Applications of Machine Learning for Prediction and Optimization of Biogas Production: A Mini-Review

Biogas production through Anaerobic Digestion (AD) is a prominent and sustainable pathway for energy recovery from organic waste; however, AD performance is regulated by complex, nonlinear interactions among substrate properties, microbial activity, and operating conditions. Conventional mechanistic models like Anaerobic Digestion Model No. 1 (ADM1) necessitate extensive parameter calibration and reliable monitoring, which constrains their applicability for real-time prediction and process optimization. Machine learning (ML) has recently emerged as a powerful data-driven approach capable of effectively capturing nonlinear system behavior without reliance on explicit mechanistic formulations. However, most ML-based approaches still inadequately represent complex microbial kinetics. Therefore, this review critically investigates and identifies gaps in the literature, synthesizing advances in ML-based prediction and optimization of biogas production, along with key influencing parameters reported between 2020 and 2026. Existing studies are often fragmented and predominantly limited to laboratory- and pilot-scale investigations that incorporated only limited parameters, for optimizing biogas production. The findings highlight persistent challenges in applying ML, including data insufficiency, limited model transferability, overfitting, and the lack of real-time deployment. Future research should prioritize explainable ML models, low-cost sensors, and real-time integrated prediction–optimization–control frameworks to enable scalable smart biogas systems.