The Payback Case Rests on Four Numbers, Not One

Most commercial and industrial (C&I) storage proposals in China are sold on a single figure: the local peak-valley spread. That is the wrong anchor. A project's internal rate of return (IRR) is decided by the spread, the number of effective full cycles per day, the round-trip efficiency (RTE), and the total installed cost per kilowatt-hour — in that order of sensitivity.
Get the spread right and the other three wrong, and a "4-year payback" pitch becomes 7 years with a degraded asset. According to the China Energy Storage Alliance (CNESA), China added roughly 43.7 GW of new-type storage in 2024, taking cumulative capacity past 70 GW, and behind-the-meter C&I systems were the fastest-growing segment by unit count — meaning competition is now on engineering quality, not on whether storage works.
The technology works. The modelling is where projects die.

Net Arbitrage, Not Headline Spread: The First Correction

Provincial time-of-use (TOU) tariffs in Zhejiang, Guangdong and Jiangsu now show peak-valley spreads of 0.70–1.00 RMB/kWh for 1–10 kV general industrial accounts, and the 尖峰 (critical peak) blocks introduced under the National Development and Reform Commission's 2021 TOU reform (发改价格〔2021〕1093号) push the top block higher still.
But you cannot bank the spread. Charging at valley price costs money, and efficiency losses mean you must buy more kilowatt-hours than you sell. Net arbitrage per discharged kWh is:
> spread × RTE − (degradation cost per kWh) − (operating and O&M cost per kWh)
At a 0.75 RMB/kWh spread and 88% RTE, gross margin is 0.66 RMB/kWh before degradation, which typically adds 0.05–0.09 RMB/kWh over a 6,000-cycle life. That is a 10–12% haircut on the revenue line, and it is the single most commonly omitted line in Chinese C&I proposals.

Installed Cost Has Fallen — But Balance-of-System Now Dominates

BloombergNEF's December 2024 survey put the global average turnkey price for a four-hour battery energy storage system (BESS) at USD 165/kWh, a 20% year-on-year decline. In China, lithium iron phosphate (LFP) cells traded near 0.30 RMB/Wh through much of 2024, with aggressive quotes below that.
Here is the trap: cell price is not system price. In a 2 MWh C&I cabinet, the cell is now under 35% of the delivered cost. The rest is power conversion system (PCS), thermal management, fire suppression, transformer and grid-connection works, permitting, and the metering and communication gateway needed for demand-charge management.
A realistic 2025 all-in figure for a 1 MW/2 MWh LFP system in eastern China, including EPC and grid interconnection, sits at 0.75–1.00 RMB/Wh. Deals quoted at 0.55 RMB/Wh usually exclude fire-code upgrades, the medium-voltage interface, or both.
| Configuration | Installed cost | RTE | Cycles/day | Simple payback* |
|---|---|---|---|---|
| 1 MW/2 MWh LFP, liquid-cooled | 0.85 RMB/Wh | 88–90% | 1 | 3.6 yr |
| 1 MW/2 MWh LFP, air-cooled | 0.75 RMB/Wh | 85–87% | 1 | 4.2 yr |
| 0.5 MW/1 MWh LFP, 2 cycles | 0.95 RMB/Wh | 88% | 2 | 3.1 yr |
| 1 MW/4 MWh LFP, 2 cycles | 0.78 RMB/Wh | 89% | 2 | 3.4 yr |
*Assumes 0.75 RMB/kWh spread, 330 operating days, 0.06 RMB/kWh degradation cost.

Sizing Is Where Projects Lose Money

Oversizing is the most expensive mistake in the category. A battery sized to the site's maximum demand rather than to its actual load profile will cycle at partial depth, wasting capital and under-delivering revenue. The correct method is a 15-minute-interval load analysis over 12 months, overlaying the TOU blocks.
Two rules of thumb hold up in field data: size the energy capacity to the discharge window, not to the day's total consumption; and size the PCS to 60–75% of site peak demand, because you are shaving blocks, not carrying the whole load.
Demand-charge management is the undervalued second revenue stream. In provinces that still levy capacity charges (基本电费) on maximum demand, a correctly sized system routinely contributes 20–30% of total project revenue by clipping the monthly peak — and it does so with the same hardware and no additional degradation.

Round-Trip Efficiency and Degradation: The Silent Margin Killer

Manufacturer datasheets quote RTE at 25°C, 0.5C, new cells. Real C&I sites run at 35–40°C in summer with intermittent partial cycles. Measured RTE on air-cooled cabinets in Chinese summer conditions frequently lands at 84–86%, not the 90% on the sheet.
Liquid cooling holds RTE within 1–2 points of datasheet and extends calendar life, which is why the cost premium — typically 8–12% — pays back in high-spread provinces where every efficiency point is worth roughly 0.011 RMB per discharged kWh. Over 6,000 cycles on a 2 MWh asset, three efficiency points is about 40,000 RMB.
Degradation modelling matters just as much. Warranties are usually written as "70% of nominal capacity at year 10 or 6,000 cycles." Ask for the cycle-life test conditions — depth of discharge, temperature, C-rate — because a 6,000-cycle claim at 80% depth of discharge and 25°C does not describe your site. For a deeper technical walkthrough of cell chemistry and warranty structures, see battery storage technology.

A Real Project: 1 MW/2.15 MWh Behind-the-Meter in Hebei

A metal-processing plant in Hebei's Langfang area commissioned a 1 MW/2.15 MWh liquid-cooled LFP system in 2024, integrated behind the meter at 10 kV alongside an existing rooftop array of solar panels.
The numbers, verified over the first 12 months of operation:

• Installed cost: 1.72 million RMB (0.80 RMB/Wh, including interconnection and fire upgrades)
• Usable energy at 90% depth of discharge: 1.94 MWh
• Measured RTE: 88.4%
• Effective spread after efficiency loss: 0.66 RMB/kWh
• Arbitrage revenue: 1,280 RMB/day × 330 days ≈ 422,000 RMB/year
• Demand-charge savings: 96,000 RMB/year
• Total annual benefit: ≈ 518,000 RMB, minus 28,000 RMB O&M
• Simple payback: 3.5 years; ten-year IRR above 20%

The plant's load profile was the deciding factor. It runs two shifts, so the battery charges in the valley block from 23:00 and discharges into the morning and afternoon peaks — a clean single cycle with no partial-cycle waste. A single-shift site with the same capacity would have returned closer to 5.5 years.
The project also demonstrated a lesson worth repeating: the original design specified 2.5 MWh. Trimming 350 kWh of capacity cut capex by 11% and reduced payback by seven months, because the extra energy had no discharge window to serve.

Policy Risk: The Tariff Structure Will Change

Every C&I arbitrage model has a policy dependency, and 2025 made that explicit. The NDRC and National Energy Administration's Document 136 (发改价格〔2025〕136号) ended mandatory storage allocation for new renewable projects and pushed storage toward market-based revenue. The effect on behind-the-meter projects was counterintuitive: with fewer forced-build front-of-meter projects, C&I systems now compete on genuine economics rather than on compliance.
The risk runs the other way too. Provinces revise TOU schedules roughly annually, and several have narrowed the peak window or shifted the critical-peak block. A financial model that assumes a fixed 0.80 RMB/kWh spread for ten years is not a model — it is a hope. Stress-test at 70% of the current spread and confirm the project still clears your hurdle rate.

Chemistry and Configuration: What to Specify

LFP dominates C&I for good reasons: cycle life of 6,000–8,000 cycles, thermal stability well above nickel manganese cobalt (NMC) chemistry, and cell costs that no alternative matches at this scale. Sodium-ion is arriving for cold-climate and cost-sensitive sites, but energy density and cycle-life data at 5,000+ cycles remain thin.
Configuration choices matter more than chemistry at this point. Liquid cooling over air. String-level or cluster-level PCS architecture over a single central inverter, because it preserves 70–80% of capacity if one module string faults. And a gateway that exposes 15-minute data to your energy management system (EMS) — without interval data you cannot verify savings, and unverifiable savings do not survive a CFO review.
If the site also has rooftop generation, co-locating storage with the array and sharing the interconnection point reduces balance-of-system cost materially. DLXN's integrated carport and storage structures are built around exactly this shared-infrastructure logic.

Five Actions Before You Sign

1. Model net arbitrage, not spread. Apply RTE, degradation cost and O&M to every revenue line. If the vendor's model shows gross spread, discount it by at least 12%.
2. Demand a 12-month, 15-minute load profile in the proposal. Any sizing that is not derived from interval data should be rejected outright.
3. Price the complete scope. Require a line-item breakdown covering fire suppression, medium-voltage interface, grid-connection fees and EMS commissioning. Compare only all-in RMB/Wh figures.
4. Verify efficiency and cycle life under your conditions. Ask for RTE at 35°C and the full test conditions behind the cycle-life warranty, not the datasheet headline.
5. Stress-test the tariff. Re-run payback at 70% of the current spread and confirm it stays inside your acceptable horizon. If it does not, resize rather than abandon — smaller systems usually clear the bar.
For sites evaluating storage alongside new generation, the product overview and inverter selection guidance cover the interconnection and conversion side of the design. Our engineering team can run an interval-data sizing study against your actual tariff schedule — start a conversation through contact.