Energy Storage Buying Guide: Eight Pitfalls to Avoid
Here is the short version: a storage system is bought on kilowatt-hours and paid back on delivered kilowatt-hours. Every gap between those two numbers — depth of discharge, round-trip efficiency, degradation, auxiliary loads — is where your return disappears. The eight sections below quantify each gap.
The Eight Decisions That Decide Payback
China's new-type energy storage market added 43.7 GW / 109.8 GWh in 2024, according to CNESA, bringing cumulative installed capacity past 70 GW by year-end, per the National Energy Administration. Growth of that speed inevitably drags a tail of badly specified systems behind it. Utility-scale tenders are written by engineers; commercial and industrial (C&I) and residential systems are often written by whoever answered the phone first.
The consequence is a persistent spread in realised performance. Systems with identical nameplate capacity routinely differ by 25–40% in annual discharge, once you account for depth of discharge (DoD), round-trip efficiency (RTE), thermal derating and parasitic loads. You are not buying a battery. You are buying an energy-delivery contract with a chemistry attached.
Choosing Between LFP and NMC Chemistries
Lithium iron phosphate (LFP) now accounts for more than 95% of new stationary storage capacity in China, per the China Photovoltaic Industry Association. The reason is not fashion. LFP tolerates 6,000–8,000 cycles to 80% capacity retention, costs less per kilowatt-hour, and its thermal runaway threshold sits near 270°C against roughly 200°C for nickel manganese cobalt (NMC).
NMC still wins on density — 250–300 Wh/kg against 160–180 Wh/kg for LFP — which matters for mobile or severely space-constrained installations. For a fixed rack in a plant room, density is irrelevant and cycle life is everything. BloombergNEF's 2024 battery price survey put average pack prices at $115/kWh, with LFP packs landing a further 15–20% below that benchmark. Cell prices in China fell near RMB 0.3/Wh during 2024.
| Parameter | LFP | NMC |
|---|---|---|
| Energy density | 160–180 Wh/kg | 250–300 Wh/kg |
| Cycle life at 80% DoD | 6,000–8,000 | 3,000–5,000 |
| Thermal runaway onset | ~270°C | ~200°C |
| Typical installed cost | 0.8–1.1 RMB/Wh | 1.1–1.5 RMB/Wh |
| Share of new C&I projects in China | >95% | <5% |
Usable Capacity: The Depth-of-Discharge Trap
A 100 kWh rack is not 100 kWh of usable energy. Most LFP warranties credit full cycle life only up to 80–90% DoD, and the tail of the discharge curve — below roughly 10% state of charge — is where voltage sags and the battery management system (BMS) trips. Derate a 100 kWh nameplate rack to 80 kWh usable, apply 88% RTE, and you deliver about 70 kWh to the meter.
Add auxiliary consumption — thermal management, controls, communications — and a liquid-cooled container can draw 2–4% of throughput on a hot day. In Guangdong summers, air-cooled cabinets frequently run their fans at full duty for six hours. Vendors who quote nameplate capacity without stating usable capacity at a defined temperature are quoting a number you cannot bill against.
Cycle Life: What the Warranty Actually Guarantees
Read the warranty in megawatt-hours, not cycles. A 6,000-cycle promise at 80% DoD on a 100 kWh rack implies roughly 480 MWh of throughput before end-of-warranty. If the warranty instead guarantees 70% retention rather than 80%, that is a 10-point haircut on residual value and typically a three-year difference in useful asset life.
Two clauses deserve scrutiny. First, the temperature window: many warranties void accelerated-degradation coverage outside 15–30°C average cell temperature, which is exactly the condition of an unconditioned plant room. Second, augmentation terms. NREL's storage cost modelling treats augmentation as a recurring capital line, not an afterthought — budget it in year five or accept degraded revenue from year six.
Power Conversion: Matching Inverter and Battery
Round-trip efficiency is set by the power conversion system (PCS) as much as by the cells. A modern bidirectional PCS runs 97–98.5% at rated load but falls sharply at low load — the same part-load cliff that affects photovoltaic inverters operating far below their maximum power point tracking (MPPT) window. A 100 kW PCS cycling at 15 kW average load will not deliver its datasheet efficiency.
DC-coupled architectures save one conversion stage and typically gain 2–4 percentage points of RTE over AC-coupled retrofits, which matters when you are cycling twice daily. Grid-following PCS units certified to IEEE 1547 are standard for grid-tied C&I work; off-grid and weak-grid sites need grid-forming capability with defined black-start behaviour. The inverter and PCS technical notes cover ride-through and reactive-power settings worth specifying before tender.
Peak-Valley Arbitrage Maths — and When It Fails
General industrial and commercial peak-valley spreads in Guangdong reached roughly 0.9–1.0 RMB/kWh in 2024, with several inland provinces above 0.7 RMB/kWh. That spread is what you sell. Annual revenue equals usable kWh × cycles per day × operating days × spread × RTE.
Take a 100 kWh usable system, one cycle daily, 330 days, 0.85 RMB/kWh spread, 88% RTE: about RMB 24,700 a year. Drop to a 0.55 RMB/kWh spread — common in provinces with compressed peak pricing — and it falls to RMB 16,000. The same hardware, a 35% revenue difference driven entirely by tariff structure.
Two failure modes recur. Sites on flat tariffs have no arbitrage case at all and should be sold on demand-charge management instead. Sites that cycle twice daily without checking the battery's daily throughput limit will burn cycles faster than the warranty allows. Verify the local tariff schedule from the provincial development and reform commission before signing anything.
Thermal Management and Safety Certification
Temperature uniformity decides pack life. A well-designed liquid-cooled system holds cell-to-cell delta within 3°C; poorly baffled air-cooled racks can see 8–12°C spread across a module string, and the hottest cell sets the degradation rate for the whole string. Ask for the thermal report, not the marketing photo.
Certification is non-negotiable. Look for GB/T 36276 for cells and packs in the Chinese market, IEC 62619 for industrial applications, and UL 9540A thermal runaway propagation testing where insurers or overseas authorities require it. NFPA 855 separation distances apply to containerised commercial systems in many jurisdictions. A supplier who cannot produce a test report has not necessarily failed testing — they may simply never have run it.
Case Study: A 20 kW / 40 kWh System in Hebei
A metal-components workshop in Hebei commissioned a 20 kW rooftop array with a 40 kWh LFP cabinet in 2023. Usable capacity was 36 kWh at 90% DoD; measured RTE was 89% in the first year. Total installed cost, including mounting, PCS, commissioning and grid-connection fees, was RMB 78,000.
The site runs one arbitrage cycle nightly plus daytime self-consumption, offsetting an average blended rate of 0.78 RMB/kWh. Measured annual benefit was RMB 12,400 in year one, against a modelled 13,100 — the gap traced almost entirely to summer auxiliary cooling loads. Payback lands at 6.3 years, comfortably inside the 10-year warranty. The owner's advice was blunt: insist on a metered commissioning test that records actual delivered kilowatt-hours for seven days. Comparable project references show the same pattern — metered data beats modelled data every time.
Five Procurement Rules Before You Sign
1. Buy usable kilowatt-hours, not nameplate. Demand a written figure for usable capacity at 80–90% DoD and a stated ambient temperature range.
2. Convert the warranty into megawatt-hours. Multiply cycles × DoD × capacity, then compare that throughput number across vendors. Retention percentage matters as much as cycle count.
3. Request a load-profile simulation with real tariff data. Feed the vendor 12 months of your interval meter data and the provincial peak-valley schedule. If they quote a payback without your load curve, they are guessing.
4. Price augmentation into year five. NREL modelling and Chinese O&M contracts both treat capacity top-up as routine. A system bought without an augmentation budget is a system that underdelivers from year six.
5. Insist on a metered commissioning test. Seven days of recorded throughput, logged RTE and peak cell temperature delta. This single document resolves most warranty disputes before they start.
For a sizing review against your own load profile and local tariff schedule, send your interval data through the contact form — and if you want to compare LFP rack options first, the lithium battery product line lists usable capacity at defined DoD rather than nameplate figures.
