What we found checking 47,000 European energy assets against published figures

A user asked a reasonable question: filter for ČEZ in the Czech Republic. The answer came back 3.9 GW against a published figure of roughly 12.6 GW. That is not a rounding problem, and chasing it down produced three findings worth publishing.

1. The same power station, counted twice

Global Energy Monitor publishes gas-fired power stations in both its Integrated Power Tracker and its Oil & Gas Plant Tracker. Ingesting both means the same physical unit exists twice under two identifiers. Počerady GT1 appeared as G100000400066 and GOGPT-2844, 847 MW each.

Across the dataset: 795 records, 114,998 MW, 7.40% of all capacity. Germany was inflated by 22.4 GW, Italy 14.5 GW, the UK 11.8 GW. Every country total, every owner portfolio and every chart built on top was wrong by that margin.

2. Ownership was missing, not wrong

The ČEZ gap was not bad data. Temelín, Dukovany, Dlouhé Stráně, Dalešice and Orlík were all present, all correct, and all carried no owner at all. An owner filter cannot see them. The same was true of EDF: 13.8 GW against a published ~110 GW, because France’s entire nuclear fleet had no ownership attached.

Reconciling owner totals against published company figures is now a build step. It reads as a scoreboard rather than a proof — the reference figures mix definitions — but it catches order-of-magnitude problems that no amount of internal consistency checking will.

3. A grid operator recorded as a plant owner

France’s open production registry has no owner field. Its only party column is gestionnaire — the distribution or transmission grid operator. Read as ownership, it attributes the Génissiat and Aston hydro plants to RTE, the French transmission system operator, which owns neither.

Related errors found in the same pass: a 400 MW Bulgarian solar farm attributed to its module supplier, and ten Romanian assets still listed under Enel and Macquarie after both had sold to PPC in 2023 and 2024.

What we do about it

• –Reconcile against published figures on every build. Country capacity against national statistics, owner totals against company reports.
• –Track source independence. Ownership recovered from a source is only worth something if that source is not the dataset whose gap is being filled. That share is measured and reported; it currently sits at 4.7%.
• –Show provenance per asset. Registry data, web-researched data and the underlying tracker are visibly different things on the page, with the citation attached.
• –Publish coverage, not just values. An owner total is only as good as ownership coverage in that country, so country pages state the coverage figure beside the numbers.

The honest position

Ownership now covers 85.7% of tracked capacity, up from 49.7%. By asset count it is 35.8% — the difference is roughly 29,000 plants under 10 MW, which together carry about 1% of the capacity in their band.

Every asset above 100 MW has an owner. Below 10 MW, most do not, and for many of them no public source records one at all. That is stated rather than papered over, because a number without its coverage is the thing that started this.