LFP Standby Aging: Why Temperature and SOC Need the Full Duty Cycle

An LFP battery held in reserve still has an operating history. Time spent charged, the temperature during that time, and the recharge after an event can contribute through different aging mechanisms. A low cycle count does not describe that history, and a temperature that looks favorable for storage need not minimize degradation during every charging duty.

A study by Michael Schimpe and colleagues at the Technical University of Munich and the National Renewable Energy Laboratory shows why these distinctions matter. The researchers combined storage experiments, controlled cycling and a separate photovoltaic-duty validation for one LFP/graphite cell design. Their model treated calendar aging and several cycling effects separately. The revealing result was that the lowest capacity loss occurred at different tested temperatures when the duty changed. Understanding that result is more useful than borrowing any of those temperatures as a cabinet setting. [1]

1. A standby requirement is more than a percentage on the SOC display

State of charge, or SOC, expresses how full the battery is relative to its capacity reference. For a reserve application, the operating requirement is the energy and power that must remain available when an event occurs. SOC is one input to that requirement. Available capacity, the permitted discharge window, conversion and auxiliary losses, temperature limits and the required output also affect what the system can deliver.

Calendar aging describes degradation associated with elapsed time under the battery’s conditions, rather than only with charge throughput. A battery can accumulate calendar exposure while awaiting commissioning, between dispatches or while maintaining a reserve. These situations are not identical: an energized installation may include balancing, auxiliary demand and replenishment charging. A laboratory cell stored at a specified SOC is therefore evidence about a component of standby behavior, rather than a complete test of an online reserve system.

The engineering problem is to connect the dwell periods with the events around them. An unplanned backup requirement may need readiness throughout the day. A scheduled peak-shaving service may have a known delivery time. The aging question is which conditions accompany those different obligations, including the current and temperature when the battery is replenished. Reducing reserve without checking the required service solves a different problem from the one the owner specified.

For the broader distinction between cycle count, energy throughput and years of service, the battery-life research guide provides the starting framework. Here the focus is narrower: what a model must retain when storage and recharge create different stresses within the same operating duty.

2. What the LFP experiment actually separated

Schimpe et al. tested a commercial 26650-format LFP/graphite cell using a 3 Ah beginning-of-life capacity reference. The study included storage conditions, constant-current cycling and constant-current/constant-voltage charging protocols. Storage SOC was varied in 12.5-percentage-point steps from 0% to 100%, with temperature coverage shown in Table II; the full matrix was not populated at every combination. Most parameterization points used a single cell, while selected cycling and validation points had replicates. (pp. A181–A182, Tables I–II). [1]

Periodic reference performance tests at 25 °C provided a common capacity-measurement basis. The researchers measured capacity using controlled charge–discharge procedures, instead of interpreting a cold cell’s immediately reduced discharge output as irreversible capacity loss. That distinction lets a storage or cycling condition be compared through a standardized later check. The tests themselves also interrupt storage; this was not a study of an untouched cabinet left for years. (p. A183). [1]

During storage at 100% SOC, higher temperature increased the rate of measured capacity loss across the tested conditions. At 45 °C, storage SOC also affected fade. The observed loss rate slowed with elapsed time, motivating the square-root time relation used for the calendar component. The measurements and fitted trend lines have different roles: the points are performance-test observations, while the lines describe the selected model form. (Figure 2, p. A183). [1]

For cycling tests, the measured loss contained both calendar and cycling contributions. The authors estimated the calendar contribution and subtracted it to obtain what they called “pure cycle” loss. This residual is a model-assisted separation, not a second direct measurement of an isolated physical mechanism. The distinction matters when comparing the temperature minima in the next section. (Equation 10, pp. A185–A186). [1]

3. Three temperature minima—and three different questions

Under the study’s 1C constant-current cycling tests, measured total capacity loss was lowest at an ambient test temperature of 25 °C. Both colder and hotter test conditions produced more loss. After the modeled calendar component was removed, the derived pure-cycle loss instead had its minimum at 35 °C. The calendar contribution became increasingly important at higher temperatures, so removing it changed the comparison. (Figure 5, pp. A185–A186). [1]

The application-oriented validation told a different story again. Under a lower-current, variable photovoltaic profile, the lowest measured loss among the tested 10, 25, 45 and 55 °C ambient conditions occurred at 10 °C. The authors explained this result through calendar aging dominating a duty with smaller cycling stresses. The PV validation did not include a 0 °C test, so the result does not identify a global minimum below the tested range. (Figure 15 and discussion, pp. A191–A192). [1]

The three results are compatible. Storage and charging do not have to respond to temperature in the same way, and total degradation reflects their relative contributions. A reserve-heavy service, a repetitive full-cycle test and a faster replenishment event can therefore expose different tradeoffs even when the chemistry label is unchanged. The useful question is which contribution dominates the proposed duty, rather than which temperature won an unrelated comparison.

This is where liquid cooling becomes relevant to a HUA-X PowerFac-261 project. Its LFP chemistry and liquid-cooled architecture establish a concrete system context; its 125 kW rated AC power and 261 kWh nominal energy identify the equipment scale. They do not specify the cell-temperature history during reserve and recharge. The thermal conditions during reserve and recharge determine how the cell-aging findings relate to the duty; liquid cooling alone does not quantify a lifetime gain. [2]

4. Recharge introduces a stress that a storage curve cannot explain

At 10 °C, the study’s calendar-subtracted cycling loss increased as the tested current rate rose from 0.25C to 0.5C and 1C. At 55 °C, the corresponding pure-cycle results showed much weaker dependence on those current rates. A single multiplier applied to all throughput would miss this difference: the influence of current was conditional on the temperature regime. (Figure 6, p. A186). [1]

The high-SOC part of charging mattered as well. At 0 °C, adding the constant-voltage charging phase produced substantially greater derived cycling loss than the constant-current protocol, particularly at the higher current. At 55 °C, the study did not show the same clear separation. The model consequently included an additional low-temperature, high-SOC charging term. Its fitted SOC threshold was a model parameter for these cells, not a recommended charge ceiling for other batteries. (Figure 7 and Equations 20–21, pp. A186–A188). [1]

The authors interpreted the low-temperature behavior as consistent with transport limitations and possibly lithium plating. They explicitly stated that no degradation analysis had confirmed or ruled out that mechanism. Their high-temperature explanation included additional SEI growth associated with cycling, also presented as an assumed mechanism. The experiments establish conditional patterns in capacity loss; identifying every internal reaction requires further diagnostic evidence. (pp. A186–A187). [1]

For standby storage, this changes how replenishment should be evaluated. The interval with little external current and the subsequent charge are different parts of the duty. An assessment that favors a colder dwell but ignores the temperature and current at recharge may leave out the condition that changes the cycling contribution. This is an engineering implication of the study, not a prescription to warm every battery or bypass the equipment’s approved charge limits.

PowerFac-522 offers an LFP, liquid-cooled configuration with 522 kWh nominal energy, 125/250 kW power options and stated 0.25C/0.5C charge–discharge options. Those options help frame a configuration-specific discussion of the required replenishment duty. They are not interchangeable with the small-cell experiment’s degradation coefficients or charge protocol. Request the applicable PowerFac-522 documents with the reserve requirement and intended recharge sequence. [3]

5. Why elapsed time and averages need careful treatment

For constant storage conditions, the calendar model takes the form Dcal(t) = kcal(T, SOC) × √t. Here Dcal is capacity loss normalized to the initial capacity, t is elapsed time, and kcal is a condition-dependent stress coefficient with units consistent with the chosen time unit. The square-root form represents a slowing loss rate; it is not a fixed percentage deducted every year. In the paper, the coefficient uses hours, so changing the time unit without changing the coefficient would change the calculation incorrectly. (Equation 2 and Table IV, pp. A183–A184 and A190). [1]

The authors modeled temperature dependence with an Arrhenius relation and SOC dependence through graphite-anode potential. They assumed separability when combining the temperature and SOC factors, and included a fitted offset in the potential-based expression. These are useful modeling choices with empirical support in the test matrix, not a claim that every LFP design has the same coefficients or that all aging mechanisms are fully resolved. The authors themselves identified the offset as requiring further investigation. (Equations 3–9, pp. A184–A185). [1]

Averages can hide which conditions coincide. A high SOC during a cooler interval is not the same exposure pairing as a high SOC during a hotter interval, even if the separate daily averages match. Similarly, an average current says little about whether the largest charge current occurred in the low-temperature, high-SOC region. A time-resolved model can retain those combinations and calculate the relevant contribution during each interval.

That does not mean averages are always unusable. To separate calendar loss from their controlled cycling tests, the researchers used average measured temperature and an approximate 50% SOC. They checked a more detailed treatment and found its influence on those particular parameterization results negligible, while it increased fitting effort. (Equation 10 discussion, p. A185). [1]

For variable operation, the paper used a rate-based integral formulation that updates the contributions over time and charge throughput. Its history is not replaced by starting a fresh storage experiment every time the operating condition changes. A dispatch model should identify that accumulation rule as well as its stress functions. The existence of a time-resolved formulation does not, by itself, prove that every possible reordering of the same exposures will produce a different physical outcome. (Equations 23–24, p. A188). [1]

6. What the validation demonstrated—and what “less than 1%” meant

The validation profile came from a residential PV-storage application and was excluded from model fitting. It had a 15.5-hour duration, 60-second resolution and an SOC range of 5.4%–80%, using the study’s initial 3 Ah reference. Average charge and discharge rates were approximately 0.24C and 0.17C in magnitude, with maximum rates of 0.75C. These conditions help explain why it did not behave like the repetitive 1C full-cycle test. (Figure 1 and Table III, p. A182). [1]

Rest periods were removed to accelerate the profile, which combined representative winter and summer usage. The longest validation lasted 348 elapsed test days, extending 114 days beyond the longest parameterization tests. Repeated compressed profiles provide useful out-of-fitting evidence, but 348 laboratory days do not become three calendar years of field standby simply because the schedule represents more operating days. Restoring the omitted idle intervals would also restore their calendar exposure. (pp. A182 and A192). [1]

At the end of the 348-day dynamic-profile tests from 10 °C to 45 °C, error was below 1% of the original cell capacity. The largest reported relative error in capacity loss was about +21% at 25 °C, using the measured capacity loss as the denominator. These are different metrics: a small error relative to the initial capacity can still be a material fraction of the smaller amount already lost. The result is not a statement of “99% lifetime-prediction accuracy.” (Figure 15 and discussion, pp. A191–A192). [1]

The 55 °C validation had only one cell and about 200 days of data; it was not part of that 348-day endpoint comparison. More broadly, the model was developed for capacities above 80% of the initial value. It did not include a later acceleration or “knee” mechanism and could not predict that transition. Dynamic ambient temperatures and non-repeating profiles were identified as further validation needs. These limits matter when translating a fitted laboratory model into years of project service. (pp. A183 and A192). [1]

7. Reserve duration and recharge change the aging balance

An always-ready reserve carries an energy and power obligation throughout its standby period. Scheduled dispatch ties that obligation to particular delivery times. A long charged dwell and a faster recharge after an event involve different exposures, even when the required delivery energy is unchanged. A lower aging estimate obtained by reducing the reserve requirement describes a different service, not a free improvement to the original one.

For PowerFac-261, the relevant quantities are the temperature experienced by the cells, the permitted SOC window and the charging limits during standby and replenishment. The study’s temperature minima describe its tested cell and profiles; they do not specify those cabinet settings. [1] [2] Configuration information is available in the model technical brief and the PowerFac-261 technical documents.

For a larger battery installation, the PowerFac-5M DC-side battery container brings the same question into a different system boundary. Its 5.016 MWh nominal energy and DC-side, liquid-cooled battery configuration describe the battery-side equipment. The proposed PCS, control scope and auxiliary supply still affect how reserve is delivered and replenished. Request PowerFac-5M project documents for that supply boundary; cell-aging coefficients cannot define a complete AC reserve service. [4]

A reserve held at high SOC during warm standby and one replenished at higher current in cold conditions involve different combinations of temperature, SOC and current. Separate averages cannot show which conditions occurred together. In the study, storage and recharge contributed differently to capacity loss; those findings explain why the combined operating conditions matter, without establishing the degradation mechanism of a particular cabinet. [1]

8. The useful conclusion is a duty-dependent model, not one ideal SOC

The important contribution of the LFP study is the separation of stresses that a single cycle counter or temperature average can combine. Calendar exposure continues between services; charging can introduce a different temperature sensitivity; a fitted calendar subtraction is not an independently measured mechanism. The comparison becomes credible when its model and its validation preserve those distinctions.

For a reserve project, the result is a better-defined engineering question: how can the proposed system meet its readiness obligation across dwell, dispatch and recharge, within its approved operating envelope? Answering that question requires the actual duty and configuration. It does not require treating a laboratory optimum as a universal setting, or confusing a small fitted capacity error with a demonstrated project lifetime.