A spiking neural network-based long-term prediction system for biogas production

Area of Science:

• Biotechnology and Bioengineering
• Computational Neuroscience
• Renewable Energy Systems

Background:

• Biomass-to-biogas conversion is crucial for sustainable energy.
• Digestor processes involve complex, multi-scale chemical dynamics.
• Accurate prediction of biogas yield is essential for process optimization.

Purpose of the Study:

• To develop a low-cost, efficient predictive model for biogas production using spiking neural networks.
• To address data generalization challenges in sensitive digestor systems.
• To capture and model the multi-scale temporal dynamics inherent in digestor processes.

Main Methods:

• Utilized the NeuCube computational framework for implementing a spiking neural network (SNN) model.
• Trained the SNN model using the initial ten days of digestor process data.
• Focused on capturing rapid early-stage and slower later-stage process variations.

Main Results:

• The SNN model accurately predicted the digestor's chemical process evolution up to the 100th day.
• Achieved high accuracy compared to experimental laboratory data.
• Demonstrated the model's ability to generalize data despite high sensitivity to initial conditions.

Conclusions:

• Spiking neural networks are effective for modeling complex information processes in digestors.
• SNNs can handle activity patterns across different time scales, crucial for multi-scale temporal dynamics.
• The developed model provides a robust tool for accurate long-term biogas production prediction.