Predicting the unpredictable: an in-depth study of meteorological data applications in electrical engineering and renewable energy

Abstract

The accelerating global transition toward renewable energy has fundamentally altered the relationship between meteorology and electrical engineering. As power systems incorporate increasing proportions of weather-dependent solar and wind generation, meteorological data have evolved from peripheral inputs into indispensable operational assets for grid planning, dispatch, resilience management, and infrastructure protection. This review article systematically examines recent literature published between 2023 and 2026 to assess the state of the art in meteorological data applications across five interlocking domains: (1) renewable energy resource forecasting for solar photovoltaic and wind power systems; (2) dynamic management of electrical transmission infrastructure; (3) weather-driven electricity demand forecasting and long-range load planning; (4) grid resilience assessment under extreme meteorological events; and (5) climate change adaptation strategies for future power systems. Innovations such as hybrid Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) architectures, AI-enhanced Numerical Weather Prediction (NWP) correction, Dynamic Line Rating (DLR) technology, and probabilistic ensemble forecasting are critically assessed. Five detailed case studies from Minnesota, Texas, Algeria, France, and Florida illustrate the tangible engineering and economic impact of meteorological data integration. The article identifies persistent research gaps and proposes directions for future investigation, particularly concerning sub-hourly localized forecasting, cross-domain data fusion, and climate-resilient power system design. In addition, in response to the rapid development of large-scale artificial intelligence, the review synthesizes the results, analysis, and outlook of large AI foundation models for weather prediction (GraphCast, Pangu-Weather, GenCast, and Aurora) and of time-series large language models as they relate to power-system applications, including quantitative benchmarking against operational numerical weather prediction. It further documents the provenance, open-access links, sample sizes, and time intervals of the principal meteorological datasets underpinning the reviewed studies, presents the typical network structures and end-to-end forecasting frameworks used across application domains, and quantifies the error levels currently achievable by state-of-the-art algorithms on shared benchmark datasets.