Correcting the effect of temperature changes in the measurement results of dielectric response in power transformers as an arrhenius type dielectric with the help of artificial neural network

Abstract Polarization and Depolarization Current (PDC) and Frequency Domain Spectroscopy (FDS) measurements are common dielectric response tests in time and frequency domain that are widely used to diagnosis insulation status in power transformer companies. Numerous factors affect the PDC and FDS test results, which the most important of them is temperature variation. To accurately interpret the results of the dielectric response, the effect of temperature on the results in time and frequency domains, must be corrected. In this paper, firstly a 200 MVA transformer is selected as a test object and FDS and PDC tests are performed on it at different temperatures. Then, based on the FDS results performed on a transformer at different temperatures, the parameters of the insulation model have been estimated by using genetic algorithm (GA). Next, with the help of artificial neural network (ANN), the parameters of the insulation model, related to the different temperatures are transferred into the reference parameters. After that, the parameters of the transferred insulation model, the FDS curves are plotted and transferred to the reference temperature curve and the effect of temperature on them is compensated. By using the correlation of the time and frequency domain results, with the help of transferred insulation model parameters, the PDC results are also plotted and transmitted on the reference PDC curve. Finally, with such a method, the effect of temperature on the PDC results is compensated.