Wind Power Forecasting by using Advanced Artificial Neural Networks

İnsel M. A., Yücel Ö., Sadıkoğlu H.

7th International Conference on Engineering Sciences, Ankara, Turkey, 9 - 10 February 2024

  • Publication Type: Conference Paper / Unpublished
  • City: Ankara
  • Country: Turkey
  • Yıldız Technical University Affiliated: Yes


Wind energy is essential for achieving a sustainable future since it offers a clean and renewable power source. It plays a significant role in combating climate change, decreasing reliance on limited fossil fuels, and promoting an environmentally friendly and robust energy infrastructure. However, the intrinsic variability of wind speeds and directions makes wind power unpredictable, which presents issues for energy planners. Fluctuations in wind can result in intermittent energy generation and significant disturbances in the stability of the power grid. Thus, accurate short-term prediction of wind power is crucial for optimizing energy grid operations and enabling efficient management of supply and demand. In this study, six-year weather and power data for a wind power facility located in İzmir/Türkiye was collected. 90% of the data was utilized as training set, and 10% was retained as test set. Advanced artificial neural network models (LSTM, NARX, and Elman) were utilized to estimate wind power from the former 8 hours weather data. All models have performed sufficiently well, while the LSTM model performed the best with a mean absolute error of 0.0495 and an R2 value of 0.9311. The proposed models here can be adapted to any wind power facility around the globe, enabling successful optimization of wind power’s contribution to the energy grid.