Neural network based programmed pulse width modulation usable on voltage source inverters (VSIs)

In this paper some three-layer neural networks are trained for execution of programmed pulse width modulation (PPWM) method in voltage source inverters (VSIs). Solving the custom difficulties of utilization of PPWM such as large size of memory, complex control circuits and controlling the magnitude of output voltage in discrete manner in addition to presentation of performance index variations vs. the time of training, spectrum frequency of output voltage, THD variations vs. the fundamental frequency and the operation of inverter at non-trained values are the main subjects of this paper