A dynamic model for a gas-liquid corona discharge using neural networks

This paper presents a novel dynamic nonlinear model for pulsed corona discharge using backpropagation neural networks. The Levenberg-Marquardt training algorithm, which is perfectly suitable for fitting functions, is employed. The developed model is based on the voltage-current characteristics of an actual hybrid-series reactor and takes the practical constrains associated with a real system into account. The validity and accuracy of the model have been tested in the Electromagnetic Transients Program, using MODELS language and a TACS-91 time-variant controlled resistor. The results clearly demonstrate that the BPNN-based model is very robust and effective in emulating the chaotic performance for pulsed corona discharge using backpropagation neural networks.