Generalizable wind power estimation from historic meteorological data by advanced artificial neural networks


Insel M. A., Ozturk B., Yucel O., Sadikoglu H.

Renewable Energy, cilt.246, 2025 (SCI-Expanded) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 246
  • Basım Tarihi: 2025
  • Doi Numarası: 10.1016/j.renene.2025.122995
  • Dergi Adı: Renewable Energy
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, PASCAL, Aerospace Database, Aquatic Science & Fisheries Abstracts (ASFA), CAB Abstracts, Communication Abstracts, Compendex, Environment Index, Geobase, Greenfile, Index Islamicus, INSPEC, Pollution Abstracts, Public Affairs Index, Veterinary Science Database, DIALNET, Civil Engineering Abstracts
  • Anahtar Kelimeler: Artificial intelligence, Artificial neural networks, Estimation, Forecasting, Wind power
  • Yıldız Teknik Üniversitesi Adresli: Evet

Özet

The unpredictability and instability caused by the widespread use of wind power (WP) pose significant challenges for ensuring the secure and consistent functioning of the electricity grid. Accurate estimation of the WP output of wind farms (WFs) will effectively mitigate these adverse effects. Thus, in this study, a comprehensive analysis of the most prevalent artificial neural network (ANN) models is conducted using the meteorological and power data of four distinct WFs situated to the west of Türkiye. The performances of these ANN models are examined by using the expanding window validation approach at three well-established WFs. The optimal ANN approach is determined based on a thorough evaluation of both performance and computational load. Then, the most effective ANN model is utilized with slight modifications to obtain the generalizable model. The generalizable model performed remarkably, obtaining high performance in both estimating WP output of different well-established WFs (R2 = 0.9294, RMSE = 7.562) and a newly-established WF (R2 = 0.9633, RMSE = 5.823). These results indicate that the model can successfully be employed in estimation of WP output of any WF within the region, and the methodology presented here can easily be applied globally, enabling anyone, including third parties like government agencies, to estimate WP output of any WF.