The transition toward a hydrogen-based energy system introduces significant challenges for infrastructure monitoring, particularly in transport networks. The hydrogen’s properties, such as high flammability, diffusivity and its propensity to cause embrittlement of materials, need robust and continuous safety monitoring. Pipelines, compressors and storage systems are subject to operational risks such as pressure fluctuations, leaks and malfunctions, which can affect both safety and system efficiency. Anomaly Detection (AD) is crucial for addressing these challenges, as it enables the early identification of irregularities and supports predictive maintenance strategies. Traditional rule-based systems often struggle with complex, multivariate patterns, making Machine Learning (ML) approaches increasingly attractive. When integrated into a Digital Twin (DT) framework, ML techniques can enhance the capability to detect subtle or emerging faults in real time. This work presents the development of a MATLAB/Simulink-based Digital Twin of a hydrogen transport network operating with methane-hydrogen mixtures. The model includes pipelines, tanks, and compressors, with embedded pressure sensors generating time-series data. The anomalies, such as compressors malfunctions or local restrictions, are artificially injected to simulate fault scenarios. Two ML-based AD algorithms, One-Class SVM and Isolation Forest, are then evaluated for their ability to detect these anomalies. The aim is to validate a flexible, simulation-driven approach to improve the safety and resilience of hydrogen infrastructures.