Wind speed forecasting using a deep learning-based hybrid forecasting model

Nowadays, the need for energy is increasing day by day. In order to meet this demand, renewable energy sources that have a more environmentally friendly structure than fossil-based sources come to the fore. Due to the many economic and environmental advantages that wind energy, which is one of the leading energy sources, has in recent years, researchers have been paying great attention to it. In particular, the most important input of wind energy cycle systems is wind speed. Due to the variable nature of the wind speed, problems are encountered with its inclusion in interconnection systems. In solving these problems, intensive studies are being carried out on accurate and reliable forecasting of wind speed. With the help of these studies, successful, efficient energy systems and optimization of these sites can be achieved by using the wind speed estimated with high accuracy. In this thesis, a hybrid model for wind speed estimation with deep learning methods is proposed. The proposed model consists of Convolutional Neural Networks (CNN) and Gated Recurrent Unit (GRU) models. The wind speed data in this thesis study were obtained from the measurement station located at the Faculty of Engineering and Architecture campus of Tokat Gaziosmanpasa University. A 3-year (2018-2020) data set has been created with December intervals of 1 hour through the sensors located at the station. The generated data set was subjected to data preprocessing by normalization process before being used. Then, the data were divided into weekly, monthly and yearly for 3 different case studies. The accuracy and reliability of the proposed model were tested by performance criteria (MAPE, R2, RMSE). In order to measure the success of the model, a comparison was made with 5 different deep learning methods (CNN-LSTM, CNN-RNN, LSTM-GRU, LSTM, GRU). It has been observed that the CNN-GRU hybrid model, which was used for the first time in the field of wind speed forecasting, achieved a high percentage of success as a result of comparisons made.