Parameter Estimation of a Distribution Transformer Model using Pseudo-Random Impulse Sequence Perturbation

This paper investigates the use of a Pseudo-Random Impulse Sequence (PRIS) as a perturbation signal for estimating the parameters of a wideband lumped-parameter transformer model for a 16kVA, 22kV/240V distribution transformer. The frequency responses of the open-circuit and short-circuit primary and secondary input impedances, as well as the forward and backward voltage transformation ratios, are simulated using PRIS perturbation signals, and compared with results published in literature. It is shown that the simulated frequency responses exhibit close correlation with the responses derived using analytical expressions, and that PRIS perturbation excites the model sufficiently to obtain accurate frequency responses up to 100kHz. The model parameters are subsequently estimated from these frequency response results and compared with results published in literature. The performance of particle-swarm optimisation and genetic algorithm for the parameter estimation processes are, furthermore, compared. It is shown that particle-swarm optimisation delivers improved parameter estimates and more efficient runtimes.