Huawei Cloud AI: Fearless in the Face of Weather Changes, Riding the Wind and Waves

2020-01-08

The importance of meteorological information in coastal cities like Shenzhen, which are prone to multiple typhoons and heavy rainstorms, is self-evident. In order to help various units plan their activities based on weather conditions, reduce casualties and economic losses caused by meteorological disasters, and safeguard the safety of every Shenzhen resident, including offshore workers, the Shenzhen Meteorological Bureau has always been committed to improving the accuracy of weather forecasts. In this battle against "Fengyun Data," the emergence of AI has given humanity new power. AI can use deep learning of a large amount of meteorological data to find clues for prediction and provide strong assistance to manual forecasting. On January 6, the Shenzhen Meteorological Bureau and Huawei Cloud announced a deep cooperation, jointly creating a "Meteorology + Cloud + AI + 5G" comprehensive cooperation model, and promoting breakthrough innovations in accurate weather forecasting for mega-cities and intelligent meteorological services for smart cities. Among them, based on Huawei Cloud AI, efforts are being made to develop more accurate models for predicting catastrophic weather. By learning from massive historical data, the changes and movement patterns of cloud clusters can be calculated, helping to improve the quality of weather forecasts in Shenzhen and promoting the development of intelligent dissemination of warnings for catastrophic weather. In the establishment of the AI model for predicting catastrophic weather, Huawei Cloud plays a crucial role in data analysis and storage. Especially in terms of data analysis, meteorological data has extremely high spatiotemporal resolution and a huge amount of data, which ordinary servers cannot handle in terms of data processing and model training. However, Huawei Cloud's Ascend cluster service can provide powerful AI computing power on demand. Coupled with Huawei Cloud ModelArts, an all-in-one AI development and management platform, the development process of meteorological prediction models can be accelerated, greatly reducing the model training cycle. It is expected that the time required for one model training will be shortened from the original 1-2 weeks to 3 days or even a few hours.

RECENT AI NEWS

RECENT AI TOOLS