Estimation of wind conditions in a wind park by using machine learning

In the wind industry, optimizing the power output of the wind park (or single turbines) in different atmospheric conditions is the key factor to achieve better performance and therefore, energy yield. For that purpose there is a need to monitor the performance of wind turbines continuously by detecting the abnormal behavior of the wind turbines at different wind conditions. The use of machine learning in predicting the different wind conditions at different locations enables to reduce the measuring costs and maintenance costs. In this work, different machine learning models (linear regression, random forest, support vector regressor, and neural network) were implemented on different types of data (SCADA, met mast and weather model data) in order to predict the wind speeds at different locations in a wind park. The thesis project will focus on the best possible use of machine learning models in predicting the wind conditions in a wind park and on comparing the different model results.