OPSD EU: use energy_source_level_3 for Fueltype classification

[MaykThewessen]

Summary

• Use the most specific energy source from OPSD's three-level hierarchy for Fueltype
• When energy_source_level_3 is set (e.g. "Biomass and biogas" for coal plants with biomass co-firing), prefer it over the top-level energy_source
• Falls back through level_3 → energy_source → level_1

This is a 3-line defensive change in the OPSD EU loader. Currently energy_source_level_3 is loaded from the CSV but silently dropped by .reindex(columns=target_columns) before it can influence Fueltype.

Context

The OPSD EU dataset uses a three-level energy source hierarchy. While level_3 currently only contains "Biomass and biogas" entries (156 plants across EU), this change ensures that if future OPSD releases or enriched data add co-firing flags for coal/waste plants, the information flows through to Fueltype correctly.

See also:

• OPSD: backup URL drops efficiency column; energy_source_level_3 (biomass co-firing) unused #286 — OPSD data quality discussion
• Use EIC codes as deterministic matching key before Duke fuzzy matching #287 — EIC-first matching (architectural proposal for preventing co-located plant confusion)

Test plan

• OPSD() loader runs without errors
• NL plants load correctly with expected fueltypes
• Full powerplantmatching.powerplants() pipeline produces valid output

🤖 Generated with Claude Code

[MaykThewessen]

Follow-up data point found while cross-checking the NL slice of the matched output against operator data:

Borssele 1 (nuclear) comes through at 515 MW, sourced from GEM (G100000500496). That figure appears to be gross capacity — IAEA PRIS lists 482 MWe net / 512 MWe gross, and operator EPZ states ~485 MW. Since the matched dataset otherwise reflects net capacities, this overstates NL nuclear by ~6%.

Smallest fix would be one manual_corrections.csv row:

projectID=G100000500496, Source=GEM, Capacity=485

Happy to push that row onto this branch as part of this PR, or open a separate one-line follow-up — whichever you prefer.

[MaykThewessen]

Gentle nudge: still mergeable; addresses the energy_source_level_3 half of #286. Could a maintainer review when convenient? (The efficiency-column-drop half of #286 is separate and still open.)

[MaykThewessen]

Added the Borssele 1 capacity correction I mentioned above as a single manual_corrections.csv row (ea84536): GEM reports it at 515 MW (gross), corrected to net 485 MW (IAEA PRIS 482 MWe net / EPZ ~485 MW), consistent with the dataset's net-capacity convention. It's a capacity-only override, so it leaves all other Borssele 1 fields untouched.

Happy to split this into its own one-line PR if you'd rather keep this one scoped purely to the energy_source_level_3 change, just say the word.