In this paper, it was investigated the characteristics of PD(partial discharge) pulse shapes in oil and the possibility of TF(Time-Frequency) analysis for the discrimination of air corona in on-site PD measurement of oil transformers. For the purpose, single and multiple discharges combined with oil discharges and air corona were generated and measured by use of artificial models through the HFPD pulse detector and the oscilloscope. PD pulse shapes were different according to the type of defects, especially including air corona- pulses. Also, through the TF analysis, data clusters could be classified each other in the TF Map, even in the case of multiple discharges. Therefore, it is known that TF analysis could be aprlied for the improvement of on-site PD measurement in oil transformers.