Forty apartments. One heating bill that nobody in the block believed. The utility pointed at the BTU meter in the basement, the residents pointed at their radiators, and the argument ran for two winters before anyone put a reference thermometer in the pipe.
The flow meter was accurate to within half a percent. The temperature sensors were individually calibrated and individually fine. Together they read 0.4 K apart from each other. On a 15 K design differential, that is a 2.7% bias on every invoice, permanently, with no fault code and no way for anyone to notice.
That is the thing about energy metering. The failure is almost never dramatic.
Three parts, one number
A BTU meter is a small arithmetic machine. Flow times enthalpy difference, integrated over time. Each input carries its own error, and they do not combine the way most buyers assume.
| Component | Typical error contribution | Where the money goes |
|---|---|---|
| Flow measurement | 1% to 2% of reading | Meter type, straight run, liner |
| Temperature pair | 0.5 K to 2 K matched pair | Platinum resistance sensors, factory matching |
| Integrator | 0.1% to 0.5% | Enthalpy tables, update rate |
Run the arithmetic on a 60/45°C circuit — 140/113°F — and a 1 K error in temperature moves the energy total by roughly 6.7%. A flow meter one percent out of calibration moves it by one percent. Budget follows accuracy, and accuracy is cheapest to buy at the sensors.
So specify a matched pair of Pt500 or Pt100 sensors, matched to within 0.1 K at the working differential, not merely individually accurate at 0°C. Two perfect sensors that read 0.3 K apart will bias the total for the life of the installation, and no flow calibration recovers it.
Picking the flow element
Electromagnetic. The default for substation primary and secondary circuits. No moving parts, no pressure loss, and it tolerates the glycol mixes and occasional dirty fill water that district systems accumulate. Needs conductivity above 5 µS/cm, which heating water always has. A flange type electromagnetic flow meter with PTFE liner covers 90/70°C and 110/70°C duties without trouble.
Clamp-on ultrasonic. The retrofit answer. On an existing substation where the pipe cannot be cut, a clamp-on ultrasonic flow meter goes on in half an hour. Lower accuracy than inline, adequate for internal allocation, and the same instrument doubles as a verification tool across a portfolio of substations.
Mechanical. Still common on household and small commercial connections where cost dominates. Shorter service life on dirty water, higher pressure loss, and a maintenance burden that shows up in year five.
Vortex. Useful on the high-temperature primary side, particularly steam circuits feeding a heat exchanger. Wrong tool for the low-flow, low-differential secondary side, where turndown matters more than temperature rating.
The substation is the hard part
A typical substation runs 110/70°C on the primary and 60/45°C on the secondary, and both sides need energy measurement if the utility bills on delivered heat and the building owner bills on consumed heat. That is two meters, two temperature pairs, two integrators — and they must not be allowed to drift apart.
Rules that come out of substation work:
- Mount the flow meter on the return line if supply runs above 120°C. Lower temperature means longer sensor life and less liner stress.
- Keep 5D upstream. A substation is a dense box of elbows and valves, and the meter usually gets whatever space is left.
- Set temperature pockets at the correct depth. A sensor in a shallow pocket reads the pipe wall, not the water — and the error is worst at low flow, exactly when the differential is smallest.
- Seal and log the integrator parameters. Most billing disputes are configuration disputes, not hardware disputes.
Chilled water is a different problem
Chilled water runs a much smaller differential. Seven degrees Celsius supply, twelve return — 45/54°F — so 5 K instead of 15 K. The same 0.3 K sensor mismatch that cost 2.7% on the heating side now costs 6%. This is why chilled water energy figures so often produce numbers nobody trusts.
Two things help. Specify a tighter matched pair, 0.05 K if budget allows. And use a flow meter with good low-end performance, because chilled water plants spend much of their life at part load with VFD-driven pumps. A clamp-on meter with a BTU option reads down to 0.1 m/s on a clean pipe, which is where a lot of chilled water actually sits.
Also check glycol concentration. A 30% propylene glycol mix changes both density and specific heat, and an integrator configured for water will be wrong by five to eight percent. Set the mixture in the calculator, not in a spreadsheet afterwards.
Where the errors hide
Sensor insertion depth. The single biggest cause of energy metering disputes we encounter. The element must sit past the pocket tip, into the flow.
Air in the circuit. A flow meter on a line with entrained air reads high. Check the air separator before blaming the instrument.
Meter on the wrong side of a control valve. Downstream of a modulating valve the velocity profile changes with every valve position. Upstream, or at least 10D away, is where it belongs.
Integrator clock drift. On monthly billing, a clock two days out over a year is a 0.5% error nobody ever finds.
A specification that holds up
For a 3 MW substation — roughly 850 tons of refrigeration equivalent on the cooling side — with 110/70°C primary and 60/45°C secondary, DN150 and DN200:
- Electromagnetic flow meters, PTFE liner, Hastelloy electrodes, ±0.5%, 4-20 mA plus pulse
- Matched Pt500 pairs, 0.1 K match, in pockets on the return lines
- Integrators loaded with enthalpy tables for the actual glycol concentration
- RS485 Modbus to the BMS, energy total as a holding register
- Annual verification of the temperature pair against a reference; five-year flow calibration interval
That build holds its accuracy for a decade. The cheaper version — unmatched sensors, a mechanical meter, an integrator set for pure water — produces numbers that get argued about every winter.
Frequently asked questions
BTU meter or heat meter?
Same instrument. BTU meter is the US term, heat meter the European one. Chilled water versions are sometimes called cooling energy meters.
Can I add energy measurement to an existing flow meter?
Yes, if the meter has a pulse or digital output and the pipe has accessible temperature pockets. The integrator is the cheap part; sensors and pockets are the work.
How accurate does it need to be for billing?
Most European regulations require class 2 or 3 under EN 1434 for custody applications. For internal cost allocation class 3 is usually enough. The price difference is modest — buy the better class if the number leaves the building.
Where the same heat ends up in a process rather than a building, the metering question shifts: sanitary flow meters for food and beverage plants covers CIP and hygienic circuits, and flow measurement in steel plants covers the steam and gas headers around a furnace.