Although the gas insulated structures have a high degree of reliability, the unavoidable defects are primary reason of their failures. Partial discharge (PD) has been regarded as an effective indication for condition monitoring and diagnosis of gas insulated switchgears (GISs) to ensure their reliable and stable operation. Among various PD detection methods, the ultra-high frequency (UHF) technique has the advantages of on-line motoring and defect classification. In this paper, there are presented 7 types of artificial electrode systems fabricated for simulation of real insulation defects in gas insulated structures. A real-time measurement system was developed to acquire defect patterns in a form of phase-resolve partial discharge (PRPD) intensity graph, using a UHF sensor. Further, the discharge distribution and statistical characteristics were extracted for defect identification using a neural network algorithm. In addition, a conversion experiment was proposed by detecting the PD pulse simultaneously using a non-induction resistor and a UHF sensor. A relationship between the magnitude of UHF signal and the amplitude of apparent charge was established, which was used for evaluation of PD using the UHF sensor.