated over the lifetime of a technology, and thus must have a capacity.
  •  TIMES therefore requires that the technology must have capacity-related costs (investment and/or fixed costs) to qualify for the levelized cost calculation.
  •  TIMES additionally requires that the technology must be output-normalized (PCG is on the output side) to qualify for the levelized cost calculation.

I tested by adding small capacity costs (NCAP_COST=0.1) for the Logit market technologies, and changed the PCG to be DEMO instead of DEMI.  In addition, I defined a fixed NCAP_AFA(FX)=1 for your "dummy" demand technologies to make sure these technologies will also be utilized when they have capacity (because levelized costs cannot be calculated unless the capacity is being utilized).

After these fixes to your model, the logit market sharing algorithm worked as expected.

[vincedh]

Dear Antti,

Thank you for your help. I applied the correction you suggested for the dummy demand processes. I also increased the dummy penalty costs, and everything worked as expected!


From my understanding, however, the fact that each dummy demand process I created has two commodity outputs —to which I’m applying the heterogeneity parameter— results in two sets of constraints on NCAP for each technology. These constraints can potentially contradict each other.

This is because, as your explained in the documentation, the constraint applies to the new capacity installations of the processes that produce the commodities with "logit market shares"—not to the contribution of each process in generating the commodities.

If my goal is to assign different heterogeneity parameters to each end-use demand, would you recommend implementing new dummy processes, each with only one output commodity? That way, I could apply the logit market shares individually to the specific commodity output I intend. Or do you have a better solution ?


Many thanks,
Vincedh

[Antti-L]

> If my goal is to assign different heterogeneity parameters to each end-use demand, would you recommend implementing new dummy processes, each with only one output commodity?

Well yes, I guess so.  While I was testing I did such a case myself, using FLO_MARK instead of FLO_SHAR, but wasn't sure it was consistent with what you had in mind.