Methodology

World Power Monitor is built on two layers of data that are never blended into one number.

• Compiled statistics. Every country page is built from Ember, an energy think tank that compiles electricity generation, demand, imports, emissions and capacity from national statistics offices, energy ministries and grid operators. These figures are annual, and monthly for 88 countries. They run weeks to months behind the present and are revised when national agencies publish final numbers.
• Operator data. US balancing authorities (through the EIA) and European transmission system operators (through the ENTSO-E Transparency Platform) publish hourly demand and generation. These readings are metered and provisional. They appear on separate grid and bidding zone pages, each with its own source line and timestamp. ENTSO-E also supplies installed capacity, cross border flows and day-ahead prices.

The two layers disagree by a few percent even when they describe the same country and year, because they measure different things (see below). Rather than choose one or average them, each page section shows one source, names it, and states the period it covers. A figure labelled "right now" or "latest hour" only ever comes from an operator feed; Ember figures always carry their year or month.

Country and world figures come from Ember's data API, which is updated more often than Ember's downloadable files. Two datasets the API does not provide, installed capacity by fuel and electricity demand per person, come from Ember's yearly data file. For 6 small territories that the API does not cover, every figure comes from that file, and their pages say so.

• Coverage. 215 countries and territories. Ember's history starts in 1985 for 22 countries and in 1990 for 30 more; most other series start in 2000. History pages show whatever depth exists and say where it starts.
• Latest year. Countries publish on different schedules. 96 countries have figures for 2025; the rest end a year or more earlier. Rankings compare countries within the same year only, and say how many countries had data for that year.
• Refresh. The pipeline checks Ember every day for the current and previous year and reloads the full history every week, so revisions to older years are picked up.
• Revisions. When Ember revises a figure, the new value replaces the old one on the site, and the previous value is kept in our database with a revision number. Nothing is overwritten without a record.

Bidding zone pages and the day-ahead prices page are built from the ENTSO-E Transparency Platform. It is a separate layer from Ember and is never added to or averaged with Ember figures.

• Time. ENTSO-E reports intervals of 15 or 60 minutes, and some zones changed from one to the other. Daily figures are summed over each zone's own local days, so a day with a clock change has 23 or 25 hours. A day with a missing interval is not published as a total.
• Prices. Day-ahead prices are shown in the currency the zone was priced in, as published, and are not converted. Only zones priced in the same currency are ranked against each other. Negative prices are real and are kept. A zone's price day runs from local midnight, and the latest delivery day is normally tomorrow's, because the auction is held the day before.
• Gaps. A zero reading of demand is treated as a missing report, not as no demand. The latest generation mix is the latest hour in which every regularly reported source is present, because some sources publish later than others. Where the reported generation adds up to well under a zone's demand, the page says so.
• Stopped feeds. Some zones no longer report, for example Great Britain after 2021 and Ukrainian generation after February 2022. Their pages show the last date held, say that they are not updated, and they are left out of rankings.
• Installed capacity is reported once a year by each operator and is not what was produced.
• Carbon. We do not publish estimated carbon intensity for bidding zones. The emission factors needed for that estimate have not yet been cleared for use here; Ember's own annual carbon intensity for each country is published on the country pages.

Every source's categories are mapped to one list: coal, gas, oil, other fossil, nuclear, hydro, wind, solar, bioenergy, geothermal, other renewables, storage and other. Imports are a flow, not a fuel. If a source introduces a category that is not mapped, the data load stops rather than guessing.

• Other fossil is Ember's category for fossil fuels other than coal and gas. It is mostly oil products, which is why it is not labelled oil.
• Other renewables covers geothermal, tidal and wave power where Ember reports them together.
• Low carbon means nuclear plus all renewables.
• Shares are percentages of total generation, as published by Ember, so they can add up to slightly more or less than 100% after rounding.
• Negative values. Occasionally a fuel's net generation is slightly negative, when its plants used more electricity than they produced in a year. These values are shown with their sign and never counted as a fuel's share or as the largest source.

• Gross or net. Some statistics count what generators produce before their own consumption; operator data usually reports what reaches the grid.
• Coverage. Factories that generate their own power, rooftop solar and small off-grid systems are included in some datasets and missing from others.
• Timing. Operator data is provisional and revised within days or weeks; national statistics are revised a year or more later.
• Carbon intensity. Ember's intensity figures count emissions per kilowatt hour generated using Ember's own emission factors. Other publishers use different factors or count only direct emissions, so their figures differ.

• Energy is shown in TWh (terawatt hours, one billion kWh), or GWh for small systems. Power is in MW, capacity in GW, and carbon intensity in grams of CO2 per kWh.
• Dates are written in full, for example 2026-11-05 (5 November 2026).
• Hourly data is stored in UTC and shown in each grid's local time with the UTC offset, so changes to and from daylight saving time are never ambiguous.
• Prices, when shown, are in the market's own currency first, with any conversion to US dollars shown separately with its exchange rate and date.

This section says plainly what is automated, what an AI tool did, and what no person has checked. Last reviewed 2026-10-18 (18 October 2026).

• The numbers are computed by scripts, not written. Every figure on the country, fuel, grid, price, record and comparison pages is calculated by a data pipeline and the site build from named, licensed sources: Ember for country statistics, the US Energy Information Administration for US grids and states, and the ENTSO-E Transparency Platform for European grids and prices. Each page names the sources it uses and links to their licences. The full list is on the sources and licences page. No figure on those pages is typed in by hand or produced by an AI model. The grid failure write ups below are the one exception, because they are written from outside reports.
• Most page text is templated. The sentences around the numbers on the thousands of country, history, fuel, comparison and grid pages come from fixed templates that insert the computed values, so two pages of the same type read alike and differ in their figures. They are not written one by one. The sentences of the daily facts and the weekly digest are built the same way from the data.
• Articles are drafted with AI assistance. The blog posts, the Learn guides and the glossary were drafted with an AI writing assistant. They carry the site name as author, not a person. Figures quoted in them are dated and name their source, and they do not update by themselves after publication. No subject expert has reviewed them.
• Notable grid failures are researched with AI assistance. Each page under Notable Grid Failures summarises one blackout or grid emergency in our own words and cites the official report or investigation it is based on, with a link. The research and drafting used an AI assistant working from those public sources. No grid engineer or other domain expert has reviewed the pages, so check the cited report for anything that matters to you.
• Eight languages are machine written. The Spanish, German, French, Portuguese, Italian, Japanese, Indonesian and Turkish pages (country pages, country lists, home pages and the consent window text) were translated by an AI and have not been reviewed by native speakers. Expect awkward phrasing and the occasional mistake. The numbers on those pages are the same computed values as on the English pages, but the surrounding words may be less exact than the English original. The English page is the reference.
• The site and its code were built with AI coding assistance. The website code and the data pipeline were written with AI coding tools working under the direction of the site's operator. Their output is tested as described below, but it was not independently audited.
• Charts are drawn by code, not by AI. Charts, maps and share images are generated at build time from the data as vector graphics or plain images. No image on the site is made by an AI image generator, and each chart has a text description and a hidden data table.
• Who runs it. World Power Monitor is published under that name by its operator and has a sister site, World Oil Monitor. No individual author, expert or credential is listed because none is claimed. Contact details are in the report line below.

How accuracy is checked. These checks exist and run before a version is published:

• Automated unit tests for the data parsers, the derived calculations, number and unit formatting, the page titles and descriptions, and the structured data.
• The data load stops if a source uses a fuel category the site has not mapped, if a file fails its schema, or if a dataset is not cleared for publication in the licence registry.
• Build audits on the finished site: internal links and orphan pages, a check that every page can be reached within three clicks of the home page, SEO and attribution rules, sitemap and translated page rules, and a style rule on copy.
• Spot checks of sample pages against the source data, made while the pages were built. These are not a systematic or independent review. The site does not publish an error rate because none has been measured.

No professional review, and not advice. No engineer, economist, lawyer or other professional has reviewed the content. It is information, not engineering, investment, trading or financial advice, and nothing here shows what a household pays for electricity.

Report an error. Email [email protected] with the page address and what looks wrong. Corrections to data pages are made at the source of the error, either in the pipeline or in the page template, so they fix every page affected.