A non-intrusive reduced order model using deep l earning for realistic wind data generation for small unmanned aerial systems in urban spaces

Code and tools used for AIP Advances Manuscript " A non-intrusive reduced order model using deep learning for realistic wind data generation for small unmanned aerial systems in urban spaces "

How to use

Input

Please use the input.yaml to change the parameters to run the code

Params

The parameters are divided between converting .nc to .mat file, data, runtime, detrending, training, misc parameters

Flags

Flags for convertion, runtime, plotting, gif generation, backend plotting are all included in the file.

Files

• init.py : File with all the initialisation
• imp_lib.py : File to import the necessary libraries
• functions.py : File with all the functions used
• main.py : File to train and run the model
• misc_plotting.py: File with functions included to plot the data

Packages used

• tensorflow
• sklearn
• h5py
• tqdm
• numpy
• netcdf4
• scipy
• hurst
• matplotlib
• time
• cmocean

Tested with

• Python version 3.6
• Python version 3.8