Abstract

Off-grid electricity in energy-constrained communities can be supported by converting organic waste into biogas and electrical energy using affordable monitoring and predictive tools. This study developed an anaerobic digestion electricity-generation framework incorporating low-cost sensor-based monitoring and artificial neural network methane prediction. The modelling dataset comprised 61 daily records combining independently measured observations with controlled-profile augmentation across the investigated operating range. A wireless microcontroller-based sensing node monitored temperature and feedstock moisture, while temperature, biogas volume, carbon dioxide fraction, pH, hydrogen sulfide, and internal combustion engine operating time were used to predict methane. Methane output ranged from 6.50 to 12.09 cubic metres per day and daily electrical energy from 10.0 to 22.0 kilowatt-hours. On held-out test data, the neural network achieved a root mean square error of 0.593 cubic metres per day and a coefficient of determination of 0.856, whereas multiple linear regression achieved 0.416 and 0.929, respectively. The implied electrical conversion efficiency averaged 15.5%. The results demonstrate predictive feasibility within the investigated range, while the structured hybrid dataset and stronger linear baseline indicate that broader generalisation requires independently measured data under more variable operating conditions.