Start with the fuel that actually belonged to the period

A generator report can improve without the engine using any less fuel. Suppose a factory records a delivery in the shift that ended just before the tanker arrived. Correcting that entry lowers reported consumption. The correction matters. Calling it an energy saving would reward the team for a change that never happened.

Generator fuel consumption needs a clear link between litres, electricity and time. A useful report names each measurement point and the loads it served. It also explains changes during the period. Only then can a manager compare performance or investigate an unusual result.

The worked factory case below is hypothetical. The readings, operating profile, fuel price and equipment rating are all invented for this example. They are not a benchmark for a particular generator. The calculations use the same assumptions throughout. This lets the reader check the fuel ledger against the electricity and load records.

Decide which performance measure answers the question

Litres per hour describes the rate of fuel use. It helps with stock planning, but it says little about efficiency unless output is also known. An engine that uses less fuel while serving much less demand may have a worse fuel rate per unit of electricity.

Two useful operating measures are:

  • kWh per litre: electrical energy delivered divided by fuel consumed. A higher value means more electricity per litre on the stated basis.

  • Litres per kWh: fuel consumed divided by electrical energy delivered. A lower value means less fuel per unit of electricity on that same basis.

These ratios are reciprocals. Choose one for the main trend and label it consistently. Switching between them without explanation makes a rising line difficult to interpret.

Neither ratio is a thermal-efficiency percentage. To calculate that percentage, fuel must also be expressed as energy. The fuel's heating value and the output boundary then become part of the calculation. The U.S. Energy Information Administration's efficiency explanation relates fuel energy to electricity. It also distinguishes net generation from the plant's own energy use.

A factory might measure output at the generator terminals. Another boundary is the point where useful power reaches its network. Write that location beside the result. A gross-output ratio and a net-delivery ratio use different bases, so a direct comparison would mislead.

Close the fuel ledger at a defined time

The example period runs from 06:00 to 18:00. The generator operates for ten of those twelve hours. It is off for the other two hours and uses no fuel during those off periods in this example.

The fuel boundary includes the storage and day tanks serving this generator. Opening and closing quantities include all stored fuel within it. Any unmeasured fuel held in connecting lines is unchanged. All volumes are on the same declared temperature basis. There are no leaks or undocumented withdrawals in the assumed case.

One documented transfer sends fuel to another use outside this boundary. That transfer must leave the stock balance without being charged to the generator.

Fuel ledger item

Litres

Treatment in the period

Opening fuel at 06:00

1,200

Available at the start

Deliveries received before 18:00

900

Add to available fuel

Closing fuel at 18:00

760

Remains in stock

Documented transfer to another use

40

Exclude from generator consumption

Generator fuel consumed

1,300

Reconciled balance

The calculation is:

Fuel consumed = opening stock + receipts − closing stock − other documented withdrawals.

For this case, fuel consumed is 1,200 + 900 − 760 − 40 = 1,300 litres.

This balance is only as sound as its boundary and records. If a second engine draws from the same tank, the residual cannot be assigned to the first engine alone. If the day tank falls while only the main tank is measured, the balance misses part of the fuel used.

The stock record should therefore show tank identities, reading times and transfers. A delivery note records a transaction. The receipt record shows when fuel entered the measured system. If those times differ, use the time that fits the physical balance and explain the choice.

A late delivery creates a false improvement

A further 200 litres arrives at 18:15. The draft report assigns it to the earlier period by mistake. Recorded receipts rise to 1,100 litres. With the other entries unchanged, it reports 1,200 + 1,100 − 760 − 40 = 1,500 litres consumed.

Removing the late delivery corrects consumption by 200 litres. Relative to the wrong 1,500-litre figure, that is a 13.33% reduction. No engine setting, load or fuel use has changed. Call this a cutoff correction and revise the affected comparison. It is not a fuel saving.

The delivery belongs in the next period's receipt record. Its later presence in stock must also agree with the next closing reading. Changing only the receipt date may leave related stock entries wrong. That would shift the error into another day.

Put fuel volumes on a common basis

A litre is a volume, so its temperature basis matters when records are compared. A supplier may report a temperature-adjusted quantity. A tank reading may instead represent the liquid at its current temperature. Adding them without checking the basis can create a difference that looks like consumption.

The 2026 NIST Handbook 44 liquid-measuring section distinguishes adjusted and unadjusted volume. It also covers how temperature-compensated readings should be marked. It is a U.S. legal-metrology reference with a defined scope. Its requirements are not presented here as the factory's local law. The practical point is to identify the volume basis before reconciling records.

Record whether each quantity is mass, observed volume or volume at a stated reference temperature. Where conversion is needed, use the applicable fuel properties and an approved measurement method. A density copied from a different fuel batch may not support a precise balance.

Tank level also needs a valid conversion to volume. Tank shape matters: liquid height and volume do not always change in the same proportion. Keep the relevant tank calibration information with the measurement record. If a reading method changes, check any step in the trend. The step may not reflect engine performance.

These checks do not require a complicated report. They require agreement on what the numbers mean. A short note can explain the units and temperature basis. Extra decimal places cannot settle an unresolved total.

Match the electricity readings to the same period

The gross electricity meter rises from 234,500 to 238,400 kWh between the two cutoffs. Its verified multiplier is already applied, and there is no reset or rollover. Gross generation is therefore 3,900 kWh.

A separate meter at the factory delivery boundary records 3,810 kWh over exactly the same period. The checked system map explains the 90 kWh difference. It covers internal use and losses between the two meters. There are no other imports, exports or storage changes across that part of the system in this example.

Do not label every difference between two meters as an auxiliary load. The meters may cover different branches or times. The electrical map must explain the difference before the report gives it a name.

Using the corrected 1,300 litres gives:

Measure

Calculation

Result

Gross electricity per litre

3,900 ÷ 1,300

3.000 kWh/L

Net electricity per litre

3,810 ÷ 1,300

2.931 kWh/L

Fuel per net unit

1,300 ÷ 3,810

0.3412 L/kWh

Wrong fuel per net unit

1,500 ÷ 3,810

0.3937 L/kWh

The final row shows the effect of the delivery error. The electrical meter did not change; the false fuel total alone worsened the result. A dashboard displaying only the ratio would hide the reason.

For routine reporting, retain both cumulative meter readings as well as their difference. Note meter changes, multipliers and missing records. Ask a qualified person to check the displayed quantity and the installation. Both must suit the reporting purpose. A current reading or a kVA rating is not a substitute for measured kWh.

When output varies, energy is accumulated over time. A single kW reading may not represent the shift average. Multiplying that reading by the whole shift could give a false result. Logged energy or suitable time-based readings provide a firmer basis.

Agree how the cutoff will be captured while the equipment is running. Fuel and electricity readings taken several minutes apart may cover different work. A timestamped logger can help, provided its clock and channels are checked. If you estimate a missing cutoff from nearby readings, explain the method. State the operating assumption that supports the interpolation. A load change during the gap may make a simple estimate unsuitable.

The record should also distinguish elapsed time from engine-running time. A shift fuel budget covers the whole shift, while an operating fuel rate may exclude stopped hours. Both can be useful. The labels prevent a change in downtime from being mistaken for a change in fuel use while running.

Use the load profile to explain the daily ratio

The ten operating hours in the case contain two load intervals. The values below are assumed means for each interval. They agree with the energy and fuel totals in the fictional log. The totals include any fuel used while starting or changing load. No extra uncounted fuel sits outside the table.

Operating interval

Hours

Mean gross kW

Mean L/h

Gross kWh

Fuel litres

Lower load

2

150

70

300

140

Higher load

8

450

145

3,600

1,160

Total

10

—

—

3,900

1,300

The total agrees with both the fuel ledger and gross electricity meter. Average fuel use during operation is 1,300 ÷ 10 = 130 L/h. Average gross output is 3,900 ÷ 10 = 390 kW.

Assume an electrical rating of 600 kW applies to this service and these site conditions. Its mean running load is then 390 ÷ 600 × 100 = 65%. The average does not show time at each load or sudden demands. It also cannot establish whether the rating suits the duty.

Actual manufacturer data should be read on their stated basis. For example, Caterpillar's C18 specification distinguishes prime and standby ratings. Its fuel figures identify load points and include the fan. Check the exact equipment and conditions behind a curve. One published curve cannot describe every generator.

In the fictional log, lower-load performance is 300 ÷ 140 = 2.143 gross kWh/L. Higher-load performance is 3,600 ÷ 1,160 = 3.103 gross kWh/L. The whole-period result is 3.000 gross kWh/L, calculated from total energy divided by total fuel.

An unweighted average of the two interval ratios would give about 2.623 kWh/L. That is the wrong daily figure because the intervals consume different fuel quantities. Sum the underlying quantities before calculating the combined ratio.

A different load mix can change the result without deterioration

Consider another hypothetical ten-hour period using the same assumed load-specific rates. It spends eight hours at 150 kW and two hours at 450 kW. Gross energy becomes 150 × 8 + 450 × 2 = 2,100 kWh. Fuel becomes 70 × 8 + 145 × 2 = 850 litres.

The resulting 2,100 ÷ 850 = 2.471 gross kWh/L is lower than the first day's 3.000. Yet the assumed fuel rate at each load is unchanged. The generator used fewer litres in total. Yet the different load mix produced a worse whole-period ratio.

This is why a daily trend needs context. Compare similar load bands, or use a justified model that accounts for the change in conditions. Keep low-load and idle periods visible. They may reflect the production schedule or the need for backup power. The maintenance team may not control those requirements.

A measurement report does not itself authorize changes to generator dispatch. The required reserve, operating limits and load response must still be assessed. Its role is to show which conditions produced the observed result.

Account for fuel returned to the tank

Some fuel systems circulate more fuel through the supply line than the engine consumes. Unused fuel returns through another line. Measuring supply alone can therefore overstate consumption.

KRAL's differential-measurement paper explains the supply-minus-return approach. It also discusses temperature differences and other operating effects. The flow boundary must include every relevant route. Return fuel inside the original tank boundary is an internal movement, not a new external receipt.

Suppose a separate meter check for the example period records 20,000 litres supplied and 18,700 litres returned. Assume both totals use a common volume basis, cover the same time and include all relevant flows. There is no change in fuel held between their measurement points. The difference is 1,300 litres, matching the stock balance.

That agreement is encouraging, but it does not prove that either measurement is accurate. The two methods may share a conversion error. Their uncertainty and independence still matter.

A small difference can have a large relative error

To illustrate the arithmetic, assume each flow total has a bounded error of plus or minus 0.5% of its reading. This is an invented error model for the example, not a claim about the instruments fitted in a factory.

The supply bound is 20,000 × 0.5% = 100 litres. The return bound is 18,700 × 0.5% = 93.5 litres. If the errors act in opposite directions, their effects on the difference add. The worst-case bound from these two terms is 100 + 93.5 = 193.5 litres.

Compared with the 1,300-litre net consumption, that is 193.5 ÷ 1,300 × 100 = 14.88%. It is not a statistical confidence interval, and it excludes other possible errors. The individual meter percentage does not describe the error in the much smaller difference.

Research by Kroner and colleagues at PTB, RISE and IB-HAWE examines fuel meters under conditions closer to actual use. Their study highlights flow range, changing flow and fuel properties. A calibration result covers stated conditions. It should not silently stand for every operating condition.

Meter selection and installation need the engine supplier and a competent measurement specialist. They must consider the intended flow range, return temperature and system restrictions. The accounting team can define the required evidence without redesigning the fuel circuit.

The size of the proposed improvement should influence that evidence request. If a trial aims to show a small change, ask whether the complete measurement system can resolve it. A highly detailed display does not establish that capability. Review the relevant calibration results, repeatability, reading resolution and unresolved corrections. Longer observation may reduce the effect of rounding. It will not remove a persistent conversion or boundary error. Measurement uncertainty may be too large to distinguish the claimed change. If so, report the result as unresolved. Explain which further check would settle it.

Convert fuel to energy only when the basis is known

A thermal-efficiency calculation needs the fuel's heating value. Assume a lower heating value of 9.8 kWh per litre for this example. Its volume basis matches the ledger. This is a chosen teaching value. An actual report needs a supported fuel property. Keep the relevant test or supplier record.

Fuel energy is 1,300 × 9.8 = 12,740 kWh of fuel energy. Gross electrical efficiency is 3,900 ÷ 12,740 × 100 = 30.61%. Net electrical efficiency is 3,810 ÷ 12,740 × 100 = 29.91%.

The different percentages reflect different output boundaries. Neither is obtained by treating litres as though they were units of energy.

HOMER's generator-efficiency documentation describes electrical output relative to chemical energy input. It also explains the need for fuel density when converting a liquid volume into mass. A fuel property in MJ/kg requires a density conversion when the ledger uses litres. The energy units must also agree: one kWh equals 3.6 MJ.

Higher heating value and lower heating value are different conventions. The DOE Alternative Fuels Data Center glossary explains the distinction. Water formed during combustion is treated as condensed in one convention and as vapour in the other. State the convention beside the efficiency result. Changing that convention changes the calculated percentage without changing the engine.

A supplier may quote brake efficiency. That output boundary differs from the factory's net electrical efficiency. Ask where the output was measured and which losses it includes before comparing performance.

Turn a suspect trend into a useful investigation

A falling kWh/L trend deserves a sequence of questions. First establish whether the fuel and electricity records still cover the same system and period. Then check whether the operating conditions changed. Resolve those issues first. The trend can then provide stronger evidence for an equipment investigation.

Keep records of running hours, load intervals and starts. Fuel batches, maintenance events and relevant ambient conditions also help explain changes. A missing meter interval should remain visible. If you estimate it, keep the method and identify the affected result. Label the estimate so readers can distinguish it from a direct reading.

An unexplained fuel difference is also not proof of theft or engine loss. It may reflect a missing transfer, a tank conversion issue, a cutoff error or a genuine physical loss. Give the unresolved quantity an owner and retain the evidence needed to distinguish those causes.

If the trend points to maintenance, review the fuel observation alongside the machine's other condition records. ApparelDiary's preventive and predictive maintenance guide helps frame that decision. A universal calendar interval cannot replace the applicable equipment instructions and duty history.

Before claiming a saving, decide how the baseline will be compared with later operation. DOE FEMP's measurement-and-verification guidance covers federal performance contracts. It addresses baseline conditions, operating hours and agreed adjustments. The relevant lesson here is to settle the measurement basis before the improvement claim. That guidance does not certify an unrelated factory trial.

Carry a defensible result into the cost report

Assume a fuel price of 120 currency units per litre. The corrected period's fuel cost is 1,300 × 120 = 156,000 currency units. Dividing by 3,810 net kWh gives 40.94 currency units per net kWh for fuel alone.

That is not the generator's complete cost of electricity. Maintenance, ownership and other costs need their own defined treatment. A tariff comparison needs a matching service and period. It must also account for taxes and demand charges.

If the factory allocates this cost to garments, show how. Link the utility cost pool to the chosen allocation basis. The garment costing breakdown discusses how factory costs reach a production price. Allocating a cost does not change the amount of fuel used.

Fuel-price changes also deserve a separate explanation. A higher purchase price can raise cost per kWh even when fuel consumption per kWh is unchanged. Show the physical ratio beside the unit price. Finance and engineering can then see what caused the change. That separation helps the team choose the right action.

The final generator record should let another person retrace the result. It should name the period, tanks and meters, then explain the output boundary and load conditions. Keep corrections separate from physical improvements. A reliable ratio begins with reconciled quantities and ends with an explanation that the next shift can test.

References