Reviewing Agreed Supply Capacity (kVA): A Practical Guide to Maximum Demand, Power Factor and DNO Applications

Agreed supply capacity is one of the few network cost inputs a business actually controls. It is the kVA figure your distribution network operator (DNO) holds against your MPAN, sometimes labelled Maximum Import Capacity, Available Supply Capacity or Maximum Power Requirement. You pay for it every day of the year, used or not, and from April 2026 the cost of getting it wrong has grown, because capacity now drives banded transmission charges as well as distribution availability charges.

Most capacity figures were set at connection, or last amended after a plant upgrade that never fully materialised. Sites that have swapped gas-fired ovens for smaller electric units, replaced lighting, closed a production line or moved to inverter-driven plant are frequently sitting on 30 to 50 per cent headroom they will never need. The review below is the one I run before recommending any change.

Why 2026 makes this worth doing now

Two charging elements attach to your kVA figure.

Distribution use of system (DUoS) includes an availability charge in pence per kVA per day, set out in each DNO's published charging statement under the Distribution Connection and Use of System Agreement. Rates vary by region and connection voltage, and DNOs publish them well ahead of the charging year, so you can model a change with confidence.

Transmission network use of system (TNUoS) is the bigger issue. Following Ofgem's Targeted Charging Review, the demand residual is recovered as a fixed, banded charge rather than through consumption, with non-domestic users banded by voltage and by agreed capacity or consumption. NESO sets out the mechanics in its own primer, TNUoS in 10 Minutes. Final 2026/27 tariffs were published on 30 January 2026, and analysis of those tariffs by Drax puts the increase in residual charges for consumers at between roughly 28 and 116 per cent, with a volume-weighted average near 64 per cent. For most consumers the residual is now the dominant part of the TNUoS bill, so the band your capacity sits in matters more than it did two years ago. Triad avoidance retains some value against the locational element, but far less than under the pre-2023 arrangements, as NESO's code modification papers on reintroducing demand locational signals discuss.

Reading maximum demand properly from half-hourly data

Start with at least 12 months of settlement data, ideally 24, so you capture a full winter and any seasonal peak. Half-hourly data is reported as kWh per settlement period. Average power in that period is the kWh figure multiplied by two, so 90 kWh in a half hour equals 180 kW average.

That gives you kW, not kVA, and it gives you a half-hour average rather than an instantaneous peak. Three points follow:

• Rank the whole year and look at the top 20 to 50 half hours, not just the single highest. A one-off peak caused by a commissioning test or a heatwave should be understood, not designed around.
• Check whether the peak is repeatable. If your top ten half hours cluster on hot August afternoons, that is your real design condition.
• Allow for short-duration starting loads and diversity that a half-hour average smooths out. Motor starts and welding plant exceed the averaged figure, as do chillers that cycle together.

Many HH meters also record reactive energy in kVArh, and some record maximum demand in kVA directly. Ask your supplier or data collector for the reactive channel alongside the active data. You are entitled to your consumption data, and the industry framework for HH settlement sits with Elexon. Without reactive data you are guessing at the conversion, and guessing is where reviews go wrong.

Power factor pitfalls

Apparent power in kVA equals active power in kW divided by power factor. At unity, 400 kW is 400 kVA. At 0.85, the same 400 kW needs 471 kVA. At 0.75 it needs 533 kVA. Applying a single assumed power factor across a year is the most common error I see in capacity reviews.

Power factor is not constant. It is usually worst at low load, when transformer and motor magnetising current dominates, and better at high load. Averaging the annual power factor and applying it to the annual peak kW can understate or overstate peak kVA by a wide margin.

Correction equipment fails quietly. Capacitor banks lose stages, fuses blow and contactors stick, and nothing on site alarms. If reactive consumption has crept up over two years with flat active consumption, suspect the bank before you suspect the load.

DUoS also carries a reactive power charge in pence per kVArh in most regions, usually applied above a threshold power factor. Poor power factor therefore costs you twice: it inflates the kVA you must reserve, and it adds a separate charge line. Fixing it before you set a new capacity figure often lets you reduce kVA further than expected. Note the opposite risk too: modern inverter-fed loads, LED drivers and some solar inverters can push a site leading at light load, which some tariffs also penalise.

One more constraint shapes how deep you cut. Since DCP 161 took effect in April 2018, exceeding your agreed capacity attracts a penal rate on the excess kVA, commonly around three times the standard availability rate, so the downside of cutting too far outweighs the saving.

Banding effects, and why small changes can be large

Because the TNUoS demand residual is banded, savings are lumpy rather than linear. Trimming 100 kVA may save only the DUoS availability charge if you stay inside the same band, while trimming 120 kVA might drop you a band and remove a much larger fixed charge. Model both charges together, band boundaries included, before advising a client. The same logic applies in reverse: a modest capacity increase for an EV charging hub or a heat pump can tip a site into a higher band and cost far more than the availability rate implies.

Check the recorded voltage and connection category as well. Sites metered at low voltage but recorded as HV, or the other way round, are banded incorrectly and charged incorrectly. A data error here is worth more than any tariff negotiation.

Approaching the DNO

DNOs accept capacity change requests through their connections teams, usually via an online form. National Grid Electricity Distribution, for example, sets out its process for applying for a reduction in capacity and will assess historic metered demand as part of that. Practical points decide these applications:

• Expect the DNO to check your own data. Submit 12 to 24 months of HH data, your calculated peak kVA, and the proposed figure with headroom stated.
• Many DNOs limit reductions to one request in any 12 month period, so make the number right first time.
• A reduction varies your connection agreement. Getting capacity back later means a fresh application and a new quote. It may also mean reinforcement or a queue position, even though Ofgem's Access Significant Code Review moved much reinforcement cost away from upfront connection charges from April 2023.
• Ask when the change takes effect for charging purposes and confirm the revised figure reaches your supplier, since availability charges are billed from the DNO record.
• Include a letter of authority, MPAN, meter serial and site contact to avoid a fortnight of email.

Where the driver is future growth rather than saving, flexible or timed connection offers are worth a separate conversation with the connections team.

What a broker can do this week

Pull the HH data for every client site above 100 kVA and build one line per MPAN: agreed kVA, top ten half-hourly kW peaks, calculated peak kVA using measured reactive data, annual power factor at peak, current availability rate, and the TNUoS band. Flag any site using under 70 per cent of its agreed capacity for 12 months, and any site with worsening power factor. Then price two scenarios: reduction to measured peak plus 15 to 20 per cent headroom, and reduction to the nearest band boundary below. Take the second only where the operational case supports it, and put the assumptions in writing to the client before you file anything with the DNO.