How battery health analysis detects capacity degradation

Battery capacity degradation can undermine energy storage performance long before a system reaches the end of its service life. Battery health analysis helps technical evaluators identify that decline by comparing available capacity with a baseline and examining operating data such as charge–discharge behavior, temperature, and internal resistance. The difficulty is that a battery can appear to have lost capacity when the real change is in the test conditions, system controls, or measurement boundary.

A defensible assessment therefore starts with a narrower question than “How healthy is this battery?” It asks what capacity is available, at which boundary, under which conditions, and how that result compares with a suitably defined reference. For a utility-scale energy storage system, the answer may be different at the cell, rack, and AC interconnection level—and each answer can be useful for a different decision.

What capacity degradation actually measures

Capacity is commonly expressed in ampere-hours at the cell or DC battery level. Available energy, usually expressed in kilowatt-hours or megawatt-hours, also depends on voltage during discharge. As a battery ages, a lower discharge voltage can reduce delivered energy even when the change in ampere-hour capacity is less pronounced. That distinction matters when a project’s obligation is defined in MWh rather than Ah.

Capacity-based state of health, or SOH, is often expressed as measured capacity divided by a reference capacity, multiplied by 100. The formula is straightforward; the reference is not always so obvious. It might be the manufacturer’s rated value, an acceptance-test result, or a measured value from commissioning. Those choices can produce different percentages for the same battery.

For example, comparing a field result against nameplate capacity may be appropriate for one contractual question, while comparing it against a commissioning test may be more useful for tracking change in a particular installation. Neither comparison is meaningful unless the test conditions and measurement uncertainty are understood.

State of charge, or SOC, should not be confused with SOH. SOC estimates how much charge remains at a given time; SOH describes condition relative to a reference. An inaccurate SOC estimate can make a battery look degraded by causing a test to start or stop at the wrong point. It can also conceal degradation until a more complete charge–discharge cycle exposes the error.

The baseline is part of the measurement

A capacity test is most persuasive when it can be compared with an earlier test performed under equivalent conditions. Evaluators should recover the original procedure, not just the headline result. Temperature, charge and discharge rates, rest periods, voltage limits, SOC limits, and the treatment of auxiliary loads can all affect the comparison.

Consider a battery commissioned under a relatively gentle discharge profile and later assessed during a high-power dispatch. A lower energy result in the second situation may reflect increased losses and earlier voltage cutoff, not the full extent of irreversible capacity loss. Similarly, a test conducted when part of the system is unavailable cannot be interpreted as a whole-system degradation result without accounting for the missing equipment.

At minimum, the test record should identify the equipment included, the measurement point, the operating limits, the starting condition, ambient and battery temperatures, and any alarms or derating. Meter calibration, sampling interval, and the method used to integrate current or power also belong in the record. If those details are missing from the baseline, the later result may still be useful, but the uncertainty should be visible in the conclusion.

This is particularly important for AC-side tests. Energy delivered at the point of interconnection includes effects from the power conversion system, transformers, cabling, and possibly auxiliary consumption, depending on the metering arrangement. An AC energy shortfall is a system-level observation; it is not automatically proof that the cells alone have degraded by the same percentage.

How operating data reveals the cause of a shortfall

A controlled capacity test provides an anchor. Continuous operating data explains what may be happening between tests and helps identify which parts of the system warrant closer inspection. The strongest interpretation comes from indicators that agree with one another, rather than a single dashboard value.

Charge and discharge behavior

When similar dispatches deliver progressively less energy, the trend may indicate reduced usable capacity. But a comparison must account for changes in requested power, SOC window, temperature, and system availability. A plant that spends more time providing short, high-power services may accumulate a very different operating history from one used for daily solar shifting, even if their calendar ages match.

Voltage behavior is often more informative than the total energy figure alone. If one cell group approaches its lower voltage limit well before the others during discharge, the battery management system may stop or limit the string to protect it. The remaining cells may still contain charge, but that energy is not necessarily usable within the configured protection limits. Cell imbalance, a weak cell group, measurement error, or increased resistance could all contribute; the voltage trace points investigators toward the problem but does not settle the diagnosis by itself.

Likewise, an early upper-voltage limit during charging can restrict the amount of charge accepted. Before attributing that behavior to aging, check balancing activity, sensor quality, charging limits, and whether the battery had enough time to complete its normal charge procedure.

Temperature and resistance

Temperature affects both immediate performance and long-term aging. A cold battery may temporarily deliver less usable energy or power because voltage limits are reached earlier under load. Persistent exposure to unfavorable temperatures can also contribute to permanent degradation, with the magnitude depending on chemistry, SOC, and operating history. Those two effects should not be merged into one capacity-loss number.

Temperature spread within a container or rack is worth examining as well as the average. A single average can hide a cooling imbalance, a sensor issue, or a localized electrical problem. Compare sensor positions and readings under similar operating conditions before deciding that one section is aging faster.

An increase in internal resistance can cause a larger voltage drop during discharge and additional heating. It may reduce usable power and make a battery reach its cutoff voltage sooner, particularly at higher currents. Resistance trends are therefore useful alongside capacity trends, but the values depend on how they were measured: pulse duration, SOC, temperature, and the equipment boundary all matter. A resistance value from a cell datasheet should not be compared casually with an estimate derived from an operating rack.

Why the battery management system estimate is not the final word

A battery management system can estimate SOH using measurements and algorithms that are practical for routine operation. That makes it valuable for screening fleets and detecting changes between formal tests. It does not mean that every reported percentage is a direct measurement of capacity.

The estimate may depend on current integration, voltage behavior, temperature, and assumptions about cell characteristics. Its performance can deteriorate when the operating profile provides limited opportunities for recalibration or when sensor errors accumulate. Firmware changes may also alter the reported value without an equivalent physical change in the battery.

When a reported SOH trend changes abruptly, review software versions, configuration changes, replacements, and data gaps before concluding that the battery has suddenly aged. Conversely, a stable SOH display should not override a repeatable decline in measured energy. The practical approach is to use the estimate as a lead, then compare it with test results and raw operational evidence.

Choosing a test without creating unnecessary risk

A full controlled charge–discharge test can provide a strong capacity measurement, but it consumes operating time and may interrupt contracted service. It also needs appropriate procedures, equipment, permissions, and manufacturer limits. For an operating microgrid, taking storage out of service may reduce resilience during the test window; for a grid-connected plant, the cost may be missed dispatch or availability exposure.

Partial-cycle analysis or a comparison of repeated operating events can be less disruptive. These methods are useful for trend detection, especially when dispatch conditions recur, but they usually require more assumptions about SOC and behavior outside the observed range. They should not be presented as equivalent to a full reference test without demonstrating that equivalence for the system and purpose.

The test design should follow the decision. A maintenance team investigating one abnormal rack may need cell-group voltage and temperature data more urgently than a site-wide energy test. A contractual performance assessment, by contrast, may require the exact boundary, instrumentation, operating profile, and correction method specified in project documents. Safety limits remain in force in either case; a capacity test is not a reason to bypass battery protection settings.

Standards and contract language: related, but not interchangeable

Standards can provide test frameworks and terminology, but the applicable document depends on the battery technology, equipment level, and purpose of the assessment. For example, IEC 62620 addresses performance requirements and tests for certain secondary lithium cells and batteries used in industrial applications. IEC 62933-2-1 addresses unit parameters and testing methods for electrical energy storage systems. Their scope and current editions should be checked before applying either to a specific project.

Safety standards answer a different question from capacity retention. UL 9540 concerns energy storage systems and equipment, while UL 9540A is a test method used to evaluate thermal runaway fire propagation. Neither should be treated as a substitute for a contractual capacity test or as a general guarantee of future battery life. Applicable requirements also depend on the jurisdiction, system configuration, and project documents.

Warranty terms deserve the same scrutiny as the technical standard. A warranty may define capacity at a particular measurement boundary, specify operating conditions, exclude certain events, or set out a prescribed verification process. A technically credible field test can still fail to resolve a warranty question if it does not follow the agreed procedure.

Turning a degradation finding into a decision

The useful output of health analysis is not merely a percentage. It is a finding that separates observed performance from interpretation: what was measured, how comparable it is to the baseline, which limitations remain, and whether the evidence points to cells, controls, thermal management, conversion equipment, or some combination.

A capacity shortfall may justify further diagnostics, balancing, repair of a cooling issue, or a review of operating limits before replacement is considered. Where deterioration is confirmed, the next question is whether the remaining capacity and power meet the actual duty, reserve requirement, and contractual obligations. A battery can be below its original capacity and still be fit for one service while no longer suitable for another.

For developers, EPC contractors, and microgrid operators comparing assessments across projects, G-EPI’s emphasis on verifiable engineering data is a useful discipline: keep the measurement boundary, test method, standard edition, and uncertainty attached to every result. If those details cannot be recovered, qualify the comparison rather than forcing a precise degradation figure from incomplete evidence.