Wind speed estimation for missing wind data with three different backpropagation algorithms

Luy M., Saray U.

Energy Education Science and Technology Part A: Energy Science and Research, cilt.30, sa.1, ss.45-54, 2012 (Scopus) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 30 Sayı: 1
  • Basım Tarihi: 2012
  • Dergi Adı: Energy Education Science and Technology Part A: Energy Science and Research
  • Derginin Tarandığı İndeksler: Scopus
  • Sayfa Sayıları: ss.45-54
  • Anahtar Kelimeler: Backpropagation, Gradient Descent, Levenberg-Marquardt, Neural networks, Resilient, Wind speed prediction
  • Sivas Cumhuriyet Üniversitesi Adresli: Hayır

Özet

In this study, wind data acquired from Tokat province located in the Central Black Sea section of the Black Sea region of Turkey were used to estimate wind speed by using artificial neural networks (ANN). A 3-layer feedback network was designed for wind speed modeling with MATLAB Neural Network Toolbox. Data used were acquired from State Meteorological station taken from a height of 10 meters. By using daily average wind speed data of Tokat province in 2010, ANN feedback network algorithms were used in order to recover any missing wind speed data. ANN feedback model Levenberg - Marquardt (LM) learning algorithm, the gradient - Descent (GD) learning algorithm and Resilient (RPROP) learning algorithm were used for randomly selected three data for each month and root mean square error (RMSE) and mean square error (MSE) values were calculated. © Sila Science.