Comparative Study of Machine Learning Techniques for Wind Energy Forecasting

Abstract

Wind energy prediction is a crucial and dynamic area within the renewable energy sector. As renewable
energy sources are integrated into existing power grids alongside traditional sources, accurately forecasting
energy production is essential for minimizing operational costs and ensuring safe grid operation. In this
context, we present a comparative and comprehensive study of various machine learning techniques,
including artificial neural networks, support vector regression, random trees, and random forest, examining
the advantages and disadvantages of each method. To verify the efficiency of the considered models, actual
measurements from wind turbines located in France, Turkey, and a dataset from Japan were used. We detail
a step-by-step process encompassing feature engineering, metric selection, model selection, and
hyperparameter tuning. We evaluate the models using specific metrics, providing a summary of optimal
results and discussing. This research aims to bridge the gap between academic studies and practical business
applications, offering detailed architectures and hyperparameters to guide wind energy professionals.