A neural network controlled unity power factor three phase current source PWM front-end rectifier for adjustable speed drives

Three-phase current source PWM rectifiers used as front-end AC/DC power converters can significantly increase the input power factor and reduce the line-side harmonics in adjustable speed drive systems (ASDs). They can therefore meet the strict specifications on the harmonics injected into the AC mains which will eventually be introduced in the near future. Three-phase current source PWM rectifiers are generally operated with offline PWM patterns. This results in nonunity input displacement factor and possible large inrush current due to the presence of the input LC filter. Moreover, the input displacement factor is dependent upon the operating point and response to transient conditions is slow. In this paper, an online neural network controller is proposed to wave shape the input line currents of the rectifier. Unity power factor operation is maintained and low frequency resonance of the input filter is effectively damped without the need for large damping resistors. The proposed controller is insensitive to load/system parameter variations, thus resulting in a robust system. The overall system exhibits good dynamics and stability. Design guidelines and steady state and operating characteristics are discussed