The economics that pushed second-life storage into the mainstream

China added 43.7 GW / 109.8 GWh of new-type energy storage in 2024, taking cumulative capacity to 73.76 GW according to the National Energy Administration — and CNESA expects annual additions to keep growing through 2030. At the same time, the China Automotive Engineering Research Institute estimates retired power batteries will reach roughly 820,000 tonnes in 2025 and about 3.5 million tonnes by 2030. That is a feedstock problem on one side and a cheap capacity problem on the other.
The price gap is what makes this commercially interesting. BloombergNEF put average global LFP pack prices near $75/kWh in 2024 after one of the steepest cost declines in the industry's history. Second-life packs sourced from fleet returns in Guangdong and Jiangsu now clear at 0.30–0.45 RMB/Wh at module level, versus 0.50–0.60 RMB/Wh for a new two-hour turnkey container. On a delivered-kWh basis, that is a 25–40% capex discount before you count the shorter warranty.

Regulation: what GB/T 34015 and the EU Battery Regulation actually require

Domestically, the framework rests on GB/T 34015-2017 for residual energy testing of vehicle power batteries, GB/T 33598 for dismantling and disassembly, and the MIIT's 2021 Measures for the Echelon Utilization of Power Batteries, which introduced the now-standard traceability code requirement. Every pack that enters a second-life channel should carry an auditable history: first-life vehicle, cycle count, thermal events, and discharge depth. If your supplier cannot produce that, you are buying scrap with a label.
Europe moves faster than most buyers expect. EU Regulation 2023/1542 mandates a carbon footprint declaration for EV batteries, imposes recycled-content thresholds from 2031 (16% cobalt, 85% lead, 6% lithium, 6% nickel), and requires battery passports. Export-oriented packagers should treat these as a design constraint today, not a 2031 problem. Much of the second-life equipment sold into Africa and Southeast Asia is now being reverse-engineered for compliance.

"70% state of health" is a threshold, not a cliff

Almost every procurement document says "second-life packs above 70% SOH (state of health)". Treat that number with suspicion. Capacity fade is non-linear, and packs sitting near a knee point can lose 5–8% in a single year, then stabilise, or the reverse. What you actually need is the degradation curve, not the endpoint.
Two specifications matter more than headline SOH. First, module-level dispersion: capacity spread within 3% and DC internal resistance within 5%, measured at the same temperature and C-rate. Second, remaining full cycles — most LFP packs retired from buses at 70% SOH still have 1,500–3,000 cycles at 80% depth of discharge, which is plenty for a two-cycle-per-day arbitrage duty. Expect round-trip efficiency of 88–92%, against 94–96% for new LFP.

Second-life LFP compared with new LFP: what the cost curve says

| Parameter | New LFP BESS | Second-life LFP modules |
|---|---|---|
| Pack/cell cost | 0.50–0.60 RMB/Wh | 0.30–0.45 RMB/Wh |
| Round-trip efficiency | 94–96% | 88–92% |
| Annual degradation | ~2% | 3–5% |
| Remaining cycles | 6,000+ | 1,500–3,000 |
| Warranty (typical) | 10 years | 3 years |
| Indicative LCOS | 0.35–0.45 RMB/kWh | 0.25–0.35 RMB/kWh |
| Insurability | Standard | Restricted |
The LCOS (levelised cost of storage) advantage of roughly 0.10 RMB/kWh is real but narrower than the capex gap suggests, because efficiency losses compound daily. Run the model at your own tariff structure — a site with a 0.64 RMB/kWh peak-valley spread tolerates the efficiency hit far better than one with a 0.35 RMB spread. Our team walks through the modelling assumptions in our battery storage technology notes.

Sorting, grading and BMS recalibration: where projects actually fail

The failures are almost never in the cells. They are in the electronics around them. Retired packs arrive with proprietary CAN (controller area network) protocols, and generic BMS boards either cannot read them or mis-map the cell voltage tables. You end up with SOC (state of charge) estimation drifting 10–15% within months, which quietly destroys the arbitrage case because the system stops short of full charge.
Three fixes are worth paying for: an independent BMS with per-module DC-DC conversion so mismatched strings do not drag each other down; a documented 72-hour capacity test at your site temperature, not the refurbisher's bench; and a module-level serial number register so you can swap a single weak module without scrapping the rack. Mixing modules from different first-life vehicles in one string is the single most common mistake we see.

Case study: a 20 kW / 40 kWh second-life system in Baoding, Hebei

A metal-parts workshop in Baoding, Hebei, commissioned a 20 kW rooftop array with a 40 kWh second-life LFP pack in mid-2024. The pack came from retired electric buses with a recorded 71% SOH and 2,100 remaining cycles. Installed capex was 42,000 RMB — 14,000 RMB for the pack at 0.35 RMB/Wh, the balance for the hybrid inverter, racking, fire detection and commissioning. A new-equivalent system quoted at 61,000 RMB.
Hebei's industrial peak-valley spread of roughly 0.64 RMB/kWh, cycling 30 kWh per day, produces about 7,000 RMB of annual savings plus a modest demand-charge reduction. That is a 6-year simple payback against 8.7 years for the new system. The honest caveat: the refurbisher's warranty covers three years, so the ten-year model includes one module refresh at an assumed 0.28 RMB/Wh. Without that line item, the business case is fiction.

Safety, insurance and the bankability gap

Second-life incidents are not dramatically more frequent than new-system incidents, but they are harder to insure. Most Chinese insurers still price second-life stationary storage case by case, and several European underwriters exclude it outright. Compliance with GB/T 42288-2022 — gas detection, per-rack thermal sensing, fire suppression and emergency ventilation — is effectively the entry ticket.
Design conservatively. Cap the charge rate at 0.3C rather than 0.5C, leave 15% headroom at the top of the voltage window, and derate nameplate capacity by 20% in the financial model. That costs you some headline kWh but keeps the pack away from the thermal edge where aged cells behave unpredictably. Some integrators now pair second-life modules with new cells in a hybrid architecture to keep the critical loads on the newer string.

Recycling technologies and why LFP changes the maths

Hydrometallurgical recycling recovers over 95% of lithium and 98% of nickel and cobalt, but it needs cobalt and nickel to pay for itself. LFP has neither in meaningful quantity, so black-mass revenue is dominated by lithium carbonate and iron phosphate — thin margins that make second-life reuse the economically preferred first stop. Pyrometallurgy is simpler and more tolerant of mixed chemistries, at higher energy cost per tonne.
Direct recycling, which restores cathode material without breaking it down to elemental form, remains at pilot scale in China and the United States, with NREL citing material-energy savings of 60–70% versus hydrometallurgy. For asset owners, the practical implication is simple: choose a recycler with documented mass-balance reporting now, because the EU recycled-content thresholds arriving in 2031 will be calculated from those records. If you want to see how we specify storage hardware for these duty cycles, our lithium battery product range lists the configurations we support.

Five procurement rules before you sign for second-life capacity

1. Demand the degradation curve, not the SOH number. Request the capacity test history at 0.5C, and reject any pack where module capacity dispersion exceeds 3%.
2. Budget a module refresh. Assume one 30% capacity refresh within ten years at 0.25–0.30 RMB/Wh, and only then compare payback against a new system.
3. Verify the traceability code. Cross-check the pack's first-life VIN (vehicle identification number) and cycle count against the MIIT platform data. Unverifiable history means no insurance and no warranty enforcement.
4. Insist on per-module DC-DC and an open BMS. Proprietary CAN mapping is the leading cause of premature replacement in second-life fleets.
5. Model at your own tariff. Second-life only works where the peak-valley spread exceeds roughly 0.45 RMB/kWh and daily cycling is at least one full equivalent cycle. Below that, buy new.
If you are specifying a C&I system and want a second opinion on whether second-life capacity fits your load profile, talk to our engineering team — we will run the LCOS comparison against your actual tariff schedule. You can also review delivered configurations in our project portfolio and the wider technical library for commissioning checklists.