Partial discharge patterns are an important tool for diagnosis of HV insulation systems. Skilled humans can identify the possible insulation defects in various representations of partial discharge (PD) data. One of the most widely used representation is phase resolved PD (PRPD) patterns. This paper describes a method for the automated recognition of PRPD patterns using a novel composite neural network system for the actual classification task. This paper elucidates the possible methods of extracting relevant features from the PRPD data in a knowledge based way i.e. according to physical properties of PD gained from PD modeling. This allows the novel complex neural network (NN) system for classification. The efficacy of composite neural network developed using original probabilistic neural network is examined. This innovative methodology of giving inputs to the composite neural network compares favorably with the traditional network architecture used previously for PD pattern recognition.