G99 Grid Connection for Data Centre Solar
Every data centre rooftop PV array above 50kW needs a G99 connection agreement from its DNO — here is how the application works, why a data centre's electrical environment makes it more complex than a standard commercial install, and how we manage the whole process end-to-end.
A G99 grid connection is the formal agreement between you and your Distribution Network Operator (DNO) that permits an embedded generator — your solar PV array — to operate in parallel with the grid. For any data centre rooftop system above 50kW, which is virtually all of them, G99 is mandatory before energisation. The application is governed by Engineering Recommendation G99, the standard that replaced G59 for all new generation, and it is approved (or refused, or conditioned) by the DNO that owns the network at your site. The statutory target for a DNO decision is 65 working days, but a data centre's distinctive electrical environment — high prospective fault levels, dense protection coordination, and frequently constrained grid zones — means a poorly prepared application can stall well beyond that. A well-structured zero-export application, by contrast, is often the fastest and lowest-risk route to energisation.
This page explains what G99 is and when it applies, why a data centre's G99 is genuinely different from a warehouse or office install, why the overwhelming majority of data centre arrays adopt a zero-export configuration, the step-by-step DNO process, the protection and witness-testing requirements at commissioning, and how we manage all of it on your behalf.
What G99 is and when it applies
Engineering Recommendation G99 is the technical framework that governs the connection of generation equipment to the UK distribution network. It sets out the protection settings, power quality limits, fault ride-through behaviour, and commissioning evidence a DNO requires before allowing your generator to run synchronised with their network. The threshold that matters for data centres is straightforward:
- Up to and including 50kW per phase (or 50kW total for single-phase): a simpler, fast-track G98 "connect and notify" route can apply for fully type-tested microgeneration.
- Above 50kW: a full G99 application and connection agreement is required before the system is energised. The DNO must assess the impact on their network and issue terms.
A meaningful data centre rooftop array — typically several hundred kW to multiple MW where roof area allows — sits firmly in G99 territory. Because on-site PV at a data centre covers only around 5–15% of the facility's annual load (the roof is the constraint, not the demand), the array is almost always smaller than the site's import capacity, which is helpful: it means the generation can frequently be accommodated without major reinforcement. But the threshold is per the generator's registered capacity, not the net site position, so the G99 obligation stands regardless.
Why a data centre's G99 is different
The DNO's core question in any G99 assessment is: what does this generator do to the network, and what does the network do to this generator under fault? At a data centre, the answers are unusual, and that is precisely why a generic commercial-solar G99 submission falls short.
High prospective fault levels
Data centres are electrically dense. Behind the meter you typically have multiple large transformers, a substantial UPS estate (often rotary or static with significant fault contribution), and standby diesel or gas generators sized to carry the entire critical load. Each of these is a source of fault current. The prospective short-circuit current at the point of connection is therefore far higher than at a comparable warehouse or office, and the solar inverters, AC combiner switchgear, and protection devices must be rated and coordinated against that elevated fault level. A fault-level calculation that ignores the standby plant and UPS contribution is not just incomplete — it can specify under-rated equipment that the DNO will reject or, worse, that fails in service.
Protection coordination across a complex topology
A data centre's distribution is built for resilience: A and B feeds, automatic transfer switches, bus-coupled boards, and segregated critical/mechanical supplies. Inserting a generator into this topology means the new G99 protection must coordinate with existing upstream and downstream protection without compromising the facility's fault discrimination or its Tier III / Tier IV resilience classification. The interface protection cannot be allowed to mis-operate during a transfer event, generator test, or upstream disturbance and trip the array — or, far more seriously, interfere with the protection that keeps the critical load alive.
Constrained grid zones
The UK's densest data centre clusters — Slough and the Thames Valley, London Docklands, West London — sit in some of the most heavily loaded distribution networks in the country. Available headroom for new generation is often limited, and DNOs in these zones scrutinise export-capable applications closely. This is one of the strongest practical arguments for the zero-export approach below: it removes the network's principal objection before it can be raised.
Power-density and load-profile context
An AI data centre running GPU compute at 40–120 kW per rack draws a heavy, flat, 24/7 baseload — which is exactly why on-site PV achieves close to 100% self-consumption with no export. That flat baseload also shapes how protection should be set: the array effectively never back-feeds the grid, which materially simplifies the case you put to the DNO.
The zero-export approach
The single most important design decision in a data centre G99 application is usually the choice to operate zero export — to formally waive the right to push any generation back onto the grid. Because a data centre's IT load runs continuously and dwarfs its rooftop generation, the array's entire output is consumed on site at every moment of the day. There is no commercial loss in waiving export, and there is a substantial procedural gain.
| Aspect | Export-capable G99 | Zero-export G99 (typical for DCs) |
|---|---|---|
| Network impact study | Full assessment of export onto constrained network | Minimal — no power flows back across the boundary |
| Reinforcement risk | Possible upstream reinforcement, long lead time | Largely removed |
| Required control | Standard interface protection | Interface protection plus certified export-limitation / reverse-power scheme |
| DNO scrutiny | High in dense zones | Substantially reduced |
| Commercial value lost | n/a | None — ~100% self-consumption anyway |
Zero export is enforced by a certified scheme — typically reverse-power or export-limitation protection that holds inverter output at or below site demand and trips the array if any reverse flow toward the grid is detected. To the DNO this transforms the application: the array can no longer be a source of export onto their constrained network, so the principal grounds for refusal or for demanding reinforcement disappear. The result is a faster, cleaner, lower-risk route to a connection agreement. Where a site later adds battery storage, the same export-limitation philosophy extends to the storage system, keeping the whole installation within a single coherent zero-export envelope.
The DNO application process and the 65-day timeline
The G99 process follows a defined sequence. The DNO's statutory target for issuing a connection offer is 65 working days from receipt of a valid, complete application — roughly three calendar months — but that clock only starts when the submission is genuinely complete. Incomplete fault-level data, missing single-line diagrams, or an ambiguous export position are the most common causes of the clock being paused or reset. Getting the submission right first time is the difference between energising on schedule and slipping a quarter.
The detailed step-by-step process is set out in the how-to steps accompanying this page. In summary it moves from a pre-application network check, through the formal application and the DNO's connection offer, to acceptance, installation, witness testing, and final commissioning. Throughout, the quality of the technical pack you submit — fault-level calculations, protection settings, single-line diagrams, and the zero-export scheme description — determines how smoothly the DNO can say yes.
Protection relays, settings, and witness testing
The interface protection is the technical heart of any G99 connection. It is the device that disconnects your array from the grid within milliseconds if the network conditions move outside the agreed envelope — loss of mains, over/under voltage, over/under frequency, vector shift, or reverse power. For a data centre, this protection has to be both reliable enough to satisfy the DNO and stable enough to ride through the site's own internal events (generator tests, transfer switching, UPS load steps) without nuisance tripping.
- Relay selection: we specify industrial-grade interface protection relays — typically SEL (Schweitzer Engineering Laboratories) or Woodward devices — chosen for their proven G99 compliance, settings flexibility, and event-recording capability for post-event diagnostics.
- Settings: protection settings are derived to the DNO's required G99 thresholds and then coordinated against the site's existing protection grading so the interface relay operates correctly without compromising fault discrimination on the critical-power path.
- Loss-of-mains protection: configured to the DNO's currently mandated method, with the relay's measurement validated against the site's fault-level environment.
- Witness testing: at commissioning, the DNO (or its approved agent) witnesses live injection testing of the interface protection — verifying that every trip function operates within its time and threshold limits, and that the array disconnects and stays disconnected as required. Only once witness testing is signed off can the connection be formally energised and the agreement completed.
How we manage the G99 end-to-end
We run the G99 connection as a managed work-stream alongside the PV and electrical installation, so it never becomes the thing that holds up energisation. Our scope covers:
- Fault-level calculations that correctly account for the data centre's transformers, UPS estate, and standby generation — not a generic commercial assumption.
- Chartered-engineer peer review of the protection design, settings, and fault-level study before anything is submitted, so the technical pack is right first time.
- Region-specific DNO liaison. We deal directly with the relevant operator: SSEN for the Thames Valley, South East and Scotland (covering Slough and the wider Slough cluster); UKPN for London, the East and South East; Electricity North West for Greater Manchester; and NGED for the Midlands, South West and Wales. Knowing each DNO's portal, expectations, and assessment quirks shortens the cycle.
- Zero-export scheme design and certification, so the application presents the DNO with the simplest possible case.
- Commissioning and witness testing management, from coordinating the DNO witness visit to producing the final compliance evidence.
This G99 work doesn't sit in isolation — it is part of the full electrical scope, which we cover in depth in our blog on data centre three-phase electrical works for solar PV. And because the connection design is settled early, it feeds cleanly into the overall cost and feasibility picture for your project, with no late surprises from the DNO.
Founded in 2012 with 350+ commercial installs and 24+ MW commissioned, our crews are BPSS-cleared as standard for work inside live data centre environments, with SC clearance available. We are MCS Commercial, NICEIC Approved and TrustMark Licensed, and every G99 protection design is peer-reviewed by a chartered engineer before submission.
Can solar power a mission-critical data centre?
\nHonestly: not on its own, and any installer who claims otherwise misunderstands the load. A rooftop array is a daytime, weather-variable source, while a data centre is a flat 24/7 baseload that cannot tolerate a flicker. On-site PV realistically covers 5–15% of annual energy, and zero kilowatts at 2am. Solar is not, and was never, a resilience source for the critical IT load — the standby generators and UPS estate remain the resilience source. That is the concession. Here is the part competitors leave out.
\nSolar earns its place when it is integrated around the resilience architecture rather than into it. The correct method keeps PV strictly on the non-critical / mechanical side of the topology — chilled-water plant, CRAC/CRAH units, pumps, lighting, ancillary loads — so it offsets the parasitic energy that drives PUE without ever sitting in the A/B critical-power path. The array, its inverters and its interface protection are downstream of the points that feed the UPS and standby plant, so a PV fault, an inverter trip or a grid disturbance physically cannot propagate to the load. This is the same design discipline behind Tier III / Tier IV resilience: the solar system is an energy-cost optimisation layer, not part of the fault-tolerant distribution.
\nPaired with battery storage the value compounds — the BESS time-shifts surplus midday generation into evening mechanical load and provides a controllable buffer, while remaining export-limited and segregated from the critical bus. The grid stays the firm supply; solar-plus-BESS simply reduces how much energy you buy at 18–32p/kWh and replaces it with on-site generation at 3–5p/kWh LCOE. The integration sequence — array sizing against mechanical load, segregation from the critical path, export-limitation, then G99 — is exactly what we set out in our data centre solar systems methodology. So the honest answer is: solar cannot power a mission-critical data centre, but designed this way it can meaningfully de-carbonise and de-cost it without ever touching its resilience.
\n\nThe G99 process step by step, with typical durations
\nThe headline 65-working-day figure is only the DNO's assessment window — it is one stage in a longer sequence. Below is the full path from first network check to energisation, with indicative durations for a well-prepared data centre application. Timings assume a complete, peer-reviewed technical pack; an incomplete submission pauses or resets the assessment clock entirely.
\n| Stage | What happens | Typical duration |
|---|---|---|
| Pre-application network check | Assess local DNO headroom, prospective fault level and import/export boundary; confirm the zero-export strategy before lodging anything formal. | 1–3 weeks |
| ENA EREC G99 application | Compile and submit the technical pack — fault-level calculations, single-line diagrams, proposed settings and the certified zero-export scheme — via the DNO portal. The 65-working-day clock starts on a valid submission. | 2–4 weeks to prepare |
| DNO assessment & connection offer | Statutory target for the operator to assess network impact and issue terms, conditions and any reinforcement requirement. | Up to 65 working days (~3 months) |
| Offer review & acceptance | Review terms against the design, challenge anything unreasonable, agree commercials, then accept. | 1–4 weeks |
| Protection / relay design finalisation | Lock relay specification (e.g. SEL or Woodward) and settings to the agreed G99 thresholds, coordinated with the site's existing protection grading. | 2–3 weeks (parallel with install) |
| Witness testing | DNO or approved agent witnesses live injection testing of every interface-protection trip function — loss of mains, voltage, frequency, reverse power. | 1 day on site |
| Energisation & agreement completion | On a clean witness-test sign-off the connection is energised and the G99 agreement is formally completed. | Immediate on sign-off |
End to end, a typical data centre G99 runs around four to six months from first network check to energisation — and the single biggest lever on that timeline is the quality of the application that starts the 65-day clock.
\n\nG99 vs G100 for data centre export
\nG99 and G100 are complementary, not alternatives, and the distinction matters for how a data centre frames its connection. G99 is the connection agreement itself — the DNO's permission for your generator to run in parallel with the network, covering protection, power quality and commissioning. G100 is the Engineering Recommendation that governs how an export-limitation scheme must perform if you rely on active limitation to cap or eliminate export. In other words, G99 says you may connect; G100 says how your export-limiting control must behave to be trusted.
\n- \n
- Hard zero-export via a fixed reverse-power relay that simply trips on any reverse flow often avoids a full G100-compliant scheme, because there is no active limiting being relied upon — the array is disconnected, not modulated. \n
- Active export limitation (holding inverter output to track site demand, allowing some flexibility) relies on G100-compliant control with defined response times and fail-safe behaviour, so the DNO can grant a higher connection capacity without reinforcement. \n
For most data centres, where the continuous IT and mechanical baseload dwarfs rooftop generation, hard zero-export is the cleanest route. Where a larger array or future battery storage makes active management worthwhile, a G100-compliant limitation scheme submitted within the G99 application keeps the whole installation export-limited and the DNO's objections minimal.
\n\nDNO vs IDNO connection routes
\nNot every data centre connects through the incumbent regional DNO. Newer and purpose-built facilities — particularly on industrial parks, regeneration sites and large campuses — are frequently served by an Independent Distribution Network Operator (IDNO) that owns the local last-mile network, even though the wider area is a DNO licence zone. Knowing which body owns the network at your point of connection determines who assesses and approves the G99 application.
\n| Aspect | DNO route | IDNO route |
|---|---|---|
| Network owner | Incumbent regional operator (SSEN, UKPN, NGED, Electricity North West) | Independent operator owning the on-site / last-mile network |
| Typical site | Established facilities on legacy distribution | New-build campuses, industrial parks, regeneration developments |
| Who approves G99 | The regional DNO directly | The IDNO, usually coordinating upstream with the host DNO |
| Process & standard | Same ENA EREC G99 standard | Same G99 standard; portal, contacts and lead times differ |
| Common pitfall | Assuming default regional contacts | Submitting to the DNO when the IDNO owns the connection point |
The practical risk is wasted weeks: an application lodged with the regional DNO when an IDNO actually owns the point of connection gets bounced. Our pre-application network check establishes ownership of the connection point first, so the G99 goes to the right body the first time — whichever network your data centre solar system physically connects to.
\n\nReverse-power and export-limitation protection
\nReverse-power protection is what makes a credible zero-export claim enforceable rather than aspirational. It is a dedicated relay function that continuously measures real power flow at the grid boundary and acts the instant any power tries to flow toward the network — disconnecting the array before export can occur. For a data centre, where export was never the point, it is the mechanism that lets you present the DNO with a generator that cannot, by design, burden their network.
\n- \n
- Measurement point: at the point of connection / main intake, sensing net flow across the boundary so it accounts for total site demand, not just the array. \n
- Action on reverse flow: trips the inverters or the array's AC interface within the DNO-mandated time, holding it disconnected until conditions are safe. \n
- Stability under site events: set to ride through internal disturbances — generator tests, ATS transfers, UPS load steps — without nuisance tripping, which is critical inside a live Tier III / Tier IV environment. \n
- Certification: documented and, where active limitation is relied upon, demonstrated to G100 performance so the DNO accepts it as a firm constraint on export. \n
This protection is verified at the witness-testing stage, where the DNO sees the reverse-power trip operate live before the connection is energised. Done properly, it removes the network operator's single biggest objection — the prospect of unmanaged export onto a constrained grid — and is the technical foundation of the fast, low-risk zero-export connections we design for data centre clients.
Frequently asked questions
Do all data centre solar installations need a G99 connection?
Effectively yes. G99 applies to any generator above 50kW connecting in parallel with the grid, and a meaningful data centre rooftop array — typically several hundred kW to multiple MW — is well above that threshold. Only very small sub-50kW systems can use the simpler G98 connect-and-notify route, which is rare for data centres. The G99 agreement must be in place before the array is energised.
What does zero export mean and why do data centres choose it?
Zero export means formally waiving the right to feed any solar generation back onto the grid; a certified reverse-power scheme trips the array if reverse flow is detected. Data centres choose it because their continuous 24/7 IT baseload consumes nearly 100% of rooftop output anyway, so no commercial value is lost. In return, the DNO application is dramatically simplified, particularly in constrained grid zones like Slough and London.
How long does a G99 application take?
The DNO's statutory target is 65 working days — about three calendar months — to issue a connection offer after receiving a valid, complete application. That clock only starts once the submission is genuinely complete, so missing fault-level data or an unclear export position can pause or reset it. A well-prepared zero-export application is usually the fastest route through the process.
Why is a data centre's G99 more complex than a normal commercial solar install?
Data centres have very high prospective fault levels from their transformers, UPS estate and standby generators, all of which contribute fault current that the protection and switchgear must be rated against. The distribution topology is also far more complex — A/B feeds, transfer switches, segregated critical supplies — so new G99 protection must coordinate without compromising resilience. Many DC clusters also sit in constrained grid zones requiring extra DNO scrutiny.
Which protection relays are used for data centre G99 connections?
We specify industrial-grade interface protection relays, typically SEL (Schweitzer Engineering Laboratories) or Woodward, for their proven G99 compliance, settings flexibility and event-recording capability. Settings are derived to the DNO's required thresholds and coordinated against the site's existing protection. At commissioning the DNO witnesses live injection testing of every trip function before the connection can be energised.
Which DNO handles my data centre's G99 application?
It depends on your region. SSEN covers the Thames Valley, South East and Scotland, including the Slough cluster; UKPN covers London, the East and South East; Electricity North West covers Greater Manchester; and NGED covers the Midlands, South West and Wales. We liaise directly with the relevant operator on your behalf, which shortens the cycle because each DNO has its own portal and assessment expectations.
Do data centres need DNO approval for solar?
Yes. Any solar PV array above 50kW connecting in parallel with the grid requires a G99 connection agreement before energisation, and a meaningful data centre rooftop system is well above that threshold. Approval comes from whoever owns the network at your point of connection — usually the regional DNO (SSEN, UKPN, NGED or Electricity North West), but on new-build campuses and industrial parks it is often an Independent DNO (IDNO). Establishing which body owns the connection point before applying avoids weeks of wasted time submitting to the wrong operator.
What is G99 for a data centre?
G99 is the Engineering Recommendation governing how an embedded generator — your solar array — connects to and operates in parallel with the UK distribution network. For a data centre it sets the protection settings, power-quality limits and commissioning evidence the DNO requires. A data centre's G99 is more demanding than a standard commercial one because of very high prospective fault levels from transformers, UPS and standby generators, complex A/B distribution topology, and frequently constrained grid zones — so the fault-level study and protection design must account for the full critical-power environment, not a generic assumption.
Can a data centre export solar to the grid?
It can, but it almost never should. Because a data centre runs a continuous 24/7 baseload that dwarfs its rooftop generation, the array's entire output is consumed on site at roughly 100% self-consumption, so there is no commercial value in export. Instead, data centres adopt a zero-export configuration enforced by reverse-power or export-limitation protection that trips the array if any flow toward the grid is detected. This removes the DNO's principal objection in constrained zones like Slough and London and delivers a faster, lower-risk connection.