High-accuracy electricity metering

A Class 0.5S meter on a Class 0.5S current transformer (CT) can read 2 % high or low on a feeder at power factor 0.5, and both devices are still inside their class limits. A wrong CT ratio in the configuration, or a CT on the wrong phase, can double the reading or remove it altogether. This guide gives the limits in the meter and CT standards, shows how they combine, and sets out how to size CTs, commission an installation and reconcile it against a reference.

Define the purpose first

Write down the use before you choose the meter. The use sets the measurement boundary, the uncertainty you can accept, the reporting interval and the approvals you need.

Savings are calculated from metered data. ISO 50015 and the Efficiency Valuation Organization's IPMVP both derive savings from a baseline period, a reporting period and adjustments. The meter's accuracy bounds one input to that calculation.

What an accuracy class says

IEC 62053-22:2020 applies to transformer-operated static watt-hour meters of classes 0.1S, 0.2S and 0.5S, and to their type tests only. For a balanced load, the Class 0.5S limits are:

In is the meter's rated current, usually 1 A or 5 A for a meter fed from CTs. The limits apply at reference conditions: reference temperature, rated voltage and frequency, and sinusoidal waveforms. Temperature, harmonics, voltage unbalance and external magnetic fields each have their own permitted variation on top.

The class covers active energy only. The kW, voltage, current, power factor and THD values on a multifunction meter fall outside it. IEC 61557-12 covers those functions in power metering and monitoring devices, and reactive energy has its own classes in IEC 62053-23 and IEC 62053-24. The 2020 edition also excludes metering systems made from several physically remote devices, and meters for low-power instrument transformers when they are tested without those transformers.

The S classes have limits down to 1 % of In. Classes 0.5, 1 and 2 are in IEC 62053-21. A note in IEC 62053-22 states the pairing: S-class meters are intended for S-class CTs, whose range starts at 1 % of rated current. A standard CT on an S-class meter is a matter for agreement between manufacturer and purchaser.

Resolution and accuracy

A register that shows 12,450.237 kWh is not more accurate than one that shows 12,450.2 kWh. Store enough digits for the calculation, and report only the digits the measurement supports.

Error and uncertainty

Error is the difference between the measured value and the true value, and in service you do not know it. Uncertainty is the dispersion of values you can reasonably attribute to the measurand, from the information you have. A class limit is one piece of that information. You turn it into an uncertainty with a model, as in the worked budget below.

The complete measurement chain

Current sensors

IEC 61869-2:2012 sets the limits for measuring CTs as a percentage of rated primary current:

These limits hold for any burden from 25 % to 100 % of the CT's rated output. A conventional 1 A or 5 A CT therefore has a burden window. The meter input and the complete lead loop must stay inside it. Use the CT ratio and burden calculator for that check. It applies to current-output CTs only, not to voltage-output CTs or Rogowski coils, and it does not check saturation, insulation or physical fit.

Phase error at low power factor

The meter multiplies voltage by current. A CT that shifts the current by an angle δ changes the measured power from V × I × cos φ to V × I × cos(φ − δ). For small δ, the relative error is about δ × tan φ, with δ in radians. The sign depends on whether the CT leads or lags and on whether the load is inductive or capacitive.

At unity power factor, phase displacement has almost no effect. At power factor 0.5, the 30 minutes a Class 0.5S CT is allowed at rated current is worth three times its ratio limit. Lightly loaded motors, compressors on unload and transformers at no load all run at low power factor.

CT sizing and low current

Size the CT for the normal load, not the breaker. For example, a 1000/5 A Class 0.5 CT on a feeder that normally draws 60 A runs at 6 % of rated current. At the 5 % test point, Class 0.5 allows ±1.5 % ratio error and ±90 minutes of phase displacement. At power factor 0.8 the phase term alone is worth up to 2.0 %. When the feeder drops to 20 A overnight, the CT runs at 2 % of rated current, where Class 0.5 has no limit.

A 150/5 A Class 0.5S CT on the same feeder runs at 40 % of rated current, inside ±0.5 % and ±30 minutes. Before you choose the smaller CT, check that the maximum load stays inside its rated extended primary current and that the fault level stays inside its short-time thermal rating.

A practical rule: keep normal load between 20 % and 120 % of the CT rating, and keep minimum load above 5 %, or above 1 % with an S-class CT.

Voltage and phase association

For a balanced three-phase load, P = √3 × V × I × PF, where V is the line-to-line RMS voltage. The meter calculates power phase by phase, so each current must be paired with its own phase voltage.

At 400 V, 100 A and power factor 0.9, each phase carries 20.78 kW. Two common wiring faults give these readings:

• One CT reversed: that phase reads −20.78 kW, and the total reads 20.78 kW instead of 62.35 kW.
• The CTs on L1 and L2 swapped: each swapped channel pairs its voltage with a current 120° away. Those two channels read power factors of about −0.83 and −0.07, L3 reads 0.90, and the total reads close to 0 kW.

In both cases every channel reads 100 A. A current check alone passes both faults. The kW, kVA, kvar and power factor calculator gives the expected kW, kvar and power factor for each phase. Also check phase rotation, and check that import and export follow the meter's sign convention.

Installation

Sensor direction, phase mapping, split-core closure, conductor position, voltage reference location and the configured ratio all change the result. No meter class compensates for a sensor on the wrong phase or a wrong ratio in software.

Data completeness and time

A lost reading does not change the energy, but it can change how the energy is allocated. Take a meter that reports its cumulative register every 15 minutes. The 10:15 reading is lost, the 10:00 reading is 12,400.0 kWh and the 10:30 reading is 12,450.0 kWh. The half hour from 10:00 to 10:30 used exactly 50.0 kWh, but the split between the two quarter hours is unknown. If the 10:15 and 10:45 readings are both lost, the two half hours cannot be separated. Only their total is known.

Flag interpolated values so that nobody bills on them as measured data. Keep the source timestamp, the clock status and the gap-handling rule with each value. The ZEM-65 keeps a real-time clock synchronised by NTP through the Gateway, holds the clock for up to 72 hours without power and logs up to 70,000 data points in non-volatile memory.

Combining accuracy classes

Add class limits only when each one bounds the same quantity under the same conditions. The worked budget below uses a Class 0.5S meter on a Class 0.5S CT, with the load at 20 % of the CT rating (so the meter is at 20 % of In), power factor 0.5 lagging and reference temperature.

The linear sum is a worst case and is a safe screening figure. The root-sum-square figure treats each limit as the half-width of a rectangular distribution (JCGM 100:2008, 4.3.7) and assumes the contributors are independent. The CT's ratio and phase errors both come from its magnetising current, so they are often correlated, and the root-sum-square result can be optimistic.

At unity power factor the same pair gives a linear sum of 1.0 % and ±0.8 % at k = 2. The phase term makes the difference. A CT test certificate gives the measured ratio and phase errors at each test point, which are usually much smaller than the class limits. Use those values when you have them.

When someone will be paid on the result, have the meter and CTs tested as a pair by a laboratory accredited to ISO/IEC 17025 for electrical energy.

Billing meters and “billing grade”

“Billing grade” and “revenue grade” have no defined limits. Billing and settlement depend on the jurisdiction, the approved instrument, its conformity route and its installation.

In the EU, Directive 2014/32/EU Annex V (MI-003) covers active electrical energy meters for residential, commercial and light industrial use, in classes A, B and C. At 5 °C to 30 °C with a balanced load, the maximum permissible error for class C is 1 % below the transition current Itr and 0.7 % from Itr to Imax. Class B is 2 %. The limits widen at other temperatures, to 2 % and 1.5 % for class C between −40 °C and −25 °C or between 55 °C and 70 °C. The power factor range is at least 0.5 inductive to 0.8 capacitive. The annex notes that a meter can be used with external instrument transformers, but it covers only the meter.

In Great Britain, government guidance states that a gas or electricity meter used for billing by a supplier or landlord must be of an approved design.

Ask for the approval certificate for the exact meter model, and for the installation and sealing rules of the supplier or network operator. A calibration certificate for another configuration, or a dashboard label, is not approval.

Commissioning a high-accuracy installation

1. Write down the measurement boundary: the incomer, feeder or equipment included, and any parallel supply, generation, bypass or unmetered auxiliary.
2. Record the chain: meter model, serial number and firmware, sensor make and rating, channel, configured ratio and datasheet revision.
3. Inspect sensor direction and closure, phase association, voltage reference location and fusing, conductor fit and terminal security against the installation instructions.
4. Read back the effective configuration from the device: CT ratio, wiring mode (3-wire or 4-wire), nominal frequency, import and export convention and reporting interval.
5. Check each phase against the reference at normal load. On an importing load, every phase must read positive active power. On a balanced load, the three power factors must be close to each other. A phase that reads negative, or far below the others, has a wiring fault until you prove otherwise.
6. Compare accumulated energy over the site's normal load cycle, such as a working week. Use the same start and stop times, time zone, boundary and energy direction.
7. Explain the difference before you attribute it to accuracy. Check boundaries, transformer losses, timing, rounding, missing data and unmetered loads first.
8. Store the readings, timestamps, reference instrument and its calibration certificate, photographs, settings, exceptions and the name of the person who accepted the installation.

Choose a reference whose own uncertainty is a quarter of the tolerance you are checking, or smaller. A clamp-on power meter with a specified active-power accuracy of ±2 % finds wiring faults, but it cannot confirm a 0.5S chain. A billing meter on the same boundary is often the best reference over a week. Its own class and CT class set the smallest difference you can resolve.

Report the signed difference:

difference (%) = 100 × (meter energy − reference energy) ÷ reference energy

For example, over seven days the reference registers 18,640 kWh and the submeter registers 18,712 kWh. The difference is 100 × 72 ÷ 18,640 = +0.39 %. With a Class 0.2S reference on Class 0.2S CTs, and the load mostly above 20 % of CT rating at a power factor above 0.9, that result is consistent with a correctly installed 0.5S chain. A difference near −67 % points to one reversed CT on a balanced load. A difference near +100 % points to a configured ratio of 200/5 on a 100/5 CT. The result is a comparison for that interval, not a calibration.

System-level accuracy with the EpiSensor ZEM

A conventional transformer-operated meter is type-tested on its own, and its CTs are specified and tested separately under IEC 61869-2. The installer then owns the pairing, the ratio setting and the burden check.

The ZEM electricity monitor is supplied with its current sensors. EpiSensor calibrates each ZEM with those sensors before dispatch, so Class 0.5S to IEC 62053-22 applies to the complete measurement: meter and sensors together. The ZEM-65 is ordered with a 120 A split-core CT or with Rogowski coils rated 300 A, 1000 A or 3000 A. None of these is a conventional 1 A or 5 A CT, so there is no burden to calculate and no CT ratio to set on site.

The site checks still apply. Use Current transformer selection to record the lowest and highest currents of interest, conductor dimensions and installation allowance. Check accuracy across that range; the tool’s optional EpiSensor matching is a catalogue comparison, not an accuracy or installation approval. Fit each sensor in the marked direction on the correct phase, then do the phase checks above. Class 0.5S is not a legal approval. If the law or a contract requires an approved billing meter, confirm the approval with EpiSensor before you install.