Abstract: Power transformer is an important equipment for energy conversion in power system. Any faults in it will seriously affect the reliability of power supply and operation stability of the power system. Therefore, the accurate identification of various types of faults inside the transformer is of great importance. The failure of the transformer will result in the generation of distinct audible sound and visible light signals. Studying the signal characteristics under different discharge forms is beneficial for the precise identification of internal faults in transformers. In this paper, three discharge models were designed to simulate the non-uniform field discharge, surface discharge and uniform field discharge of the inner pole of the transformer. The audible sound and visible light signals were measured and obtained under various defects, and the differences and patterns between the signals were compared and analyzed. It is found that in terms of audible acoustic signals, the main frequency of extremely uneven field discharge was 1980 Hz, the main frequency of surface discharge was 200 Hz, and the main frequency of uniform field discharge was 611 Hz. In addition, the visible light wavelength of surface discharge was primarily concentrated in the range of 360-540 nm. Moreover, the visible light energy of discharge under extremely uneven field and uniform field was predominantly focused around 482.8 nm and 660.8 nm. It is impossible to realize fault identification by visible light signal alone. The main frequency characteristics of audible acoustic signals for the two types of defects differ, however, By combining these two types of signals, fault identification can be achieved. The research results of this paper provide data support and guidance for transformer online monitoring.