Distribution Transformer Oil Age Prediction Using Neuro Wavelet

Abstract

The distribution transformer is one of the vital components in the power system distribution, which deliver electricity power to the consumer. Various disturbances on the transformers can cause a decrease of their performance, so that they cannot reach the operation life. This study proposes a simulation study to predict the transformer oil age by using wavelet transform and backpropagation neural network. Transformer's current measurement was carried out in North Surabaya with a rating of 20 KV/380-220V and capacity of $100~\mathrm {k}\mathrm {V}\mathrm {A}$. The secondary current of the distribution transformer has been processed using the haar wavelet to obtain the detail coefficients, which is used to calculate the energy and PSD (power spectral density) value. Energy value and PSD are the input data on training and testing of back propagation neural network, while the output (target) is the transformer oil age. The simulation results show that the proposed method can predict the transformer oil age with an accuracy rate of 89.5795%.