A synthetic dataset of French electric load curves with temperature conditioning

2025/04/18 by Tahar Nabil, Nabil, Tahar, Ghislain Agoua +7 · 1 citation
Engineering · #Artificial Intelligence (cs.AI) #Energy Load and Power Forecasting #FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · pdf · doi:10.48550/arxiv.2504.14046

openalex publication_date 2025/04/18 · openalex created_date 2025/09/25 · openalex updated_date 2026/07/28

Abstract

The dataset contains 10 000 synthetic time series representing 30-minute individual electric consumptions for one year. The samples are generated by training a deep learning model (Latent Diffusion) conditionnally on local outdoor temperature, contracted power (6kVA, 9kVA or 12kVA) and time-of-use plan. The dataset captures the correlation between cold temperature and electric consumption that can be found in certain countries, e.g. in France.

Citations

• Defining 'Good': Evaluation Framework for Synthetic Smart Meter Data
• Faraday: Synthetic Smart Meter Generator for the smart grid
• TimeXer: Empowering Transformers for Time Series Forecasting with Exogenous Variables
• Unified Training of Universal Time Series Forecasting Transformers
• BuildingsBench: A Large-Scale Dataset of 900K Buildings and Benchmark for Short-Term Load Forecasting
• A Time Series is Worth 64 Words: Long-term Forecasting with Transformers
• High-resolution synthetic residential energy use profiles for the United States
• High-Resolution Image Synthesis with Latent Diffusion Models
• High-Resolution Image Synthesis with Latent Diffusion Models
• Membership Inference Attacks From First Principles
• Sig-Wasserstein GANs for Time Series Generation
• PSA-GAN: Progressive Self Attention GANs for Synthetic Time Series
• Denoising Diffusion Probabilistic Models
• Optuna: A Next-generation Hyperparameter Optimization Framework
• Tackling Climate Change with Machine Learning
• Unsupervised Scalable Representation Learning for Multivariate Time Series
• Neural Discrete Representation Learning
• A Test of Relative Similarity For Model Selection in Generative Models
• Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation
• Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation

Cited by

• Are Time-Indexed Foundation Models the Future of Time Series Imputation?

Related

• Faraday: Synthetic Smart Meter Generator for the smart grid
• Dataset on residential electricity load profiles in Switzerland
• Forecasting Anonymized Electricity Load Profiles
• Privacy Preserving Demand Forecasting to Encourage Consumer Acceptance of Smart Energy Meters
• EnergyDiff: Universal Time-Series Energy Data Generation using Diffusion Models