Machine learning-aided inverse design for biogas upgrading through biological CO₂ conversion

Area of Science:

• Biotechnology and Bioengineering
• Renewable Energy
• Environmental Science

Background:

• Biogas upgrading via CO2 bioconversion to CH4 is key for energy decarbonation.
• Optimizing operational parameters like CH4% and H2 utilization is crucial but hindered by inconsistent lab conditions.
• Existing models struggle to simultaneously optimize multiple operating conditions.

Purpose of the Study:

• To develop a robust, data-driven method for optimizing biogas upgrading processes.
• To overcome limitations of existing models in simultaneously optimizing multiple operational parameters.
• To enhance efficiency and consistency in sustainable methane production.

Main Methods:

• Developed two XGBoost models (R²=0.779, R²=0.903) using literature data.
• Embedded models into a multi-objective particle swarm optimization algorithm.
• Validated predictions against experimental results under optimized conditions.

Main Results:

• Achieved high model accuracy for predicting biogas upgrading performance.
• Experimental validation showed low relative errors: 8.50% for CH4% and 2.95% for H2 conversion rate.
• Demonstrated successful simultaneous optimization of multiple operating parameters.

Conclusions:

• The proposed data-driven approach effectively streamlines biogas upgrading.
• This method enhances efficiency and consistency in methane production from biogas.
• Offers a pathway towards more sustainable and reliable renewable energy solutions.