Coal fired power plant database: every generating unit of 30 MW and up, delivered as rows

It is the asset register of the fuel that still anchors a dozen grids: one row per boiler, carrying its capacity, its operating status, its owner and the ground it sits on. Everything else in coal analytics hangs off that spine. A national statistic says a country consumed so many million tonnes; a unit-level database says which boilers would stop burning if one utility repriced its portfolio, which ones are still being built, and which went cold last spring.

The category splits cleanly in two, and the split decides most purchasing decisions before anyone opens a file. Asset registers inventory the machines - the Global Coal Plant Tracker is the reference, one row per generating unit of 30 megawatts and larger. Statistical ledgers track what the machines consume - receipts, deliveries, stocks and prices - compiled by the federal and European statistical systems cataloged beside it. Serious work needs both halves, because a fleet without a burn ledger cannot price fuel risk, and a ledger without a fleet cannot attribute demand to anything specific.

The question underneath most searches for a coal fired power plant database is not really where the file lives. It is which grain answers the question - unit, plant, company, state or country - and whether the rows arrive ready to join instead of needing a rebuild first. The table below is the practical map of the shelf; each profile behind it carries sample rows and the complete field dictionary.

The Global Coal Plant Tracker (GCPT) is the closest thing the space has to a canonical fleet register, and Datadory carries it as a product rather than an errand. The July 2026 release inventories 14,674 generating units across 111 countries, every one of 30 MW or larger - the operating fleet, units proposed since 2010, units retired since 2000 - with more than 3,000 distinct plant owners attached. Capacity-wise the rows describe roughly 2,200 GW operating alongside about 760 GW still under development, each unit holding its own status instead of dissolving into a national total.

Two properties separate it from everything else on the shelf. First, the grain: one row per generating unit, rollable to plant, country or region because the identifiers survive the aggregation - a 660 MW cancellation never blurs into a 25 MW retirement. Second, the corporate chain: direct owner and ultimate parent ride on every row with percentages, so assets roll up to corporates without a reconciliation project. On Datadory's rubric it scores 10 out of 10 - the only quality-10 record among the 17 primary Coal & Consumable Fuels datasets, against a catalog-wide average of 7.81.

Exactly as a sample arrives - three of the 14,674 units, chosen because they sit in three different corners of the status enum:

# unit-level records as delivered (3 of 14,674)
unit-id=G100000113225  status=operating
  name=Huaibei Pingshan power station / Phase II Unit 3
  capacity_MW=1350  start-year=2022
  owner=Huaibei Shenneng Power Generation Co Ltd [100%]
  parent=Shenergy Co Ltd [100.0%]
  country=China  subnational=Anhui  region=Asia
  lat,lon=33.83392,116.831102

unit-id=G100000101858  status=operating
  name=Cumberland Steam Plant / Unit 1
  capacity_MW=1300  start-year=1973
  owner=Tennessee Valley Authority [100%]
  parent=Tennessee Valley Authority [100.0%]
  country=United States  subnational=Tennessee  region=Americas
  lat,lon=36.389947,-87.651636

unit-id=G100000107640  status=cancelled
  name=Ontimavadi power station / (whole-plant record)
  capacity_MW=6300  start-year=-
  owner=GMR Energy Ltd [100%]
  parent=GMR Power and Urban Infra Ltd [69.6%]
  country=India  subnational=Andhra Pradesh  region=Asia
  lat,lon=16.945181,82.238647

Two operating giants and one dead proposal, all in the same schema. Cumberland is four decades of baseload in seven lines. Ontimavadi is the reason the status enum exists at all: a 6,300 MW ambition recorded as cancelled, kept in the file so pipeline analysis never mistakes announcements for steel.

Seventeen verified core fields, written against the published schema rather than guessed. Identification comes first: unit-id and project-id are stable keys for the unit and the plant it belongs to, followed by the English station name, the unit name within it and the local-language name. Physical attributes are capacity in gross (nameplate) megawatts and the nine-value status enum - announced, pre-permit, permitted, construction, operating, mothballed, shelved, cancelled, retired.

Corporate attribution is the part most copies of this data lose: owner carries the direct owner(s) with the stake in brackets, and parent the ultimate parent(s) the same way - including multi-chain cases like the cancelled Indian proposal split across three parents. Geography stacks four deep - country or area, subnational unit, continent-level region, latitude and longitude - while start-year and end-year bracket the operating life. Coal type, combustion technology, heat rate and the per-unit CO2 estimate ride along as additional fields, which is what turns the register into a financed-emissions starting point.

The dictionary below is the verified core; every delivery ships with it attached.

• United States, living series. EIA Coal Data spans thousands of time series across production, distribution, receipts at electric power plants, consumption by sector, stocks, prices by rank and sector, and reserves - cut to national, state, county and individual-mine resolution, with mine-level survey detail reaching into the 1980s.
• United States, annual census. The EIA Annual Coal Report ships about forty-six table entries per edition, its summary spine running from 1949 through a final 512.5 million short tons for 2024, with producing mines counted beside the tonnage.
• Fifty states since 1960. The EIA State Energy Data System holds annual coal production, consumption, prices and CO2 for every state, DC and the territories inside one consistent state-by-source-by-sector matrix exceeding 2.5 million records.
• Europe. Eurostat's solid fossil fuel statistics follow hard coal and brown coal through inland consumption, import dependency and deliveries to power plants and coke ovens across EU aggregates, EFTA and candidate countries, from 1990 through 2025.
• The world, harmonized. The UN Energy Statistics Database carries roughly 6,000 series keys across 200-plus countries and territories - hard coal, anthracite, coking coal, lignite and coal products through production, trade, stock changes, transformation and final consumption.

One seam deserves naming before anyone joins the two halves: registers and ledgers identify plants differently. A tracker row carries a stable unit identifier and coordinates; a statistical series often names nothing finer than a receiving sector or a delivery category. Where identifiers do not exist, match on geography and capacity and treat the result as an estimate rather than a fact. The head-to-head between the fleet register and the American census argues this properly in GCPT versus the Annual Coal Report.

Yes, and the join runs in two hops. GCPT carries direct owner and ultimate parent with percentages on every unit, so assets aggregate to corporates directly - Huaibei Pingshan's Unit 3 rolls up to Shenergy Co Ltd at one hundred percent without interpretation, and the file's 3,000-plus parents make a ranked corporate view a group-by rather than a research project.

For screening those corporates, urgewald's Global Coal Exit List profiles roughly 3,000 companies - more than 1,500 parents plus their subsidiaries, headquartered across over eighty countries - which urgewald states hold more than 90 percent of world thermal coal production and coal-fired capacity. Each firm carries coal share of revenue, coal share of power production, installed capacity, annual production and three expansion flags covering power, mining and infrastructure, so a screen reads direction of travel and not just size. The Metallurgical Coal Exit List adds the steelmaking side: 145 mining companies pursuing more than 250 met coal expansion projects across twenty countries, a pipeline urgewald estimates would lift global annual met coal production by about half again.

The two grains do different jobs and stay distinct: the tracker counts megawatts, the exit lists screen companies. Roll unit capacity up the parent chain first, then test the resulting corporates against the flags. That sequence survives diligence; stopping after the first hop leaves a competitor list with no direction of travel attached.

There is, though none reaches unit level. The World Bank Energy Data Portal aggregates 1,191 energy-sector packages from 31 partner organizations across 193 countries - coal-relevant plant inventories with facility coordinates and capacities, extractive-industry layers and quarterly IEA coal statistics in one multi-publisher catalog. The IEA Energy Statistics Data Browser carries six Coal Information files spanning a full coal energy balance across 93 flows, with World Coal Supply reaching from 1971 to 2024 plus a provisional 2025. For India, the Ministry of Coal's statistics hold the official supply account - company-by-company production against target, despatches, coking and non-coking splits and the National Coal Index - for the world's second-largest producer.

These are balances, not boilers: they resolve a country-year cleanly and a specific power station not at all. Teams that need both buy the register once and hang the balances off it.

Market researchers and consultants turn a headline gigawatt figure into a ranked customer or competitor list - 14,674 units, 111 countries, 3,000-plus owners - and track which developers are still adding capacity. Worked flows continue on the market sizing page.

Investors and quant researchers roll capacity up parent-company chains to price transition exposure where it actually sits, then read the status enum as a leading indicator: construction slowing, shelving rising. The playbook lives at investors and quants.

ESG and climate teams start financed-emissions work from per-unit CO2 estimates built on capacity, coal type and combustion technology, and screen counterparties against exit-list expansion flags; see esg & emissions analysis.

Credit and risk analysts read status as collateral state - operating, mothballed and shelved units answer differently, and the difference is one enum value away; see credit risk screening.

Data scientists and ML engineers inherit a typed, geocoded panel that joins on unit-id and maps without a geocoding step; notes for the group sit at data scientists.

Journalists, academics and students cite a footnoted evidence trail behind every plant, with project history and financing attached; see journalists and academics.

Four disciplines separate a defensible fleet analysis from a chart.

Match the clock to the claim. The register publishes in dated editions - July 2026 is the current reference - and a proposal recorded as permitted stays permitted until the next edition revises it. Pin every extract to its release label so a count of units or megawatts reproduces line for line later.

Keep revisions as vintages. Statuses change - proposed units become operating, operating units retire - and the honest way to handle it is a new vintage keyed on the same unit-id, never a silent overwrite. Backtests built on overwritten files quietly flatter their own accuracy.

Respect the coordinate caveat. Latitudes and longitudes point at the plant, exact where confirmed and approximate where only the vicinity is known. Radius screens at a hundred meters are fiction; screens at grid-node, basin or airshed scale are exactly what the geometry is for.

Do not ask the register for tonnage. It describes assets, not fuel. Burn, receipts, stocks and prices come from the statistical ledgers cataloged above, and the join between them is an estimate wherever plant identifiers do not carry across - flag it as such and the analysis holds.

Files, feeds, or straight into your warehouse. Daily, weekly, or hourly - your call.

Pick the channel your stack already speaks; the rows arrive identical either way - one record per generating unit with its identifiers, status, capacity, ownership chain and coordinates - extracted so nobody on your team reassembles a fleet register by hand.

Name the countries, statuses, capacity bands or parent companies you actually touch when you request a sample, and the extract returns shaped like the question - field dictionary attached, code vocabularies resolved to typed enumerations rather than free text, coordinates flattened to columns beside any geometry. The schema in the sample is the schema every ongoing delivery ships against.

Start with the product page behind this piece - Global Coal Plant Tracker (GCPT) - for the full field dictionary, coverage chips and sample rows. The trade-off against the federal view is argued in GCPT versus the Annual Coal Report, and the Coal & Consumable Fuels Data Guide ranks every pooled dataset in the slice.

Two sibling guides extend threads started here: state coal consumption by year works the demand side of the American ledger from 1960 forward, and newcastle coal price history adds the benchmark that turns burned tonnage into revenue exposure.

For breadth, browse the coal-consumable-fuels data hub or jump straight to the best coal-consumable-fuels datasets ranking - then request the sample cut to the units you actually work with.