Machine learning-driven optimization of plasma-coupled photocatalytic CO₂ reduction system for syngas production

Dielectric barrier discharge (DBD) plasma-coupled photocatalysis offers a promising route for CO₂-to-syngas conversion, but rational optimization remains challenging because operating variables are strongly coupled. Here, we develop a small-sample machine-learning framework to predict and optimize a plasma-coupled photocatalytic CO₂ conversion system using a Cu-Pd/TiO₂ photocatalyst. Using 120 experimental runs, five operating parameters, including discharge power, catalyst dosage, gas flow rate, relative humidity, and light intensity, were evaluated against CO₂ conversion, CO yield, and H₂ yield. Among the evaluated models, a shared-weight GBR-KRR hybrid model achieved the best predictive performance, with test-set R² values of 0.947, 0.950, and 0.954 for CO₂ conversion, CO yield, and H₂ yield, respectively. Permutation importance and SHAP analyses identified relative humidity, light intensity, and discharge power as the most influential variables and revealed distinct model-predicted trends across the three outputs. The trained model was further used for constrained optimization under target H₂/CO ratios of 1 and 2, followed by experimental validation of the selected operating conditions. Overall, this work establishes a data-driven strategy for interpreting nonlinear plasma-photocatalytic CO₂ conversion and provides practical guidance for syngas-ratio regulation under experimentally relevant conditions.