Electricity submetering in commercial buildings

The submeters on a commercial site rarely add up to the main meter. Cable and transformer losses, circuits without a meter, and the permitted error of each meter and CT all open a gap between them. Installation faults open another, and only that one is a defect. A submetering scheme works when you know the size of the first three well enough to see the fourth.

Main meters, submeters and equipment data

A commercial site usually has three sources of electricity data.

The three can disagree for good reasons. An inverter reports power at its own terminals, and some models report DC power. The main meter measures at the grid connection, so its reading includes the site's cable and transformer losses, and it sees only the net flow when the site generates. Draw the circuits before you treat a difference as an error.

Why sites submeter

A main meter can show that the overnight load has risen. Submeters show whether refrigeration, ventilation or lighting caused it. At the time of the monthly peak, they show which systems were running, and the load factor calculator compares that peak with the average.

In England, Approved Document L Volume 2, paragraph 5.17, applies to new buildings, and to existing buildings when fixed building services are provided or extended. End uses such as heating, lighting and cooling should be submetered so that at least 90% of the annual consumption of each fuel can be assigned to an end use. The metering should also let the owner compare in-use energy with the design-stage forecast, monitor each tenant separately and monitor renewable systems separately. Buildings with a total useful floor area over 1,000 m² should have automatic meter reading and data collection. For the method, the paragraph points to CIBSE TM39. Approved Document L lists the 2009 edition. CIBSE replaced it in 2026 with TM39 Building metering and monitoring, which adds guidance on meter accuracy classes, MID-approved meters and the reconciliation of submeters with the utility meter.

Choose the points

Write down the questions the data must answer and who will act on each answer. Then choose the smallest set of points that answers them.

Rank the loads by estimated annual energy: nameplate kW × typical load fraction × run hours per year. A 90 kW chiller plant that runs at 50% load for 3,000 hours uses about 135,000 kWh a year. A 5 kW lift motor that runs at 20% for 2,000 hours uses about 2,000 kWh. Meter down the ranked list until the metered loads cover the share you need. For Approved Document L, that share is 90% of each fuel.

Some points need a meter whatever their size. Meter each system with a schedule, a set point or a control input, because that is where a saving can be made and then proved. Meter each tenant or cost centre that you must recharge. Meter solar PV, batteries and generators, because they change the flow through the main meter and the site totals cannot be explained without them.

Draw the metering tree

Draw the main meter at the top, each distribution board below it, and each submeter in its place. Mark the circuits without a submeter, and any meter on the other side of a transformer. The tree shows what each meter includes, and it lets you reconcile the numbers:

main meter = sum of the submeters directly below it + unmetered loads + losses

Most of the losses are in the cables between the parent meter and each board, and in any transformer between them. Resistive loss rises with the square of the current, so as a fraction of the power carried it rises in proportion to the current. At unity power factor, a sub-main designed for a 2% voltage drop at full load loses about 2% of its power as heat at full load, and about 1% at a steady half load. Estimate the loss from interval data. The monthly average understates it: a board that spends half the month at 20% load and half at 80% has the same average, but loses about 1.4% of its energy.

Where the site has PV or a battery, the main meter sees only the net flow. Meter the PV inverter output and the battery on the AC side, with separate import and export registers, and write the sign convention on the tree. Then:

site consumption = grid import − grid export + PV generation + battery discharge − battery charge

The negative power and meter reconciliation guide explains how to check the sign of each meter and compare meters over the same period.

Set a reconciliation tolerance

Set the tolerance for each parent meter before you look at the result. Base it on the accuracy class of each meter and each CT at the current it actually carries, the cable and transformer losses, and the uncertainty of each estimate for an unmetered circuit.

A CT's permitted ratio error widens as the current falls below its rating. IEC 61869-2 allows a Class 1 CT ±1% at 100% and 120% of rated current, ±1.5% at 20% and ±3% at 5%. A Class 0.5S CT holds ±0.5% down to 20%, then ±0.75% at 5% and ±1.5% at 1%. A 400 A Class 1 CT on a board that normally carries 40 A runs at 10% of its rating, between the ±1.5% and ±3% limits. Size the CT to the real load, or specify an S class where the load stays low for long periods. The CT ratio and burden calculator checks the operating point.

For example, a parent meter records 42,000 kWh in a month. Its four submeters record 38,900 kWh between them, about 9,700 kWh each. All five meters are Class 1 to IEC 62053-21, which allows ±1% over the normal current range, on Class 1 CTs that run between 20% and 100% of their rating. A lighting board without a meter is estimated at 1,800 kWh, and the fire and security panels at 300 kWh. The sub-mains run at a steady half load and were designed for a 2% voltage drop, so the losses are about 1%, or 420 kWh. The remainder that nothing explains is:

42,000 − 38,900 − 1,800 − 300 − 420 = 580 kWh, or 1.4% of the parent

Now build the error budget. Each meter with its CTs can be wrong by up to √(1² + 1.5²) ≈ 1.8%. That is about 760 kWh on the parent and about 175 kWh on each submeter. Give each estimate its own uncertainty: ±20% on the lighting (360 kWh) and the panels (60 kWh), because they come from connected load and hours, and ±50% on the losses (210 kWh). The errors are independent, so combine them as a root sum of squares:

√(760² + 4 × 175² + 360² + 60² + 210²) ≈ 940 kWh, or 2.2% of the parent

The 580 kWh remainder is inside that band, so it is not evidence of a fault. The worst case, with every error in the same direction, is about 2,090 kWh. A remainder of 4,000 kWh is outside both. It points to a missing circuit or an installation fault from the next section. Compare cumulative counters over the same start and end times, and track the remainder each month. A remainder that grows usually means a new load was connected without a meter.

Find installation faults

Each of these faults leaves a signature in the data. Check for them at commissioning, before the first reconciliation:

When a meter is replaced, record the time, the old meter's last counter reading and the new meter's first reading. The energy for a period that spans the swap is the old meter's advance up to the swap plus the new meter's advance after it. When a CT is replaced with one of a different ratio, store the new ratio with the date it took effect, so that the data before that date keeps its old scaling.

Reuse the meters you have

Main switchboards, chillers and generator sets often have multifunction meters with an RS-485 port that was never wired. List each meter with its model, its interface, its CT ratio and whether another system already reads it:

Modbus RTU allows one master on each RS-485 bus. If a BMS already polls the meter, agree who reads it, and how, before you connect.

Billing tenants from submeters

In Great Britain, a gas or electricity meter used for billing, by a supplier or by a landlord, must be of an approved design. For electricity, the approval is under the Measuring Instruments Regulations 2016 or the Meters Regulations 1998 (Office for Product Safety and Standards). In the EU, the Measuring Instruments Directive covers active energy meters in Annex V (MI-003). It defines classes A, B and C. Where a member state requires measurement for commercial or light industrial use, it must allow any Class B meter, and it may require Class C for specified purposes. The annex covers the meter only. External CTs are outside it.

A Class 0.5S rating to IEC 62053-22 is a performance class, not an approval. Ask for the approval certificate of the exact meter model before you bill from it. The high-accuracy metering guide explains the difference.

Ofgem's maximum resale price direction applies when electricity is resold for domestic use or for use in any form of accommodation, including holiday lets. The reseller may then charge no more per unit than they paid. The standing charge is shared pro rata to the units each user buys, and a reseller that occupies part of the premises or runs common services there takes its own share.

The direction does not cover resale to commercial tenants, so the lease sets the charge. Write the method into the lease: the unit rate, how the standing and capacity charges are shared, and how the energy of common areas is recovered. Give the common areas their own meter where you can. A common-area figure calculated as the remainder also carries every meter error and every unmetered tenant load.

Wired or wireless

Most of the cost of a wired submeter is in its data cable: the route, the containment, the fire-stopping at each compartment wall and the time to install it in an occupied building. The meter is a small part. The wired vs wireless guide compares the two, including RS-485 bus limits and radio capacity.

A wireless meter removes the data cable. The electrical work stays. To measure kW, the meter needs a fused voltage connection on each phase, and a qualified electrician must install it.

Survey the radio before you order. EpiSensor's Zigbee devices work at 2.4 GHz over up to 50 m indoors between devices, and every mains-powered device relays for its neighbours, so each installed meter extends the mesh. A closed steel board or a plant room behind a steel door can stop the link completely. Test at the real mounting position, with the board and plant-room doors shut. If a plant room fails, put a mains-powered device where it has a link to both the plant room and the nearest working device, then test again.

Choose the data interval

Match the interval to the use of the data:

Store the cumulative kWh counter as well as interval values. The counter lets you recalculate any interval later, and it survives a gap in communication. Record a missing reading as missing. Never record it as zero.

Timestamp every value in UTC from a synchronised clock, and state whether each interval is labelled by its start or its end. Align the intervals with the supplier's settlement periods, which are 30 minutes in Great Britain. An interval offset by a few minutes will not line up with the peak on the bill. The telemetry time guide explains source time and clock synchronisation.

Worked example: an overnight load

A building uses 18 kW overnight. With the building empty, the operator expects 10 kW. The difference is 8 kW for 10 hours, or 80 kWh a night. At £0.25/kWh, that is £20 a night:

Submeters on the main boards put most of the difference on the ventilation board. The air-handling units run all night because a schedule was overridden and never restored.

After the fix, measure the saving on the ventilation board meter. The main meter would mix in every other change in the building. Take the median overnight energy over two comparable weeks before and after the fix, with the same day types and similar outdoor temperatures. If the median falls from 95 kWh to 23 kWh, the saving is 72 kWh a night, or about 26,300 kWh and £6,570 a year. Report that figure. The first night after the fix, multiplied by 365, is not a measurement. The overnight baseload guide covers baseload in more detail.

Start with a pilot

Choose a pilot area that contains the difficult cases: a CT sized well above its load, a board fed through a long sub-main or a transformer, a split-core CT fitted on an existing cable, and the worst radio position on the site, which is often a basement plant room. Commission every point from the conductor to the report. Disconnect the communication of one meter for an hour, then restore it. The history must show a gap, and the counter must resume without a step. The commissioning checklist lists the records to keep.

Expand when the pilot boards reconcile with their parent meter within the tolerance you set, for at least four weeks, with every gap flagged. The guide to choosing an energy monitoring system lists the questions to ask each supplier before the pilot.

Submetering with EpiSensor

A ZEM electricity monitor ships with its current sensors connected and calibrated to it. EpiSensor rates the meter and its sensors together as Class 0.5S to IEC 62053-22, so the CT term in the error budget above is already inside the meter's class, and there is no CT ratio to set on site. Each sensor must still go on the conductor of its own phase. The ZEM sends its readings over the Zigbee mesh. Its accuracy class is not a billing approval, so apply the tenant billing rules above before you recharge from it.

A ZMB Modbus interface, a ZHM M-Bus interface or a ZPC pulse counter reads a meter that is already installed. All of them report to a Gateway running Edge, which keeps the history on site.

In Edge, a calculated device can hold the reconciliation. With the parent meter as the first input, the expression $1-($2+$3+$4) publishes the remainder under that parent as its own sensor, which you can chart and export. Set its stale action before you rely on it. With the default, Use the last value, a silent submeter keeps counting at its last reading, so the remainder looks normal. With Treat as zero, the remainder jumps by that meter's load and looks like a new unmetered load. Set When an input is stale to Emit null instead of computing. The remainder then publishes nothing until every input reports again, and the history shows a gap. The stale-input threshold is five minutes by default. The tenant submetering application shows a complete system, and the metering and CTs series covers CT selection and meter accuracy in more depth.