A priori metrics to select best training dataset of top-oil temperature models of power transformers

Abstract

In order to use top-oil temperature models in an on line thermal monitoring system, the unknown parameters of the models should be determined. Moreover, the values of these parameters differ for different transformers and different datasets; therefore, these parameters should be considered as Empiric Factors which should be estimated based on measured data. The estimation process is called training here. The contribution of this paper is to present some a priori indices which are the judge of selecting the best dataset among available dataset to train top oil temperature models. On the other words, the question is that, among different available datasets to train the top oil temperature model which one will present lower prediction error when the model is used in an On-line monitoring system.The methodology will be applied on two transformers whose available datasets contain ambient temperature, load current, and measured top oil temperature during four months with sampling step of 15 minutes. The investigated models in this paper are IEEE clause 7, lEe 60076-7, and Linear model.

Transformer hot spot temperature estimation through adaptive neuro fuzzy inference system approach

“…It was further confirmed through applications that solar radiation can lead to transformer top oil temperature rise by about 3.7 °C, [ 17 ]. Other variables considered by different researchers in thermal dynamic model modifications for TOT and HST estimation to improve on adequacy and accuracy are not limited to tap-changer position, oil viscosity, harmonics in loading, insulation water content, constant or variable thermal resistances taking into account the type, design and capacity of the transformer [ 5 , [16] , [17] , [18] , [19] , [20] , [21] ].…”