Breaking the noise barrier: A filtered framework for decomposition-based renewable energy forecasting
Proceedings of the Institution of Mechanical Engineers, Part A: Journal of Power and Energy, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Basım Tarihi: 2026
- Doi Numarası: 10.1177/09576509261470907
- Dergi Adı: Proceedings of the Institution of Mechanical Engineers, Part A: Journal of Power and Energy
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Applied Science & Technology Source, Compendex, Greenfile, INSPEC, Engineering Source (EBSCO), Materials Science & Engineering Collection (ProQuest), Pharma Collection (ProQuest), Technology Collection (ProQuest)
- Anahtar Kelimeler: breaking the noise barrier, decomposition-based forecasting, hybrid models, noise-filtered, renewable energy forecasting
- Bilecik Şeyh Edebali Üniversitesi Adresli: Evet
Özet
This study introduces a novel framework, Breaking the Noise Barrier, for enhancing renewable energy forecasting through decomposition-based models. By systematically excluding the first Intrinsic Mode Function (IMF) component, which may contain high-frequency noise depending on the decomposition method, a Noise-Filtered IMF (NF-IMF) approach is proposed to improve prediction accuracy. Hybrid models integrating five signal decomposition techniques, such as Empirical Mode Decomposition (EMD), Ensemble EMD (EEMD), Complete Ensemble EMD with Adaptive Noise (CEEMDAN), Variational Mode Decomposition (VMD), and Swarm Decomposition (SWD), with Long Short-Term Memory (LSTM) network model were developed and evaluated across three datasets: wind, solar, and wave energy. Experimental results demonstrate that the proposed NF-IMF framework outperformed Baseline IMF counterparts in 12 out of 15 test cases, achieving high predictive performance with R2 values reaching 0.986 on wind data and 0.959 on wave data. These findings underscore the effectiveness of selective noise reduction in decomposition-based forecasting, particularly for wind and solar energy applications. Beyond comparing the proposed NF-IMF approach with its Baseline IMF counterparts, comparative experiments are also conducted against direct LSTM, Gated Recurrent Unit (GRU), and Least Squares Boosting (LSBoost) models applied to raw signals, thereby providing a more comprehensive evaluation relative to both conventional machine learning and standard deep learning baseline models.