Impact of Electric Vehicles on Distribution Transformers: Influence of Modeling Errors

Abstract

This paper addresses different strategies for modeling the variables (ambient temperature and electricity demand) involved in a thermal model used to calculate the operating temperature of a distribution transformer, which affects the service life of its cellulose-based insulating material. Since both variables exhibit seasonality that roughly coincides with the seasons of the year, they are divided into four parts and modeled using nonlinear regression, varying the number of terms in the resulting model. The annual results are then compared with those from the stratification, and the optimal number of terms is determined based on the coefficient of determination and the standard deviation. The distribution of errors and outliers is analyzed, followed by the application of an iterative method to generate the thermal model results. It is concluded that considering ten terms with stratification considerably reduces the dispersion of the results, and the application of an iterative method such as the Monte Carlo method prevents the selection of unlikely outliers.