Local-Global methods for Generalised Solar Irradiance Forecasting

This repository contains the implementations for the methods and supplementary material for paper "Local-Global methods for Generalised Solar Irradiance Forecasting"

Paper

@article{,
  title = {Local-Global methods for Generalised Solar Irradiance Forecasting},
  author = {Cargan, T and Landa-Silva, D and Triguero, I},
  year = {2023},
}

All the model code can be found in the folder /expers:

Model	File
CNN	simple_cnn.py
DNN	simple_dnn.py
LSTM	simple_lstm.py
Transformer	simple_transfomer.py

Results

The raw results from the experiments outlined in the paper can be found the /results folder.

Running Experiments

Installing

The easiest way is to clone the repo and run flowing commands to install everything needed to run the experiments:

pip install --upgrade "jax[cuda12_pip]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
pip install chemise @ git+https://github.com/TimCargan/chemise.git@master
pip install hemera @ git+https://github.com/TimCargan/hemera.git@master
pip install -e .

Then to run some models

python ./expers/simple_conv.py simple_conv.py --batch_size 64 --learning_rate 3e-4 --num_epochs 20

Data

The satellite image data can be downloaded from EUMETSAT using our downloader scripts: EUMETSAT-downloader All irradiance and weather data is available on request, we are working to have permission to upload it to hugging face.