Computational efficiency and stability of physics-informed neural networks (PINN) in modeling charge dynamics of corona discharge plasma circuit

This study proposes a Physics-Informed Neural Network (PINN) approach for modeling the charge dynamics of a corona discharge plasma (CDP) circuit that is governed by a second-order ordinary differential equation (ODE). The training data were obtained from a combination of experimentally derived plasma charge measurements and analytical solutions of the circuit differential equation, with added noise to reflect realistic discharge conditions. By embedding the governing physical law directly into the loss function, the PINN accurately reproduced the damped oscillatory charge response of the plasma system. The model showed strong agreement with both experimental observations and analytical solutions, achieving an R-squared value of 0.93 and a mean squared error of 0.016 after sufficient training. To the best of the authors’ knowledge, this study represents one of the earliest applications of PINNs to the dynamics of corona-discharge plasma generators. Another novelty of this research lies in the deployment of artificial intelligence to computationally resolve and simulate charge dynamics within plasma generator circuits. These findings have practical applications, allowing engineers and researchers to optimize the design and fabrication processes of plasma-generating systems.