DEEP RECURRENT NEURAL NETWORK MODELS FOR FORECASTING SHORT-TERM WIND SPEED

Abstract

Wind speed/power has received increasing attention worldwide due to its renewable nature and environmental friendliness. Wind power capacity is rapidly increasing with the global installed, and the wind industry is growing into a large-scale business. We are looking for wind speed prediction to use wind power better. In this research, Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Simple Recurrent Neural Network (Simple RNN), and LSTM-GRU in the subset of artificial intelligence algorithms are used to predict wind speed. The data used in this study are related to the 10-minute wind speed data. In the first study on this dataset, we obtained significant results. To compare the deep recurrent models created, we implement four neural network models: Stacked Auto Encoder, Denoising Auto Encoder, Stacked Denoising Auto Encoder, and Feed-Forward presented in the research of others on this dataset. According to the RMSE statistical index, the LSTM network is worth 0.0222 for a short time and performs better than others in this dataset.