Partial discharges measurements are an important tool for the determination of the degradation level of high voltage apparatus. If partial discharge is found in insulating systems then it is important to identify its character, i.e. internal discharge, surface discharge, corona etc. The aim of this paper is present a methodology for the generation of partial signals with different characteristics using two developed partial discharge cell, that allows the generation of well defined PD pulses, that can be used in creation of patterns that allow the automatic (aid computer) defects classification based on the employment of artificial intelligence techniques.