Evidencing battery warranty compliance from the owner's side

A battery warranty is usually built around capacity retention: the supplier commits that the system will retain some stated fraction of its rated capacity at defined points in time or after defined throughput, tested under a defined procedure.

Wrapped around that commitment is a set of conditions on how the asset is operated. Temperature limits, charge and discharge rate limits, depth of discharge, time spent at high or low state of charge, throughput ceilings, and requirements on maintenance and on the operating environment.

The structure means the evidence burden falls on the owner. The supplier's obligation is contingent on the owner demonstrating that the asset stayed inside the envelope, and demonstrating it requires data the owner has to have been keeping from day one.

The executed contract governs in every case, and warranty terms vary enough between suppliers that generic advice about what they say is worth less than an afternoon spent reading the one you signed.

Temperature is usually the strictest and the most consequential. Both the limits and the time spent near them matter, and cell temperature is what the warranty means even where the available measurement is module or ambient temperature. Where those differ, the difference is worth quantifying before a dispute rather than during one.

Rate limits, expressed as C-rate, bound how hard the pack may be charged and discharged. A market duty that occasionally calls for a fast response can breach these in short bursts that an averaged record will not show.

Depth of discharge and state of charge dwell. Deep cycles are usually counted differently from shallow ones, and prolonged time at high state of charge has its own effect, so a record of the state of charge distribution matters as much as the cycle count.

Throughput, in energy delivered over the life, which is frequently the binding term for a heavily cycled asset and the one that quietly determines when the warranty ends.

Each of those is a measurement with a resolution requirement attached, and the resolution has to be fine enough to show a short excursion rather than an average that hides it.

Cycle counting definitions vary enough that the same duty produces materially different counts under different contracts, which makes this one of the first things to pin down.

Equivalent full cycles, where partial cycles are accumulated by energy throughput, is the most common and the most forgiving of a duty with many shallow cycles. Rainflow counting, borrowed from fatigue analysis, decomposes a complex profile into equivalent cycles of varying depth and is more representative of what the cells experience. Simple event counting, where any discharge above a threshold counts as one, is the crudest and can be either generous or punitive depending on the duty.

The definition matters more on a battery than on most assets because market duties produce complex, shallow, frequent cycling that the definitions treat very differently.

Whichever applies, the counting has to be done continuously and from commissioning. A count reconstructed later from a historian that compressed the profile is not evidence, and compression is exactly what destroys the shallow cycles the count depends on.

The commitment is on capacity retention and the retention is established by a test, so the test procedure decides the result more than the battery does. It is also, in many contracts, the most underspecified part.

The parameters that determine the answer: the rate at which the test is conducted, the temperature and how it is controlled, the state of charge window, the rest periods before and between steps, how many repetitions are averaged, who witnesses it, and how the result is corrected if conditions departed from the specification.

A test run at a different rate or temperature from the one the warranty assumed produces a different number from the same pack, and both parties can be arguing in good faith about a result neither procedure fully defines.

The preparation worth doing is to run the test procedure once, early, while nothing is in dispute. It establishes that the procedure is executable at this site, produces a reference result, and surfaces the ambiguities while they can still be clarified cheaply.

Many storage assets are designed to be augmented, with additional capacity added as the original degrades. That is sensible engineering and it complicates every later comparison.

After augmentation the system contains cells of different ages and possibly different specifications. A pack-level capacity measurement no longer describes a single population, and the warranty on the original cells has to be tracked separately from the warranty on the added ones.

The treatment has to be agreed before the first augmentation, not after: how retention is measured for each vintage, how the combined system is tested, how throughput is allocated between vintages, and what happens to the original commitment.

This is where cell-level or at least string-level records become necessary rather than merely useful, because an aggregate that mixes vintages cannot be decomposed afterwards.

Resolution fine enough to show a short excursion. Retention set to the warranty term rather than to a historian default. Provenance that is separable from the party being measured, which is the same structural point as in reporting to lenders after commercial operation and applies for the same reason.

Traceability from any figure in a claim back to measurements, without a spreadsheet in the middle that nobody can reproduce.

A measurement inventory, written down: which parameters, at which points, at what resolution, retained how long. It answers most of the questions that get asked and its absence signals that nobody thought about the boundary.

And the estimation method for state of health, stated, because retention is going to be assessed against an estimate and the method behind that estimate is contestable. That is covered in estimating state of health from operational data.

Map every defined term in the warranty to a measured signal. Terms that map to nothing are the ones that will be argued about, and finding them at signature is far cheaper than finding them at claim.

Run the capacity test procedure once, early, and record everything about how it was run.

Set retention deliberately, at commissioning, against the warranty term and any dispute window.

Establish whether cell or module-level data is available and at what granularity, because that is fixed by procurement and cannot be added later without opening enclosures.

And keep the operating record continuously from the first day of commercial operation, because a warranty argument asks about the whole term, and no instrument installed today can describe the year behind it.