Research on Machine Learning-Assisted Pulse Discharge Plasma Catalysis for Methane-Carbon Dioxide Simulation

Due to the substantial complexity of plasma numerical simulation processes and the significant time required for operation, this paper establishes a dedicated dataset. It investigates deep learning-assisted modeling methods for pulsed discharge plasma catalysis. By employing machine learning, we assess regression algorithms using K-fold cross-validation alongside the coefficient of determination (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>) metrics. We select the most effective Artificial Neural Network (ANN) to assist in plasma catalysis modeling. Notably, the ANN demonstrates multiple advantages, including improved efficiency and adaptability. To enhance prediction accuracy, we implement Bayesian Optimization (BO) for tuning the ANN parameters. Experimental results indicate that the outcomes derived from the ANN, optimized through Bayesian methods, align well with the numerical results from dynamic models.