One way of spotting whether services or equipment are running unnecessarily anywhere in a building is to monitor power levels at quiet times. For many kinds of enterprise, 2 a.m. to 4 a.m. would be a good window to monitor. Any unexplained increase in power drawn outside working hours suggests in itself that something is running that shouldn’t be; and there is a possibility that it may also be running continuously during working hours when it isn’t needed.
In an organisation where such issues are a significant risk, it would be possible to define ‘quiescent thresholds’ for individual metered circuits, with alerts automatically generated when out-of-hours demands exceed those levels. As ever, though, we need to guard against creating small nuisance alerts; that means not applying the method indiscriminately but restricting it to significant energy users.
The quiescent-threshold method naturally fits operations with fixed daily time schedules, but it could potentially be applied to irregular intermittent loads. The kind of situation I have in mind is a batch manufacturing process with lots of ancillary equipment that might accidentally be left running during idle periods. I would approach this task by scrutinising past data to establish the lower decile half-hourly demand (say) and then deciding how long a period to tolerate with demand above that level. For example I might want to verify that a particular load never stays above 15 kW for longer than 7 hours at a time.