Prediction technique for transformer oil breakdown voltage via multi-parameter correlation based on grey theory and BP neural network

Prediction of breakdown voltage of transformer oil has a great significance to the fault diagnosis and daily maintenance of transformer. Based on the correlation of performance parameters of transformer oil and the prominent fault-tolerance, non-linear approximation, and self-learning capabilities of BP neural network, this paper constructed a prediction method of breakdown voltage via multi-parameter correlation under the development environment of Matlab. Through examining the routine monitoring data of transformer oil by gray correlation analysis, this paper excavated some parameters which have strong correlation with breakdown voltage. Then a relational model of breakdown voltage and those parameters was further constructed using BP neural network. The clustering centers used to train network were acquired through clustering the original monitoring data samples with fuzzy C-means clustering algorithm. This method which uses clustering centers to train network can resolve natural problems of neural networks caused by large sample capacity, such as complication of net structure, inferior astringency, poor generalization ability, and so on. Test results show that the relative errors between predicted values and real values are all less than 10%, which indicates the significant practical values of this model.