Methane Concentration Prediction in Anaerobic Codigestion Using Multiple Linear Regression with Integrated Microbial and Operational Data

Anaerobic codigestion of organic residues is a proven strategy for enhancing methane recovery. However, the complexity of microbial interactions and variability in operational conditions make it difficult to estimate methane concentration in real time, particularly in rural contexts. This study developed a multiple linear regression model to predict methane concentration using operational data and microbial community profiles derived from 16S rRNA gene sequencing. The system involved the codigestion of cassava by-product and pig manure in a two-phase anaerobic reactor. Predictor variables were selected through a hybrid approach combining statistical correlation with microbial functional relevance. The final model, trained on 70% of the dataset, demonstrated satisfactory generalization capability on the other 30 test set, achieving a coefficient of determination (R²) of 0.92 and a mean relative error (MRE) of 6.50%. Requiring only a limited set of inputs and minimal computational resources, the model offers a practical and accessible solution for estimating methane levels in decentralized systems. The integration of microbial community data represents a meaningful innovation, improving prediction by capturing biological variation not reflected in operational parameters alone. This approach can support local decision making and contribute to Sustainable Development Goal 7 by promoting reliable and affordable technologies for clean energy generation in rural and resource-constrained settings.