Researchers from Peking University and Alibaba’s DAMO Academy have used AI and satellite imagery to produce a high-resolution inventory of China’s wind and solar infrastructure. The dataset identifies 319,972 photovoltaic facilities and 91,609 wind turbines across 1,915 counties.
The project processed 7.56 terabytes of sub-metre imagery, creating a nationally consistent view that earlier energy studies lacked. Its importance goes beyond mapping assets: researchers used the inventory to examine how geographically distributed wind and solar generation can balance each other.
As data centres drive electricity demand upward, knowing where renewable capacity exists and how it behaves together becomes a strategic operating capability.
Geospatial AI creates a national energy inventory
Solar panels and wind turbines appear across diverse terrain, scales and image conditions. Deep learning allowed the team to identify installations ranging from urban rooftops to large facilities on the Mongolian plateau.
Traditional infrastructure records can be fragmented across owners and jurisdictions. Satellite analysis offers a consistent external method for locating assets and tracking change, although operators still need to validate attributes such as capacity and connection status.
The resulting inventory turns a vast physical system into data that planners can analyse at county and national levels. The researchers have released the dataset and code through Zenodo, allowing others to inspect the method or adapt it to additional countries.
Public availability matters because the value of the work depends on reproducibility and updates. Renewable infrastructure expands rapidly, so a static map will lose operational relevance without a repeatable pipeline.
Geographic coordination reduces renewable variability
Wind and solar output fluctuate with weather and time. Previous studies often assessed complementarity through hypothetical deployment models rather than the actual location of infrastructure.
The new analysis shows that wind and solar can offset each other more effectively as the geographic area of coordination expands. Cloud over solar facilities in one region may not coincide with weak wind generation hundreds of kilometres away.
China currently coordinates much of its electricity system at provincial level. A wider national approach could pair complementary resources, stabilise supply and reduce curtailment—the disposal of renewable electricity that the grid cannot absorb or transmit.
National coordination still requires transmission capacity, market rules and fast operational data. A map reveals where assets are located; it does not itself move electricity across congested lines or resolve incentives between provinces.
AI demand makes grid visibility more urgent
Chinese data services and large computing facilities consumed 22.9 billion kilowatt-hours in the first quarter of 2026, a 44% year-over-year increase according to the China Electricity Council.
New data centres are concentrating in northern and western provinces where land, electricity, wind and solar resources are more available. Those same regions show strong potential for complementarity, creating an opportunity to align computing loads with renewable generation.
Globally, the International Energy Agency expects data-centre electricity consumption could approach 1,000 terawatt-hours by the end of the decade. Other economies face the same need to understand generation, transmission and demand as an integrated system.
China’s clean-energy sector produced an estimated 15.4 trillion yuan in economic output in 2025. Managing an asset base of that scale without consistent national visibility limits planning, maintenance and investment.
The geospatial model offers a template, but replication elsewhere depends on satellite access, local validation and transparent governance. High-resolution maps can support public planning while also raising questions about sensitive infrastructure data.
The project illustrates one of AI’s most concrete infrastructure roles. Rather than merely increasing electricity consumption, it can reveal the physical energy system needed to support that demand.
The “god’s-eye view” becomes useful only when operators connect it to transmission decisions, forecasts and market coordination. If China can translate national visibility into national operation, it may reduce wasted renewable power while serving rapidly growing computing loads more reliably—a lesson other grids will find difficult to ignore.
