Combining CFD and ML to optimize biogas power plant agitation efficiency

Abstract Effective agitation is critical for the performance of agricultural biogas power plants, while quantifying mixing efficiency for a given system remains challenging. Direct measurement of substrate flow fields in full-scale digesters is costly and intrusive, and relying on computational fluid dynamics simulations (CFD) alone can be computationally expensive for routine design exploration. This work addresses this gap by integrating CFD with supervised machine learning to optimize mixing efficiency through strategic rotor placement and configuration. We simulate a diverse range of biogas digester scenarios, with varying rotational speeds and agitator locations, capturing how these factors influence the mixing/agitation of the substrate. We then train an artificial neural network to predict the agitation efficiency, for previously unseen configurations. During validation, the proposed model indicates a mixing energy level below the EPA guideline range (5 W m −3 – 8 W m −3 ). Overall, the hybrid CFD-ML methodology demonstrated here provides a scalable pathway for data-informed optimization of mixing in biogas plants, with potential to improve energy efficiency and support design and retrofit decisions in renewable energy infrastructure.