A Comparative Study on RNN-LSTM-GRU Ensemble Model for Short-Term Wind Speed Prediction
Year: 2026
Type: article
Abstract: Precise forecasting of wind speed is essential for the efficient operation of wind power systems and the optimal planning of power grid generation. This study seeks to simplify the process of predicting short-term wind speeds, a crucial step for planning and conducting feasibility assessments of wind farms. In this paper, a short-term wind speed forecasting model based on recurrent neural networks (RNN), long short-term memory (LSTM), gated recurrent units (GRU), and hybrid deep learning and ensemble models is proposed, utilizing data from wind speed prediction in Turkey obtained by SCADA systems. These models use a range of activation functions, including hyperbolic tangent (tanh), inverse hyperbolic tangent (archtanh), complex hyperbolic tangent (hardtanh), exponential linear unit (ELU), Gaussian error linear unit (GELU), parametric exponential linear unit (PELU), scaled exponential linear unit (SELU), rectified linear unit (ReLU), leaky rectified linear unit (Leaky ReLU), parametric rectified linear unit (PReLU), randomized leaky rectified linear unit (RReLU), and sigmoid-weighted linear unit (SWISH). These models were trained in time series data, with 75% of the data allocated for training and 25% for testing. The performance of deep learning, hybrid deep learning, and ensemble learning models is measured using a variety of metrics such as mean square error, normalized mean square error, root mean square error, normalized root mean square error, relative root mean square error, mean absolute error, symmetric mean absolute percentage error, mean bias error, and the coefficient of determination. The outcomes show that the ensemble learning models integrating RNN, LSTM, and GRU, utilizing RReLU and PReLU activation functions, are best at predicting wind speed, with a coefficient of determination ranging from 0.9684 to 0.9688.
Authors: Sathyaraj J, Sankardoss V
Institution: Vellore Institute of Technology University
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,075
OpenAlex ID: W7130535789