H2-SimNet Dataset: A Simulation-based Dataset for Anomaly Detection in Hydrogen Blend Transport Networks

This dataset provides multivariate time series generated from a MATLAB/Simscape model of a hydrogen blend transport network, including both normal operating conditions and anomalous scenarios. It captures transient and steady-state behaviors typical of industrial infrastructures, with anomalies such as gas leaks, compressor faults, startup delays, and overlapping anomalies. The simulation model was developed using MATLAB R2025a and the Simscape Gas Fluids library, with sensors distributed throughout the network measuring pressure and mass flow at multiple points. Anomalies are injected in a controlled manner across three operational phases: initial ramp-up, transient, and steady state. Both weak and strong anomalies are included, with multi-label annotations for detection, classification, and localization. Gaussian noise is added to replicate realistic measurement uncertainties. The dataset includes: Original MATLAB/Simscape simulation files for reproducing scenarios and generating custom variants. CSV and Parquet formats for time series data. Organized directories for anomalous scenarios and normality scenarios. Dataset Use CasesThe dataset is suitable for: Development and validation of anomaly detection algorithms in hydrogen networks. Predictive maintenance and compressor management studies. Investigating overlapping anomalies and network resilience. Benchmarking supervised, semi-supervised and unsupervised methods. Optimization of sensor placement and measurement strategies.