— y
A free, coarse model of the three ways a lithium cell loses capacity — calendar ageing, cycle wear and lithium plating. This is an estimate, not a measurement. It is at its best comparing one choice with another.
Three loss mechanisms, added: calendar ageing (Arrhenius, square root of time), cycle wear (Wang et al. 2011 throughput model) and charge-side lithium plating — the term that turns the curve back down when cells run cold. Every constant is labelled SRC (sourced) or ASM (assumption, calibration pending) in the model source. The model runs on the server; this page receives only the resulting coefficients and draws everything from them.
calendar: LFP Ea 20.6 kJ/mol, z 0.5 (SRC Schimpe 2018 / Naumann 2018) · NMC Ea 58, z 0.75 (SRC Schmalstieg 2014) · cycling: Wang et al. 2011 (SRC), with the calendar share of its temperature dependence removed — Wang’s cells aged on the calendar while they cycled, so his figure used as published counts temperature twice + deep-cycle stress (ASM) · cold branch of cycling below 25 °C — Schimpe 2018’s net shape, both of his terms (SRC) · plating threshold anchors 1C@25 / 0.3C@10 / 0.05C@0 °C (ASM) · anchored to the first-year calendar loss (MB31: from the spec storage bound, ASM) · cycle term scaled so the full model reproduces the cell’s rated cycle test · calendar averages g over the 24-hour SOC profile · uncertainty band = corners of the SRC ranges for Ea × z; the ASM constants are not varied, so the true band is wider · NMC: coarse variant
What it has been checked against. Checked against measured cycle-ageing data for a cell its maker rates at 6000 cycles to 80 %. Given that rating, the model overestimates wear by about 1.25× (median life error about 30 %) and ranks conditions correctly in 9 of 10 comparisons — and that part barely depends on the datasheet number; asked how much worse a change makes things, it is within roughly 12–15 %. Comparisons are its strong suit — a number of years is an estimate. Without a datasheet the tool assumes a cautious 3500 cycles, which for that same cell overestimates wear about 1.7× — so the datasheet number matters most.
Temperature: what goes in. The temperature you enter is what the pack is cooled to — the room, the cabinet, the coolant. Self-heating is added on top of it, because heat leaves the cells rather than entering them: in steady state no cell is colder than its surroundings, so the coldest cell is the surroundings themselves and the hottest one runs above them by as much as the cooling allows (a BMS shows that difference as max − min). A pack reaches end of life when its first cell does. Above the optimum that is the hottest cell; below it, the cell at the surrounding temperature, because charging a cold cell plates lithium. The page always shows the worse of the two.
A year is not one temperature. For a place that follows the weather, the model is given the temperature that ages a cell as fast as the whole year does — not the year’s arithmetic mean. The two differ because ageing is exponential in temperature: a year swinging between −3 and +21 °C wears a cell faster than a year held at its 9 °C average, so the plain average would flatter every outdoor installation. Months are weighted by their length and by the calendar activation energy of the chemistry in question. Two things this cannot express, and the page says both separately: the cycling term has its own much weaker temperature dependence, and lithium plating is a threshold rather than a rate — a January below freezing damages a pack that is charged in January, however mild the annual figure looks.
Limits. Sliders hold the model’s ranges; below 15 °C the tool still answers, with the plating banner, because refusing there would hide the turn. Past 15 years it says “more than 15” instead of a number: the calendar term is a square-root law fitted to studies a few years long. Throughput is normalised to the 2.2 Ah reference cell of the Wang fit. Profile presets are shapes (ASM). Datasheet cycle ratings are continuous 25 °C tests, often to 70 % SOH — at one cycle a day the rated count would take 20+ years, so field packs are usually calendar-limited; the tool anchors to the rating and computes what your conditions do to it. One coupling it ignores: high C-rate also heats the pack, so the self-heating you enter should be the one it reaches at that current.
What a “pack” is here. The model follows one cell, so the pack voltage never enters it: 12 V, 24 V and 48 V packs built from the same cells, at the same temperature, charge level and C-rate, age the same way. C-rate is power over capacity — a 2 kW load on a 2 kWh pack is 1 C at any voltage. What the model cannot see is how well the cells in a pack are made and matched. A pack is only as good as its weakest series group; LFP cells from two makers stored side by side for 30 months lost 0.3 % and 6 % at 10 °C (Gailani et al. 2020); and a BMS that is never taken through a full cycle can misreport the capacity left. Portable power stations add one more unknown: charger, inverter and cells share one case, so a unit kept plugged in as a backup can hold its cells warmer than the room. Kept at 100 %, a pack ages at the rate this tool shows for 100 % — in measured LFP storage data 1.5 to 2.5 times as fast as at half charge (Naumann 2018).
A battery needs a space that stays above freezing all year. A cabinet in the open air is not one.
Cold does not damage a battery that is only sitting there. Calendar ageing is a chemical rate, so it slows as the temperature falls: a pack parked at −10 °C ages more slowly than the same pack at +25 °C. Capacity lost to ageing while resting at 50 % charge, with nothing charged and nothing cycled:
| stored at | after 1 y | after 5 y | after 10 y |
|---|
Three limits, stated rather than buried. The calendar term behind these numbers is anchored to storage data at 25 and 40 °C — below 0 °C it extrapolates an Arrhenius slope rather than reporting a measurement. Datasheets bound long storage more tightly than this arithmetic does: the EVE MB31 spec allows −20…+45 °C for a month but only 0…+35 °C for a year, and that bound is about electrolyte and seals rather than fade. And self-discharge is not modelled at all — it is what actually ruins a pack left alone for a year, because the pack goes flat, and a deeply discharged cell sitting cold fails a different way. Store at about half charge, check on it, and bring it above 0 °C before charging it.
Working on it, a few seconds.
The cell that governs here is the cold one, and that side of the curve is published shape rather than measurement: it follows Schimpe 2018, fitted to one cylindrical LFP cell, and the blind check behind this model has measured data at 25 and 40 °C only, nothing colder. Trust the direction — cold charging is not free — more than the number of years.
This is an estimate, not a measurement. Checked against measured cycle-ageing data for a cell its maker rates at 6000 cycles to 80 %. Given that rating, the model overestimates wear by about 1.25× (median life error about 30 %) and ranks conditions correctly in 9 of 10 comparisons — and that part barely depends on the datasheet number; asked how much worse a change makes things, it is within roughly 12–15 %. Comparisons are its strong suit — a number of years is an estimate. Without a datasheet the tool assumes a cautious 3500 cycles, which for that same cell overestimates wear about 1.7× — so the datasheet number matters most.
Assembling prismatic cells? What the datasheets say about clamping them →
Against the one cycle-ageing dataset it has been checked on, this model overestimates wear by about 1.25× when it is given the cell’s datasheet rating, and about 1.7× on its cautious no-datasheet default — and it has never seen a large prismatic cell. If your pack disagrees with the number above, that is useful — especially if you have capacity measurements.
The selected cell on one fixed duty — one cycle a day, 80 % DOD, 0.5 C charge and discharge, a daily peak-shaving profile — held at three cell temperatures. End of life at 80 % is where warranties and augmentation budgets live. For a cell that works every day most of the wear comes from the work itself, so the curves sit closer together than the old “ten degrees, half the life” rule would have them. That rule describes storage.
What the numbers above are worth, in the two places that follow directly from your inputs.
The red wave is a thermal runaway propagating from cell to cell. That is abuse and safety physics — different mechanisms, different data. This estimator covers ageing only and says nothing about runaway.