Aplicación de herramientas de aprendizaje de máquina en la determinación de fuerzas electromagnéticas en los devanados de transformadores de distribución para casos con corriente de excitación despreciable: diseño de un modelo basado en árboles aleatorios para la determinación de las fuerzas electromagnéticas en los devanadosde transformadores de distribución para casos con corriente de excitación despreciable.

Abstract

This document proposes a method based on random trees for the determination of electromagnetic forces that occur in the windings of distribution transformers under normal operating conditions and when short circuits occur, which takes less time to find the solution than the Method of Finite Elements. The necessary steps are detailed to carry out the design of a transformer, focused mainly on the behavior of the magnetic field. This procedure is applied for the design of a 500 kVA, 13.8 kV/208 V transformer, in Dy11 connection, with a leakage reactance of 12.47%. Once the dimensions of the iron core and the discs of each of the high and low voltage windings have been determined, the transformer is implemented in the FEMM (Finite Element Method Magnetics) software. Simulations were performed for values of currents ranging from 1 p.u. up to 8.019246 p.u. which is the value of current that would occur when a three-phase short circuit occurs in the transformer terminals. Three instants of the period were considered, the instants in which each of the phases is at its maximum value. An analysis of the results obtained by varying the design parameters of the random forest model is shown. Finally, a comparison of the values of the electromagnetic forces produced in the windings of distribution transformers due to short-circuit currents obtained by the Finite Element Method with those obtained by a random forest model is presented.