The invention discloses a predistortion processing method based on a parallel structure. The method includes the following steps: (1) inputting data into a system to obtain a universal predistortion coefficient; (2) when the input data is stable, enabling the system to obtain a predistortion coefficient, and turning off an adaptive predistorter branch switch; (3) when the input data is stable, enabling an adaptive predistorter to update the predistortion coefficient; when the input data is unstable, turning off the adaptive predistorter branch switch; and when the input data is a wide-band high-power signal, returning to the step (1); (4) when the input data is a narrow-band low-power signal, turning off the adaptive predistorter branch switch, and returning to the step (3); and (5) when the input data in the step (4) is a wide-band high-power signal, returning to the step (1). According to the scheme of the invention, the problems existing in a traditional predistortion method can besolved, more excellent predistortion performance can be realized, and the power amplifier output has better linearity.