Data archived, but no answers: which trends are needed for HVAC analysis

Asked for the operating history, the facilities team sends a temperature graph for the week. The line fluctuates, but it does not show who changed the setpoint, whether the unit was running or what was happening in the associated system.

Such an archive is not useless. However, a single parameter is not enough to reconstruct the cause of an event. Analysis requires a history of interrelated signals.

Log for a specific scenario, not indiscriminately

To check supply air heating, we would compare the setpoint, temperatures before and after treatment, valve command, fan operation and available heating medium parameters. For compressor short cycling, we would examine its actual status and transition times.

BMS data combined with interval meter readings help identify scheduling problems and inefficient operating modes. This is how DOE describes the preliminary assessment of building re-tuning potential. ¹

Once the question is defined, selecting the right set of points becomes easier. This is more useful than storing hundreds of signals whose purpose nobody has explained.

The logging interval can hide an event

Suppose a unit runs from 10:02 to 10:08. If only its instantaneous status at 10:00 and 10:15 is logged, both values may be ‘off’. The six-minute cycle will disappear from this archive.

If the electrical power draw during this time was 20 kW, the energy consumed is:

E=20×6/60=2 kWh

It has not physically disappeared — the logging method missed it. For states, consider logging transitions; for analogue parameters, use an appropriate sampling interval or change-of-value logging. Such modes are available in Metasys, for example. ²

Meanwhile, the 15-minute increment of a cumulative meter will correctly capture these 2 kWh. It simply will not show the exact timing and duration of operation within the interval. It is therefore important to know what has been logged: an instantaneous, average or cumulative value.

Power cannot be summed as energy

For a sequence of average power values over intervals:

E=ΣᵢP̄ᵢΔtᵢ

If power is given in kW, time must be in hours. Instantaneous measurements taken at irregular intervals require a justified numerical integration method. Where a cumulative energy meter is available, use the difference between its readings, checking for resets and gaps.

These are different ways of obtaining a result. Labelling a graph ‘kWh’ does not correct a faulty calculation.

Time and quality are part of every value

The logging design should distinguish between measurement time and server receipt time. After a connection failure, older data may arrive later; without correct timestamps, the sequence of events will be distorted.

A quality status is also required. The last stored value cannot be treated as current indefinitely. Reliability indicators in trend objects allow a valid signal to be distinguished from a source problem. ³

For signals that rarely change, change-of-value logging should be combined with source availability monitoring. Otherwise, a long flat line does not tell you whether the system is stable or communication has been lost.

Check how long the history is actually retained

A controller buffer may overwrite old records or stop logging when full. This behaviour depends on the configuration and is explicitly described in the Johnson Controls documentation. ⁴

Before handover, therefore, check transfer to long-term storage, retention after a restart and data export. Choose the logging frequency according to process dynamics and network capabilities, not on the assumption that more frequent is always better.

A good archive lets you answer what happened, why, and which action preceded the outcome. Useful analytics start with this kind of history.