Hourly Wind Generation Data Europe Data: Dataset Structure and Field Coverage

The OPSD Time Series Package (Load, Wind & Solar Hourly) is the reference answer. Version 2020-10-06 covers 32 European countries from 2015 through mid-2020 at hourly resolution, with 15-minute and 30-minute files for selected regions, and aligns six variable families in one table: actual load, day-ahead load forecasts, day-ahead prices, actual wind and solar generation, installed capacity, and capacity-normalized profile shares.

Provenance is transmission-grade. The package consolidates ENTSO-E Transparency Platform publications with national sources — BNetzA/netztransparenz.de for Germany, BEIS for Great Britain, BFE for Switzerland, Energinet.dk/Energistyrelsen for Denmark and Vindbrukskollen for Sweden — so every megawatt in the wind columns traces to a system-operator publication rather than a model estimate. That distinction is exactly what dispatch, balancing and grid studies require.

Columns follow a predictable {REGION}_{variable} pattern, so a country pull is a string filter away. The wind family is what modellers come for:

• {REGION}_wind_generation_actual — total actual wind generation in MW where no technology split is published
• {REGION}_wind_onshore_generation_actual and {REGION}_wind_offshore_generation_actual — the split where reporters provide one
• {REGION}_wind_capacity, _wind_onshore_capacity and _wind_offshore_capacity — installed capacity by technology, available mainly for the DE and GB_GBN zones
• {REGION}_wind_profile and siblings — capacity-normalized generation shares usable directly as proxy capacity factors

Timestamps arrive twice, as utc_timestamp and cet_cest_timestamp, so daylight-saving artefacts can be sidestepped by working in UTC alone. Each region also carries {REGION}_price_day_ahead in EUR and {REGION}_load_actual_entsoe_transparency in MW beside its day-ahead forecast. Because wind output, load and prices occupy one 300-column frame at identical hourly stamps, a residual-load or merit-order calculation needs no join logic — which is why the package became the standard European modelling input.

Resolution varies because the underlying publications do. Beside the 60-minute base, separate 15-minute files cover the AT, BE, DE, HU, LU and NL regions where original data is finer, and 30-minute files serve CY, GB and IE. Aggregation runs at three levels: whole countries, the four German control areas (50Hertz, Amprion, TenneT and TransnetBW), and bidding zones such as DE-LU, DK1/DK2, IE-SEM, the Italian zones, Norway's NO1-NO5, Sweden's SE1-SE4 and the GB zones.

Sizing follows resolution. The 60-minute singleindex CSV weighs 124 MB against 841 MB stacked, the 15-minute pair 107 MB versus 669 MB, and the 30-minute pair 37 MB versus 219 MB. Non-tabular routes ship alongside — a 216 MB SQLite database, a 198 MB XLSX workbook and a datapackage.json describing every resource and column schema — while three CSV layouts (singleindex for human-readable pipelines, multiindex for GAMS-style tooling, stacked for long-format storage) let the same data drop into different stacks unchanged.

Measured OPSD columns report what turbines actually fed into grids, so they embed curtailment, outages and reporting gaps — ideal for price and balancing work, but silent before 2015 and frozen after mid-2020. Simulated Renewables.ninja output instead converts weather reanalysis into power using documented models: GSEE for PV and the virtual wind farm R package for wind, driven by MERRA-2 on a 0.625 x 0.5 degree grid and SARAH on 0.05 x 0.05 degrees for solar.