Machine-level energy metering: what submetering buys you

The hardware case is straightforward and rarely the obstacle. Current transformers and a logging meter on a machine's supply are inexpensive relative to the machine, installation is a shift's electrical work, and the data comes out at whatever interval is configured. What follows is the part that determines whether the money was well spent.

Three questions a submeter can settle

The first is allocation. A plant that makes several products on shared infrastructure cannot cost energy per product from a site bill, and estimates by machine rating are wrong in a direction nobody can predict, because rated power and drawn power differ by duty cycle. Per-machine measurement turns energy into a cost that can be attributed, which changes product costing and occasionally changes which products a plant wants.

The second is standby. Equipment left energised between shifts is a line item that appears in no budget, and it is visible in submeter data immediately as a floor under the consumption curve. This is the finding that most often pays for the installation, and it is almost always larger than the people running the plant expect.

The third is condition. A rising baseline on a machine doing unchanged work is a mechanical signal before it is an energy signal — drag, misalignment, a filter nobody changed. Energy data is a crude condition monitor, but it is one that is already being collected for another reason, which makes it the cheapest one available.

The standards that make the numbers comparable

ISO 50001 specifies requirements for an energy management system, and its contribution here is structural rather than technical: it requires that energy performance be measured against something, reviewed, and acted on by somebody named. The companion document ISO 50006 deals with the measurement itself, covering energy baselines and energy performance indicators — how to establish what normal is, and how to express performance so that two periods can be compared.

That is the step plants skip. Consumption per month is not a performance indicator, because production varies; consumption per unit of output is one, but only if output is measured over the same boundary as the energy. Getting that boundary right is most of the work, and it is the reason two plants in the same company report incomparable figures while both believing they follow the same standard.

What submetering cannot do

It cannot allocate utilities. Compressed air, chilled water and extraction are generated centrally and consumed locally, and a meter on a machine's electrical supply sees none of it. On compressed-air-heavy processes this is not a rounding error; it can be the larger share of the true energy cost of the operation, and attributing it requires flow measurement that is considerably less convenient than a current transformer.

It also cannot survive reorganisation on its own. Meters are installed against a plant layout, and plants move. A submetering scheme with no documentation of which meter is on which asset becomes, after two production changes, a set of time series with no referents — data that is technically present and practically lost. Keeping that register current is a clerical task that nobody is assigned and everybody needs.

A modest recommendation

Meter the few assets that plausibly dominate, pair each with an output measure, review weekly for a quarter, and act on the standby finding that will appear in the first month. Then decide whether to extend. Plants that start with site-wide instrumentation and a dashboard project usually end with site-wide instrumentation and a dashboard; plants that start with four meters and one question tend to end up with twenty meters and a habit.