A Systematic Review of Machine-Learning Solutions in Anaerobic Digestion.

Area of Science:

• Environmental Science
• Biotechnology
• Data Science

Background:

• Machine learning (ML) is increasingly utilized in anaerobic digestion (AD) to interpret complex operational parameters.
• Effective ML integration in AD promises enhanced process optimization and efficiency.
• Challenges exist in applying ML models across diverse AD systems and scales.

Purpose of the Study:

• To systematically review the current applications of ML in anaerobic digestion.
• To identify implementation challenges and benefits of ML in AD.
• To explore future research directions for ML in AD.

Main Methods:

• Systematic literature review of ML applications in anaerobic digestion.
• Analysis of ML techniques, focusing on artificial neural networks (ANN).
• Examination of data sources (lab vs. industry scale) and their impact on model performance.

Main Results:

• Artificial neural networks are the predominant ML technique used in AD.
• Models trained on lab-scale data often fail to generalize to industrial-scale operations due to system variability.
• ML offers potential for real-time predictive modeling to ensure AD stability and efficiency.

Conclusions:

• ML integration in AD faces challenges related to data heterogeneity and model scalability.
• Addressing these challenges can lead to significant improvements in biogas production and waste treatment.
• Further research into ML techniques and broader domain applications is needed for robust AD optimization.