Round-trip efficiency: what the 92% figure actually measures
Two suppliers quote 92% and 88% round-trip efficiency for the same 100 kWh cabinet. Neither number is necessarily wrong. They were measured at different points on the same circuit, and until you know where the meter sat, the two figures cannot be compared or priced.
Round-trip efficiency (RTE) is the ratio of energy delivered out of a storage system to the energy put into it over one full cycle. It sounds unambiguous and is not. The answer moves by several percentage points depending on the measurement boundary, the power level, the ambient temperature and whether the auxiliary loads were running during the test. For a commercial and industrial (C&I) system cycling once a day, that spread is worth more than the difference between two competing quotations.
This page covers the boundary itself: what each quoted figure includes, what the standard requires, and the two losses that no RTE number captures. It follows our page on usable versus nameplate kWh, which decides how much energy the cabinet holds. This one decides how much of it reaches the meter.
The short answer
An RTE figure is only comparable when four conditions are stated alongside it: the measurement boundary, the charge and discharge power, the ambient temperature, and whether auxiliary consumption is inside or outside the measurement. In commercial offers, three boundaries are quoted, and they are not interchangeable.
| Boundary | What the number includes | Indicative range | Where it usually appears |
|---|---|---|---|
| Cell | Coulombic and internal-resistance losses inside the cell only | 97–99% | Cell datasheets — never a system figure, though it is sometimes presented as one |
| DC block | Adds module interconnect, DC cabling, contactors and BMS measurement | 90–96% | Battery datasheet, often labelled "DC round-trip efficiency" |
| AC at the converter terminals | Adds the charge and the discharge conversion stages | 86–92% | The figure in most C&I offers, quoted without a boundary statement |
| AC at the point of connection | Adds the transformer, AC cabling and all auxiliary loads | 84–90% | What a meter at the site actually records, and what the bill reflects |
The ranges are indicative and move with rated power, storage duration and cell chemistry. The point is the gap between the first row and the last: up to fifteen percentage points can sit between a cell datasheet and a site meter, on identical hardware.
What the standard actually specifies
IEC 62933-2-1 defines round-trip efficiency as total output energy divided by total input energy over one charge and discharge cycle at rated input and output power, measured at the point of connection, with the energy consumed by the auxiliary subsystem included in the calculation. The test procedure starts from a fully discharged system, runs at least two complete cycles at rated power, and averages the results.
Three consequences follow, and each one is a place where an offer can drift away from a verifiable claim.
- The test is at rated power, not at your operating point. Auxiliary consumption — dominated by the cooling system — is at its maximum during the test. That makes the number conservative on losses for part-load operation, but it also means the figure does not describe the losses your site will actually experience.
- Auxiliaries are inside the boundary. A cooling system drawing 2 kW while the system discharges at 100 kW removes 2% of output on its own, before any conversion loss is counted. A test that quietly disconnects the thermal management reports a better number than the site will ever see.
- The boundary is a location, not a percentage. Move the measurement point across a transformer and the efficiency changes while the hardware does not. Two suppliers can both be honest and still quote numbers that differ by four points.
A supplier quoting "92% RTE" without naming the standard, the boundary and the treatment of auxiliaries is quoting a marketing figure. It may be a true figure. It is not a specification, because nothing in it can be tested at acceptance.
Where the losses sit, and which ones your number includes
The clearest way to read any efficiency claim is to name the loss elements and ask which of them are inside the measurement. The last two rows are the ones that go missing most often.
| Loss element | Loses energy when | Inside a converter-terminal figure | Inside a point-of-connection figure |
|---|---|---|---|
| Cell internal resistance and coulombic losses | Charging and discharging | Yes | Yes |
| DC cabling, contactors and BMS shunt | Charging and discharging | Yes | Yes |
| Converter conversion losses, both directions | Charging and discharging | Yes | Yes |
| Transformer and AC cabling | Charging and discharging | No | Yes |
| Cooling, pumps and heaters | While the system is active | Only if the test keeps them energised | Yes |
| Standby and idle draw | When the system is neither charging nor discharging | No | No |
Why four points is not a rounding error
Take a 100 kWh cabinet with a 90% depth of discharge, so 90 kWh of usable capacity, cycling once a day for 350 days a year. Compare a system measured at 92% with one measured at 88%, on the same boundary.
- At 92%: 82.8 kWh delivered per cycle, about 28,980 kWh a year
- At 88%: 79.2 kWh delivered per cycle, about 27,720 kWh a year
- Difference: 3.6 kWh a day, about 1,260 kWh a year
If the delivered energy displaces a peak tariff of USD 0.15/kWh, that gap is roughly USD 189 a year, or about USD 1,890 over ten years before discounting — around USD 1,270 in present value at an 8% discount rate. Against a 100 kWh cabinet costing USD 250–350/kWh, or USD 25,000–35,000, four points of efficiency is therefore 5–8% of the capital cost. It is larger than the difference most buyers negotiate on price per kilowatt-hour.
The arithmetic also runs the other way. If the site needs a fixed amount of delivered energy rather than a fixed nightly charge, the penalty lands on the input side: the lower-efficiency unit draws about 4 kWh more per night for the same output, and every one of those kilowatt-hours also counts against the throughput cap in the warranty. Efficiency and warranty are not independent line items.
The load that no efficiency figure counts
The standard’s measurement window is open while the system charges and discharges, and closed the rest of the time. A C&I cabinet on one cycle a day is idle for most of the twenty-four hours, and during that time the BMS, the EMS, contactor control, anti-condensation heaters and any cooling held in standby continue to draw power.
| Continuous standby draw | Annual energy, 24 h a day | Annual energy, 20 h idle a day |
|---|---|---|
| 80 W | about 700 kWh | about 584 kWh |
| 150 W | about 1,314 kWh | about 1,095 kWh |
Set that against the 1,260 kWh a year that four points of round-trip efficiency is worth in the example above. A cabinet with 150 W of parasitic load can cancel most of the efficiency advantage of a competitor, and the loss will never appear in an RTE figure — because the measurement window was closed while it happened.
This is not a corner case. One cycle a day with long idle periods is the normal duty of peak-shaving and solar self-consumption systems, which is precisely where standby consumption does the most damage per kilowatt-hour delivered. Two questions cover it: what is the standby consumption in watts, and are the heaters thermostatically controlled or permanently energised.
Part load is where the datasheet number stops applying
The IEC test is run at rated power. A commercial site rarely runs there: peak shaving and self-consumption typically operate at 30–60% of rated power for a few hours, then stop. Conversion efficiency curves are flattest in the mid-to-upper load range and fall away at low load, while cooling power does not scale down in proportion to throughput. Two systems with identical rated-power efficiency can therefore differ meaningfully at the point the site actually operates.
The remedy is a curve rather than a number. Ask for efficiency at 25%, 50%, 75% and 100% of rated power, and for the round-trip figure at the expected average discharge power. A supplier with only the rated-power number is giving you an upper bound, not a prediction.
What to write into the specification
Six conditions turn an efficiency claim into something that can be tested on site. The clause below is deliberately explicit about the boundary and about the auxiliaries, because those are the two variables a supplier can move without changing a single component.
System round-trip efficiency shall be not less than 88%, measured at the AC terminals of the battery converter at the point of connection, at rated charge and discharge power, at 25 °C ± 2 °C ambient, over a full cycle from minimum to full available energy and back, with all auxiliary loads energised, in accordance with IEC 62933-2-1. Auxiliary loads shall not be disconnected for the test. The manufacturer shall state efficiency at 25%, 50%, 75% and 100% of rated power, the continuous standby consumption in watts, and the annual efficiency degradation assumed over the warranty period.
Then decide how it will be verified:
- The meter. Charge and discharge energy measured with a calibrated meter of accuracy class 0.5 or better, on the same boundary the guarantee is written against.
- The moment. Acceptance test after commissioning, with the thermal management running as it would in service, at the ambient temperature the guarantee names.
- The remedy. A minimum floor below which the supplier must correct or compensate. An efficiency figure with no consequence attached is a description, not a guarantee.
Two of those conditions are worth more than the number itself: auxiliaries energised, and the boundary named as a location. A test that docks the cooling system, or measures at the DC terminals while the buyer pays for AC energy, will produce a passing result for a system that underperforms in service.
How efficiency fits the quote comparison
Our quote comparison worksheet normalises offers on price per usable kWh, which is a capacity measure and deliberately excludes efficiency. Efficiency belongs in the remarks column, and in a second calculation that most buyers skip:
cost per delivered kWh = price per usable kWh ÷ system round-trip efficiency
On the example above, USD 300 per usable kWh at 92% is USD 326 per delivered kWh, and the same USD 300 at 88% is USD 341 — a 4.6% difference that no capacity-based ranking will show. Add the standby consumption over the warranty period and the ordering can change again. The 12 fields in the RFQ checklist do not include efficiency as a priced column for the same reason it is easy to game; it is a gate, and it is checked against the market documentation in the compliance document index.
Frequently asked questions
Is 92% round-trip efficiency realistic for a 100 kWh cabinet?
It depends entirely on the boundary. At the converter’s AC terminals, an LFP system with an integrated converter typically lands in the high 80s to low 90s, so 92% is plausible for a well-matched system at rated power. At the point of connection with a transformer and auxiliaries included, 92% would be an unusually good result and should be substantiated with a test report rather than a datasheet.
Why is the measured efficiency lower than the datasheet after commissioning?
Four usual causes: the meter sits at a different boundary than the guarantee, the system runs at part load rather than rated power, the ambient temperature is above the 25 °C test condition so cooling works harder, or the auxiliary loads were excluded from the datasheet figure. All four are visible in the first week of operating data if the boundary is known, which is why the specification has to name it.
Does DC coupling really beat AC coupling on efficiency?
Generally yes, because a DC-coupled system avoids one conversion stage when the battery is charged directly from the PV array. The advantage is several percentage points in a solar-charging duty cycle and close to nothing in a grid-charging one, where both topologies convert twice. The comparison is only meaningful for the charging pattern the site will actually run.
Should the transformer be inside the efficiency guarantee?
If you own the transformer and it sits between the battery and the meter you are billed on, then yes. Excluding it moves roughly one to two points of loss outside the guarantee and onto your operating cost, which is a legitimate commercial arrangement only when both parties know it is happening. Name the boundary as a physical location in the contract.
Does round-trip efficiency degrade over the warranty period?
Yes. Rising cell internal resistance increases losses in both directions, and a decline of roughly 0.5 to 1 percentage point a year is a common modelling assumption for LFP systems. Ask the supplier for the figure it expects rather than assuming a flat value, and check whether the warranty has anything to say about it — most capacity warranties do not cover efficiency at all.
Modelling a project on an efficiency figure?
Send the load profile and the tariff structure. We reply within 24 hours with the configuration we would propose, the efficiency basis we would guarantee at the point of connection, and the standby consumption to expect.
