Deep Learning-Based Wind Speed Prediction for Optimized Wind Turbine Operation

Abstract

Wind energy has been a promising source of clean energy that does not negatively affect our environment. Because of the fluctuations in wind speed, it is crucial to predict its values for wind turbines to have the maximum effective power output. This project aims to develop a way for short-term wind speed prediction based on deep learning technologies, such as CNN, LSTM, RNN, and GRU models, alone and in combination. Through iterative experimentation and evaluation, we develop ten final models and assess their performance based on Mean Squared Error (MSE), score, and computational efficiency. Our findings reveal that the GRU model achieves the highest performance with a MSE of 0.00238 m/s and R2 score of 0.8796. Additionaly, the similarly structured LSTM model demonstrates superior computational efficiency along with high R2 value, outperforming GRU model. By examining the performance of multiple deep learning architectures, the project seeks to identify the most suitable approach for wind speed prediction, thereby facilitating more efficient and sustainable utilization of wind resources for power generation.