SocioDiff Codebase

May 12, 2025 · View on GitHub

This repository contains the source code for our paper:

SocioDiff: A Socio-aware Diffusion Model for Residential Load Data Generation
(Under review)

SocioDiff is a conditional diffusion model that generates realistic electricity consumption time series conditioned on structured socio-demographic information. It is specifically designed to address data scarcity and representational bias in underrepresented communities, and supports fairness-aware data generation for smart grid research.

📌 Key Features

  • Conditional diffusion framework integrating social attributes
  • Two-step generation mechanism: global trend + household-specific refinement
  • Fairness-aware adversarial training for disadvantaged groups
  • Compatible with real-world load datasets (e.g., Irish CER dataset)

📁 Repository Structure


├── src/                     # Core model components and training pipeline
│   ├── models/              # Diffusion and denoising networks (4SD)
│   ├── utils/               # Preprocessing, metrics, and functions
│   └── train.py             # Training entry point
├── configs/                 # Training configuration files
├── scripts/                 # Shell scripts for running experiments
├── requirements.txt         # Required Python packages
└── README.md

🧪 Getting Started

1. Clone the repository

git clone https://github.com/Intelligame/SocialDiff_code.git
cd SocialDiff_code

2. Create environment and install dependencies

conda create -n sociodiff python=3.10
conda activate sociodiff
pip install -r requirements.txt

3. Prepare dataset

  • The synthetic dataset used for evaluation can be found at 🔗 https://github.com/Intelligame/SocialDiff
  • To use other datasets, ensure they are formatted as .csv files with socio-demographic fields and normalized load sequences.

4. Run training

python train.py --config configs/sociodiff.yaml

📊 Evaluation

We provide evaluation metrics including:

  • Maximum Mean Discrepancy (MMD)
  • Context-FID Score
  • Discriminative Score
  • Downstream forecasting performance (MSE, MAE, R²)

Pre-trained models and evaluation scripts will be released upon acceptance.

🔓 Release Policy

This repository is currently private. We will release the full codebase publicly upon acceptance of the paper. If you are a reviewer and need access, please contact us directly.

📜 Citation

If you use this framework or dataset, please cite:

@article{SocioDiff2025,
  title={SocioDiff: A Socio-aware Diffusion Model for Residential Load Data Generation},
  author={Chen, Weilong and others},
  journal={IEEE Transactions on Smart Grid},
  year={2025},
  note={Under review}
}

📬 Contact

For questions or collaborations, feel free to contact: Weilong Chen Email: chenweilong921@gmail.com