Organisation Name

IBERDROLA

Industry

Utilities (water, gas, electricity)

Organisation Website

https://www.iberdrola.com/

Country

Spain

Sustainable Development Goals (SDGs)

General Description of the AI tool

MeteoFlow is an advanced AI-powered forecasting tool to optimize renewable energy production. It combines numerical weather prediction (NWP) models using evolutionary algorithms for input selection, deep learning networks to extract spatial information, and machine learning (ML) to extract temporal information. This tool generates highly accurate forecasts for wind, solar, hydroelectric, and offshore energy installations, enabling teams to plan maintenance, manage energy dispatch, and reduce market penalties due to inaccurate forecasts.

Relevant Research and Publications

Prieto-Godino et al. (2025) – Predicting weather-related power outages with deep learning ensembles
This paper introduced ensemble deep learning models to forecast outages caused by extreme weather. It directly influenced MeteoFlow’s extreme weather module, helping anticipate operational risks and improve grid resilience.

Salcedo-Sanz et al. (2009) – Hybridizing mesoscale models with neural networks
A foundational work that demonstrated how to combine physical weather models with neural networks. It shaped MeteoFlow’s hybrid forecasting architecture, enabling more accurate short-term wind speed predictions.

Dorado-Moreno et al. (2020) – Multi-task learning for wind power ramp events
This study applied deep learning to predict multiple aspects of wind ramp events. It enhanced MeteoFlow’s ramp event forecasting, supporting better decision-making in energy dispatch and grid stability.

Salcedo-Sanz et al. (2018) – Feature selection in ML systems for renewable energy
Provided methods to identify the most relevant input variables for prediction models. This improved MeteoFlow’s model efficiency and accuracy, especially when scaling across diverse renewable assets.

Vega-Bayo et al. (2024) – Generative data augmentation for extreme wind prediction
Introduced techniques to improve model generalization using synthetic data. It strengthened MeteoFlow’s forecasting of rare and high-impact events, such as storms and gusts, where real data is limited.

Needs