A tank level gauge and a fuel flow meter answer different questions. The gauge says how much fuel is in the tank at the moment of reading; the meter says how much passed a point in the line. Neither one, on its own, proves how much the generator consumed. Reconciliation is the discipline that connects them: it accounts for every delivery, transfer, return and drain, compares the result with the meter, and explains the difference before it becomes an outage or a stock loss.
A tank level is a stock figure, a meter is a flow figure
A level reading is a snapshot of volume. It is affected by the shape of the tank, the temperature of the fuel, the tilt of the base, the amount of water and sludge at the bottom and the position of the sensor. A flow meter is a record of movement. It is affected by the fluid profile at the measuring point, entrained air, the direction of flow and the condition of the meter itself. Treating either instrument as a direct consumption figure removes the evidence that would show which one is wrong.
The distinction matters most on a standby site, because the engine is not the only thing that moves fuel. A delivery adds volume, a transfer to a day tank removes it from bulk storage, a polishing loop circulates it, a drain removes water and a test run burns it. If the inventory equation includes only the engine, every other movement appears as an error. If it includes all of them, the remaining difference becomes a measurement worth investigating.
The basic equation is simple: opening stock plus deliveries, minus closing stock, equals the fuel that left the tank. Subtract transfers out and add transfers in where the boundary of the calculation includes more than one tank. Compare that result with the net meter total over the same period. The gap is not automatically a leak, and it is not automatically a meter error; it is the quantity that the reconciliation process exists to explain.
That explanation has practical value. Autonomy depends on usable litres, not on nominal tank size, and a leak, a stuck level device or a drifted meter all show up first as a gap between two records that should agree. The earlier the gap is seen, the cheaper it is to resolve.
Four measurement points and what each can prove
Most sites already have more measurement than they use. The generator controller may record fuel rate, the tank may have a level transmitter, the output may have a kWh meter, and the fuel line may have a meter installed for another purpose. The task is to give each signal a defined role and a common time base, so that a comparison means something.
| Measurement point | What it measures | What it can prove | What it cannot see |
|---|---|---|---|
| Tank level, mechanical or electronic | Volume in the tank at tank temperature | Stock before and after a delivery, transfer or run | Density, tilt, foam, water and unusable heel |
| Supply flow meter | Gross volume leaving the tank or entering the engine | Fuel movement during a run, including leaks downstream | Return flow, so it overstates net consumption unless paired |
| Return flow meter | Volume returned from the engine or injection system | Net consumption when subtracted from supply | Low flows, air slugs and reverse flow at shutdown |
| Engine fuel rate from the controller | Calculated burn from injector or ECU data | Trend and comparison with the manufacturer’s curve | Leaks, transfers and any fuel that never reaches the engine |
| Generator kWh meter | Electrical energy delivered at the terminals | Litres per delivered kWh when paired with a fuel total | Fuel volume, density and tank stock |
The points should share a clock. A tank reading at 08:00, a meter total at 09:30 and a delivery at 10:15 cannot be reconciled without adjustment. Close the period at a fixed time, read every instrument in a defined order and record that order. With remote monitoring, the same rule applies to historian time stamps.
Each point also needs a stated basis. A level transmitter may report millimetres, a percentage or litres after a strapping table; a flow meter may report gross litres, net litres or mass. Two numbers with the same unit are not necessarily the same quantity. The reconciliation sheet should name the unit, the reference temperature and the source of the conversion, not only the value.
The reconciliation equation, worked once
Set the boundary before writing the equation. It may be the bulk tank alone, the bulk tank plus the day tank, or the whole fuel system including the polishing loop. Each boundary needs its own opening and closing readings, and every crossing of the boundary needs a record. A transfer between two tanks inside the boundary is not a consumption event; a transfer out of the boundary is.
The arithmetic is easier to trust when it is written as a sequence rather than a formula. Start with the opening stock, add every delivery, add any transfer into the boundary, subtract any transfer out, and subtract the closing stock. That gives the indicated consumption for the period. Then take the meter total, subtract the return total where a return meter exists, and compare the net figure with the indicated consumption.
An illustrative case makes the scale clear. A 1000 kVA set running at about 800 kW for 24 hours uses about 200 litres an hour, or about 4800 litres, and delivers about 19,200 kWh. At 4800 litres that is about 0.25 litres per delivered kWh, inside the 0.24 to 0.29 band used across this site. Suppose the tank opened at 12,000 litres, received a 5,000 litre delivery and closed at 12,200 litres. The indicated consumption is 12,000 + 5,000 – 12,200 = 4,800 litres. If the net meter total is 4,720 litres, the difference is 80 litres, or about 1.7 percent of the indicated figure. The example is arithmetic, not a site record; its purpose is to show that the gap is small enough to be missed and large enough to matter over a year.
A gap of that size may sit inside the tolerance written into the project specification, or it may be the first sign of an unrecorded transfer. The number only becomes meaningful when the tolerance and the flow map are agreed in advance. A site that adjusts the meter to match the tank every month has stopped measuring; a site that investigates the same gap every month is learning where its fuel actually goes.
| Line | Quantity | Sign in the equation | Source of the record |
|---|---|---|---|
| 1 | Opening stock | Add | Level reading converted through the strapping table |
| 2 | Deliveries received | Add | Delivery ticket and before-and-after level readings |
| 3 | Transfers into the boundary | Add | Transfer log and receiving tank reading |
| 4 | Transfers out of the boundary | Subtract | Transfer log and source tank reading |
| 5 | Closing stock | Subtract | Level reading at the period close |
| 6 | Indicated consumption | Result of lines 1 to 5 | Calculated and checked by a second person |
| 7 | Supply meter total | Starting point for net | Meter register at period open and close |
| 8 | Return meter total | Subtract from supply | Meter register over the same period |
| 9 | Net meter total | Line 7 minus line 8 | Calculated |
| 10 | Difference | Line 6 minus line 9 | Reported with sign and percentage |
| 11 | Delivered kWh | Comparison basis | Generator kWh meter over the same period |
| 12 | Litres per delivered kWh | Line 6 divided by line 11 | Calculated against the 0.24 to 0.29 band |
Meter types and placement for diesel fuel
Diesel is a viscous, lubricating fuel that can carry air, water and particulate matter. That combination rules out some meters and makes placement as important as the measuring principle. A meter installed where it sees a stable, full pipe is more useful than a more expensive meter installed after a flexible hose and an air pocket.
The first placement decision is whether the site needs one meter or two. A meter on the supply to the engine measures gross fuel draw. If the engine returns surplus fuel to the tank, the supply figure is not consumption. A return meter, matched to the same time base, allows the net figure to be calculated. Where only one meter is practical, it should be on the engine supply after the last branch, with the return route documented and the gross figure labelled as gross rather than net.
The second decision is where the polishing loop sits. A polishing pump that circulates fuel out of the tank and back into the same tank can pass the same litre through a meter many times. Give the loop its own meter, or record its operation and exclude it from engine consumption. Mixing the two flows produces a number that looks precise and means nothing.
| Meter principle | How it measures | Strength | Limit to check before selection |
|---|---|---|---|
| Positive displacement | Counts fixed volumes passing through a chamber | Works over a wide viscosity range and tolerates modest dirt | Pressure loss, wear and the effect of air on the count |
| Turbine | Infers volume from the speed of a rotor | Compact and fast response at steady flow | Bearing wear, low-flow cutoff and sensitivity to viscosity |
| Coriolis | Measures mass flow from tube vibration | Direct mass, independent of density and temperature | Cost, pressure loss and installation orientation |
| Ultrasonic | Measures transit-time difference across the pipe | No moving parts and low pressure loss | Straight-run requirement, lining and entrained air |
| Differential pressure | Infers flow from a pressure drop across a restriction | Simple and well understood | Rangeability, fouling of the restriction and permanent loss |
| Engine controller estimate | Calculates burn from injector or ECU data | No additional pipework and useful for trend | Calibration, and blind to leaks or transfers outside the engine |
Every meter has a K-factor or calibration constant, and every constant changes with wear, contamination or a replacement part. The meter record should carry the constant, the date it was last verified and the method used. A meter that has never been checked is a trend indicator, not a custody instrument. The interval and the acceptable drift belong to the manufacturer’s data and the project specification, not to a rule copied from another site.
Temperature, density and the litre that is not a litre
A litre is a volume, and the volume of a given mass of diesel changes with temperature. Diesel expands by roughly 0.1 percent per °C, so fuel delivered at 45 °C into an external tank is around 3 percent less dense than the reference figure. The tank may receive the same number of litres as the ticket states and still hold less energy than the previous delivery. A meter on the line and a level gauge in the tank may both be working correctly while reporting volumes at different temperatures.
The 40 °C reference ambient used throughout this site’s guidance is a planning basis for the generator, not a fuel specification. A tank in direct sun or in an unventilated room can be hotter than the air around it, and a return line that runs beside a hot exhaust can raise the fuel temperature further. The reconciliation record should therefore carry the fuel temperature at the meter and at the tank, or the density used to convert volume to mass. Without that, a three percent seasonal difference looks like a metering fault.
The practical choice is to reconcile in one consistent basis. Either convert every volume to a stated reference temperature using the supplier’s density data, or reconcile in mass and convert to litres only for the report. Mixing a meter total at line temperature with a tank volume at tank temperature and a delivery ticket at the supplier’s reference is a reliable way to create a gap that has no physical cause.
| Source | Why it appears | How to treat it in the record |
|---|---|---|
| Thermal expansion | Fuel volume changes with temperature between delivery, storage and the meter | Record temperature and density; convert to a common reference or reconcile by mass |
| Return flow | The engine returns surplus fuel, so supply volume is not consumption | Subtract the return meter total, or label the supply figure as gross |
| Entrained air | Air from suction leaks, a low tank or a polishing loop passes through the meter | Check suction joints, vent and pump priming; record the event, not only the total |
| Water and sludge | Bottom volume is not usable fuel and displaces the level reading | Drain and measure water; inspect the tank; do not count the bottom volume as stock |
| Tank tilt and strapping | The level-to-volume table assumes a level base and a known geometry | Verify the strapping table at several levels and record any base or support change |
| Dead stock | The suction cannot reach the bottom of the tank | Define the unusable heel and subtract it from gross volume before calculating autonomy |
| Meter drift or replacement | The K-factor changes with wear, contamination or a new meter body | Record the constant and the verification date; investigate a step change in the ratio |
| Unrecorded transfer or drain | Fuel leaves the boundary without a matching entry in the log | Reconstruct the movement from transfer records before adjusting any instrument |
Turning tank level into usable inventory
A level sensor reports a position, not a volume. The conversion depends on the tank geometry, the orientation of the sensor and the strapping table that was produced for that tank. A float, a radar head, a pressure transmitter and a manual dipstick can all be valid, but each has a different failure mode. A float can be held down by falling fuel or made buoyant by foam; radar can be confused by build-up or a disturbed surface; a pressure transmitter can be fooled by a blocked impulse line; a dipstick depends on the person reading it and on the tank being level.
The strapping table should be verified at more than one level. A gauge that reads correctly at half level but wrongly at the low-level alarm is worse than no gauge, because it gives confidence at the point where the engine is about to lose fuel. The same applies to the high-level end: a reading that is correct in normal operation but optimistic near the fill limit can turn a routine delivery into an overflow.
Usable volume is smaller than nominal volume. The fuel pick-up sits above the tank floor, a water layer may occupy the bottom, and the high-level margin has to remain available for expansion and for the fill procedure. The fuel consumption guidance on this site notes that a tank is not usable to its nominal volume: a 1,000 litre tank gives about 850 to 900 litres of running fuel once the unusable heel and the fill margin are allowed for. That is the figure to use for autonomy, not the nameplate volume.
The scale matters. A 1000 kVA set at about 200 litres an hour needs about 4800 litres for a 24-hour run. A 1,000 litre tank is a buffer or a day tank in that context, not a 24-hour reserve. A 500 kVA set uses 101 litres an hour at full load, 80 at 75 percent and 60 at 50 percent, so the same tank gives a very different number of hours depending on the load. The reconciliation should show usable litres and the load assumption beside the hours, otherwise the autonomy figure is not reproducible.
| Inventory item | Definition | Evidence to keep |
|---|---|---|
| Gross volume | Volume from the level reading through the strapping table | Level, temperature, table version and time of reading |
| Water bottom | Free water at the lowest point, measured by drain or interface probe | Drain record, quantity and date |
| Dead stock | Fuel below the suction that cannot be drawn in normal operation | Tank drawing and pick-up elevation |
| Usable volume | Gross volume minus water and dead stock | Calculated and checked against the table |
| High-level margin | Space kept for expansion and for a safe delivery stop | Fill procedure and alarm setpoint record |
| Running autonomy | Usable volume divided by consumption at the stated load | Load assumption and litres per hour |
| Day tank buffer | Local volume between the low-level alarm and the high-level stop | Level sequence and transfer-pump record |
Reading the difference: what each discrepancy pattern means
A single difference number is not a diagnosis. The pattern over time is. A tank that falls faster than the meter every month points in a different direction from a meter that reads high only after a filter change. The first step is always to reconstruct the movement record, because a missing transfer or a delivery entered against the wrong tank can explain a gap without any instrument being wrong.
The second step is to check the instruments against a known quantity. A measured container, a proving run or a comparison with a second meter can establish whether the meter is reading correctly. The tank can be checked by a physical dip or by a level-to-volume comparison at a stable condition. Neither check requires a change to the operating setpoints, and neither should be used to force the two records to agree.
The third step is physical. A leak, a weeping joint, a stuck float or a drain left open is a site condition, not a metering error. The reconciliation points to the area; the inspection confirms it. A site that starts by adjusting the meter can hide a leak for months, and a site that starts by assuming a leak can spend a lot of time on a delivery that was simply recorded against the wrong tank.
| Pattern | Likely direction | First check |
|---|---|---|
| Tank falls faster than the net meter | Unrecorded transfer, leak or meter under-reading | Transfer and drain log, then visible pipework and joints |
| Meter total higher than tank loss | Return flow not subtracted, air through the meter or meter drift | Return meter, suction joints and meter K-factor |
| Both records high against delivered kWh | Load higher than assumed, engine efficiency loss or ambient derating | Load profile, kWh meter and the 40 °C reference conditions |
| Delivery gain lower than the ticket | Temperature difference, tank shape, water or a short delivery | Before-and-after level, fuel temperature and delivery record |
| Step change after meter or filter work | New K-factor, reversed connections or air in the line | Work record, meter constant and a proving run |
| Step change during polishing | Polishing loop counted as engine consumption | Polishing meter and loop operating log |
| Gauge stuck at a plausible level | Float, radar or impulse-line fault | Physical dip and sensor response at two levels |
Send the tank data, meter schedule and load profile
Tell us the tank geometry, the meters already installed, the delivery and transfer records you keep, and the load profile the site actually runs. The reply sets out the reconciliation boundary, the readings to close each period and the points to verify before the next fuel delivery.
Can a tank level gauge replace a fuel flow meter?
No. A gauge reports stock at one moment; a meter reports flow through a point. Reconciliation needs both, together with delivery and transfer records, because the engine is not the only thing that moves fuel.
Where should a diesel fuel meter be installed?
On a stable, full pipe after the last branch, with a matching return meter where the engine returns fuel. Straight runs, orientation and accuracy class come from the manufacturer’s data and the project specification.
How often should tank inventory be reconciled?
At every delivery and transfer, and at a fixed monthly close. Close the period again after a long outage, a meter change or any work on the fuel line, so the next comparison starts from a known record.
Why does the meter read more fuel than the tank lost?
Common causes are return flow that was not subtracted, air passing through the meter, an unrecorded transfer or meter drift. Check the movement records and the installation before changing the K-factor.
How do I calculate usable fuel in a tank?
Deduct water, dead stock and the high-level margin from the gross volume. Use the result with the load-specific litres per hour, not the nameplate tank volume, when you state autonomy.
How does the annual load test fit the fuel record?
The annual test at 50 to 100 percent of rating gives a known load and duration, so the fuel used can be compared with the meter and the tank. The 10 to 20 percent band is continuous running and does not replace that test.
KVA Grid sales engineering desk. The figures used here are published elsewhere on this site: the annual load test at 50 % to 100 % of rating and the 10 % to 20 % light-load band from the maintenance schedule; 0.24 to 0.29 litres per delivered kWh, and 101, 80 and 60 litres an hour for a 500 kVA set at full, 75 % and 50 % load, from fuel consumption; the 1000 kVA example at about 200 litres an hour and about 4800 litres for 24 hours, the UPS role between mains loss and generator load, and the PUE 1.4 example from data center backup power; the 40 °C reference ambient from hot and humid climates; tank sizing from generator fuel tanks and autonomy; day tank levels from day tank layout and return-line sizing; transfer-pump protection from transfer-pump dry-run protection; delivery levels from vent sizing and overfill protection; fuel condition from fuel polishing; load-bank practice from load-bank testing; and commissioning records from the commissioning checklist. Meter accuracy, correction factors, reconciliation tolerances, tank strapping and alarm setpoints are set by the manufacturer’s data and the project specification. Reviewed 7 November 2026.
