Performance analysis of three evolutionary algorithms: the case for improvement on quality of power supply

Evaluates the predictability of BP, GA and combined GA&BP in terms of control program design, convergence behaviour, learning capability and harmonic compensation efficacies, as applied to the filtering of load generated current harmonics in a DC variable-speed drive. Using the same measured current harmonic data, our performance evaluations confirm that the BP has a faster learning convergence whereas GA provides a more robust control system. A Combined BP and GA algorithm offers good reduction in harmonic content and thus improves the quality of power supply