IT Home on October 17 reported that, according to the Hong Kong Wen Wei Po, the “Fengqing” artificial intelligence global short and medium-term combined weather forecast system V1.0 has recently passed the commercial evaluation and is officially put into operation in national meteorological services. This system was developed using the AI4S supercomputing cluster with 100,000 cores at the core node of the National Supercomputing Tianjin Center for full process training and iteration. This is China’s first business-level meteorological AI combined forecast model that achieves complete autonomy and control over data, models, and computing power. It marks an important progress in Tianjin’s supercomputing capabilities in supporting the development of meteorological artificial intelligence technology in China.
According to the introduction, the “Fengqing” system relies on the computing infrastructure of the National Supercomputing Center in Tianjin. It uses a large-scale, high-parallel AI4S scientific computing cluster to complete the training of massive meteorological data, model optimization, and accuracy iteration. It is reported that the core computing power, underlying algorithms, and data sources of this system are all domestic, enabling the establishment of an independent and controllable meteorological intelligent forecasting technology system. This fills the technical gaps in China’s global intermediate ensemble intelligent forecasting field.
According to IT Home, the “Fengqing” system is developed by the National Meteorological Center in collaboration with multiple universities and research institutions. It integrates the Visual Transformer architecture with meteorological physics constraint algorithms, enabling it to capture the uncertainty of atmospheric movements and enhancing the probability forecasting capabilities for disastrous weather events such as heavy rain, typhoons, and strong convection.The global effective forecasting horizon of this system can reach 11 days. Many core forecasting indicators, probability forecasting accuracy, and discrete degree error control capabilities are superior to the latest artificial intelligence ensemble forecasting system of the European Centre for Medium-Range Weather Forecasts (ECMWF).。
Compared to traditional numerical weather prediction models, this system can utilize the computing power of Tianjin supercomputing center to complete the simulation of global meteorological elements within a few minutes. It balances forecasting speed, accuracy, and stability, enhancing the ability to predict extreme weather conditions and medium- to long-term weather trends.
After the system is put into operational use, it will serve various fields such as national meteorological disaster prevention and mitigation, transportation, energy scheduling, ecological protection, and people's livelihood support. The subsequent research and development team will continue to optimize the accuracy of the model and the forecasting timeliness, and further improve the domestic intelligent meteorological forecasting system.
In recent years, Tianjin has been continuously promoting the integrated development of supercomputing, artificial intelligence, and meteorological science and technology. Leveraging the computing power of the National Supercomputing Tianjin Center, it provides support for scientific research in fields such as atmospheric science, Earth system simulation, and intelligent weather forecasting. The “Fengqing” system has been officially put into operation. It is a achievement of Tianjin’s supercomputing efforts in supporting national-level meteorological science and technology innovation, and it will also provide support for China’s independent innovation in meteorological science and technology as well as the development of global intelligent weather forecasting technologies.