Google has launched WeatherNext 3, an AI weather forecasting model designed for the energy sector. The model incorporates new variables such as wind speed at 100 meters, cloud cover, and solar radiation levels to assess power generation risks.
The new version offers improved resolution and latency compared to WeatherNext 2. It provides global updates every hour, with a 5 km resolution for surface variables and 11 km for precipitation, whereas the previous version updated every six hours with a 25 km grid.
The model is trained on meteorological station data and satellite imagery, reducing data latency from 7 hours to 3–4 hours. It is integrated into Google Search, Gemini, Maps, and the Weather API, with corporate access available via BigQuery, Earth Engine, and Cloud Storage.
While Google did not disclose subscription costs, the tool aims to optimize energy procurement and grid management. Accurate forecasting is critical for energy companies to avoid costly last-minute purchases or grid curtailment when generation estimates are inaccurate.
The energy market for weather data includes competitors such as Vaisala, Solcast, DNV, IBM, and Jua. Google also utilizes the model to manage its own clean energy consumption for computing clusters, directly impacting its operational expenses.
