Why Peak-Valley Arbitrage Cleared the Payback Hurdle in 2025

The economics flipped on the cost side, not the revenue side. Turnkey four-hour battery energy storage system (BESS) prices averaged USD 165/kWh globally in BloombergNEF's 2024 survey, down roughly 40% from 2022 levels, and Chinese C&I systems tracked that curve even harder. Domestic 2-hour containerized systems now quote at RMB 0.75–0.90/Wh including engineering, procurement and construction (EPC), compared with RMB 1.8–2.2/Wh in 2022.
Meanwhile provincial peak-valley spreads widened. Guangdong's general industrial tariff for 1–10 kV customers can exceed RMB 1.1/kWh between sharp-peak and valley periods in summer, and Zhejiang's spread sits near RMB 0.85/kWh. A system that earns RMB 0.75 per kilowatt-hour discharged and costs RMB 0.85/Wh installed pays back in about three years with two cycles a day. That arithmetic simply did not exist in 2021, when CAPEX was triple and spreads were 30% narrower.

China's Provincial Peak-Valley Spreads: Where the Money Is

Geography decides returns more than any equipment choice. According to provincial development and reform commission tariff notices, the usable spread for a 1–10 kV industrial customer after efficiency losses looks like this:
| Province / Region | Peak-valley spread (RMB/kWh) | Typical payback, 2 cycles/day | Notes |
|---|---|---|---|
| Guangdong (Pearl River Delta) | 0.95–1.15 | 2.5–3.2 years | Sharp-peak pricing in Jul–Sep |
| Zhejiang | 0.80–0.90 | 3.0–3.8 years | Two-part tariff with demand charge |
| Jiangsu | 0.65–0.80 | 3.5–4.5 years | Strong demand-charge savings |
| Shandong | 0.55–0.70 | 4.5–6 years | Deep midday valley from solar |
| Hebei / Henan | 0.40–0.60 | 6–8 years | Marginal without demand-charge relief |
The trap is a province with a headline spread that only exists for three months. Always model an annual weighted-average spread, month by month, using the actual transformer-level tariff rather than a marketing figure.

The Real Cost Stack Behind a RMB/kWh Arbitrage Yield

Gross spread is not net yield. Three deductions destroy naive business cases. First, round-trip efficiency: a modern LFP system delivers 88–92% AC-to-AC, so every 1 kWh discharged requires 1.09–1.14 kWh of cheap charging energy. Second, degradation: expect 2–2.5% capacity loss a year in the first five years, reaching 80% state of health (SOH) at 6,000–8,000 cycles. Third, parasitic load — thermal management, fire suppression and communication systems draw 1.5–3% of rated capacity.
Stack these and a RMB 0.85 headline spread becomes roughly RMB 0.62–0.68 of bankable yield. Operators who skip this step routinely overstate first-year revenue by 20%. Request the efficiency curve at 25°C, 35°C and 45°C, not a single nameplate number — summer derating in a non-air-conditioned container is real.

One Cycle or Two Cycles a Day?

Sizing Rules That Decide Returns
Two cycles a day roughly halves payback, but only if the load profile supports it and the tariff has distinct peak and sharp-peak windows. A 2-hour system cycling twice needs about 4.5 hours of charge-discharge window plus a contingency margin; industrial plants running 8:00–20:00 shifts often cannot fit it.
Over-sizing is the more common failure. Installing 2 MWh when the site can only absorb 1.2 MWh of discharge means paying for cells that never cycle. The right sequence is: pull 12 months of 15-minute interval data from the meter, identify the actual peak window, cap discharge power at the minimum base load during that window, and then size energy. Details on how the DC side interacts with site load are covered in our battery storage technology overview.

Case Study: A 1.2 MW / 2.4 MWh Textile Plant in Zhejiang

A Ningbo-area textile finishing plant with a 1.6 MW base load installed a 1.2 MW / 2.4 MWh LFP system in early 2024, dispatching twice daily — once into the morning peak, once into the evening peak. The site's realized average peak price was RMB 1.05/kWh and its valley price RMB 0.31/kWh.
Annual figures: 1,536 MWh discharged (2 cycles × 320 days), revenue of RMB 1.61 million, charging cost of RMB 0.54 million, gross margin RMB 1.07 million. After applying 93% availability-and-degradation adjustment and deducting RMB 55,000 of operations and maintenance (O&M) plus insurance, net pre-tax benefit landed near RMB 940,000 against a RMB 2.3 million investment — a 2.4-year simple payback and roughly 25% IRR. Post-tax and including a mid-life augmentation reserve, the realistic payback is about three years.

Battery Chemistry, Degradation and the Warranty Fine Print

LFP dominates C&I for good reason: 6,000–10,000 cycle life, thermal runaway onset above 200°C, and no cobalt exposure. Cell prices fell to roughly RMB 0.30–0.35/Wh for storage-grade LFP in 2024, according to China Photovoltaic Industry Association data, and cell chemistry is now a smaller share of project cost than the enclosure, thermal system and power conversion hardware.
Warranty language matters more than the chemistry label. Look for an energy-throughput guarantee (megawatt-hours delivered over ten years) rather than a capacity percentage alone, an explicit definition of end-of-warranty SOH, and clarity on who pays for augmentation. A ten-year, 70% SOH warranty with a throughput cap of 1.5× nameplate energy per cycle is worth far less than it appears.

Demand Charges, Ancillary Services and Stacked Revenue

China's two-part industrial tariff includes a demand charge — often RMB 30–45 per kVA per month in Zhejiang and Jiangsu — that storage can shave without cycling. A 1.2 MW system clipping 300 kW of monthly peak demand saves RMB 108,000–162,000 a year on its own, and that revenue is nearly free because it uses a fraction of one cycle.
Additional layers exist but should be treated as upside, not underwriting: virtual power plant (VPP) aggregation payments, participation in provincial demand response events, and peak-shaving incentives. In Guangdong's 2024 demand response pilots, event compensation reached RMB 3–5/kWh in extreme grid-stress hours, but only for a handful of hours a year. Finance the project on arbitrage alone, and the rest becomes margin.

Procurement Checklist: Five Rules Before You Sign

1. Model the tariff, not the brochure. Build a 12-month, hour-by-hour model from the site's own interval meter data. Demand a weighted average spread figure and use 90% of it.
2. Specify round-trip efficiency at 35°C. Anything below 88% AC-to-AC under real ambient conditions should be rejected or re-priced.
3. Cap power at minimum base load. Discharge power must never exceed what the site can absorb, or the system will be forced to export at unattractive rates.
4. Require throughput-based warranties. Replace vague "10-year warranty" language with guaranteed MWh delivery, defined SOH testing protocol, and a named augmentation cost.
5. Verify the DC and AC side integration. Use tier-one inverters and LFP battery modules with documented site references, and inspect an operating reference project before signing — see our project portfolio for examples of how a poorly integrated system underperforms in year three.
For a site-specific feasibility study with an hour-by-hour arbitrage model, talk to our engineering team.