AI weather forecasting enters the energy market as Google targets grid operators with WeatherNext 3
ainews ·
Google has introduced WeatherNext 3, an AI weather forecasting model aimed at the energy industry, offering hourly wind, cloud, and solar forecasts at up to five-kilometre resolution for grid operators and renewable developers. The model, released by Google DeepMind and Google Research on September 3, updates much more frequently than its predecessor and also powers consumer products like Google Search and Maps. It is designed to help predict renewable generation and match it with demand, a need driven by rising solar and wind capacity and electricity demand from data centres. Google is entering an established market occupied by firms like Vaisala and Solcast, but its advantage is distribution through BigQuery, Earth Engine, and Google Cloud. The architecture trains on live observations rather than just simulations, reducing data lag from about seven hours to three or four, though dependence on numerical models is not fully removed. Accuracy claims include improvements of up to 60% against certain baselines, but these are best-case figures without independent third-party validation. Pricing for enterprise access has not been announced, and Google has a commercial stake because its own data centres and renewable agreements depend on accurate weather predictions.
ainews