Budget-constrained Collaborative Renewable Energy Forecasting Market

This repository contains the core implementations and datasets used in our paper:

"Budget-constrained Collaborative Renewable Energy Forecasting Market"
Carla Gonçalves, Ricardo J. Bessa, Tiago Teixeira, João Vinagre
Published in IEEE Transactions on Sustainable Energy, 2025

Table of Contents

1. Introduction
2. Project structure
3. Running the code
4. License
5. Citation and contact

1. Introduction

This project presents a budget-constrained collaborative forecasting market, where data owners can monetize their features, and buyers can optimize forecasting performance while staying within a budget.

1.1. What is the main purpose of this project?

This project implements a collaborative forecasting market that allows data sellers to monetize their data while data buyers purchase relevant features under a given budget constraint. It uses Spline LASSO regression to improve the accuracy of renewable energy forecasting (e.g., wind or solar).

• Collaborative forecasting model: Spline LASSO regression
• Bid-based data market where:
  • Data sellers set prices for their features.
  • Data buyers specify a budget or price related to forecast accuracy improvements.
• Incentive Mechanism: Estimates the more accurate LASSO B-spline model within the budget

1.2. Which forecasting tasks does this code target?

We primarily target wind power forecasting, though the underlying models can be adapted to other use cases (solar, load forecasting, etc.) if you have suitable time series data.

1.3. Where can I find the related paper?

🔗 Read the full paper: ArXiv or IEEE

---

2. Repository Structure

The codes provided are organized as follows:

├── data/                                          # Datasets used in experiments
├── src/                                           # Main implementation files
│   ├── proposal.py                                # Algorithmic solution 
│   └── custom_pipeline.py                         # Feature selection filters
├── examples/                                      
│   ├── 1-a-generate-basic-synthetic-data.py       # Basic synthetic setup 
│   ├── 1-b-run-markets-basic-synthetic-data.py    # Run the proposal
│   ├── 2-advanced-synthetic-run-data-markets.py   # Advanced synthetic setup
│   ├── 3-a-gefcom2014-create-data-folders.py      # Wind power setup (GEFCom 2014) 
│   ├── 3-b-gefcom2014-compare-models.py           
│   ├── 3-b-gefcom2014-draw-plots.py               
│   └── 3-c-gefcom2014-run-data-market.py          
├── LICENSE                 # License file
├── requirements.txt        # Dependencies
└── README.md               # This file

2.1. Are the datasets fully included in the repository?

• Synthetic Datasets: Example synthetic data is stored in the data/ folder, generated via scripts in the examples/ directory.
• Real-world Datasets: We provide scripts to process the GEFCom2014 wind power dataset. You need to manually download the original GEFCom2014 (link here).

---

3. Running the Code

Clone the repository

git clone https://github.com/INESCTEC/budget-constrained-collaborative-forecasting-market.git
cd budget-constrained-collaborative-forecasting-market

Install dependencies

pip install -r requirements.txt

Dependencies include:

• numpy
• pandas
• scikit-learn
• scikit-optimize
• scipy
• sympy
• plotnine

---

The results can be obtained by running the scripts in the 'examples' folder, e.g.:

Generate the basic synthetic dataset

python examples/1-a-generate-basic-synthetic-data.py

Run the market and obtain the results in Table I of the paper

A folder results/ will be created to save a .csv with Table I results.

python examples/1-b-run-markets-basic-synthetic-data.py

5. License

This project is licensed under the AGPL v3 license - see the LICENSE file for details.

---

**Disclaimer: ** This code contains parts adapted from the original implementation provided in Yu, G., Fu, H., & Liu, Y. (2022). High-dimensional cost-constrained regression via nonconvex optimization. Technometrics, 64(1), 52-64. which was published and released as open-source by the authors. This version contains modifications, improvements, and deviations from the original code.

---

6. Citation and contact

If you use this code, please cite our work:

@article{goncalves2025budget,
  title={Budget-constrained Collaborative Renewable Energy Forecasting Market},
  author={Gon{\c{c}}alves, Carla and Bessa, Ricardo J and Teixeira, Tiago and Vinagre, Jo{\~a}o},
  journal={IEEE Transactions on Sustainable Energy},
  year={2025},
  publisher={IEEE}
}

---

For questions or collaborations, feel free to reach out:

• Carla Gonçalves - carla.s.goncalves@inesctec.pt
• Ricardo J. Bessa - ricardo.j.bessa@inesctec.pt

---