OpenSTEF

OpenSTEF (Open Short-Term Energy Forecasting) is an open source Python package that provides automated machine learning pipelines for short-term energy load forecasting.

Explore Documentation GitHub Repo Meeting Calendar

OpenSTEF provides automated machine learning pipelines that forecast energy load at specific locations up to two weeks ahead.

OpenSTEF is collaboratively developed by a growing community of grid operators, technology vendors, researchers, and energy-sector experts, and is hosted at LF Energy. It generates probabilistic forecasts for hours to days ahead for any given energy signal, enabling grid operators, energy companies, and researchers to anticipate congestion, support grid safety analysis, and optimize flexible assets. The latest release, OpenSTEF 4.0, introduces a fully redesigned modular architecture that broadens the range of forecasting applications and reduces implementation effort.

Probabilistic Forecasting

Load and generation forecasts with uncertainty quantification

Self Correcting

The ML pipelines ensure that forecasts adapt as load patterns evolve

Production-Proven

Deployed at scale in production at multiple organisations

Broadly Compatible

OpenSTEF can be integrated into any IT infrastructure environment

Multiple Applications

Supports electricity, thermal, and potentially other energy systems

Fully Open Source

MPL 2.0 license, which enables commercial use and customization with lightweight open-source requirements.

Features

OpenSTEF combines a complete forecasting library with proven presets and a modular, extensible architecture designed for operational energy environments.

Comprehensive machine learning pipelines

Data preprocessing, feature engineering, model training, inference and evaluation
Probabilistic output: forecasts include uncertainty bands, not just point estimates
Self-correcting: the forecasts adapt as grid conditions and consumption patterns evolve
Includes effective backtesting tooling through OpenSTEF-BEAM

Modular and flexible setup

Plug-and-play architecture: extend or replace components without rebuilding the full pipeline
Compatible with any IT infrastructure environment
High forecasting performance including ensemble and foundation models

Broad application range

Grid operations: congestion management, grid safety analysis, asset optimization
Trading, flexibility optimization, and other forecasting-driven use cases
Thermal systems: district heating and combined heat and power

Why OpenSTEF?

Electrification and distributed renewable generation are pushing grid capacity toward its limits. Accurate short-term forecasting is now a core operational requirement, not a nice-to-have.

Anticipate congestion before it happens. Probabilistic forecasts for grid loads give operators the lead time to act, not just react.
Make the most of existing assets. Accurate load forecasts reduce the need for costly grid reinforcement by enabling smarter use of available capacity.
Stop building the same tool twice. OpenSTEF is a shared, community-maintained framework. Alliander, RTE, Sigholm, SIA and others contribute improvements back, so every organization benefits from the collective investment.
Production-proven at scale across organisations. The same forecasting framework that is used for congestion management at the largest DSO in the Netherlands also powers district heating optimization across the Nordic region.
Fits your environment. Modular architecture and MPL 2.0 licensing means you can integrate OpenSTEF into your environment regardless of your IT setup.

What’s New in OpenSTEF 4.0?

OpenSTEF 4.0 introduces a fully redesigned, modular setup that puts flexibility and adaptability at its core.

The new design makes it easier to tailor forecasting pipelines to your specific use case. Whether that’s adapting to your data, integrating with your IT environment, or experimenting with different modeling approaches. You can easily change, add, remove, adapt pre-processing and other steps of the pipeline.

OpenSTEF 4.0 gives users full freedom to shape their own forecasting setup. Integrate your domain knowledge on feature engineering with ease. And whether you prefer to use classical machine learning, ensemble modelling techniques or a foundational model, openSTEF supports it all. It also provides the tooling around those ML components to ensure everything works smoothly in production. From researching new ideas to deploying to production at scale, OpenSTEF 4.0 contains all modules needed to support your workflow.

OpenSTEF in the Field

Alliander

Alliander, the largest distribution system operator in the Netherlands, uses OpenSTEF in production at thousands of grid locations. These forecasts are used as input for congestion management, the communication of energy exchange with the transmission system operator and for estimating grid losses.

Read Case Study

Sigholm

Sigholm implemented OpenSTEF as the forecasting engine behind its Aurora production optimization platform for district heating and combined heat and power systems. The platform now generates approximately 20,000 forecasts per week, completing each forecasting cycle in about 4 seconds. Aurora optimizes approximately 40% of Sweden’s district heating production.