A neural network controlled unity power factor three phase PWM rectifier

The simplest method of operating three-phase PWM rectifiers is based on the use of offline PWM patterns. However, in this scheme the input power factor is less than unity and varies with the rectifier operating point due to the presence of the input LC filter. Furthermore, the response to transient conditions is slow and large current oscillations may occur due to the resonance of the filter. In this paper an online neural network (NN) controller, proposed to waveshape the input line currents, forces unity power factor operation and damps the low frequency resonance of the input filter. The proposed controller is insensitive to load/parameter variations thus resulting in a robust system. The performance of the proposed NN controller is verified by simulation.<>