Energy-TTM is a domain-specific Time Series Foundation Model (TSFM) for energy meter data analytics, pretrained on large-scale real-world smart meter data from the EnergyBench corpus. Built upon IBM Research's Tiny Time Mixer (TTM) architecture, it learns transferable representations of electricity consumption patterns across diverse residential and commercial buildings.
The pretrained model is designed for zero-shot and transfer learning across heterogeneous buildings, regions, and operational contexts, while remaining lightweight and computationally efficient. It is optimized for short-term load forecasting, enabling accurate day-ahead energy demand prediction with minimal task-specific adaptation.
Pretraining Dataset
Energy-TTM is pretrained on EnergyBench, a large-scale real-world smart meter
dataset available on Hugging Face:
The scale and diversity of EnergyBench enable EnergyFM to learn daily, weekly,
and seasonal consumption patterns and to generalize robustly to unseen
buildings and regions.
Default configuration:
Context length: 168 hours
Prediction horizon: 24 hours
Available Checkpoints
Model Variant
Context
Horizon
Pretrained On
Intended For
Energy-TTM-168-24(Default)
168
24
Real-world energy meter data from diverse residential and commercial buildings
General-purpose energy forecasting
Energy-TTM-168-24-comm
168
24
Synthetic commercial building data (ComStock)
Commercial buildings
Energy-TTM-512-96-comm
512
96
Synthetic commercial building data (ComStock)
Commercial buildings
Energy-TTM-168-24-res
168
24
Synthetic residential building data (ResStock)
Residential buildings
Energy-TTM-512-96-res
512
96
Synthetic residential building data (ResStock)
Residential buildings
Supported Tasks
Load Forecasting
Energy-TTM supports short-term electricity load forecasting under both
zero-shot and fine-tuning regimes. It demonstrates strong generalization
across residential and commercial buildings and outperforms traditional
machine learning baselines and generic TSFMs in out-of-distribution settings.
To compare EnergyFM against other state-of-the-art Time Series Foundation Models
for energy analytics tasks, please visit our public benchmark leaderboard:
The leaderboard provides standardized evaluations across forecasting,
anomaly detection, and classification tasks, enabling direct comparison
under consistent experimental settings.
Limitations and Intended Use
EnergyFM is intended for energy meter analytics and has been pretrained
on electricity consumption data. Performance may degrade when applied to
unrelated domains or data with significantly different temporal
characteristics.
Citation
If you use EnergyFM in your work, please cite:
bibtex
1@inproceedings{energyfm2026,
2author = {Arjunan, Pandarasamy and Srivastava, Naman and Kumar, Kajeeth and Jati, Arindam and Ekambaram, Vijay and Dayama, Pankaj},
3title = {EnergyFM: Pretrained Models for Energy Meter Data Analytics},
4year = {2026},
5url = {https://doi.org/10.1145/3744255.3798119},
6doi = {10.1145/3744255.3798119},
7booktitle = {Proceedings of the 17th ACM International Conference on Future and Sustainable Energy Systems},
8pages = {556–568},
9series = {E-Energy '26}
10}
