Machine learning vs. ADM1: Reliable biogas prediction with minimal data requirements in full-scale plants

Area of Science:

• * Agricultural science and renewable energy engineering.
• * Bioprocess engineering and computational modeling.

Background:

• * Anaerobic digestion converts organic waste into biogas, a renewable energy source, but feedstock variability in agricultural settings challenges stability and yield.
• * Mechanistic models like the Anaerobic Digestion Model No. 1 (ADM1) are accurate but require extensive data, limiting their real-world application.
• * Optimizing biogas production is crucial for enhancing agricultural sustainability and renewable energy integration.

Purpose of the Study:

• * To evaluate the predictive accuracy of a simplified ADM1 and machine learning models (random forest, LSTM) for biogas and methane production.
• * To compare the performance and computational efficiency of these models in a full-scale agricultural biogas plant.
• * To explore hybrid modeling strategies for real-time monitoring and process optimization.

Main Methods:

• * Implemented a simplified version of the Anaerobic Digestion Model No. 1 (ADM1).
• * Utilized machine learning algorithms: random forest and long short-term memory (LSTM) networks.
• * Validated models using daily biogas and methane production data from a full-scale plant (2023-2024).

Main Results:

• * All models achieved high predictive accuracy, with Nash-Sutcliffe efficiencies above 0.78.
• * Random forest model performed best when including feedstock quantities and maize silage volatile solids.
• * LSTM demonstrated effectiveness with minimal input data but had significantly longer training times compared to ADM1.

Conclusions:

• * Simplified ADM1 and machine learning models offer comparable accuracy to complex mechanistic models for biogas prediction.
• * Hybrid modeling approaches can balance predictive precision with data requirements for practical applications.
• * These findings support enhanced real-time monitoring for optimizing biogas production and promoting agricultural sustainability.