eGon — Industrial Gas Demand (CH4 & H2) for Germany

The hour-by-hour demand ledger behind Germany's energy-transition scenarios. The eGon-data pipeline, built by the openego research consortium, models the coupled German electricity and gas system under two futures: eGon2035 and the fully renewable eGon100RE. Its industrial gas demand component converts the FfE eXtremOS collection into hourly methane and hydrogen load curves for roughly 400 German NUTS-3 districts, resolving regional figures to district level by matching region IDs to NUTS codes and centroid geometries from the vg250_krs district table.

The scenario math is where this stops being a spreadsheet and becomes an argument. Project-documented totals put eGon2035 at 195 TWh of industrial methane and 16 TWh of hydrogen, against 105 TWh methane and 42 TWh hydrogen in eGon100RE - hydrogen nearly triples while industrial methane falls to zero, and the hydrogen profile is rescaled to PyPSA-Eur-Sec totals including Fischer-Tropsch and methanation demand. Every load lands assigned to a gas bus in the PyPSA grid model, written to the pipeline's load and timeseries tables, which means the series arrive grid-aware rather than orphaned.

Within Datadory's catalog of 1,744 datasets across 159 viable industries, this record scores 7/10 and holds a unique slot in the eight-record industrial gases slice: the pool's only hourly gases-demand series. Get a sample of this dataset cut to your regions, carriers and years.

Two captured series headers for the same NUTS-3 district, same scenario year and same weather year - differing only in carrier:

field                     row 1              row 2
------------------------------------------------------------------
id_opendata               66                 66
id_region_type            38 (NUTS-3)        38 (NUTS-3)
id_region                 4000001            4000001
year                      2035               2035
year_weather              2012               2012
internal_id [sector,car]  [2, 11]            [2, 335]
value (annual total)      621,826.13         63,937.26
values_hourly_first3      [65.43, 61.47,...] [6.68, 6.43,...]

# internal_id_1 = 2 marks the industry sector
# internal_id_2 = 11 is natural gas in the pipeline's documented
# carrier mapping; 162 is hydrogen. Code 335 above is a further
# carrier code captured in the slice - decode confirmed at sampling.
# 'values' carries all 8760 hourly points; three shown here.

Read what the pair settles immediately. The same district, same 2035 horizon and same weather-year shape carry two distinct carrier curves, and the natural-gas series runs nearly ten times the annual total of its neighbor - the ratio between today's incumbent fuel and the next fuel's early footprint, visible inside a single district. The first three hourly values hint at the wider structure: a weather-normalized annual curve, 8760 points deep, not twelve monthly averages pretending to be a profile.

One decoding rule belongs next to the rows. The sector half of the internal-id pair is stable - code 2 is industry everywhere - while the carrier half reads off the pipeline's documented mapping, natural gas and hydrogen among them. Carrier codes outside the documented pairs appear in captured slices; your set gets decoded and confirmed with the sample rather than guessed here.

Nine fields define every series, each checked during the August 2026 research pass. The identity stack does the heavy lifting: id_opendata names the industry-sector load-curve family, id_region_type fixes the geography scheme (38 is NUTS-3), and id_region pins the individual district. year carries the scenario horizon and year_weather the historical meteorology shaping the curve - 2035 on weather year 2012 in the captured slice. The internal_id pair separates sector from carrier, so methane and hydrogen never blur into one gas column.

The measurement pair closes the loop: value holds the annual demand total for the region-year-carrier combination, and values holds the full 8760-point resolved profile. Together they mean you can validate any hourly claim against its annual sum on arrival - the kind of internal consistency check most demand collections make impossible.

Four extensions ship as additional fields on request: human-readable region labels joined from the vg250_krs district reference table onto every row; tidy long-form normalization that unfolds the 8760-value arrays into timestamp-per-row shape keyed on region, carrier and hour; carrier-label decoding beyond the documented natural-gas and hydrogen pairs, confirmed against your requested slice; and scenario-total rollups by carrier and year for benchmarking regional shares against the project's published TWh figures.

The unit flag on value is cataloged as MWh and gets pinned against upstream documentation at sampling rather than asserted here - a one-line check that saves a tenfold error downstream. Name the extensions you want when you request the sample and they arrive applied, not appended.

Geography - Germany at NUTS-3 district resolution, roughly 400 districts strong, which is fine enough to separate Ruhr industrial clusters from their rural neighbors. The underlying FfE collection extends Europe-wide at the same NUTS-3 level, so a German-first panel can widen later without a schema change.

Temporal - scenario years 2020 through 2050 in five-year steps, seven horizons per scenario family. Each series is one full 8760-hour annual profile shaped by a historical weather year - 2012 in the captured slice - so extreme-week behavior is embedded in the curve instead of smoothed away.

Granularity - one hourly time series per region per energy carrier per scenario year: about 400 regions x scenario years x 2 carriers x 8760 hourly values, tens of megabytes per scenario year. Nothing aggregates away the district you care about unless you ask it to.

API, files, or your warehouse. Daily, weekly, or hourly.

Name the districts, carriers and scenario horizons when you request the sample - one region's full 8760-hour panel, or every district at 2035 and 2050 - and it arrives shaped to that scope with the complete nine-field dictionary attached. The sample ships first either way; the ongoing feed lands on whatever cadence your models run at.

An hourly scenario panel earns its keep in five specific jobs:

• Hydrogen-transition stress testing - scenario demand paths to 2050 in five-year steps quantify how fast industrial methane erodes while hydrogen scales, with eGon100RE taking methane to zero.
• Demand-profile model training - 8760-hour CH4 and H2 series per NUTS-3 region give supervised models real annual shape instead of flat averages.
• Regional siting and screening - district-level annual totals rank German industrial regions by gas appetite, sharpening electrolyzer, storage or supply coverage decisions.
• Grid and storage planning context - loads arrive pre-assigned to gas buses in the PyPSA grid model, so network studies start from connected demand.
• Scenario-assumption benchmarking - test your own plan's regional assumptions against the open modeling community's published math, district by district.

Each job maps back to the industrial gases data hub, where this record anchors the slice's demand-modeling layer beside trade and production alternatives.

Data scientists and ML engineers get 8760-hour CH4 and H2 training panels per district under both scenarios, on a nine-field spine that never changes shape (relevance 3). Journalists, academics and students get citable scenario profiles with documented methods - the version that survives peer review (relevance 3). Investors and quant researchers stress hydrogen-transition exposure with demand paths to 2050 in five-year steps (relevance 2). Market researchers and consultants quantify German industrial gas demand by region and carrier for studies that need district-level numbers (relevance 2). Developers building data products wire hourly profiles into PyPSA-style planning tools on a stable schema (relevance 2).

Start from the best industrial gases datasets ranking to see where this sits in the pool, then pair it with Eurostat Comext when modelled demand needs realized trade flows beside it.

Provenance note - produced by the eGon-data pipeline of the openego research consortium, which models the coupled German electricity and gas system under its eGon2035 and eGon100RE scenarios. The demand series draw on the FfE eXtremOS collection, resolved to NUTS-3 districts and wired to gas buses in the PyPSA grid model before publication.

Methodology note - these are modelled scenario quantities, not metered consumption: the profiles describe what the scenario math says demand would be, shaped by a historical weather year. Treat the two scenarios as a bracket on the future rather than a forecast line, and reconcile against measured consumption where your analysis can.

Completeness note - Datadory scores this record 7/10 against a catalog mean of 7.81 across 1,744 datasets, carried by the uniqueness of hourly gases-demand depth. Region display names, carrier-code decodes beyond the documented pairs and the unit pin on value are confirmed at sampling rather than claimed here.

Where to go next - the rail below collects the trade, production and portal companions in this slice, plus the vocabulary pages that make the hourly-profile dimension legible.