Wind Speed Time Series Prediction with Deep Learning and Data Augmentation

Flores A.; Tito-Chura H.; Yana-Mamani V. · Lecture Notes in Networks and Systems (2022) · DOI: 10.1007/978-3-030-82193-7_22

Resumen

This paper presents a hybrid model based on recurrent neural networks known as Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) for the prediction of daily wind speed time series in the Moquegua region of Peru. The proposal model called GRU LSTM GRU LSTM + DA consists of an architecture of 4 hybrid sequential layers and works on a normalization scale of −1, +1 instead of the traditional 0, +1 scale, it also uses data augmentation (DA) to improve the process training and prediction results of the model. The results of the proposal are compared with 4 benchmark models (GRU GRU GRU GRU, LSTM LSTM LSTM LSTM, GRU LSTM and GRU LSTM GRU LSTM), showing that the proposal in terms of RMSE, RRMSE and MAPE by far exceeds the benchmark models. In the same way, the results achieved in terms of RMSE are compared with the results of related work, showing the superiority of the proposal model in this study.

Instituciones: Universidad Nacional de Moquegua