STATISTICAL ANALYSIS OF TEMPERATURE RISE TEST FOR INCREASED ACCURACY ON X- AND Y- EXPONENTS

Abstract

On a large number ( > 50 ) of identical large power single-phase autotransformers ONAF temperature rise test have been performed as part of the customer specification. Part of the 24 hour test is a 125% overload for 8 hours. Based on this data one can statistically evaluate aspects as average value, standard deviation and correlation on many parameters. Measurement tolerances have a large influence on the top oil exponent x and the winding exponent y. The calculation of x and y is determined by the ratio of two temperature differences due to a load difference, which can result in larger errors than expected. In two transformers fibre optic ( FO ) sensors were installed in the common winding to measure the hot-spot, not only during steady state, but also during transient conditions. For ONAF cooling, a step increase of load takes a relatively long time before the oil flow reaches a steady state, as is described in the loading guide. The overshoot in the gradient between the hot-spot temperature and the top oil temperature is demonstrated and can influence the hot-spot gradient exponent z. Based on the test results and a statistical simulation one can conclude that the normal tolerances in the temperature- and resistance measurements, result in a large standard deviation in the exponents x and y. Use of the exponents x and y, based on a heat run of one single transformer, should be handled with extreme care. In case of doubt, the use of the exponents given in the loading guide result in a safe margin when determining the overload capabilities of a transformer.