Paper code for "An adapted convolutional RNN model for spatio-temporal prediction of wind speed extremes in the short-to-medium range for wind energy applications"
Model and training code can be found in \ConvLSTM_PyTorch_master.
Example model forecasts can be found in \example_forecasts.
All figures can be reconstructed in visualisation_notebook.ipynb, except for the forecast visualisations, which require that the models in question have been trained and saved using \ConvLSTM_PyTorch_master/main.py.
All scores were computed with save_scores.py and have been saved in \saved_scores.
Clone repository:
git clone https://github.com/dscheepens/Deep-RNN-for-extreme-wind-speed-prediction.git Wind speed data was obtained from the Copernicus Climate Data Store: https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-pressure-levels?tab=form.
Used were 'U-component of wind' and 'V-component of wind' at pressure levels 1000, 925, 850 and 775 hPa (seperately) with an hourly interval for a 10 year duration and between 3-18.75 longitude, 40-55.75 latitude.
data_loader.py requires the data to be in the .npy format, rather than .grib. This can be achieved with the file grib_to_numpy.py.
preprocessing.py standardises the data and saves it as era5_standardised.npy into the specified data root.
The first row from the top displays the 12 input frames, the second row the succeeding 12 target frames and the following rows the 12 predicted frames of the models. T refers to the index of the frame (in hours), with T=0 denoting the last input frame and T=+12 denoting the final target and prediction frames. The final row shows the averaged forecast of an ensemble of the W-MAE, W-MSE and SERA-trained models.