The directory contains main steps for wind power forecasting applications and is a part of a Master thesis at KU Leuven, 2022. The dataset of interest is La Haute Borne wind farm in France. The repository details all preprocessing and forecasting steps - data analysis, imputation, feature engineering and modeling. The notebooks are used to compare various forecasting methodologies - Persistence, ARIMA, LSTM or ensembles (SVR or Decision trees).

• solver - directory containing all required functions used in the notebooks
• notebooks - directory containing all workflow for wind power forecasting
• script - directory with external files important for forecasting

notebooks
   |-- 1_data_analysis.ipynb
   |-- 2_outliers_removal.ipynb
   |-- 3_feature_engineering.ipynb
   |-- 4_imputation.ipynb
   |-- 5_ARIMA_forecasting.ipynb
   |-- 5_COND-LSTM_forecasting.ipynb
   |-- 5_LSTM_forecasting.ipynb
   |-- 5_Persistence_forecasting.ipynb
   |-- 5_ensembles_forecasting.ipynb
script
   |-- get_hyperparameters_LSTM.py
   |-- get_hyperparameters_ensemble.py
solver
   |-- arima.py
   |-- ensembles.py
   |-- lstm.py
   |-- persistence.py
   |-- processing.py
wind_power37.yml
.gitignore
LICENSE
README.md
setup.py

Use the conda environment file wind_power37.yml to install the required wind_power37 environment and its modules.

conda env create -f wind_power37.yml

Activate wind_power37 conda environment:

conda activate wind_power37

From the root directory, create solver package which can be accessed from all the notebooks:

conda develop .