The ultrasonic method (AE) is more and more widely used in partial discharge (PD) pattern recognition, however, it is difficult to detect the PD level by AE method. So it is necessary to find the relationship between the acoustic waves and the PD level for different PD patterns. In this paper, an approach to detect the degree of PD by AE method based on feature extraction and correlation degree analysis has been proposed. In this work, three typical discharge models were designed, then the discharge amount (Q) from electrical signals waves, and ultrasonic amplitude (A) duration time (DT) and ultrasonic energy (E) from ultrasonic signals waves are extracted as feature parameters. In the experiment, for each PD model, 50 signals of PD are recorded to find a relation between ultrasonic signal and electrical signals. Through analysis, it can be found that there are close mathematical relationships between the discharge amount Q and three characteristic parameters (A/DT/E) of ultrasonic signals. This provides a method for further research on the quantitative monitoring of partial discharge by ultrasonic method.