Design and implementation of neural networks for digital current regulation of inverter drives

A discussion is presented of the design of one, two, and three layer neural networks for digital current regulation of the inverter drives. The learning requirements of various designs are evaluated by developing two different learning techniques for such inverter current regulators. The models and learning techniques have been investigated by simulation. These simulation results along with design considerations are used to determine the network best suited for this application. The implementation of neural networks is described, and experimental results are given.<<ETX>>