BESS Battery Life: Beyond Cycle Count and Capacity Curves

A battery can complete more cycles yet deliver less lifetime energy than a larger cell with a lower cycle count. It can also lose capacity while sitting idle, or retain stored energy that it cannot deliver at the required power. These are different aspects of battery life. A capacity curve describes one measured response under particular conditions, rather than a complete prediction of an industrial storage system's service.

Two experimental studies make the distinction concrete. Preger and colleagues varied cycling conditions and compared normalized cycles with discharged energy. Keil and colleagues examined aging during storage and linked its SOC dependence to graphite-electrode behavior. Their findings explain why energy throughput, temperature, reserve dwell and conversion losses all matter when moving from a cell to a liquid-cooled cabinet such as HUA-X PowerFac-261.

1. What a battery-life curve actually measures

Capacity retention is the capacity measured at a later check divided by a stated reference capacity, multiplied by 100. A different test temperature, discharge rate or cutoff can change the measured result even without an equivalent irreversible loss. Comparing measurements made under different conditions can therefore mix the battery's aging with the conditions under which its remaining charge was accessible.

The horizontal axis also changes the question. Elapsed time describes how long the battery has existed under the reported conditions. Counted cycles describe repeated test events. Equivalent full cycles normalize throughput against a capacity reference, but the counting convention still needs to be stated. A shallow charge-discharge event and a full excursion cannot be compared simply by counting each as one cycle.

The familiar 80% capacity threshold is useful as a shared comparison point when a study specifies it. A project, however, may lose its ability to complete an assigned duty before that point, or continue serving a less demanding duty afterward. Retirement, safety, warranty and analytical comparison therefore require their own definitions; one threshold cannot silently stand in for all four.

2. A measured lifetime endpoint differs from an extrapolated one

Preger et al. (2020) tested 18650 LFP, NMC and NCA cells with varied temperature, SOC and discharge rate, 0.5C charging and at least two replicates per condition. The matrix was not fully factorial (Tables I–II). [1]

The experiment's varied conditions explain why no single bar can represent an entire chemistry. Temperature comparisons used matching SOC windows and rates; other comparisons changed the discharge window or rate. A result reflects the tested combination. Replicates reveal variation between cells under that combination, although the small groups cannot describe the lifetime distribution of an industrial fleet.

Preger’s Figure 1 separates capacity-check circles from visual guide lines. Figure 2(a) distinguishes measured 80%-capacity crossings (+) from linear extrapolations; most LFP samples had not reached 80%. [1]

In the original study's Figures 1–2, a measured crossing means the cell passed the 80% benchmark during observation. An extrapolated crossing extends the current fade rate beyond the observed range. That is useful for comparing incomplete tests, but its endpoint remains conditional on the future curve continuing in the assumed way.

The uncertainty grows when a small early slope is extended far beyond the test duration. A nearly flat line can imply a very long projected life, yet the measured portion does not establish whether later degradation will accelerate. Preger's discussion distinguishes initial loss, an approximately linear middle region and a later rapid-fade region found in prior aging models. Linear continuation describes one possible future path, not proof that the middle region persists indefinitely. [1], general analysis, p. 4

3. Why more cycles need not mean proportionally more delivered energy

Preger’s LFP advantage narrowed when comparing cumulative energy instead of EFC, reflecting unequal cell capacities and voltages (Figure 2). [1]

An EFC is normalized to a reference capacity. It is valuable for following a cell through an aging experiment, yet it does not put unlike cells on an equal energy scale. Comparing only cycle count resembles comparing trips without asking how much was carried on each trip. Energy throughput adds that missing scale, while still requiring the same endpoint and a clearly stated measurement location.

For a project, lifetime discharged energy is the sum of the energy delivered over all discharge intervals. Where power varies, each interval contributes measured discharge power multiplied by its duration. Summing those intervals preserves changing output and operating windows. Multiplying an initial nameplate energy by a headline cycle count instead assumes a repeated delivery quantity that the calculation has not established.

The energy metric also has a limit: equal cumulative kWh need not produce equal project value. Delivery during a constrained peak, absorption of otherwise curtailed solar generation and energy held in reserve serve different objectives. The useful engineering comparison therefore links the aging metric to the assigned service and the interval schedule. It does not select a battery solely because one laboratory axis produces the largest number.

4. Temperature, SOC and discharge rate interact

At 0–100% SOC, 0.5C charging and 1C discharging, raising temperature from 15 to 35 °C accelerated LFP fade, slowed NMC fade and weakly affected NCA. Narrower SOC windows at 25 °C benefited NMC/NCA more than LFP (Figure 3(a–f)). [1]

The result is an interaction between cell design and duty. The cycling comparison cannot, by itself, prescribe the temperature of an idle cabinet: cycling and storage expose cells to different mechanisms. Nor does the chemistry label identify all details of the electrode and electrolyte design. A temperature trend observed in one sample family is not a universal operating optimum.

An SOC window has both a width and a position. Preger's cycling comparison varied the width around a common midpoint; it did not isolate the effect of moving an unchanged window upward or downward. Keil's storage experiment, discussed below, addresses why time spent at different SOCs can matter even without large charge-discharge excursions. [1], depth-of-discharge discussion, p. 6; [2]

NCA faded more slowly per EFC at higher discharge rates but returned lower efficiency; the cause remained unresolved (Figures 2(c), 3(i)). [1]

That counterexample shows why a single 'best' setting is difficult to defend. Improving capacity retention per EFC can coincide with worse energy efficiency. Faster discharge also changes the clock time needed to accumulate a given EFC count, so per-cycle and per-year comparisons need not tell the same story. The unresolved NCA result does not establish that higher discharge power extends battery life. [1], Figures 2(c), 3(i) and discharge-rate discussion

Liquid cooling influences the cell's thermal environment but does not remove the energy tradeoff. Cell temperature, outdoor temperature and the electricity used by cooling describe different quantities. A cooling architecture can manage conditions; its name alone gives neither the temperature history inside the battery nor a quantified gain in lifetime.

5. Calendar aging explains why the idle hours matter

Keil et al. (2016) stored graphite-anode 18650 cells with LFP, NMC and NCA cathodes at 16 SOCs and 25, 40 and 50 °C. After approximately 9–10 months, hotter storage increased fade; SOC dependence showed plateaus (Figure 2). [2]

Electrode measurements and differential-voltage analysis, which tracks voltage changes as charge is added, linked fade mainly to cyclable-lithium loss and shifted electrode balancing. The authors attributed the main effect to lower graphite potential favoring electrolyte reduction and growth of the solid electrolyte interphase (SEI), a passivating layer. Capacity fade and resistance growth were not directly correlated (Figures 2(d–f), 3–6). [2]

Cyclable-lithium loss removes lithium available to shuttle between electrodes, without requiring wholesale electrode destruction. Graphite’s anode potential stays near plateaus across broad SOC intervals, helping explain similar aging within those intervals. Full-cell SOC and anode potential are not proportional. [2]

Resistance affects how the cell's voltage responds under current. Since those changes are not proportional, the same capacity retention can coexist with different power capability or energy loss. Keil's separate capacity and resistance results are a direct example of why a single health percentage cannot describe every performance change. [2], Figure 2

Keil’s SOC transitions differed among cell designs. These results do not establish numerical SOC settings for modern large-format LFP cabinets. Accelerated storage over months also cannot directly establish years of field service (Figures 5, 8 and conclusions). [2]

For industrial storage, the idle period is part of the duty. Energy reserved for an unpredictable outage may need to remain available for a long time; energy intended for a known peak may not. Holding the latter target much earlier than necessary adds dwell time, but lowering an outage reserve can remove the readiness the system is meant to provide. Aging exposure and service availability therefore interact, even when daily cycle counts are identical.

The SAM battery-life documentation represents calendar and cycle degradation as distinct components, with combination rules that depend on the selected model. This is why dividing a cycle count by assumed cycles per day cannot determine years of service: the result omits aging during the hours between cycles and assumes the modeled duty repeats unchanged. [3]

6. Cell aging and a liquid-cooled cabinet operate at different scales

The HUA-X PowerFac-261 uses 3.2 V / 314 Ah LFP cells, liquid cooling, 125 kW rated AC power and 261 kWh nominal battery energy. Its recommended depth of discharge is 90%; output derates above 45 °C. [4]

A cabinet contains more than a scaled-up laboratory cell. Its cell design, thermal environment, control limits, converters and auxiliaries all affect how stored charge becomes usable AC energy. The 18650 samples in the studies and the 314 Ah cells in PowerFac-261 differ in physical implementation. Sharing an LFP label does not make their measured life curves interchangeable.

PowerFac-261's 125 kW rating describes AC power, whereas 261 kWh describes nominal battery energy. Their ratio is a nameplate energy-to-power ratio, not a guaranteed discharge duration. Available AC energy also depends on the SOC interval, remaining capacity and losses; the duration of service additionally depends on the load. The 90% recommended discharge depth describes an operating window and is not itself a measure of delivered AC energy. [4]

PowerFac-261's output derating above 45 °C makes the distinction between capacity and power practical. A hot-site operating period can restrict output even while energy remains stored. Cooling demand also consumes electricity that does not reach the external load. A change in delivered energy during that period can therefore reflect temperature, power limits and auxiliaries as well as electrochemical aging. [4]

The PowerFac-261 technical-document request is the route to model-specific usable-energy conditions and derating information for a proposed duty. Those quantities determine how the cabinet can serve a particular load; the small-cell aging studies explain mechanisms rather than supplying its service-life guarantee.

7. Less delivered energy does not always mean the same amount of aging

A capacity test and an operating day's electricity output measure different things. A capacity test uses defined endpoints and conditions to characterize the battery. Daily output also depends on when discharge begins, where it stops, the requested power and the energy consumed inside the system. A narrower SOC window can reduce delivered kWh without representing the same fractional loss in the cell's full capacity.

Efficiency can change while capacity changes. Preger calculated cell round-trip efficiency from discharged energy divided by charging energy and found it depended on rate, temperature, SOC and rest time; efficiency also declined as the tested cells aged. That cell-level ratio excludes a cabinet's converter and cooling loads. Equal remaining ampere-hours therefore need not imply equal energy recovered at an AC meter. [1], Figure 2(c) and round-trip-efficiency discussion, pp. 4–5

Power limits create a further separation. A system that reaches a voltage, current or temperature limit may be unable to sustain the requested load even though unused energy remains. Conversely, a lower-power duty may still use that energy over a longer interval. The relevant lifetime is consequently tied to the service: retaining enough capacity is necessary for an energy-intensive task, while a high-power task can become constrained for another reason.

The distinction also explains the limits of an 80% threshold. It is a useful laboratory benchmark, but not a universal point at which every storage application stops working. A duty with little spare energy may become infeasible before that benchmark; a less demanding duty may remain feasible afterward. Research endpoints, usable power and the service required by the site answer separate questions.

8. What changes the meaning of a battery-life comparison

The research reveals several reasons why two apparently similar lifetime claims can describe different outcomes. These differences remain important even when both claims use the same cycle-count or capacity-retention label.

Battery life is best understood as performance under a duty, rather than a cycle number detached from its conditions. Preger's results connect throughput and operating conditions; Keil's results add the significance of idle time and electrode behavior. At cabinet scale, those insights meet the required load, reserve, thermal conditions and conversion losses. Together they explain why useful energy and service capability can evolve differently from a single capacity curve.