Most first monitoring sites stall on three unknowns: which circuit to measure, at what interval, and who receives the data at the far end. This worksheet makes the facilities team, the installer and the owner of the receiving system answer those questions on paper before anyone opens a panel.
Download the editable worksheet (.txt). No email or sign-in is required. Fill it in a text editor, or print it for the site walk. Your answers stay on your computer.
Write "unknown: survey" against any value you cannot confirm. Leave passwords and access tokens out, and name the person who controls access instead.
1. Name the decision
Write down the site, the question, the person who will use the answer and the action they can take. "Monitor energy" names no decision. A useful question looks like this:
The facilities manager needs to know why the refrigeration feeder draws 18 kW overnight when the store is closed, so that the refrigeration contractor can check the defrost timing and the door seals.
That sentence sets the boundary (the refrigeration feeder), the quantities (kW and kWh), the interval (fine enough to see compressor cycles) and the owner (the facilities manager). The rest of this worksheet follows from it.
Monitoring shows what changed. If the result must also justify a spend, agree the measurement and verification method before the work starts. For one sub-metered feeder, IPMVP Option B (retrofit isolation, all parameters measured) fits. Measure a baseline period, then adjust it for the variables that drive the load. For refrigeration, these are usually outdoor temperature, expressed as cooling degree-days, and throughput.
2. List the measurement points
For each point, record the circuit or equipment boundary, the quantity and unit, any existing source, the interval, the accuracy and the owner. The downloadable copy has blank point records that you can duplicate.
Interval
Use the coarsest interval that still shows the behaviour behind the decision.
| Interval | What it shows | Typical use |
|---|---|---|
| 30 minutes | Energy in each settlement period | Checks against supplier and network data in Ireland and Great Britain, which is half-hourly |
| 15 minutes | Average demand in each period | Maximum-demand checks and many demand-response programmes |
| 1 minute | Equipment cycles, starts and baseload | Refrigeration, compressed air and HVAC scheduling |
| 1 second or faster | Response to a grid-frequency event | Frequency-response services |
The refrigeration question needs one-minute kW. Suppose the feeder uses 126 kWh between 23:00 and 06:00. Every half-hourly total is 9 kWh, which is an average demand of 18 kW:
The half-hourly record is flat. The one-minute record shows the compressors at 26 kW for 40 minutes in every hour, and the feeder at 2 kW for the other 20. That duty cycle, with the store closed, is what the contractor must explain. Half-hourly data cannot show it.
Record energy as a cumulative kWh register, and record power as kW at the fine interval. A cumulative register survives a gap: subtract the reading before the gap from the reading after it, and you recover the missing energy. Energy integrated from kW samples loses whatever happened during the gap. The kW to kWh calculator shows the conversion.
Two register faults make false intervals. A meter replacement or a reset makes the next difference large and negative. A counter that reaches its maximum wraps to zero: a 32-bit unsigned register that counts Wh wraps after 4,294,967 kWh. Flag any negative or implausibly large interval for a person to check. Do not fill it automatically.
Accuracy
Record the accuracy class and its standard, and include the current sensors. A meter's class normally covers the meter alone. IEC 61869-2 limits a Class 1 CT to ±1% ratio error at rated current and allows a larger error at light load, and the two errors add. Overnight loads often sit at a small fraction of the feeder rating, which is where CT error is largest. The high-accuracy metering guide explains the classes.
A sub-meter reading is management information. In Great Britain, a meter that a supplier or landlord uses for billing must be of an approved design. In the EU, national rules for trade metering apply the Measuring Instruments Directive (Directive 2014/32/EU, Annex V, MI-003). The ZEM datasheet lists CE, RoHS and WEEE, not MID approval. If a reading will go on an invoice, get the requirement from the billing owner before you choose the meter.
3. Match each point to an interface
The survey must record the facts that decide the hardware.
| Point | EpiSensor interface | Record at the survey |
|---|---|---|
| New three-phase electrical point, 110 to 480 V AC line to line | ZEM electricity monitor | Per-phase current, conductor or busbar size, supply arrangement (3-wire or 4-wire) |
| Existing meter with a pulse output | ZPC pulse counter | Output type, pulse weight, maximum pulse rate |
| Existing meter with Modbus RS-485 | ZMB Modbus interface | Meter model, firmware, register-map revision, configured CT ratio |
| Existing heat, water or electricity meter with wired M-Bus | ZHM M-Bus interface | Meter model, primary or secondary address, data record units |
A ZEM ships with its current sensors fitted and calibrated to the meter. Class 0.5S to IEC 62053-22 applies to the meter and sensors together, and there is no CT ratio to set on site. Build the current-range and dimensional brief with Current transformer selection, then use its optional EpiSensor matching to investigate published options. Confirm installation access and performance at the lowest load before choosing one. When you reuse an existing meter with conventional 1 A or 5 A CTs, record the CT ratio from the CT nameplate, not from a drawing, and check the secondary circuit with the CT ratio and burden calculator.
A qualified electrician fits the sensors and the voltage reference. Record what they need to plan the work: the conductor size, the protective-device rating, the free space in the panel, and when the circuit can be isolated. The installation instructions for each model are in the documentation library.
4. Agree the data destination
Record whether the first site will use Edge locally, an external platform, or both. Name the owner of the receiving system and the person who controls the interface and its authentication.
Write a point map for three or four points before anyone configures an interface:
| Point ID | Source | Quantity | Unit | Type | Interval | Example source timestamp |
|---|---|---|---|---|---|---|
| REFRIG_FDR.kWh | ZEM, refrigeration feeder | Active energy import | kWh | Cumulative | 15 min | 2026-09-22T02:15:00Z |
| REFRIG_FDR.kW | ZEM, refrigeration feeder | Active power, three-phase total | kW | Instantaneous | 1 min | 2026-09-22T02:16:00Z |
| REFRIG_FDR.PF_L1 | ZEM, refrigeration feeder | Power factor, phase L1 | None | Instantaneous | 1 min | 2026-09-22T02:16:00Z |
| MAIN_INC.kWh | Existing main meter, Modbus RTU | Active energy import | kWh | Cumulative | 30 min | 2026-09-22T02:30:00Z |
Send timestamps in UTC. Keep the time of measurement separate from the time the platform received the reading. Local-time timestamps make a 23-hour day in March and a 25-hour day in October, and the repeated hour in October can overwrite one set of readings. Agree how a missing, invalid or delayed reading appears: a gap, or a null with a status. A zero is a valid reading and must not mean "no data".
Then make three checks separately: the reading exists in Edge, the transport reports success, and the receiving application stores the correct value against the correct point. A send can succeed while the third check fails. For example, the broker acknowledges a QoS 1 MQTT publish of 18.2 kWh, but the platform reads the value as Wh. The feeder then appears to use one thousandth of its real energy. Only the third check finds this, so compare one value in the platform with the same value in Edge.
5. Define interruption behaviour
For each of these scenarios, write the required behaviour, a safe test and the person who accepts the result:
- A measurement source stops reporting.
- The upstream connection becomes unavailable.
- Service returns after an interruption.
- A value is outside its expected range, such as negative kW on a feeder with no generation.
Know what each layer holds. A ZEM logs up to 70,000 data points in non-volatile flash, time-stamped to the second, and the Gateway collects them when the radio link returns. Capacity depends on the number of feeds and the interval. Six feeds at one-minute reporting make 8,640 data points a day, so the flash holds about eight days. At 15-minute reporting, it holds about four months.
Upstream of the Gateway, Edge keeps local history. For supported delivery paths, Edge writes a failed send to a disk-backed outbox and replays it later with the original timestamp. A generic MQTT or HTTP flow that you build yourself needs its own buffering, and you must test it through a restart. The data storage guide explains how to size local history and outage capacity.
Test the upstream case directly. Disconnect the Gateway's WAN for two hours, then reconnect it. The platform must show 120 one-minute kW readings for the gap, with their original timestamps and no duplicates. The kWh register difference across the gap must match the energy from those readings.
If the project will also control equipment, agree a separate control and safety scope with the asset owner. Acceptance of the monitoring does not cover control.
6. Set first-site acceptance
On a new electrical point, most faults are installation faults. Test for each one before you accept the site.
| Fault | What you see | Check |
|---|---|---|
| CT reversed on one phase | That phase shows negative kW. On a balanced load, the three-phase total reads about one third of the true value. | Every phase shows positive kW on a feeder with no generation. |
| CT on the wrong phase for its voltage | Two phases show a power factor near zero or negative. | Each phase shows a lagging power factor that fits the load. |
| Wrong CT ratio on a reused meter | Readings are wrong by the ratio: a 200/5 CT set as 1:1 reads one fortieth of the true value. | Per-phase current matches a calibrated clamp-on meter. |
| Wh sent as kWh, or the reverse | Readings are wrong by a factor of 1,000. | One value agrees in Edge and in the platform. |
| Local time in timestamps | A missing or doubled hour at each clock change. | Timestamps end in Z (UTC). |
| Register reset or rollover | One interval is very large or negative. | The platform flags the interval for review. |
The ZEM's calibrated sensors remove the ratio fault, but the installer still sets the orientation and the phase of each sensor.
Compare each new point with a reference. Spot-check voltage, current, power factor and kW on each phase with a calibrated power analyser. Then compare the kWh over at least 24 hours with the energy the analyser logs on the same feeder. Set the tolerance from the stated accuracy of both instruments before the test, not after it.
The site is accepted when you have this evidence:
- a point map checked against the physical boundaries, units and scaling;
- a reference comparison within the agreed tolerance;
- readings in the receiving application with their source timestamps;
- a passed interruption test;
- named owners for handover, configuration records and support.
Record the date, the acceptance owner, the open questions with their owners, and what must be true before you add a second site.
Next steps
If you are still comparing suppliers, the guide to choosing an energy monitoring system lists ten questions to ask of each one.
Build your monitoring system to choose hardware for the points on your worksheet.
Discuss this site plan if unknowns remain.