Nonparametric Kernel Density Estimation Model of Transformer Health Based on Dissolved Gases in Oil

Abstract

In this paper, a health status calculation method based on nonparametric kernel density estimation of dissolved gas in oil and association rules for transformer is proposed. Firstly, the online monitoring data of dissolved gas content which collected from multiple identical transformers are analyzed by nonparametric density estimation to obtain the health probability distribution function of various gases. Then, by mining the correlation between various gases and transformer fault conditions, an association rule method to calculate the weight coefficient of each gas is introduced. Finally, the weighted method is applied to calculate the health probability of transformer when the health probability and weight of each gas are obtained. The method introduced in this paper is validated by the state parameter data of transformer in a substation. The example shows that the health status of the transformer can be obtained in real time and this method is completely based on data driven, which is important to ensure the safety of grid.