Neural Algorithm Aided Operation of CO₂ Electrolyzers.

Area of Science:

• Electrochemistry
• Catalysis
• Machine Learning

Background:

• Electrochemical carbon dioxide reduction (CO2R) is crucial for sustainable chemistry but faces challenges in long-term stability, selectivity, and energy efficiency.
• Lab-scale CO2R studies are often limited in duration and parameter space, hindering process optimization.
• Complexity of cell operation parameters complicates achieving optimal performance.

Purpose of the Study:

• To introduce a high-throughput methodology for testing CO2 electrolyzer operation.
• To develop a machine learning (ML) model for data evaluation and process optimization.
• To enable adaptive optimization for stable and efficient CO2 electrolyzer operation.

Main Methods:

• Implementation of a high-throughput cell operation testing methodology.
• Autonomous operation of a test station for extensive data collection.
• Development and application of an artificial neural network (ANN) model for predictive analysis.

Main Results:

• The ANN model accurately predicts CO2 electrolyzer performance under various conditions after training on a subset of data.
• Precise predictions are achievable for newly assembled cells and extrapolated parameter settings.
• The ML-based approach facilitates holistic data evaluation for process optimization.

Conclusions:

• The developed methodology and ML model significantly advance the potential for long-term stable operation of CO2 electrolyzers.
• Adaptive optimization of process conditions through ML-based data evaluation is demonstrated.
• This approach overcomes limitations of traditional lab-scale testing for CO2 reduction technologies.