No abstract

Harmonics produce nonlinearity, compromising system stability. Numerical approaches are effective yet time-consuming for solving dynamic nonlinear harmonic issues. This paper suggests employing a mix of the Recurrent Neural Network and Newton-Raphson (RNN-NR) techniques to address nonlinear harmonic problems. Inspired by the brain, RNNs are powerful problem prediction and pattern modelling algorithms. RNNs are able to effectively handle newly collected inputs even in the absence of NR-based mathematical models. Harmonics can be eliminated by Sigma-Delta Selective Harmonic Elimination (SD-SHE) PWM, which is used by Cascaded H-Bridge Multilevel Inverters (CHB-MLI). SD-SHE PWM notches the cascaded multilevel inverter's output waveform at varying angles to decrease harmonics. The neural network is trained using the MATLAB RNN programme. To implement its stages, Xilinx Vivado transfers the RNN to a Field Programmable Gate Array (FPGA). The nonlinearity problem is solved when the neural network is operated on an FPGA, and the design can be easily modified to meet the requirements of different applications.