Predicting the CFO voltages of distribution lines insulation using radial basis functions neural network

This paper presents an artificial neural network (ANN) based approach to predict the total critical flashover (CFO) voltages of two series dielectrics used on the constructions of the distribution lines. Multiple radial basis functions neural networks (RBFNN) have been used to predict the total CFO voltages for various groups of combinations of two series dielectrics. The CFO voltage of the composite components was predicted from the corresponding CFO voltage of each individual dielectric. The results show that the proposed approach is very promising and encouraging for fast prediction of the total CFO voltage of a composite insulation and with very high accuracy. In addition, it has the ability to map any arbitrary complex nonlinearity of the experimental data.