On-line and off-line control design in power electronics and drives using genetic algorithms

This paper describes an automated control system design strategy for power electronics and drives. The design method employs a genetic algorithm (GA) based optimization routine that can be implemented off-line and on-line. A suitable user definable fitness function assesses the performance of several control parameters, and selects which values will evolve to the next generation. Evolution over many generations evaluates a wide range of controller structures and values, and ultimately determines which one best achieves the control goals. Simulation and experimental results carried out using as applications an active filter and an induction motor drive show that the method produces an excellent controller design, but equally importantly, the design is carried out with very little user interaction. As such the design method is attractive to commercial manufacturers, as it can significantly reduce commissioning times.