The accuracy of pattern recognition of PD is closely related to the completeness of discharge samples. At present

most of the existing defect sample libraries are obtained from the discharge data of a single and typical void defect in the laboratory and it is difficult to cope with the complex conditions on the site

which will lead to the poor effectiveness in the diagnosis and identification for void defects on the field. Therefore

the effect of the variety of void defects on the discharge is considered in this paper and two kinds of void defects are designed in this paper including 3 kinds of single-void defects with different shapes and 3 kinds of multi-void defects with different numbers and sizes. The discharge signals of two types of void defects model are collected and processed by ultra-high frequency（UHF） detection system of PD. The results show that there are differences in the discharge characteristics of different types of void defects

the influence of factors such as shape and size on the discharge characteristics can be reflected in the characteristic spectra. The discharge characteristics such as phase and amplitude of single void defects with different shapes have obvious differences

and the discharge spectra of multi-void defects with different numbers and sizes are also different. Adding the discharge characteristics of void defects considering multiple factors into the partial discharge library is beneficial to the subsequent development of pattern recognition of PD.