From Dense to Sparse: Event Response for Enhanced Residential Load Forecasting (TIM 2025)

IMPORTANT
Thank you for your interest in our project! If you find this work useful, please give it a ⭐ Star on GitHub to show your support.

Table of Contents

1. Introduction
2. Quickstart
3. Execution Example
4. Data and Models
5. Citation
6. Acknowledgement
7. Contact

---

Introduction

This repository contains the implementation for our IEEE TIM 2025 paper "From Dense to Sparse: Event Response for Enhanced Residential Load Forecasting". We propose a novel framework that improves residential load forecasting accuracy by leveraging sparse event responses to capture appliance usage patterns.

Quickstart

To set up the environment and dependencies like build.sh.

Execution Example

Train MSP model (model 2),Enhance RLF with MSP (model 3).

python knowledge4tsf/main.py \
    --model 3 \
    --device cuda:0 \
    --tsf_model patchTsMixer \
    --pred_len 1 \
    --status_file status_model1_predL_1_horizon_1_topk_22_umass3.pth \
    --data umass3 \
    --data_dim 22 \
    --topk 22 \
    > /tmp/umass3_1_patchTsMixer.log

Data and Models

Google Drive Download

Citation

@article{cao2025erkg,
  title={From Dense to Sparse: Event Response for Enhanced Residential Load Forecasting},
  author={Cao, Xin and Tao, Qinghua and Zhou, Yingjie and Zhang, Lu and Zhang, Le and Song, Dongjin and Oliver Wu, Dapeng and Zhu, Ce},
  journal={IEEE Transactions on Instrumentation and Measurement}, 
  volume={74},
  pages={1-12},
  year={2025},
  doi={10.1109/TIM.2025.3544349}
}

Acknowledgements

We acknowledge these open-source projects:

• RevIN - Reversible Instance Normalization
• ETSformer - Time Series Forecasting Transformer
• AttrFaceNet

Contact

For questions and collaborations:

Xin Cao: caoxin9629@gmail.com