Applying artificial optimization methods for transformer model reduction of lumped parameter models

Detailed R-C-L-M models of power transformers, which are based on lumped parameters, are used extensively not only for transient analysis of power transformers to determine electrical stresses in windings, but also for studying transients in power systems. Models with few elements are generally more practicable for power system studies but at the expense of accuracy. The use of artificial methods to reduce an R-C-L-M model is the main contribution of this paper. Advantages of the suggested method include: (1) a reduced loss of accuracy compared with the original model and (2) the flexibility to choose the number of model elements to achieve the desired model depending on size and accuracy. The ability of three different artificial methods, genetic algorithm, particle swarm optimization, and bacterial foraging algorithm, to model reduction is evaluated using measurements on an actual 400 kV test object and the results are compared with those obtained by common analytical formulae.

► Transformer model reduction is investigated in details.
► Three different artificial methods are discussed for model reduction.
► Artificial methods improve transformer modeling in transient states.
► A suitable fitness function is suggested for three optimization algorithms.