A Multi-Scale Model Based on the Long Short-Term Memory for Day Ahead Hourly Wind Speed Forecasting

•A Multi-scale model is developed for hourly wind speed forecasting 24 steps ahead.•The model uses dense layers to build coarser timescales and LSTMs to process them.•An experiment shows the long term dependencies learning capabilities of our model.•Experiments are on real wind speed datasets from Chile, with data from 2 to 8 years.•Performance is better when using information from multiple timescales explicitly.