Industrial Energy Audit: Where to Start Measuring Before a Retrofit

Without a baseline, a retrofit cannot be accepted

If the contract lists new machines, variable-speed drives and a monitoring system, but never says which period and which production regime to compare against, acceptance turns into a dispute over numbers. A baseline is not a report written afterwards. It is the reference both parties confirm before work starts.

• Output rises and the absolute electricity bill rises with it, the retrofit is swamped, and the buyer concludes that nothing changed
• Outdoor temperature shifts HVAC and chilling load, and the two sides argue about weather versus equipment
• Only the retrofit month is compared with the month before, so maintenance, full production and low production are missing and the window is not representative
• Only the incoming electricity meter is watched, so changes at the compressor room, the chiller plant and process bays cannot be tied to a point
• A reduction target is set from experience, with no measured support, and cannot be repeated by the same method

All five share one cause: the object of comparison was not locked before the retrofit. ISO 50001 treats the energy baseline as an input to the energy-management system, not as an optional annex. In general manufacturing plants the same rule holds: measure first, then calculate, or there is no shared ruler at acceptance.

Which utilities to meter, and to which level

The baseline must cover the media the retrofit will actually touch, not only the main electricity meter. Factory and building energy SCADA commonly combines sub-metering of electricity, water and gas with the chiller/heat plant and the compressed-air station — two areas that often take a large share of utility electricity and are frequent retrofit targets.

Metering depth follows the layer the retrofit will change. The site incomer cannot explain what happened after compressors were replaced. If process machines are outside the scope, there is no need to meter every asset for the baseline.

• Electricity: the incomer, the hall or substation, and the energy-intensive machines planned for replacement or speed control — compressors, chillers, pumps, furnaces
• Compressed air: compressor-room electricity, discharge flow and pressure, main user zones; night flow shows whether leakage is already inside the baseline
• Water: incoming volume, process and circulating water, reconciliation with effluent; this circuit is required when pumps or water treatment are in the retrofit
• Gas and steam: supply to the boiler or furnace, main heat-using bays; a heat-source retrofit without measuring the heat side distorts the electrical baseline
• Chilled and hot water: plant-room electricity, machine-level meters, supply/return temperature and flow; HVAC zones are split only when they sit inside the retrofit

The measurement window: representative production states

How long to measure depends on whether production has a stable cycle, not on whether a calendar month has passed after the meters went in. The baseline window must cover the states that move consumption in a material way, or later benchmarking and acceptance will be biased.

• Shifts near the load limit: lock specific energy when both consumption and output sit in the upper part of the working range
• Consecutive working days on the normal production plan: exclude holiday stops and concentrated maintenance; this becomes the main line
• Planned downtime or low-output periods: record idle, hold-temperature and hold-pressure base load separately so it is not later counted as a retrofit effect
• Season-sensitive load: if HVAC and chilling track outdoor temperature, the window must span a sufficient temperature range, or the line is built by temperature bands
• In the same window, record output, shift, product mix and outdoor temperature in step with energy — without these covariates the baseline cannot be normalised

What to measure, sampling rules, and recurring mistakes

The table lists the items most often measured incorrectly at the baseline stage. Sampling that is too coarse, a window without representative regimes, and cumulative totals without a load curve all strip later benchmarking of its basis.

Where an electricity metering system already exists, electrical channels can often be reused. Compressor rooms, chiller plants, water and gas usually lack readings synchronised with process quantities — those are added during the audit, not after the new equipment is running.

How to write the baseline into the technical specification

A baseline binds acceptance only when it is written into the specification or attached as a measurement report. A clause such as “average consumption over the three months before the work” is not enough: a three-month average flattens output and season and is still a target set by guesswork, only written in another form.

• Name the media, metering points, instrument serial numbers, polling interval, and the system that issues the data — existing electricity metering, energy SCADA, or both
• Name the baseline window, the production states covered (full output, normal plan, low output), and the method used to normalise by output and temperature
• State that acceptance uses the same points and the same normalisation, and that switching to absolute incomer totals is not allowed
• State the primary benchmarking frame: one machine before and after, like-for-like machines side by side, or deviation from the line — pick one as the main method
• Do not pre-write a reduction figure that has no measured support; lock targets after the baseline report is confirmed

AI load forecasting and operating suggestions based on that forecast are a research and joint-validation direction. They are not part of the standard scope of baseline measurement or retrofit acceptance. The audit stage needs repeatable actuals, not forecast values.