Deep diveEnergy and Materials
Second-life battery assessment: grading retired EV packs for reuse or recycling
A second-life battery assessment decides whether a retired EV pack, or its modules, can safely store energy in a less demanding job, or should be recycled. It rests on more than one health number: remaining capacity, resistance growth, module spread and usage history. This deep dive defines the metrics, compares fast test methods, shows how BMS data and machine learning triage packs, and outlines the safety and EU rules that apply.
On this page
- Four routes for a retired pack, and why grading decides between them
- Health metrics and test terms used when grading used EV batteries
- Five ways to measure a used pack, from data read to full cycle
- A grading funnel that spends test time where it changes the route
- Using BMS data and machine learning to triage before testing
- Why module heterogeneity, not average health, sets second-life value
- Safety evaluation and the EU rules for repurposed batteries
- Repurpose or recycle: routing rules by what the grading shows
- Questions and answers
- Sources
Four routes for a retired pack, and why grading decides between them
A retired electric vehicle battery still holds most of its useful chemistry but no longer meets the vehicle's range or power needs. It can follow four routes. Reuse puts it back into the same kind of vehicle. Remanufacturing restores it for its original purpose; the EU Battery Regulation defines this as evaluating every cell and module and restoring capacity to at least 90 % of the original rating, with cell state of health differing by no more than 3 %1. Repurposing moves it to a different application, usually stationary storage. Recycling recovers its materials.
Grading is the evidence behind that choice. A pack sent to a storage project that later fails costs more than the recycling value it displaced, and a healthy pack sent to the shredder wastes the reason it was collected. The aim is enough confidence, per pack and per module, to route it and stand behind the route.
Health metrics and test terms used when grading used EV batteries
Graders, buyers and certifiers use these terms loosely. Agreeing definitions in the purchase contract avoids disputes later.
- Capacity state of health (SoH-C)
- Usable charge or energy measured under a defined test, divided by the rated value when new. It answers how much energy the battery can still hold, not how well it delivers power.
- Resistance state of health (SoH-R)
- Internal resistance compared with its value when new. Resistance growth reduces power, raises heat under load and often signals ageing mechanisms that capacity alone does not show.
- Module spread
- The difference in capacity or resistance between the best and worst modules or cells in a pack. In a series string the weakest unit limits the whole, so spread can matter more than the average.
- Electrochemical impedance spectroscopy (EIS)
- Applying a small alternating signal across a range of frequencies and measuring the response. It separates ohmic, charge-transfer and diffusion effects in minutes, but results only compare at fixed temperature and state of charge.
- Pulse test
- Short charge and discharge current pulses at several states of charge, from which resistance and power capability are calculated. Hybrid pulse power characterization is a common structured form.
Five ways to measure a used pack, from data read to full cycle
Each method buys information with time and equipment. A practical program layers them rather than choosing one.
| Method | What it reveals | Time and equipment | What it misses |
|---|---|---|---|
| BMS history read | Logged capacity estimates, cycle and fast-charge counts, temperature exposure and fault codes | Minutes per pack, given read access and a decoder for the BMS data format | Anything the BMS never measured well; estimates may drift from true capacity |
| Full reference cycle | Directly measured capacity and energy under controlled conditions | Hours per unit plus cycler channels sized for pack or module voltage | Little on capacity, but it ties up equipment and says less about ageing mechanisms |
| Partial cycle with curve analysis | Capacity estimate plus clues to the dominant ageing mode, read from how charge changes with voltage | Shorter than a full cycle; needs a good model of the chemistry | Accuracy falls if the chemistry or cell format is unfamiliar |
| EIS | Resistance components and early signs of degradation mechanisms | Minutes per module with an impedance analyzer and fixed temperature | Capacity is only inferred, so it relies on a calibrated model |
| Pulse test | Resistance and power capability at chosen states of charge | Minutes per module using standard cycler hardware | Little about capacity; results depend on pulse length and temperature |
A qualitative comparison: set test durations from your own validation data.
A grading funnel that spends test time where it changes the route
- Intake and identity
Record identifier, chemistry, origin and any passport or service records.
- Safety screen
Visual, thermal and isolation checks; anything damaged leaves the reuse path immediately.
- BMS history triage
A model ranks packs from logged data and flags missing or suspicious history.
- Targeted electrical tests
Impedance, pulse or partial-cycle tests on the packs whose route is still uncertain.
- Grade and route
Each pack or module gets a grade, a confidence level and a destination.
- Grading record
Results, methods and test conditions are kept for buyers, certifiers and regulators.
Using BMS data and machine learning to triage before testing
The battery management system has watched the pack for its whole first life. Cycle counts, time at high state of charge, fast-charging frequency, temperature extremes and logged faults all predict how it will age next. In the EU, Article 14 of the Battery Regulation requires EV battery management systems to hold data for determining state of health and expected lifetime, with read-only access for the buyer and for operators preparing the battery for repurposing or remanufacturing1. Elsewhere, access depends on the vehicle maker.
A triage model trained on packs that were later fully tested can predict capacity, resistance and likely spread from that history, with an uncertainty band. Confident predictions far from a routing threshold need only a light confirmation test; packs near a threshold, or with gaps in their history, go to detailed testing. Validate the model per chemistry and vehicle platform, because manufacturers' BMS estimates differ. ColdAI's energy and materials practice lists state-of-health monitoring, degradation prediction and second-life assessment among its use cases3; the same predictive monitoring methods apply once a repurposed system is in service.
Why module heterogeneity, not average health, sets second-life value
Packs age unevenly. Modules near a cooling inlet or hot spot drift apart from the rest over years of use, and in a series string the weakest module reaches its voltage limits first. A pack with a good average and one weak module behaves like a weak pack. Grading at module level and regrouping similar modules can recover more value, but it adds disassembly labor, new busbars and enclosures, a BMS that understands the new layout and a fresh safety evaluation. The right choice depends on how much spread varies within a batch and on the integrator's ability to re-certify the result.
Safety evaluation and the EU rules for repurposed batteries
Repurpose or recycle: routing rules by what the grading shows
- If
Capacity and resistance are healthy, spread is narrow and the history is complete.
ThenRepurpose the whole pack for stationary storage with a light confirmation test.
Disassembly adds cost and risk without improving a uniform pack.
- If
Average health is good but one or more modules are clearly weaker.
ThenGrade at module level and regroup matched modules, or sell the pack as a module source.
The weak module would set the performance of the whole string.
- If
Resistance has grown sharply even though capacity looks acceptable.
ThenLimit it to low-power duty, such as long, slow discharge, or route it to recycling.
High resistance means heat under load and often signals accelerating ageing.
- If
BMS history is missing, inconsistent or shows faults.
ThenTest in full before routing, or recycle if testing costs more than the reuse value.
A short test cannot rule out hidden damage.
- If
The pack shows physical damage, swelling, leaks or thermal events.
ThenRemove it from the reuse path and handle it as damaged for storage and transport.
No electrical result outweighs a safety finding.
Questions and answers
What state of health makes a retired EV battery suitable for stationary storage?
There is no universal threshold. Suitability depends on the new duty cycle: a slow, shallow daily discharge tolerates lower capacity and higher resistance than fast frequency response. Set thresholds per application from the power and energy the project needs, with margins for test uncertainty and further ageing. Include module spread and resistance, not just capacity.
Can machine learning replace physical testing of used batteries?
Not entirely. A model trained on BMS history can rank packs and identify the clear cases, which cuts how many need long tests. It cannot detect physical damage, and its accuracy depends on chemistry, vehicle platform and the quality of the logged data. Use it to decide where test time goes, and keep confirmation tests and safety inspections for every pack.
Who is responsible for a repurposed battery under EU rules?
Under the EU Battery Regulation, the operator that places a repurposed or remanufactured battery on the market takes on the manufacturer's obligations for it, including conformity and producer responsibility, and the battery receives a new passport linked to the original once passports apply. Contracts between the original maker, the repurposer and the integrator should state who holds which records. Confirm the details for your case with legal advisers.
Sources
- Regulation (EU) 2023/1542 concerning batteries and waste batteries — EUR-Lex · checked 10 October 2026
- ANSI/CAN/UL 1974:2023, Standard for Evaluation for Repurposing or Remanufacturing Batteries — Standards Council of Canada · checked 10 October 2026
- Energy and Materials: use cases and delivery process — ColdAI