Enhancing biogas production predictions using ARIMAX models on mixed silages

Biogas production as a renewable energy source is gaining more attention from different actors in the energy sector due to the use of different residual products for its generation. This interest also comes from the agricultural sector. A typical crop used for biogas production is maize, which poses environmental challenges related to soil erosion and nutrient depletion. Furthermore, land use changes can also reduce biodiversity and attract pests. An increasing number of strategies to diminish these issues rely on combining maize with other leguminous plants, improving the nutritional silage profiles, and potentially enhancing biogas production. Nonetheless, adopting these new approaches remains limited since the farmers hesitate to invest in new technologies without clear and quantifiable improvements. In this regard, in this study, we propose time-series-based models to predict biogas and methane production based on the silage features of crops and the time-series data. In particular, we fitted models based on Autoregressive Integrated Moving Average with eXogenous variables (ARIMAX) to capture the temporal dependencies, aiming to characterize the methane volume and methane concentration accurately. We used a previously validated measurement campaign, which included other anaerobic digestion variables like volatile solids, crude protein, cellulose, and hemicellulose, among others, from crops of maize and mixed maize-legume silages, along with the production of biogas and methane, with a sample period in days. The reactor was a 5 L fermenter operated at 40 °C with manual mixing daily. It used inoculum and silage, with a 21-day delay before measurement. Biogas volume was recorded using a measuring cylinder, and composition was analyzed with a Dräger X-am 8000. We tested our ARIMAX-based models regarding their goodness of fit using the determination coefficient R2 and the Root Mean Square Error (RMSE). In the case of the methane volume, we obtained an R2 of 0.92 and an RMSE of 0.001 liters, and for the case of methane concentration, our models exhibited an R2 of 0.908 and an RMSE of 0.85%. Our promising models help farmers, researchers, and policymakers to accurately characterize and forecast biogas and methane production as promising renewable energy generation technologies.