Modelagem térmica de transformadores – método de regressão linear múltipla para previsão de variáveis

Abstract

For the proper operation of transformers, the temperature control is fundamental, since temperature is the main limiting factor for the load of the transformer, directly affecting its lifespan. In this context, this work aims to study the thermic behaviour of transformers using data acquisition, using the database generated by the supervisory system and climate data collected by a meteorological station installed in the substation area, using linear regression analysis tool, and searching for correlations between the temperature of the transformer and other operational parameters. The correlation analysis aims to verify the statistical inferences of the linear association parameters such as the multiple correlation coefficient, which measures the “strength”” or “degree” of linear relation between a variable and a set of other variables. This way, it is possible to generate a tendency for the temperature rise curve of the transformer and parameters such as its load. Therefore, an application to predict the thermal variables of a three-phase 230/138/13.8, 150 MVA transformer using oil, windings and room temperatures and active power collected in intervals of 5 minutes was developed. After treatment and tabulation of the collected information, various correlation situations were studied in each analysis, were one parameter was chosen as main or dependent. By the end of the simulations, the r-square value for each scenario was found, and consequentially the equations that represent the dependent variable. By recalculating the dependent variable through these equations, it was noticed that the best equation for the prediction of load and temperature situations was the one that used more independent variables, being the one closest to the value measured by the SAGE supervisory. Ultimately, through system modelling, this tool can be incorporated to the SCADA system in the operation centre of the power distribution companies to assist on equipment temperature prediction analysis on critical situations, or as an assist tool for data refreshment in case of a failure on the acquisition equipment.