Estimation of the eddy current losses in a dry-type 3000 KVA transformer with machine learning

Abstract

This paper explores the design parameters of high-voltage coils of three-phase transformers and the effects of eddy current losses caused by energy conversion. The commercial software, ANSYS-Maxwell, was employed to conduct the simulation of electrical and magnetic fields. The design parameters of the high voltage coil include the leg distance of core, the height and the block thickness of primary windings, the height of primary coils, and the height of secondary windings. The specification of the three-phase transformer of this study are 3000 kVA with rated voltage of 6600 V and corresponding current of 151.5 A. Base on thirty cases of simulation, machine learning, artificial neural network, was utilized to predict the extra loss due to eddy current in the clamps and the windings. The prediction accuracies are 0.72 and 0.86 for primary and secondary windings, respectively. Keywords: Dry-type transformer, Eddy current loss, Numerical simulation, Machine learning