A lithium iron phosphate battery does not fail the way people expect a battery to fail. There is no warning light, no dead cell, no morning when the system will not start. What happens instead is quieter and more expensive: the pack slowly holds a little less than it did, year on year, while the dispatch plan written for its original capacity keeps asking it for energy it can no longer store. By year three this gap has grown large enough to matter, and on most sites nobody is reading it. That is the failure mode worth understanding, because it is the one your installer almost certainly did not warn you about.
This is not a criticism of installers. Capacity fade is invisible at handover, it breaches no specification on day one, and it does not become material until two or three years after the truck has left, by which point the relationship that built the asset is long over. Somebody still has to read the curve. This is a short guide to what that curve is, how to read it, and what it should change about the way the asset is run.
State of Health is a measurement, not a feeling
State of Health, or SoH, is the single number that tells you how much of a battery’s original capacity it can still hold. A pack at 100 percent SoH stores its full nameplate energy. A pack at 90 percent stores nine tenths of it. The number only ever falls, and the rate at which it falls is what separates an asset that meets its model from one that quietly drifts below it.
The problem is that SoH is not handed to you. On most behind-the-meter systems it is not published as a clean telemetry stream at all, so it has to be derived from how the battery has actually been used rather than read off a gauge. The honest way to do that is to reconstruct it from state-of-charge history: count the cycles the pack has done, weight each one by how deep it went, and compare that accumulated wear against the chemistry’s known cycle life. That is the calculation Soluno runs as part of its battery health stewardship, using rainflow cycle counting against per-chemistry depth-of-discharge tables to turn raw charge and discharge history into a wear figure and an effective cycle count. An unattended asset produces all the same data and nobody ever turns it into a number.
What the LFP degradation numbers actually say
Lithium iron phosphate is a durable chemistry, which is exactly why people stop paying attention to it. Capacity typically declines on the order of one to four percent per year, driven by temperature, charge and discharge rate, total throughput and how deeply the pack is cycled. Warranty end-of-life thresholds generally land somewhere in the 60 to 80 percent State of Health range after roughly ten years, depending on the manufacturer.
Take the middle of that fade band, a nominal two to three percent a year, and the arithmetic is sobering in a slow way. By the end of year three a pack is down by roughly 5 to 9 percent of its capacity. That is not a failure, and not even close to warranty replacement. But it is the inflection point, for three reasons at once:
- Usable energy starts to materially undershoot the dispatch plan that was written against the original nameplate.
- Cell-to-cell divergence widens, so how the battery management system balances the pack starts to matter more than it did when everything was fresh.
- The levers that extend remaining life, namely depth-of-discharge limits, thermal exposure and avoiding deep cycling, have their largest effect precisely now, while there is still life to protect.
Year three is where active stewardship earns its keep, not because the battery is in trouble, but because this is the first point at which how hard you work it visibly changes how long it lasts.
Cycle life is bought with depth of discharge
The reason year three behaves this way comes down to how cycle life is spent. “Cycle life” is conventionally defined as the point where usable capacity drops to 80 percent of the original rated capacity, the standard end-of-life benchmark. How many cycles you get to that point is not fixed. It is bought, and the currency is depth of discharge.
A representative LiFePO4 manufacturer table makes the trade explicit. At 100 percent depth of discharge a pack delivers on the order of 2,000 cycles, around 5 to 6 years of service. Ease that to 80 percent and it returns roughly 3,000 cycles, 7 to 8 years. At 50 percent the same pack yields about 5,000 cycles and 10 or more years, and at 30 percent it can reach 6,000 to 8,000 cycles and well past 12 years. The shallower each cycle, the more cycles you get, and the relationship is steep rather than linear.
The implication for an asset run for energy-cost arbitrage is direct. A battery cycled hard and deep every day to chase the spread between peak and off-peak tariffs ages faster than the same battery cycled gently. There is real money in that arbitrage, so the answer is not to stop cycling. It is to keep depth of discharge inside the band that preserves the cycles you are counting on, tracked against the warranty curve for that specific make rather than a generic assumption.
Degradation quietly rewrites your dispatch plan
Here is where the two halves come together, and where the cost lands. A dispatch plan is a set of instructions telling the battery when to charge and when to discharge, sized around how much energy the pack can hold. Set that plan against a fresh battery and it is correct. Leave it unchanged while the pack fades, and every year it becomes a little more wrong.
Consider an evening peak. In South Africa the high-demand season weekday evening peak runs 17:00 to 20:00, and in the low-demand season it shifts to 18:00 to 21:00, under the structure NERSA approved in February 2025. A dispatch plan assumes the battery can carry the site through that whole peak window on stored energy charged earlier at off-peak rates. If the pack has quietly lost 8 percent of its capacity, the plan still tries, the battery runs flat partway through the window, and the site draws the back end of its evening peak straight from the grid at the most expensive rate of the day. Nobody notices, because the lights stay on. The only evidence is on the bill, and only if someone is reading the bill against what the model promised.
This compounds with everything else moving underneath the asset. The cumulative direct Eskom tariff increases over 2023 to 2026 ran 18.65 percent, then 12.74 percent, then 12.74 percent, then 8.76 percent, compounding to roughly 64 percent over four years. So the value of every kilowatt-hour the battery can no longer deliver into the evening peak is rising at the same time as the battery’s ability to deliver it is falling. A degrading pack and a steepening tariff push in the same direction, and a frozen dispatch plan loses on both.
The discipline that fixes this is not a hardware swap. It is treating dispatch as a question to be re-answered as the pack ages. Soluno’s optimiser re-solves the next 25 hours of charge and discharge at 30-minute resolution and re-runs that calculation across every site on a schedule, so the plan is refreshed rather than frozen at commissioning. The wear curve derived from cycle history is what tells your consultant when the capacity assumptions behind that plan have drifted, so dispatch can be re-sized to the energy the pack actually holds today rather than the energy it shipped with.
What reading the curve looks like in practice
Stewardship of a battery past year three is not exotic. It is a small set of recurring disciplines, done consistently:
- Derive SoH from real state-of-charge history rather than waiting for a gauge that, on most systems, does not exist.
- Track accumulated cycles and working depth of discharge against the specific warranty curve for that make of pack, not a generic assumption.
- Keep depth of discharge inside the band that preserves the cycle life the project IRR was built on, easing it where backup duty allows.
- Re-size the dispatch plan to current capacity, so the battery is never asked to deliver energy it no longer holds.
- Read the gap between modelled and actual savings every billing period, in rand, and treat a widening gap as the signal it is.
None of this requires new cells. It requires somebody whose job is the asset’s performance across its whole life rather than its first day, and the instrumentation to turn raw charge history into a wear curve.
The question worth asking
If you hold a battery asset more than two years old, the question for whoever oversees it is simple: what is this pack’s State of Health today, and when did our dispatch plan last get re-sized to it? If the honest answer is that SoH has never been read and the plan has run unchanged since commissioning, the battery is almost certainly being asked for energy it can no longer give, and that cost is landing on the evening peak where energy is dearest. The fade itself cannot be reversed. The loss it is quietly creating in your dispatch plan can be, and recovering it costs a fraction of the cells already in the ground.
Sources and further reading
Want to know what your solar asset could save?
How Much Do I Save?