Run one number before you pick a model: the share of the array's output your site can consume directly. At 60% and above, build distributed, because each self-used kWh wipes out a ¥0.6 to ¥1.0 purchase while the same kWh exported earns the floor price of ¥0.15 to ¥0.4155.[1]
Below 60%, more of the output slides to the cheap side. On the same 1 MW array, dropping self-use from 60% to 40% cuts the annual total from ¥624,000 to ¥516,000, a 17% loss. With no usable roof at all, only one model is left, and that one needs contiguous land plus a corridor to carry the power out.
A 1 MW roof at 1,200 full-load hours makes 1.2 million kWh a year. Keep 60% and you take ¥504,000 off your bill. Sell the remaining 480,000 kWh for ¥120,000, and the year lands at ¥624,000. Export all 1.2 million instead and you get ¥300,000, a gap of ¥324,000. The capex behind the distributed case breaks into modules, inverters, racking and grid paperwork, and payback arrives in 4.2 years at ¥2.59/W.
Where those figures come from: ¥0.7/kWh for self-use and ¥0.25/kWh for surplus export, at 1,200 equivalent full-load hours a year. Both prices sit inside the official ranges above.
The biggest distributed project allowed is 50 MW. The everyday commercial rooftop tops out at 6 MW. Against a 1,000 MW central base that is 167 times smaller; against the 50 MW ceiling, 20 times. The "Three-North" wind and solar bases alone add 370 GW over the 15th Five-Year Plan.[2]
China added 71.77 GW of solar in the first half of 2026. Distributed took 42.22 GW; central took 29.55 GW. Rooftops came out 43% ahead.[3]
The IEA expects 4,600 GW of new renewable capacity by 2030, with solar PV taking almost 80% of it. China accounts for nearly 60% of that growth, and it is on track to hit its 2035 wind and solar target five years early.[4]
Where the Models Split
| Item | Distributed | Centralized |
|---|---|---|
| Where the power goes | Your plant first, surplus to the grid | All of it to the grid |
| Land | About 14,000 sqm of roof per MW | 25 to 40 mu per MW, or 16,700 to 26,700 sqm |
| What sets the price | The tariff you avoid, ¥0.6 to ¥1.0 per kWh | Floor price to market, ¥0.15 to ¥0.4155 per kWh |
| Where it breaks | Grid capacity, roof load, lease length | Corridor capacity, curtailment |
By end-June 2026 the country held 576 GW of distributed solar and 696 GW of central, a 120 GW gap. Distributed is 45.3% of the 1,272 GW total.[3]
Developers front-loaded to beat the June 1, 2025 reform cutoff, and 212 GW went in during the first half of 2025. The full year closed at 317 GW, split 164 GW central and 153 GW distributed. The first half of 2026 came in 66% below the same period a year earlier.[5][6]
How Much Roof Is Left
Rural China has over 27 billion square meters of rooftop. In theory that holds about 1,900 GW, one and a half times the entire installed solar fleet.[7]
One kilowatt wants about 14 sqm of roof, so a megawatt wants 14,000. The modules themselves cover only 5 to 6 sqm per kW. The rest goes to shading, row spacing and walkways.
Tsinghua researchers folded urban roofs in and came out higher. Technical potential runs 2,785 GW; take away what the grid can absorb and what pays for itself, and deployable capacity falls to 1,173 GW. The grid and the economics take 58% off the top. That is twice today's 576 GW distributed fleet, and it would generate 463.1 TWh a year.[8]
Industrial and public buildings carry more than half of it, above 586 GW, since they draw power exactly when the sun is out. Residential and rural rooftops hold the balance.[8]
When Distributed Wins
The national rules on distributed PV split projects into four types, and they set both your system size and your connection voltage. Capacity caps run from 6 MW up to 50 MW, and your export options shrink from three down to one.[9]
- Residential, owned by a person. Connection at 380 V or below. Any of the three modes works. Take full export, full self-use, or self-use with surplus.
- Residential, owned by a company. Connection at 10 (20) kV or below, up to 6 MW. Same three modes.
- General commercial. Connection at 10 (20) kV or below, up to 6 MW in principle. Self-use only, or self-use with surplus.
- Large commercial. Up to 20 MW at 35 kV, up to 50 MW at 110 (66) kV. In principle, self-use only.
The Ratios That Bind You
For direct green-power supply, self-consumption has to reach 60% of available generation and 30% of your total consumption, rising to 35% by 2030. Exports are capped at 20% of available generation. Generation and load have to sit on the same land plot.[10]
Multi-user direct supply has 24 provinces issuing supporting rules and 100 projects cleared, together 34.1 GW, or 341 MW apiece.[11]
When Centralized Wins
Land is the gatekeeper here. A 10 MW array needs 250 to 400 mu, which works out to 25 to 40 mu per MW, or 16,700 to 26,700 sqm. Over the 14th Five-Year Plan the desert-and-gobi programme alone booked 22 million mu, and stacking crops or grazing under the array is how developers claw back part of that footprint.[12]
Water avoids the land bill entirely. The fishery-plus-PV layout puts the array over an existing pond and keeps the fish underneath, so one surface carries two revenue streams.
As of end-July 2026 China had 46 ultra-high-voltage channels running, moving more than 340 GW from west to east. Losses average 1.5% per 1,000 km at ±1,100 kV, 2.79% at ±800 kV and 6.94% at ±500 kV.[13][14]
Stretch that across 3,000 km and you are looking at 4.5%, 8.4% and 20.8%. The ±1,100 kV line sheds about 22% of what the older ±500 kV line does.
- Land in one block. Lock 250 to 400 mu for every 10 MW of array, and plug into a corridor that already exists.
- Win the floor-price auction. Solar cleared between ¥0.15 and ¥0.4155 per kWh.
Costs and Revenue
IRENA puts the global levelized cost of solar at $44/MWh for 2025, flat on 2024. China is the cheapest market in the same dataset, at $36/MWh for utility-scale solar and $27/MWh for onshore wind, roughly ¥0.26 per kWh at 7.1 to the dollar.[15] Solar PV has shed 89% of its cost since 2010, and China alone accounted for USD 177 billion of the USD 377 billion in fossil-fuel purchases avoided across 20 major economies, about half that total.[16]
Four-hour battery systems dropped almost 30% in 2025 to $140/kWh, roughly 95% below their 2010 level, and they run under $70/kWh inside China. About one in four new central plants went in with batteries attached.[15]
Industry ranges, useful for scale only. EPC bids in the first half of 2026 averaged ¥2.59/W for large distributed and ¥2.63/W for central. Payback ran about five years on commercial rooftops and six on homes. That covers modules, inverters, racking, labour and grid paperwork. It does not cover storage, roof reinforcement or 25 years of upkeep.
EPC prices differ by ¥0.04/W. Tariffs differ by ¥0.45/kWh. Most of the gap between commercial and utility-scale economics sits on the second line rather than on hardware, since rooftop, carport, floating and ground-mounted layouts all draw on the same module supply. On high-efficiency modules, though, distributed pays more. Central deals clear at ¥0.68 to ¥0.70 per watt, distributed buyers land at ¥0.76 to ¥0.83, a gap of ¥0.06 to ¥0.15 per watt.
Distributed claws back the module premium by skipping the booster station, the collector lines and the land bill. The finished EPC number still comes in lower, ¥2.59 against ¥2.63, a 1.5% edge.
Clean-tech manufacturing investment halved from a USD 70 billion quarterly peak in 2023 to USD 35 billion by the end of 2025, and IRENA expects rising component prices to push installed costs up through 2026.[16]
Shandong's January 2026 settlement put solar at a reference price of ¥0.1448/kWh and wind at ¥0.2055. The spread against the floor price works out to ¥0.0576 per kWh, and every commercial user in the province quietly picks up that tab.[17]
Where the Grid Says No
Once an area crosses an 80% reverse-load ratio on a transformer, it goes red and new distributed hookups stop. Nationwide, more than 50 GW of projects sat outside that door. In the countryside transformers run 200 to 400 kVA, so one village could take about ten households' worth.[18]
Fifty gigawatts is roughly a third of all the distributed capacity added in 2025. Late December 2025 brought DL/T 2041-2025, effective June 18, 2026. The red, yellow and green zones now run on open capacity instead of one hard number. Shandong, Hebei, Henan and Guangdong alone can release north of 10 GW.[18]
Utilization ran 95% in 2025 and 91.4% in the first half of 2026. Curtailment went from 5% to 8.6%. Against 655.5 TWh of generation, the loss works out to about 23.6 TWh in six months.[6][3] The IEA lists rising curtailment in China, Germany, Brazil, Chile, the UK and Ireland, and surging negative-price hours across multiple markets, so this is not a local quirk.[4]
The 2030 confidence-capacity target for wind and solar together is 8%, which breaks down to about 6% for solar and 11% for wind. Put 1,000 MW of solar on the grid and you can count on 60 MW at the evening peak, and no more.[2]
How the Floor Price Splits Old and New
Document 136 pushed all renewable volume into the market and left the floor price, the mechanism tariff that tops up market revenue, to settle the difference. Round one cleared 251,200 projects. Distributed took about 250,000 of them, which is 99.5% of the project count, but only 9% of capacity and 7% of the energy.[19][20]
Shandong has already shown distributed the door. General and large commercial systems, plus residential projects owned by companies and commissioned after June 1, 2026, no longer qualify for the floor price. Henan, Guangxi, Jiangxi and Anhui have done the same for new commercial systems.[20]
| Item | Distributed | Centralized |
|---|---|---|
| Existing projects covered by the floor price | 80% to 100% | 10% to 100% |
| Where new projects get their price | Excluded for new commercial distributed in most provinces | Provincial auction, ¥0.15 to ¥0.4155 per kWh |
| Projects cleared in round one | About 250,000 | About 1,200 |
The floor price runs 10 to 12 years, and 14 to 15 for offshore wind. Solar cleared at its ceiling of ¥0.4155 in Shanghai and its floor of ¥0.15 in Xinjiang.[1]
Nationwide, the average renewable price including the mechanism fee came to about ¥0.3/kWh in the first half of 2026, or 80% of the average coal benchmark. The mechanism fee made up 40% of it.[21]
What Green Certificates Add
Distributed systems can trade green power once certificates are issued.[22] Assume, conservatively, that only surplus power qualifies and the premium runs ¥0.05/kWh. On 480,000 kWh the premium brings in ¥24,000 a year, or ¥600,000 across 25 years.
Direct-supply projects meter both directions at every gateway, matching generation to consumption hour by hour, all 8,760 hours a year. Exporters can use the record to account for their power emissions.[10]
Common Misreads
Your Own Roof Still Has Gates
Miss any of these three gates and the project stops short of the meter. Get a roof load report. Put roof rights in the lease if you rent. And for anything at 10 kV or above, get the grid's interconnection opinion before you order steel.
That Floor Price Will Not Hold
It steps out after 10 to 12 years. Run the numbers over 25 years and the last 13 ride on market prices. If those prices land ¥0.1/kWh below the floor, a 1 MW project gives up ¥1.56 million.
Quick Answers
Can Distributed Still Export Everything
Projects commissioned on or before April 30, 2025 and under 20 MW can keep exporting all of it. From May 1, 2025 it depends on size. Residential below 6 MW still exports everything; commercial systems are limited to self-use, or self-use with surplus.[23]
How Much Does a Rooftop Array Generate
At 1,000 to 1,500 equivalent full-load hours, 1 MW puts out 1 to 1.5 million kWh a year. Shandong sits near 1,300 hours, the Sichuan Basin around 1,000. Over 25 years the array delivers roughly 30 million kWh before module degradation, worth about ¥15.6 million at the prices above.
Do You Need Batteries
Only if the load runs past sunset. Storage system bids in China averaged ¥541.3/kWh for four-hour systems and ¥599.3/kWh for two-hour systems in the first half of 2026, both up year on year.[24] A solar-plus-storage plant holding 95% reliability at a high-quality site now delivers firm power under $85/MWh.[15]
Match Your Situation
| Your situation | What to build | Capacity and connection | What has to be true first |
|---|---|---|---|
| You own the plant and run it 8am to 6pm | Commercial distributed, self-use with surplus | 100 kW to 6 MW | Self-use reaches 60% of generation |
| You lease the plant, ten years or more left | Commercial distributed, energy management contract | 100 kW to 6 MW | The lease spells out roof rights |
| Two or more firms in one park want green power | Multi-user direct supply | Sized to the load | Combined load absorbs 60% of output |
| Your own house roof | Residential distributed | 10 to 30 kW at 380 V or below | Solid roof, clean title |
| A block of desert, tidal flat or old mine | Central plant or a base add-on | From 100 MW | Inside a provincial base plan, with corridor access |
| High power bills and carbon border exposure | Distributed with storage, or direct supply | Work back from the load | Certificates issued, hourly metering in place |
Every extra 20 points of self-consumption is worth ¥108,000 a year on a 1 MW system. At 40% you make ¥516,000, at 80% you make ¥732,000. That is why the ratio decides the model and the roof area does not.
Send us your roof area or land plot, your daytime load profile, and the grid connection point you already have. We will size the array, run the self-consumption ratio against your own tariff, and tell you which of the two models clears your hurdle rate.