Industrial scale biogas reforming modelling and validation

This study investigates the industrial data from a biogas reforming plant for data reconciliation (DR) and process modelling to enhance syngas production. With biogas emerging as a vital renewable feedstock, effective data management and optimization strategies are essential for advancing sustainable chemical synthesis. In this work, we assess plant data quality from a demo-scale biogas steam reforming unit (BSR), identifying bias and control loop inconsistencies that impact measurement reliability. Among the outlier methods implemented, Isolation Forest (IF) achieves a 93% raw data variance reduction, simultaneously isolating anomalies and detecting steady states. Further analysis of the steam-to-carbon (SC) ratio reveals a control loop error, increasing confidence in output measured variables. A DR problem is then solved, enabling closure of the mass balance, with a 4% adjustment of input mass flowrates, highlighting process data discrepancies. Considering the reconciled data, a catalytic tube model is applied to predict BSR product distribution and ensure thermodynamic equilibrium. A Monte Carlo sensitivity analysis shows the importance of methane tenor in biogas and input temperature for optimal reactor performance. The syngas quality, defined by the hydrogen and carbon content, experimental results highlight the potential for enhanced syngas efficiency, particularly for fuel synthesis applications, requiring a stoichiometric ratio of 2. This methodology establishes a framework for improving biogas plant reforming operations, offering a foundation for control strategy advancement and future process optimization. • The results from the operation of a biogas steam reformer are provided. • Outliers detection methods are applied to identify nominal steady-state data. • An extensive data reconciliation procedure is applied to obtain significant process data. • The reconciled data are used to validate the developed process model. • The combination of experimental and modelling data are used for further optimization of the unit.