Energy monitoring: turning meter readings into cost per unit produced

A site meter tells you almost nothing

One incoming meter gives a total and a rough profile, which is enough to know the bill is large and not enough to act. Submetering assigns consumption to the systems that use it, and the useful first targets are rarely the production machines: compressed air, refrigeration, process heating, ventilation and lighting frequently account for more than the equipment people assume is responsible. Meter at the level where somebody could actually change something, which usually means a system or an area rather than an individual machine. Metering everything produces a data burden and a maintenance obligation nobody funded.

Consumption only means anything against a driver

Absolute consumption falls when production falls, which tells you nothing about efficiency. The figure that carries information is energy against an appropriate driver: per unit made, per batch, per tonne processed, or adjusted for outside temperature where heating and cooling dominate. Choosing the driver is a judgement about what genuinely varies the load, and getting it wrong makes improvement look like mix change and vice versa. The driver data has to come from production records, which means the monitoring system needs an interface to whatever knows what was made and when.

The baseline and what disqualifies it

Every savings claim is a comparison against a baseline, and every baseline is arguable. Product mix, throughput, weather, shift pattern and equipment changes all move consumption independently of any improvement. Agree the model and its adjustment rules before the project starts, not when the first claim is challenged by someone who dislikes the answer. ISO 50001 provides the management framework organisations adopt for this and is issued by the International Organization for Standardization. Where savings support a contract or an incentive claim, the measurement approach usually has to be agreed with the counterparty in advance.

Non-production load is where the straightforward savings sit

Plot consumption across a period when nothing is being produced and the result is often uncomfortable. Compressors running to feed leaks, ovens held at temperature over a weekend, extraction serving an idle area, machines left in a standby state that draws nearly as much as running. This base load requires no process change to address, only a shutdown routine and something that confirms it happened. It is also the most persuasive evidence when asking for investment, because it is visible in the data before and after, and nobody can attribute it to production mix.

Where the meters live and who owns the data

Meters usually sit on plant networks, and the monitoring application usually sits on the business side, so the data crosses a boundary that also carries security implications. Read-only flows through a controlled gateway are the normal arrangement, and the segmentation should not be weakened to make reporting convenient. Ownership matters too: meters drift, communications fail silently, and a channel returning a plausible constant is worse than one returning nothing. Someone has to own meter verification and a check that every expected reading actually arrived, or the dataset degrades without any visible symptom.

Frequently asked questions

Do we need submetering, or is the utility bill enough?

The bill is enough to establish that energy is worth attention and useless for deciding what to do. Without submetering, every improvement claim is a comparison of total site consumption in periods with different production, weather and mix, which nobody can defend. Start with a small number of meters on the largest and most controllable systems rather than a comprehensive installation, and expand once the first ones have produced a decision somebody acted on.

Where should the first meters go?

On the systems that consume most and can most readily be changed, which in a typical plant means compressed air generation, process heating or cooling, and any single large piece of equipment. Add a meter that isolates non-production periods, because base load is usually the quickest available saving. Avoid starting with a meter on every machine: the installation cost is high, the data volume is large, and machine-level consumption rarely changes anybody's behaviour on its own.

Can monitoring data support a certification or a reporting obligation?

It can, provided the measurement approach, the boundaries and the data handling were designed with that in mind rather than retrofitted. An energy management standard expects consumption data tied to significant uses and a documented method behind any performance claim. External reporting schemes typically want auditable readings with a clear trail from meter to published figure. Retrofitting provenance to data collected casually is far more work than collecting it properly from the outset.