AAA_renewable names from one shared technology identity

GGG_renewable, INVDATASET_renewable, and AAA_renewable are three different Balmorel set domains (per-vintage generators, investment-option groups, and region-technology pairs) that legitimately need different string spellings for the same underlying technology — but build_GGG, build_INVDATASET, and build_AAA in additional_inc.py each independently hardcoded their own f-string template for that same (turbine/PV model, resource grade, onshore/offshore, region) tuple, with no shared source of truth. This already caused one real bug (a SP316-HHHH155 vs SP316-HH155 spelling divergence), and is the direct cause of INVDATA linking an area computed under one spelling (IE_VRE-ONS_SP335-HH100_RG1) to an investment option built under another (GNR_WT-SP335-HH100_ONS_RG1_Y-2020) for what is meant to be the same technology.

We decided to extract a single per-technology identity helper (_template_for plus the _WIND_TEMPLATES/_SOLAR_TEMPLATES tables) that each of the three builders calls to produce their (deliberately different) name strings, rather than accept the risk of three independently hand-maintained templates drifting apart again.

We initially also planned to consolidate build_INVDATA_renewable’s AAA_renewable-based regex-substring bridge to INVDATASET_renewable onto build_ALLOWEDINV’s GGG_renewable-based matching, on the assumption that it duplicated the same linking task. Reading the actual implementation showed this assumption was wrong: build_AAA’s templates always nest the INVDATASET_renewable string directly inside the corresponding AAA_renewable string by construction (region + “_” + that string for wind, region + “_VRE-” + that string for solar), so build_INVDATA_renewable’s plain containment check is already exact and correct — it isn’t duplicating build_ALLOWEDINV’s logic, it’s solving a different problem (linking every real region to an investment option, vs. build_ALLOWEDINV finding which literal generator members belong to one representative region). We left this matching mechanism as-is and only removed the dead, always-no-op .replace("ONSVRE_", "").replace("OFFSVRE_", "") call and switched str.contains from regex=True to regex=False, since no metacharacter-sensitive matching was ever needed.

Considered options

Keep three independent templates, add a cross-consistency test asserting they stay in sync. Rejected — this treats a structural duplication as a testing problem rather than removing the duplication itself; a test only catches drift after the fact instead of making it impossible.
Rebuild build_INVDATA_renewable on build_ALLOWEDINV’s GGG-based matching. Rejected after reading the actual code — the two functions solve genuinely different linking problems (see above), and forcing them onto one mechanism would only add complexity without fixing anything that was actually broken.

Consequences

GGG_renewable, INVDATASET_renewable, and AAA_renewable keep their distinct spellings (GNR_WT-..., VRE-ONS_..., <region>_VRE-ONS_...) — this is not a move toward one shared string across all three sets, only toward one shared source of truth for the parts of those strings that should never disagree.