Artificial neural network modeling of partial discharge severity in XLPE cables under various soil conditions

Partial discharge (PD) activities are one of the major causes of the insulation failures in underground power cables. Nevertheless the effect of soil conditions on PD severity in power cables is not yet well established. In this paper, an experiment was carried out to study the effect of soil thermal resistivities (/spl rho/) and loading of the cable (T) on various PD parameters e.g. PD magnitude (Q), PD counts (n), and inception voltage (V) in an underground power cable. The sample of the cable studied is 11 kV, 240 mm/sup 2/, XLPE, single core cable. The experimental results clearly indicate that the soil thermal resistivity and cable loading have a significant effect on the PD parameters. It is well known that Artificial Neural Network (ANN) works well as a function estimator to model accurately real world complex relationships. In this paper an attempt has been made to model (Q,n,V)=f(/spl rho/,T) for estimating the PD severity under various soil conditions using multi-layer feed-forward neural network with back propagation technique. Once trained, the proposed ANN model (Q,n,V)=f(/spl rho/,T) is then capable of predicting PD severity, under any given set of /spl rho/ and T with a mean absolute error (MAE) less than 5%.