The Use of Machine Learning and Performance Concept to Monitor and Predict Wind Power Output

Abstract

Monitoring and predicting wind power output more precisely can be very beneficial for an increasingly competitive Wind Power industry. Although many advances have been made throughout the last decades, the production forecast is still based mainly on the manufacturing power curve and wind speed. Even though this approach is very useful, especially during the design phase, it does not consider other factors that affect production, such as topography, weather conditions, and wind features. A more precise prediction model that is able to recognize production fluctuation and is tailored using current operational data is proposed in this paper. The model analyzes the performance through Meteorological Mast Data (Met Mast Data) and then uses it as an input to monitor and predict power output. As a result, the model proposed achieves high accuracy and can be key to understanding the wind turbine asset's behavior throughout its lifespan, assisting operators in decision making to increase overall power production.