Engineering practice · Automation and BMS
Article 03 / 10
A sensor is one degree out. Why does it matter to the whole system?
The most obvious sensor fault is when it stops transmitting values. A much more troublesome case is when the value looks plausible but is wrong.
The BMS sees +23 °C, maintains this setpoint and raises no alarm. Meanwhile, the room is cooler than required. The controls can execute an algorithm flawlessly while relying on a false signal.
A one-degree offset means a one-degree change in the effective target
Consider a simplified steady-state condition with no deadband. The sensor reads 1 °C too high:
If the controller maintains +23 °C according to its input signal, the actual temperature is +22 °C. This is a direct consequence of the assumed offset, not an assessment of a real office.
However, this does not imply ‘6% savings’ or any other universal percentage. An energy assessment requires heat losses, weather conditions, duration and equipment operating mode. Correcting the sensor may even increase consumption if the room was previously underheated.
Calibration cannot correct poor placement
A room sensor must represent the zone it controls. A sensor in a supply air jet or near a local heat source may measure the local temperature correctly while poorly representing conditions for occupants.
Ducts, pipework and rooms require appropriate sensor designs and installation methods. Belimo specifically identifies location, sensing element type, range and output signal compatibility as selection parameters. 1
Before applying software compensation, we would therefore check the physical installation, contact with the measured medium and whether the input configuration matches the sensor. A fixed correction must not conceal an installation problem.
A small temperature difference amplifies the effect of measurement error
For a water circuit, thermal output is often estimated using the formula:
where flow rate is in m³/h, temperature difference is in K and the result is in kW. The coefficient is rounded for water under normal conditions; other heat transfer fluids require their own properties.
Assume a flow rate of 10 m³/h and an actual temperature difference of 3 K. The result is approximately 34,9 kW.
If each sensor has a permissible absolute error of ±0,3 °C, the worst-case combination of opposing errors gives ±0,6 K for the difference. The calculated output could then range from 27,9 to 41,9 kW — a deviation of ±20%, even without flow meter error.
This is an arithmetic limit for the stated conditions, not a statistical assessment of a particular instrument. It explains why the accuracy of energy analytics cannot be promised merely on the basis of decimal places.
CO₂ sensors also require careful selection
For some CO₂ sensors, automatic baseline correction relies on periodic exposure to reference fresh-air conditions. If a facility is continuously occupied, the suitability of this algorithm must be checked. Belimo explicitly distinguishes these scenarios when selecting sensor technology. 2
This does not mean that every sensor with automatic correction is inaccurate. It means that the measurement method must suit the operating conditions.
Checking the entire measurement chain
A useful check compares a reference measurement, the sensor output, the value in the controller and the BMS display. Sometimes the physical sensor is sound, but the error lies in scaling or units.
For important parameters, the verification date, applied compensation and quality status should be recorded. If a signal is unreliable, it must be flagged and must not be used unconditionally for optimisation; a BMS can support separate data reliability indicators. 3
Before training a model on data, make sure that the data describe the building rather than its sensor errors.