BBL-WPP

This code solves short term wind power prediction problem using multiple sampling resolution data. For more details, please see our paper A Bilateral Branch Learning Paradigm for Short Term Wind Power Prediction with Data of Multiple Sampling Resolutions which has been accepted at JCLP. If this code is useful for your work, please cite our paper:

@article{liu2022bilateral,
  title={A bilateral branch learning paradigm for short term wind power prediction with data of multiple sampling resolutions},
  author={Liu, Hong and Zhang, Zijun},
  journal={Journal of Cleaner Production},
  pages={134977},
  year={2022},
  publisher={Elsevier}
}

Dependencies

• python = 3.6.3
• NumPy
• Scipy
• PyTorch = 1.7
• xgboost
• scikit-learn
• optuna
• pyearth

Data

Since the github cannot process the data larger than 4GB, please download the files 'data_wf1_7s' from this google drive link and 'data_wf1_10min' from this google drive link, and place them to the folder 'wf1'.

Quick Start

For using LASSO as the feature selection and prediction model, and single sampling resolution data:

python code/run.py --fs lasso --fp lasso

For using random forest as the feature selection and decision tree as prediction model, and multiple sampling resolution data:

python code/run.py --fs rf --fp dt --need_fh