Causal Learning at ESS: A Time Series Benchmark

The dataset is a benchmark intended for causal learning. It comprise control system data from the accelerator cryogenics plant at the European Spallation Source. The data was taken during three periods when the system was in steady state operation. There are 233 different process variables represented in the dataset. These arise from sensor measurements of physical quantities. Each measurement is indexed by a timestamp and have several metadata fields in addition to the value field. If you use the dataset, please cite this dataset as well as the paper S.W.Mogensen, K.Rathsman, P.Nilsson, Causal discovery in a complex industrial system: A time series benchmark, in Proceedings of the 3rd Conference on Causal Learning and Reasoning (CLeaR), 2024, Available: https://doi.org/10.48550/arXiv.2310.18654 The paper also contains a detailed description of the data and of the underlying system. More description, code, and help to get started can be found at https://soerenwengel.github.io/essdata

Note: Downloading the dataset can be slow. Therefore, we have published a reduced dataset containing one hour of operation data for testing and viewing the data. The reduced dataset is available from 10.5281/zenodo.10679737.