Mapping Knowledge Structures in Deep Learning For Wind Energy Forecasting: A Technical and Bibliometric Review
Keywords:
wind-energy forecasting; deep learning; wind-speed forecasting; wind-power forecasting; bibliometric analysis; knowledge mapping.Abstract
Accurate wind-energy forecasting is essential for grid reliability and renewable-energy integration. This review aims to synthesize deep-learning methods for wind-speed and wind-power forecasting while mapping the field’s knowledge structure. A technical review was integrated with bibliometric analysis of 324 Scopus-indexed publications from 2014–2024 using OpenRefine, BiblioMagika, VOSviewer, and Biblioshiny for data cleaning and co-authorship, co-citation, and keyword analyses. Results show that LSTM, Bi-LSTM, decomposition-assisted hybrids, and attention-based models dominate the literature; China leads in attributed publication output, while several lower-volume countries achieve higher citation rates per paper. Emerging themes include Transformers, federated learning, explainable artificial intelligence, and physics-informed forecasting. Persistent limitations involve data standardization, reproducibility, chronological evaluation, uncertainty calibration, and cross-site validation. The review concludes that progress depends on transparent benchmarks, robust probabilistic evaluation, and deployment-oriented reporting rather than model complexity alone.
