Fixed and variable energy
A plant's energy use is a fixed load, paid whether or not the line runs, plus a variable load that follows production. Written as a line, energy = base load + slope × output. Fitting that line to measured data separates the two parts, gives an honest expected consumption for any volume, and shows the moment performance changed, which a kWh-per-unit ratio cannot do.
The model
Monitoring and targeting practice has used the same equation for decades: energy over a period equals a constant plus a coefficient times the production in that period. The constant is the base load, the energy the plant uses at zero output: compressors idling, chillers holding temperature, lighting, extraction, machines powered and waiting. The coefficient is the variable load, the extra energy each additional unit costs. Plotted with output on the horizontal axis and energy on the vertical, the periods fall around a straight line; the intercept is the base load and the gradient is the slope. The plant's efficiency question splits in two: is the base load as small as it could be, and is the slope?
Source: Bureau of Energy Efficiency (India), energy monitoring and targeting guide; ISO 50006:2023 on model-based energy performance indicators
Two base loads
Augos fits two intercepts rather than one, because a plant has two kinds of idle. Non-operating base load is what the site draws when no shift is scheduled: nights, weekends, shutdowns. Operating base load is what it draws during paid time when the lines are stopped or running empty: the standing load of a shift in progress. The two have different owners and different fixes. Non-operating load is a facilities and controls question, what is left on overnight. Operating load is a production question, how much of the shift the equipment spends powered but idle, which is the same time the Owner column already attributes. The shift plan decides which half hours are paid time, and because production is sometimes recorded outside the plan, the boundary is editable after the fact.
Fitting the line
The fit is ordinary least-squares regression of energy on output across periods of equal length, with the operating and non-operating periods treated separately. The output has to be measured, not planned; Augos uses the counts from the production counters or the Augos Operator Terminal on each line, normalised to one site-wide unit (bags, cases, kilograms) so that lines making different products can be added. The energy comes from the site's mains meter, gross import by default. R² reports how much of the variation in energy the output explains; a low R² during operating periods is itself a finding, usually that the operating base load dominates and output barely moves the meter.
A real month
One anonymised flexible-packaging plant, one month in 2026, daily totals: electricity from the mains meter, output from the production counters.
| Fit | Days | Result |
|---|---|---|
| All 30 days | 30 | energy = 100 kWh + 0.0033 kWh per bag; R² 0.83 |
| Production days only | 20 | slope about 0; R² 0.01 |
| Non-production days | 10 | mean 94.9 kWh per day, range 85 to 125 |
The whole-month fit looks convincing and says almost nothing: its R² comes from the on/off difference between production days and shutdown days. Restricted to production days, the slope disappears. Electricity on a production day did not depend on how many bags were made; it depended on what was switched on. The non-operating base load is clear at about 4 kW (95 kWh a day). The operating base load is the bulk of a production day, and it changed in the last ten days of the month: daily electricity fell from 591 kWh to 379 kWh with output unchanged at about 128,000 bags a day. The fixed-plus-variable model locates the change where it belongs, in the fixed part, and dates it. A ratio would have called it a 36% efficiency gain.
Source: Augos platform, one anonymised flexible-packaging plant, mains electricity and production counters, a 30-day month in 2026. Daily fits; half-hourly fits are the module's working resolution.
When performance changed
Once the line is fitted to a baseline period, every later period has an expected consumption for its actual output, and the difference between actual and expected is the period's saving or overspend. CUSUM, the cumulative sum of those differences, turns noise into a trend: a flat CUSUM means performance is holding at baseline, a falling one means sustained saving, and a kink marks the date something changed. On the packaging plant the kink would fall on a single date. Finding out what happened that day is the plant's job; finding the day is the method's.
Source: monitoring and targeting practice, CUSUM analysis (Bureau of Energy Efficiency, India)
Why the half hour
Monthly data gives twelve points a year and hides everything inside a month. Daily data separates production days from shutdowns but blends shifts, breaks and changeovers into one number, which is why the production-day fit above is flat. Half-hourly data, the resolution South African electricity is billed and settled at, gives 48 points a day and lets the model see the ramp-up at the start of a shift, the plateau while the lines run, the dip at a break and the tail after the last line stops. On the packaging plant's mains meter, a single production day reads as 2 kWh per half hour through the night, a ramp from 06:00, 33 to 38 kWh per half hour from 07:00 to 13:30, 24 to 26 through the afternoon and back to 2 kWh by 17:00. Matched to the production counters half hour by half hour, that profile is where the operating base load and the per-unit slope separate. The Augos module works at this resolution and rolls up to day, week and month views.
Source: Augos platform, one anonymised flexible-packaging plant, mains meter, two days in 2026, half-hourly
Common questions
What is base load in a factory?
The energy the site uses at zero output. Augos splits it into non-operating base load (nights, weekends, shutdowns) and operating base load (paid time with lines stopped or idle), because they have different owners.
How do you separate fixed from variable energy?
By regression of metered energy on measured output over equal periods: the intercept is the base load and the gradient is the energy per unit. Operating and non-operating periods are fitted separately.
Why not just track kWh per unit?
Because the ratio spreads the base load over whatever volume was made, so it falls when output rises and rises when it falls, regardless of efficiency. See Specific energy consumption.
What is CUSUM?
The cumulative sum of the differences between actual and expected consumption. A flat line means baseline performance; a change of direction marks the date performance changed.
Does this need a sub-meter on every machine?
No. The site mains meter and the production counts the lines already produce are enough to fit the site-level model, which is where Augos starts. Sub-metering refines the picture later.