Unified SSP techno-economic data

[lecfab]

Purpose of this PR

It was hard to maintain different generisdata_tech files different SSPs:

• https://github.com/remindmodel/development_issues/issues/414

Major elements of this PR:

• go through all the technologies to choose how their inco0 (and other parameters) should change with SSP narratives. This google sheet shows the reasoning.
• turn all SSP-specific technoeconomics into in-code relative modifications and remove files generisdata_tech for SSP1, 3, 5
• rehabilitate cm_VRE_supply_assumptions: it affects inco0 and floorcost for storage as well as learning rate for storage and VRE:
  • 1 optimistic improves values by around 10%
  • 2 pessimistic worsens values by around 10%
  • 3 very pessimistic worsens values by around 30%

Minor elements:

• merge generisdata_flexibility into generisdata_tech
• replace incolearn by floorcost
• remove absolute in-code modification of inco0, learn etc
• rename the values of c_techAssumptScen to SSP1..5 instead of numbers without meaning
• add gasftcrec to teCCS as it was missing
• forbids CCS technologies to be cheaper than non-CCS equivalent (using new set te2teCCS)
• document https://github.com/remindmodel/development_issues/issues/486#event-16297434367 (side modification)
• remove old code involving ccscomp, ccspipe etc
• clean stuff; please review in "Hide whitespace mode":

Type of change

• Bug fix
• Refactoring
• New feature

Checklist:

• My code follows the coding etiquette
• I performed a self-review of my own code
• I explained my changes within the PR, particularly in hard-to-understand areas
• I checked that the in-code documentation is up-to-date
• I adjusted the reporting in remind2 where it was needed
• I adjusted forbiddenColumnNames in readCheckScenarioConfig.R in case the PR leads to deprecated switches
• I checked the log.txt file of my runs for newly introduced summation, fixing or variable name errors
• All automated model tests pass, executed after my final commit (FAIL 0 in the output of make test)
• The changelog CHANGELOG.md has been updated correctly

Comparison of results

Compare scenarios are here: /p/tmp/fabricel/202502-remind-learningDoc.
The former techno-economic assumptions had a lot of bugs due to missed update etc. Hence significant differences after the update:

• SSP1: more gas, less biomass and hydrogen
• SSP2: few differences, less wind
• SSP3: gas replaces coal, lower emissions
• SSP5: less coal, rerouting of gas, lower emissions

SSP1: more gas, less biomass and hydrogen

SSP2: few differences, less wind

SSP3: gas replaces coal, lower emissions

SSP5: less coal, rerouting of gas, lower emissions

[tabeado]

thanks a lot for accepting to keep the ccsinjeCost switch as it was and improving the description!

[cchrisgong]

Thanks a lot!
Why is there a lot of coal to liquid without CC in SSA under SSP3? There is less after the change but rather strange...

[lecfab]

Thanks a lot! Why is there a lot of coal to liquid without CC in SSA under SSP3? There is less after the change but rather strange...

Indeed, it's a separate known issue, but this PR can influence it because the Fischer Tropsch costs changed significantly.

• https://github.com/remindmodel/development_issues/issues/433

[robertpietzcker]

did you have a specific reason to merge the flexibility file with generisdata_tech?

I would prefer to keep them separate, as the stuff in generisdata_tech is really fundamental techno-economic properties, while the values in _flexibility are hand-tuned parameters to create reasonable technology mixes for a variety of power system configurations.

so if you don't have a very strong preference, I would propose to keep it split, @lecfab

[lecfab]

Sorry Robert I had merged already as I thought you had finished commenting.

For flexibility data, I found it clearer to add it into the general tech file, as it did not add too much clutter there. I have to admit that I still don't fully understand this parameter.
So it's perfectly ok with me if you re-separate them, and maybe add some documentation to the file?

[robertpietzcker]

wow, what a great effort to make the model more systematic, @lecfab - thanks a lot!

I have added two specific comments below with potential minor changes;
there is one further issue: do you have an idea why there are changes in SSP2 results in the power sector?
from what I see, you didn't implement any changes to input data in the power sector in SSP2, only in SSP1/SSP3?

[lecfab]

why there are changes in SSP2 results in the power sector?
from what I see, you didn't implement any changes to input data in the power sector in SSP2, only in SSP1/SSP3?

That's a puzzle for me too, and I started digging into it: it could come from using floorcost instead of incolearn.
At some point there is a formula $$inco0' = \alpha * inco0$$ where $$\alpha = 1 + tkpremused$$.

• Formerly, incolearn was fixed, so we had $$floorcost = inco0' - incolearn = \alpha * inco0 - incolearn$$
• Now, floorcost is fixed using the former value of incolearn: $$floorcost = inco0 - incolearn$$

So this $$\alpha$$ could be responsible for different dynamics for learning technologies in the new version. However, I have a feeling it should actually increase their deployment. So this needs further investigation!