Analysis of the influence of service life on the value of oil indicators in 330 kV autotransformers

Abstract

The results of two-factor crossed dispersion analysis of the influence of duration and operating conditions on the values of transformer oils in the tanks of 330 kV autotransformers are presented. On the basis of analysis of known mathematical models of oxidation processes of transformer oils on the basis of oxygen absorption, it was concluded that there is a significant influence of oxidation duration on the amount of absorbed oxygen, and, consequently, on the values of indicators of oils. To verify the conclusions made, the results of periodic tests of transformer oils condition for 49 autotransformers of 330 kV voltage were used. The total volume of the analysed sample amounted to 3654 values for 6 indicators of oils. Before the direct performance of the analysis, the initial data array was subjected to two-stage statistical processing, the purpose of which was to remove stationary and distorted by the sign of the angular coefficient of the regression model time series of oil indicators, with the subsequent formation of data arrays with a close rate of drift of indicators in time. The solution of the problem of the first stage was carried out on the basis of testing the statistical hypothesis of equality to zero of the angular coefficient of the linear regression of the oil indicator on the duration of operation. To solve the problem of the second stage, the criterion of maximum correlation ratio was used, which allowed to form data arrays with a close rate of drift of indicators in time, in conditions of either complete or partial absence of a priori information about the modes of operation of equipment of electric networks. Since for many indicators the presence of areas of intersection between the values of different data sets was revealed, along with testing the hypothesis about the influence of the duration of operation on the values of oil indicators, the hypothesis about the significance of differences between the values of indicators from different data sets was tested. For this purpose the model of two-factor crossed variance analysis was used. According to the results of the analysis it was found that there was a significant influence of the duration of operation on the values of oil indicators. In addition, significant differences between the values of indicators from different data sets, as well as non-additivity of the effects of changing the levels of factors, that is the effect of changing the level of exposure to one factor leads to a change in the effect of the level of exposure to another factor. The obtained results allow us to develop models for early diagnosis of transformer oil condition using variation boundaries of indicators, as well as to predict and manage the residual life of transformer oil.