Refactoring of 45_carbonprice incl. postsolve algorithms

laurinks commented 3 weeks ago

Purpose of this PR

Two main purposes of this PR:

Add further flexibility for carbon pricing in peak budget runs (regional carbon price differentiation, regional flexibility for convergence end year, regionally flexible interpolation between carbon prices from NPi run and peak budget run to avoid large jumps, setting carbon prices from NPi as lower bound) (see also Issue#295)
Refactoring of 45_carbonprice. Mainly: (a) moving postsolve algorithms to 45_carbonprice, (b) deleting unused postsolve algorithms, (c) combine diffLin2Lin and diffExp2Lin in one realization called "functionalForm" (see also Issue#295)

Type of change

[x] Refactoring
[x] New feature
[x] Minor change (default scenarios show only small differences)
[x] This change requires a documentation update

Checklist:

[x] My code follows the coding etiquette
[x] I performed a self-review of my own code
[x] I explained my changes within the PR, particularly in hard-to-understand areas
[x] I checked that the in-code documentation is up-to-date
[ ] I adjusted the reporting in remind2 where it was needed
[x] I adjusted forbiddenColumnNames in readCheckScenarioConfig.R in case the PR leads to deprecated switches
[x] All automated model tests pass (FAIL 0 in the output of make test)
[x] The changelog CHANGELOG.md has been updated correctly

Further information:

Default REMIND runs are only affected in one way, namely that carbon price trajectory from path_gdx_ref is used as lower bound for carbon price trajectory in current run (e.g. PkBudg run cannot have lower carbon price than NPi run). This is introduced due to relatively high CO2 price for EUR in 2030 which could be overwritten by lower values in climate policy scenarios such as PkBudg1000).
Decisions on default settings need to be taken based on NPi2025 realization and validation.
Test runs showcasing functionalities added in realization functionalForm are in folder p/tmp/laurinko/hpc/remind (started on 18.11.2024), and config file used for testing is attached. scenario_config_functionalForm.csv
Screenshots from compareScenario PDFs: Example showing effect of lower bound (from path_gdx_ref which is NPi2025 in this case) in SSP2-EcBudg400: SSP2-PkBudg650 experiments: All behave as prescribed in scenario_config_functionalForm.csv. Examples are linear increase for EUR until 2060 in h, detailed convergence of SSA in e, linear curve going through the point (2010,0) in c SSP2-PkBudg1000 experiments: All behave as prescribed in scenario_config_functionalForm.csv. Lower bound for EUR well visible in all plots (except i, where is is deactivated) Examples are linear increase for EUR until 2060 in h, detailed convergence of SSA in e, linear curve going through the point (2010,0) in c

laurinks commented 1 week ago

Process to merge this PR:

[x] Wait for NPi2025 PR #1888 of Rahel to be merged.
[x] Make adjustments to scenario_config.csv so that it includes NPi2020 and NPi2025 scenario for each SSP (with calibration on NPi2020, and no cascade, i.e. NPi2025 starts also in 2005). Also adjust CO2 price in start year as needed.
[x] Merge this PR

@LaviniaBaumstark , @robertpietzcker , @Renato-Rodrigues : The adjustment of the scenario_config.csv contains the changes to include SSP2-NPi2025 (and same for other SSPs) and fix PkBudg runs to NPi2025. Please have a close look if this is as intended (no cascade, i.e. NPi2025 starting in 2005, path_gdx_policyref still set to old NPi). Should be rename the old NPi to NPi2020 to avoid confusion? @LaviniaBaumstark : It also contains the changes of PR #1884 which had gotten lost in a previous merge.

laurinks commented 1 week ago

@LaviniaBaumstark, @robertpietzcker: I just had a look at the updated scenario_config.csv (of this PR) with Lavinia. It includes SSPx-NPi and SSPx-NPi2025 for all SSPs. As of now, settings are as follows:

[x] SSPx-NPi2025 starts in 2005, i.e. no cascade. SSPx-NPi and SSPx-NPi2025 should be compared in validation to decide if differences are too large so that cascade would be needed.
[x] path_gdx_ref is set to SSP2-NPi2025 for all scenarios which are not NPi or NDCsy2025
[x] path_gdx_bau is set to SSPx-NPi for each SSPx-NDC and SSPx-NDCsy2025 scenario.
[x] path_gdx_refpolicycost is set to SSPx-NPi for each SSPx climate policy scenario (NDC, PkBudg).

Additional question: Do we want to keep old NPi for all SSPs in AMT runs / runs for validation? Or remove after next validation, i.e. once we are happy with NPi2025 results?

robertpietzcker commented 1 week ago

Additional question: Do we want to keep old NPi for all SSPs in AMT runs / runs for validation? Or remove after next validation, i.e. once we are happy with NPi2025 results?

my understanding is that we currently still calibrate to NPi, so we definitely need to keep NPi as default runs to look at. This may change over the next half year, though.

remindmodel / remind