Maximizing Biogas Production Efficiency: A Comparative Analysis of Predictive Models for Operational Anaerobic Processe

Abstract

Industrial-scale cattle farms and poultry husbandry meet the growing food demands but generate vast quantities of animal excrement, leading to environmental pollution and greenhouse gas emissions. To address this, converting organic waste into biogas offers a sustainable solution. However, inefficient management often hinders biogas production. This study explores the application of AI strategies to enhance biogas production efficiency. Utilizing two datasets, the study examines input variables (CO2, O2, pH, H2S, temperature, fluid/cattle manure, poultry manure, clean water) measured daily on a particular day and output methane gas amounts measured after 30 days, while the second dataset includes more input variables measured over a period of 30 days. In this study, the predictive accuracy of methane gas alone, as well as in combination with oxygen and carbon dioxide is examined. A variety of statistical and machine learning models were used, including Artificial Neural Networks, Support Vector Regression, Polynomial Regression, Random Forest, Gradient Boosting, Ridge, LASSO, and ElasticNet Regression. ANN produces better results than statistical and machine learning models, exhibiting RMSE of 0.1815, MAE of 0.1412, and MSE of 0.0329. Due to the large number of parameters, ANN performs well with the first data set but struggles with the second data set. Applied models excel at predicting methane gas alone compared to CH4, CO2, and O2 together. Furthermore, the study analyzes the input parameters' effects using the model results. The most effective parameters in the biogas formation process were examined and determined using the results obtained from the models. The first five parameters are positively correlated: CO2, fluid manure input, cattle manure, poultry manure, and activated sludge output. Among those with a negative correlation are input volume, dry substrate, oxygen, clean water, and oxidation. It has been verified how effective the substances fed to the reactor are for biogas production.

More by Dr. Zarina Oflaz

• Optimal adaptive control of methane production in anaerobic digestion via reinforcement learningJournal of Process Control · 2026
• Development of machine learning models for intelligent prediction of workability, mechanical, and durability characteristics of cement mortar made of recycled sandConstruction and Building Materials · 2026 · Citations: 1
• Leveraging eco-friendly pozzolans and machine learning for reduced cement use in soil improvement of road base layersEuro-Mediterranean Journal for Environmental Integration · 2026
• Comparative Analysis of Prediction Accuracy in Clean Energy Forecasting: Implications for Pricing and Profitability in Wind and Biogas Energy SystemsSpringer proceedings in earth and environmental sciences · 2026