A novel approach for partial discharge pattern recognition based on FP3CM-ENN

Partial Discharge (PD) pattern recognition is a main tool for assessing the insulating ability of high voltage (HV) equipment. A new PD identification method found on the fuzzy-possibilistic product partition c-means (FP
³
CM) — extension neural network (ENN) for HV equipment is proposed in this paper. A PD detector is familiar to assess the PD patterns. The Statistical feature extraction method is adapted to take out the features for every PD pattern. This advanced method combines the FP
³
CM clustering algorithm with the ENN architecture. The FP
³
CM is framed by the product of probabilistic and possibilistic fuzzy partitions instead of linear combination. This clustering model produces very accurate partitions in the existence of noise data. The centers of the FP
³
CM are used as the initial centers of the ENN. The FP
³
CM-ENN algorithm adopts extension distance to compute the resemblance among tested data and cluster centers. This proposed FP
³
CM-ENN algorithm procures the benefit of high precision, robustness and less memory consumption, that are helpful in PD recognition. To verify the efficacy of the suggested method, similar studies with a MLP, ENN, fuzzy c means (FCM) — ENN, Possibilistic FCM (PFCM) — ENN are carried out for different samples with rather encouraging results.