Forecasting Wind Energy Production: Analysis of Meteorological and Temporal Variables Using Optimized Regression Modeling

In order to meet the power demand in energy generation systems, it is of great importance to predict the production potential in advance. In energy systems such as wind energy, whose generation potential is directly dependent on atmospheric conditions, it is possible to make short or medium-term forecasts based on meteorological data. In addition, the safe operation, efficient planning, operation, and control of wind turbines are also facilitated by predictability. The proposed study aims to improve the accuracy of wind energy forecasts through regression models supported by an optimization algorithm. In the proposed study, the potential energy production of a wind turbine is modeled using genetic algorithm-optimized artificial neural networks (GA-ANN) and gradient boosting (GA-GB) techniques. In the proposed study, one year of energy production data for Nordex N117/3600 model wind turbines is used in combination with NASA's hourly meteorological conditions data for the same period. The success of the applied techniques is analyzed in terms of performance metrics. The R2=0.971 value of the GA-GB model shows the superiority of the model over both the ANN considered in the proposed study and the deep learning methods performed with the same dataset in the existing studies.