GridTelemetry FAQ

GridTelemetry uses five live public feeds plus published project-register snapshots. The Carbon Intensity API provides carbon emissions, regional mix and forecasts. Elexon BMRS supplies generation, demand, interconnectors, frequency, prices, balancing, margin and outages. NESO Open Data provides renewable forecasts, network-constraint forecasts, balancing economics and the Transmission Entry Capacity register. Sheffield Solar PVLive supplies embedded-solar outturn estimates. Octopus Energy supplies Agile tariff rates for regional price windows. The DESNZ Renewable Energy Planning Database supplies the future-project map.

GridTelemetry combines these sources and calculates some comparisons, records and outlooks. Derived figures and interpretive labels are identified where they appear; the future-project registers are dated snapshots rather than live operational feeds.

Measured values come from meters or monitored system data. Declared values are an operator’s stated plan, such as a power station’s intended output for a half-hour, and can differ from the later meter reading. Estimated values are calculated from source data where no direct national measurement exists. Forecast values describe a future period and will not exactly match the eventual outturn.

Published means GridTelemetry is displaying a figure released by the named source. “GridTelemetry interpretation” marks a threshold, grouping or plain-language status calculated by this site; it is not an official warning. Official NESO notices are labelled separately and retain their original authority. A missing or unavailable value is shown as unavailable or with a dash, never converted to zero.

The scenario explorer uses the mapped projects of 10 MW or more retained from the DESNZ Renewable Energy Planning Database. Visitors can compare status, technology, region, minimum-capacity, storage and recorded-stage-date filters. Totals are gross published project capacity; they are not delivery forecasts, and GridTelemetry does not assign completion probabilities. When a date horizon is active, records without the relevant current-stage date are excluded and counted rather than assumed to be inside the horizon.

The connection outlook is a separate view of NESO’s Transmission Entry Capacity register. Its effective-year horizon applies to the annual TEC aggregate, while the REPD technology, region and status filters do not. NESO warns that capacity may be split across technology or connection-stage rows, so GridTelemetry removes only exact duplicate rows and does not guess that similarly named projects are the same asset. Project counts are distinct within each year and can repeat across years.

Scenario choices use public filter identifiers in the URL so comparisons can be shared. They contain no postcode, account or private watchlist information.

The underlying feeds refresh at different rates: generation by fuel type roughly every five minutes, system frequency about every fifteen seconds, and carbon intensity and wholesale prices once per half-hourly settlement period.

Live operational pages re-poll once a minute, while the 1-day frequency view refreshes every 30 seconds. The “Live” badge shows how far through the latest measured period the data covers, rather than merely when the API was last called. Carbon intensity, generation and prices still move at their source publication cadence, so faster requests would usually return the same figures.

When you open the panel, GridTelemetry requests one representative dataset from each live provider. “Fresh” means the check succeeded and its source timestamp is within that feed’s expected publication window. “Stale” means the provider answered but the representative reading is older than expected. “Unavailable” means the check failed or timed out. Sources such as Octopus that do not expose a timestamp can be checked for availability but not freshness.

This is a GridTelemetry connectivity check, not an official status for the provider as a whole. One dataset can fail while other data from the same provider continues to work.

The data-quality centre reports what GridTelemetry can observe. “Current availability” is one representative live request to each provider. “Uptime” is the share of expected half-hour periods retained in GridTelemetry’s own history; it is not an official availability figure or service-level measure for the provider. A source that GridTelemetry does not retain, such as Octopus Agile tariff data, is shown as not retained rather than as 0% uptime.

Ingestion lag is how far the latest retained reading sits behind the newest completed half-hour. Fallback age is the age of the newest retained reading available to a page that supports stored fallback; it reports readiness and does not imply that fallback is currently active. An incident is one or more contiguous missing expected periods. Time before GridTelemetry first began retaining a dataset is excluded rather than mislabelled as an outage.

A missing history store or unavailable dataset makes the overall picture incomplete. Known zero gaps may still be shown, but the page will not turn incomplete evidence into an all-clear.

Choose “Explain this reading” on a Live Grid metric tile. The panel stays attached to that displayed value and source timestamp, and explains its calculation basis, provider inputs, reading age, fallback state and any component omitted from that response.

The matching data-quality link opens the 24-hour quality view with the relevant provider and retained-dataset rows highlighted. Those checks add current source health and historical coverage context; they do not replace the selected reading’s own timestamp or prove that every dataset from a provider is healthy.

No. It’s an independent project built on openly published data, and it isn’t affiliated with or endorsed by the electricity system operator, Elexon, or any energy company.

It’s built for general interest and education. Live data can lag, be revised later, or briefly drop out when an upstream source is unavailable, and some figures are estimated (see the drill-down question below). For operational or commercial use, always go to the primary sources directly.

A megawatt (MW) and gigawatt (GW) measure power: the rate at which electricity is being generated, transferred or used. One gigawatt is 1,000 megawatts. A megawatt-hour (MWh) and gigawatt-hour (GWh) measure energy delivered over time. For example, 100 MW sustained for two hours supplies 200 MWh.

Hertz (Hz) measures grid frequency, which is kept close to 50 Hz. Grams of carbon-dioxide equivalent per kilowatt-hour (gCO₂/kWh) describe the estimated emissions associated with each unit of electricity used. Wholesale prices are normally shown in pounds per megawatt-hour (£/MWh); a supplier tariff in pence per kilowatt-hour (p/kWh) is labelled separately.

A timestamp has a role as well as a clock time. Observations describe when a reading applied; completed half-hour charts use the period start while freshness can use its end. Forecasts keep the target period separate from the time the provider published the forecast. Labels distinguish measured, declared, estimated, forecast and modelled values wherever that changes how the number should be read.

Total supply adds signed output and net imports. Pumped storage consumes electricity while pumping uphill, so a negative reading reduces the total; it adds supply when generating. Exports also reduce total supply. Fuel percentages describe only positive supply sources, so pumping and exports have zero share.

Two reasons. First, they are genuinely different quantities: Great Britain also imports and exports power over interconnectors and loses a little in transmission, so generation and demand rarely line up to the megawatt.

Second, the raw feeds are “transmission-level”, which excludes smaller generators connected to local distribution networks. GridTelemetry adds estimates for embedded solar and wind back on, so both figures sit on a comparable national basis. Solar uses a modelled outturn; embedded wind uses NESO’s forecast estimate and is labelled as such.

Transmission figures cover only what flows across the high-voltage grid the system operator meters directly. National figures also include “embedded” generation connected further down, in local distribution networks — especially solar and smaller wind farms.

National is the more intuitive “what the country is producing and using” number, so that’s the basis the dashboard shows.

The main live generation feed measures wind farms connected to the high-voltage transmission system. Smaller wind farms connected to local distribution networks are not visible in that total.

GridTelemetry adds NESO’s embedded-wind forecast for the matching half-hour. It is an estimate, not a live meter reading. The site shows the forecast period and source update time; if a matching estimate is unavailable, it keeps the measured transmission wind and says that the national total is incomplete rather than substituting zero.

Most solar is embedded, so it doesn’t appear in the transmission generation feed at all. GridTelemetry uses Sheffield Solar PVLive’s national embedded-solar outturn estimate and adds it to the transmission figures. If PVLive is temporarily unavailable, the live grid can fall back to an estimate derived from the Carbon Intensity API’s fuel-mix percentage.

Historical Generation Mix charts use retained PVLive observations where available. Solar remains a modelled estimate rather than a meter reading, so it is labelled separately from transmission-connected generation.

Carbon intensity is the grams of CO₂ emitted per kilowatt-hour of electricity generated. The coloured band is a simple qualitative scale — very low, low, moderate, high, very high — matching the Carbon Intensity API’s own thresholds, so a glance tells you how clean the grid is right now without reading the exact number.

Interconnectors are the subsea and cross-border cables linking Great Britain to France, the Netherlands, Belgium, Norway, Denmark and Ireland. “Import” means power is flowing into GB over a link; “export” means it’s flowing out. The net figure sums simultaneous imports and exports across every link, so it is not the gross volume traded over all cables.

On the chart, positive values are net imports and negative values are net exports. The flow-pattern card summarises regularly spaced periods in the selected window; values within ±25 MW are treated as near balanced. Connected-market rows group parallel links to the same country, so opposing flows within that corridor offset.

It’s the half-hourly Great Britain wholesale reference price in pounds per megawatt-hour, taken from Elexon’s Market Index Data (the APX/EPEX index). It reflects the short-term market price of electricity — not a retail or household tariff.

The day-ahead price, or Market Index Price, is the volume-weighted price traded for each half-hour on Great Britain’s short-term electricity markets.

The imbalance price, also called the cash-out price, settles differences between the electricity generators and suppliers contracted and what they delivered or used. It is published later than the day-ahead price and may be revised by subsequent settlement runs.

The grid runs at a nominal 50 hertz. Frequency drifts slightly above 50 Hz when generation exceeds demand and below it when demand exceeds generation, so it acts as a real-time heartbeat of grid balance. The operator continuously works to hold it within a tight band around 50 Hz.

De-rated margin is NESO’s estimate of the generation available above forecast demand, adjusted for the likelihood that each source will be available. GridTelemetry shows the forecast for each half-hour over roughly the next two days.

Loss of load probability is the modelled chance that available generation will fall short of demand in a settlement period. It is not a blackout forecast: NESO can use reserves, imports and demand-side actions, and it issues formal Electricity Margin Notices when it needs a market response.

Labels such as “Comfortable” are GridTelemetry interpretations, not official NESO status labels. Any active official notices appear on the Live Grid and Grid Margin pages.

The outage page uses operator reports published through the European wholesale-energy transparency system, commonly called REMIT. A notice says that some capacity at a physical asset is unavailable or expected to be unavailable. It does not, by itself, mean Great Britain is short of electricity or that customers will lose supply.

GridTelemetry keeps the latest valid revision of a notice, excludes dismissed, superseded and closed notices, and avoids double-counting overlapping notices or mirrored interconnector directions. “Notice open; no capacity currently reported unavailable” does not contribute to outage totals or alerts. If an asset cannot be mapped or the source is stale or incomplete, that limitation is shown rather than treated as an all-clear.

A network constraint occurs when the transmission system cannot safely carry every planned electricity flow, even when Great Britain has enough generation overall. NESO may pay generators or flexible demand to change output so a boundary stays within its operating limit.

The boundary chart is NESO’s day-ahead forecast of flow and limit, not measured outturn. The 1D view includes the next 48 hours; historical views compare the day-ahead snapshots GridTelemetry stored. Forecast utilisation is the absolute flow divided by the published limit.

NESO changed the calculation on 22 April 2024. Earlier published flows included expected mitigation and optimisation actions; later flows do not, so comparisons across that date need care. The balancing-cost figures are separate historical outturn datasets and may be revised after the event.

Balancing actions are instructions accepted by the system operator to change a Balancing Mechanism unit’s level. “Raise” and “lower” describe the direction of the instruction; the profile shown is an instruction, not metered output.

A system flag means the action may relate to a network constraint or another system-operating need. The published feed does not provide a plain-English reason for each flag. Historical balancing costs and volumes are separate datasets and may be revised.

For a network boundary, GridTelemetry finds a published day-ahead period above the operating limit and compares its negative transfer headroom with the first later period forecast at or below the limit. The change is a scenario from NESO’s flow and limit data, not a forecast of extra generation, avoided cost or the actions NESO will take.

For outages, the scenario applies operators’ latest REMIT end dates to the current unavailable physical capacity. Overlapping notices are clamped to the asset’s reported normal capacity and mirrored interconnector directions count once. “Due back” means no longer reported unavailable; it does not mean the asset will generate or import at that level.

Boundary transfer headroom and outage capacity are deliberately not added together. Stale, unavailable, incomplete and open-ended outage data stays visible, and a missing value is never replaced with zero. The selected boundary code and return date may be shared in the URL; outage notice details are not added to it.

The detailed 48-hour view combines actual wind and embedded solar with the latest available forecasts. The 14-day view uses daily summaries for a simpler look ahead. A gap means a component is not available; it is not treated as zero.

Forecast capacity factor is expected output as a share of the capacity NESO publishes for the same settlement period. It is context for the forecast, not a measurement of every renewable asset in Great Britain.

For most fuels, yes. Gas, coal, nuclear, biomass, wind, hydro, pumped storage and batteries all drill down to real named stations carrying their own output, taken from Elexon’s per-unit data. The imports drill-down likewise shows each interconnector’s real measured flow.

What “output” means there is worth being precise about: it is the operator’s own declaration for the current half-hour, not a meter reading. Metered per-station figures aren’t published until five working days later, so anything shown live is necessarily a declaration.

Solar is mostly embedded in distribution networks and “other” is a mixed national category, so neither has a trustworthy live per-station split. Their drill-down therefore shows one clearly labelled national estimate with installed-capacity context; it does not allocate output to named sites or infer a load factor.

Great Britain is divided into fourteen regions for the purpose of reporting carbon intensity, and Your Region shows you the one you live in: how clean your local electricity is right now, how that compares with the national average, and what is generating in your area.

You can enter a full postcode or its outward half. GridTelemetry’s API reduces it to the outward code — the part before the space — before asking the Carbon Intensity service. The application does not persist the postcode; only the resulting region number can be remembered in your browser. See the privacy answer for request-log details.

Because the regional generation mix is not an inventory of what stands nearby. It is a model of the electricity being supplied to your region, and it accounts for power flowing in from neighbouring regions and across the transmission network.

So a fuel can hold a large share in a region that contains no plant of that kind at all. Large gas stations in particular are connected to the national transmission network and serve the whole of GB rather than the patch of ground they sit on. Read the mix as “this is the character of the electricity arriving here”, never as “these are the power stations down the road”.

You can compare two to four of the fourteen distribution-network regions using current and forecast carbon intensity, the past day’s actual high, the next 48 hours’ forecast high, modelled regional supply mix, and the cleanest and cheapest complete three-hour windows. The GB difference is a GridTelemetry calculation against the national reading for the same snapshot.

The cheapest windows and displayed rates use Octopus Agile: one supplier’s regional retail tariff including VAT, not a wholesale price or a claim about every household tariff. Missing carbon, national or tariff data is shown as unavailable, never as zero.

A comparison link contains only public Carbon Intensity region IDs. It never contains the postcode originally used to find a region. The supply-mix comparison describes electricity supplied to each region and must not be read as a list of generation assets physically located there.

The opposite question, deliberately. That list is what physically sits inside your region — real named stations, with the output their operators have declared for the current half-hour. It answers “what is generating here”, while the mix above answers “what is powering here”. The two genuinely disagree, and that is the point of showing both.

Offshore wind farms are tagged as such: they are out at sea and connect to the grid in the region, rather than sitting in it. Some regions have no transmission-connected generation at all, which is a real answer rather than a failure.

They are the electricity distribution licence areas — the historic patches served by each regional network operator — not counties, postcodes or anything a map of England would show you. The Carbon Intensity API reports against them, so the site follows the same boundaries.

That produces occasional surprises which look like bugs and are not: Iron Acton in South Gloucestershire falls in the West Midlands area, and Coventry falls in the East Midlands one. The licence areas simply do not follow the names you would expect.

Almost. Roughly 99.6% of the transmission-connected fleet by capacity is placed in a region; a handful of small units are not yet located and are left out rather than being put in the wrong place. The fleet also grows on its own as new units connect, so a very new site may take a while to appear.

The Asset Explorer also retains sites with sourced closure history. Retirement is recorded against each Balancing Mechanism unit with an effective date and primary source, so a mixed site can keep its operating units without counting the capacity or fuel of retired ones. Units without a sourced event remain operating in the current catalogue; a live non-zero declaration overrides a retirement event pending review.

Embedded generation never appears here at all — rooftop solar, small wind and anything else connected to the local distribution network is invisible to the per-unit feed, however much of it there is.

It lists grid-scale batteries and pumped hydro, kept separate from the generation table because they are not generating: they return energy taken off the grid earlier, and spend a good part of their lives consuming it. A negative figure means the unit is charging.

Your Region compares complete three-hour windows across the rest of today and tomorrow. “Cleanest” uses the lowest average forecast carbon intensity, “Cheapest” uses the lowest average Octopus Agile rate, and “Best balance” combines the two. The single cleanest half-hour is shown separately.

Agile is one supplier’s regional retail tariff, including VAT, so it is a useful price signal rather than a forecast of every household’s bill. If a required forecast or price is missing, GridTelemetry does not fill the gap with zero.

For an EV or home battery, enter battery capacity, current charge, target charge, charger power and charging efficiency. The planner derives how much electricity must come from the charger and the minimum number of half-hours needed. You can keep charging continuous or let it split across cheaper or cleaner half-hours before the deadline. Split charging never exceeds the entered power, switches only on half-hour boundaries and stays within the maximum number of sessions you choose. Appliances and heat pumps always keep their continuous cycle.

The planner ranks complete windows or split schedules by marginal cost, operational grid carbon or an equal-weight balance of both. Its “starting now” comparison is one continuous run. A split calendar export contains a separate event for each charging session.

Optional household solar uses NESO’s national embedded-solar capacity-factor forecast, scaled to the kWp you enter. This is a broad GridTelemetry estimate rather than a roof-specific forecast: it does not know your location within the region, roof direction, shading or ordinary household demand. The energy-flow chart shows how much of the planned load is assigned to forecast solar and how much remains a grid import.

Marginal cost means import charges plus any export payment you give up by using forecast solar yourself. Choose no export credit, enter one flat rate, or import half-hourly export rates. It is not a whole-home bill forecast and does not schedule battery discharge or electricity export.

For import prices, choose the live Octopus Agile regional rate, enter one flat unit rate, enter separate day and night rates with your off-peak hours, or import a custom CSV. The planner uses Europe/London time for a day/night schedule, including daylight-saving changes. Enter the unit rates including VAT from your bill; standing charges are excluded because they do not vary by planned time.

A custom CSV can contain timestamp, import_p_per_kwh and/or export_p_per_kwh; Octopus-style valid_from and value_inc_vat import files are also accepted. Missing solar, carbon or tariff periods are never treated as zero.

Device, planning and entered flat or day/night tariff choices are shareable in the URL and can be exported to a calendar. Imported tariff rows stay only in versioned browser storage: they are not uploaded or included in a shared link, so another device must import its own copy.

They switch the time window. “1D” is the live view of roughly the last 24 hours; the longer ranges show averages over a week, month, year, five years, or all stored history, each at a progressively coarser resolution.

On Generation Mix, the live snapshot remains fixed at the top and the range buttons apply only to the stacked historical chart. On supported analysis pages, range and comparison choices are retained in the page URL so the view can be shared.

Yes, where a page has selectable analysis state. Carbon Intensity, Frequency, Prices and Interconnectors retain range and comparison choices in the URL; Carbon Intensity also retains its regional table or map choice. Outage search and filters are shareable in the same way.

Prices, wind forecast revisions and Outages can freeze a view at a selected “known at” time. Forecast and outage links use retained publications and revisions available by that cutoff. Price history may be restated after settlement, so a frozen price link is labelled as a reconstruction from today’s revised record rather than an exact copy of what the page displayed then.

Download CSV exports the points or filtered notices behind the current view. Download chart exports the visible SVG where offered. These files are generated in your browser and do not upload the view or its data anywhere.

On the day view, the solid line is the latest available outturn or estimate and the dashed tail is a short forward forecast (for demand and carbon intensity), with a “now” marker separating them. The longer ranges are historical only and carry no forecast.

Explore can compare up to four measures, including demand, individual generation types such as wind, solar and nuclear, carbon intensity, price and net imports. Because their units differ, each line is indexed to 100 at its first available reading; hover or focus a point to see the original value. This shows relative movement, not equal physical quantities.

The daily-pattern heatmap is different: it shows the underlying daily averages by hour and date rather than an index.

Accuracy views compare forecasts made at fixed lead times with the later outturn. Mean absolute error shows the typical size of a miss; bias shows whether forecasts tended to be high or low. Results are only shown where both forecast and outturn exist, and summary figures stay unavailable until enough periods overlap—missing history is never presented as a zero error.

Revision or “vintage” views compare forecasts published at different times. Their ranges describe how much retained forecasts changed, not a probability that the future value will fall inside them. Some long-range history only accumulates from the date GridTelemetry began retaining vintages because the upstream source does not publish a historical archive.

The wind confidence band is a separate GridTelemetry interpretation based on forecast-versus-outturn errors. For each source and fixed lead time, the earliest 80% of paired periods fits a central 80% outcome range and the latest 20% tests it. Observed coverage is the share of held-back outcomes inside the range; calibration gap is its distance from the 80% target in percentage points, so smaller is better. At least 672 half-hour pairs—about 14 days—are required before a band can be assessed. GridTelemetry withholds a band when held-out coverage misses the target by more than ten percentage points, while retaining the accuracy and calibration scores for inspection. Missing periods are never filled with zero.

It places sourced warnings, outages, demand, wind and balancing facts on one chronology to make a period easier to investigate. Nearby events are context, not proof that one caused another; follow the source links before drawing an operational conclusion.

The Asset Explorer lets you save a browser-only watchlist, monitor it as a fleet and compare up to four assets across the available history. The fleet workspace combines current operator declarations, retained metered history and published REMIT outage notices. Missing readings stay blank rather than being counted as zero.

The saved list is not placed in the address bar automatically. “Share fleet” first shows every asset ID that would enter the URL and adds them only after you confirm. Opening that link displays a shared fleet without saving it to the recipient’s private watchlist. Renamed, retired or missing catalogue entries remain visible until you remove them.

The bell can check official NESO notices and thresholds you choose for expected spare capacity, frequency, wholesale price, current carbon intensity in your remembered region, upcoming three-hour low-carbon or Octopus Agile cheap windows, watched-asset availability and major unplanned outages. Agile is one supplier’s retail tariff, including VAT—not a GB-wide or wholesale price. Official NESO notices are labelled and styled separately; the other conditions are GridTelemetry interpretations of published data.

Your choices, which currently active alerts you have marked as seen, and when this browser first detected each alert use versioned browser-only settings. They are not linked to an account or synced to another device. Regional alerts store only the public region ID; the postcode used to find that region is not retained. Forecast and tariff availability are checked separately, and missing or insufficient look-ahead coverage prevents an all-clear rather than becoming a zero.

Delivery is in-app only: it works while GridTelemetry is open and does not ask for browser notification permission. Reliable alerts while the site is closed would require a separate explicit opt-in, service worker and server-side push subscription, with a separate privacy and operational decision.

Carbon intensity reaches back to 2017; the Elexon-derived series (generation, demand, interconnectors and price) to around 2019; and the regional series — regional carbon intensity and regional generation mix — to 2018. Because those start dates differ, the far-left edge of a “Max” chart can look ragged where one series begins before another.

Each record is calculated from the half-hourly observations GridTelemetry holds for that series. The cards show when the available series begins, so these are records for the stated data period rather than claims about all time.

Implausible readings are excluded before records are calculated. Runs use consecutive half-hours, and any gap in the data ends a run rather than assuming the missing period continued it.

Dated milestones are sourced historical events rather than numerical records or a leaderboard. They are grouped separately so the two are not confused.

Grid frequency is excluded because only a rolling 60 days are retained, which is not enough history for a meaningful long-term record.

There are no accounts or sign-in, and GridTelemetry does not profile or sell your activity.

If you use Your Region, the postcode you enter is sent to GridTelemetry’s own API. The server reduces it to the outward half — the part before the space — and sends only that less precise value to the Carbon Intensity service to find one of the fourteen regions. The application remembers the resulting public region number, not the postcode. Your region number, asset watchlist, alert settings, planner preferences and imported tariff rows are kept in your browser; clearing local browser data removes them.

Share links can include selected assets, a region number or non-sensitive planner settings so another browser can reproduce the view. Imported tariff rows are deliberately excluded: a link can say that custom mode was selected, but the recipient must import their own file. A fleet share includes the exact watchlist asset IDs only after a confirmation panel shows them; ordinary Asset Explorer links do not include the private watchlist. Share links never contain a postcode.

The API counts requests per IP address so one heavy caller cannot overwhelm the upstream feeds. Its ordinary request log records requested URLs, so the postcode entered for a region lookup or a watchlist request’s public asset IDs can appear in the site’s size-limited container logs. Error reports are handled separately by a tracker on the same host: query strings are removed before those events are forwarded. Custom tariff rows are never sent to the API or monitoring service.

The underlying data belongs to its providers and is published under their own terms. Check the relevant provider’s licence before reusing it, especially commercially.

Use the share button in the header to copy a link to the exact page you’re looking at, and send it along with a note about what you saw. A precise link makes issues much easier to reproduce. On this page it is more precise still: opening a question puts it in the address bar, so the link you copy points at that individual answer.