Neural network-based controllers for voltage source PWM front end rectifiers

In order to meet more stringent distortion and power factor requirements in AC/DC rectifiers, PWM voltage source topologies have been proposed. However, the characteristics of these converters are nonlinear and the conventional PI-type regulators cannot be optimized for all operating conditions. Neural network-based controllers among others can handle such nonlinearities. This paper proposes an investigation of suitable neural network control structures to achieve constant output voltage and unity power factor. Requirements, performance and implementation aspects are investigated. It is shown that these controllers can adequately control the rectifier over a wide range of operating conditions