S4Casting

by Alliander

AI Application

Low Confidence

S4Casting is an open-source forecasting toolkit published by Alliander, built around deep-learning models such as state space models (S4, Mamba-style selective SSMs) and transformer variants. It targets medium-voltage power forecasting, typically up to two days ahead. The repository contains training, evaluation and benchmarking pipelines, MLflow and Weights & Biases tracking, inference scripts and ONNX export. Alliander states that future S4Casting models will be integrated into LF Energy's OpenSTEF. The repository does not say whether the models run in Alliander's operations.