International Journal of Academic Engineering Research (IJAER)   Year: 2023 | Volume: 7 | Issue: 4 | Page No.: 1-3 Wind Speed Forecast Tool for Wind Power Plants in Madagascar Download PDF ANDRIAMAHITASOA Bernard Andriamparany, RAKOTOARIMANANA Liva Graffin, MANDIMBY Junior Zoë Jean Tigana, RAFANJANIRINA Eulalie Odilette, RANDRIAMANANTANY Zely Arivelo, RAKOTOMALALA Minoson Abstract: Since wind energy is an exploitable source, its exploitation requires knowledge about reserve of load capacities. Upcoming load capacities can be available thanks to the data from measuring stations in the region of Antananarivo. This paper aims at presenting non-linear short-term forecasting techniques of wind speed. It proposes new design model based on Artificial Neural Networks (ANNs) techniques for wind speed forecast. For the wind speed forecast, Classical neural networks, Bayesian neural networks and Gaussian process models is used. Each model runs with two year-wind speed data training and one-year data for testing. The Gaussian process model has shown the best performance and it can predict 3-day-horizon wind speed during one year with less than 29% accuracy. This model can be a useful tool to develop wind power plants, particularly in Madagascar.