On December 19, 2025, the conference and signing ceremony for the demonstration application of the artificial intelligence power production model jointly hosted by Huairou National Laboratory and Southern Power Grid Corporation was held in Beijing. On this day, the world's first native power professional big model that innovates from the bottom architecture source - NWHR Power Production Big Model AI EPS V1.0- was officially unveiled, equipping the new power system with a "smart brain".  

This large model has the ability to complete power grid state inference within 35 milliseconds and has been successfully demonstrated and operated in Dali, Yunnan. It generates over 120 million kilowatt hours of green electricity annually and reduces carbon dioxide emissions by about 60000 tons annually, providing a system level solution for the consumption of new energy and helping to achieve the "dual carbon" goal.

The architecture revolution from "adaptation" to "native"

Unlike the "adaptation" path of general large-scale models, the NWHR power production large-scale model is the world's first native large-scale model that innovates from the underlying architecture and is designed specifically for the characteristics of the power system. The R&D team focused on the core requirements of China's power system, based on the unique operating characteristics and real-time changing status of the power grid, and created the NWHR power production big model architecture, which for the first time achieved deep integration and coordination of the power system in terms of measurement data, physical laws, and operating procedures.  

This achievement is a typical practice of Huairou National Laboratory and Southern Power Grid Corporation in implementing the new national system and conducting organized scientific research. In 2023, Huairou National Laboratory, together with China Southern Power Grid Corporation, will establish the first research and development institution jointly established by central enterprises and national laboratories in China, and set up a research and development team for the power model. Led by Li Lizheng, an academician of the CAE Member, the leader of the scientific research team of Huairou National Laboratory, and the honorary chairman of the expert committee of China Southern Power Grid Corporation, the team will gather all forces of universities, enterprises, and national scientific research institutions to carry out the "battle of conferences" in scientific research.  

Breakthrough in three core competencies

Real time breakthrough: The R&D team has designed a spatiotemporal graph state space model, which can quickly deduce the state of the power grid in only about 35 milliseconds, achieving real-time perception of the "nerve endings" of the power grid and shortening the generation time of scheduling plans from hours to minutes.  

Precision breakthrough: Achieving a "deep understanding" and "effective application" of the physical laws of the power grid by AI, translating them into a "function set" that can be understood and executed by AI, and deeply embedding them into the model framework, enabling calculations, analysis, and decision-making to automatically meet the requirements of safe operation of the power grid, significantly improving numerical calculation accuracy. With only a few training sessions, it is possible to maintain the same level of accuracy in billions of unseen operating conditions.  

Intelligent breakthrough: The R&D team has designed a heterogeneous spatial semantic alignment unit, establishing a core center for accurately understanding the technical guidelines of the power system, greatly improving the accuracy of complex risk assessment of the power grid, and effectively supporting the safe, efficient, and autonomous operation of the power grid under the high proportion of new energy access.

The demonstration application has achieved significant results

The NWHR power production model released this time has been successfully tested in the Dali Power Supply Bureau of Southern Power Grid in Yunnan. Actual operational data shows that the system has shortened the generation time of scheduling plans from the hourly level to the minute level, increased the average consumption rate of new energy by about 25%, increased green electricity by over 120 million kilowatt hours annually, and reduced carbon dioxide emissions by about 60000 tons annually.  

In Xinjiang, through data fusion and machine learning, the annual increase in green electricity is 3.19 billion kilowatt hours, equivalent to the annual electricity consumption of a prefecture level city. Through data sharing and intelligent analysis, optimize 15 grid connection processes and improve grid connection efficiency by 30%.  

Future deployment and policy support

During the 15th Five Year Plan period, the R&D team will continue to iterate and optimize the NWHR large model according to the technological evolution route from "city level provincial level regional level" power grid, in order to achieve coordinated scheduling and optimization of larger scale and more types of energy resources, and fully guarantee that every kilowatt hour of green electricity can be "generated, connected, and used well".  

The National Energy Administration has promoted the commencement of more than 30 key new energy supporting power grid projects and established a project list to accelerate the preliminary and construction work. The policy emphasizes optimizing the network connection process, simplifying the review process, further improving efficiency, providing system level solutions for the consumption of new energy, and laying a key technological foundation for the future billion level distributed new energy to achieve "plug and play" and "unconditional" friendly access. Keywords: New Infrastructure News, Infrastructure Engineering Information

The release of this breakthrough achievement marks significant progress in promoting the deep integration of energy and electricity with artificial intelligence, empowering the construction of new power systems, and providing important support for the safe and efficient integration of high proportion new energy into the grid. It will effectively promote the construction of China's new power system and the implementation of the strategy of building an energy strong country.Editor/Yang Meiling