Abstract: The signal bandwidth, modulation and peak to average power ratio (PAPR) of the fifth generation mobile communication system new radio (5G NR) are much higher than those of 4G signal. The nonlinear distortion of the radio-over-fiber (RoF) fronthaul system used in small cell or micro cell base stations is more serious. In order to make 5G NR signal transmit linearly, this paper uses the memory polynomial (MP) and generalized memory polynomial (GMP) models to construct digital predistorters to correct the nonlinearity of 5G RoF fronthaul system. In this paper, a RoF fronthaul system with 2km fiber is built, and the linearization test is carried out with a 100 MHz bandwidth 5G NR signal. The experimental results show that using the MP and GMP digital predistorters for linearization, the adjacent channel power ratio (ACPR) of the 5G RoF fronthaul system can be improved by more than 13 dB and 17 dB respectively, which indicates that the MP and GMP digital predistorters can significantly suppress the nonlinearity of 5G RoF fronthaul system. Therefore, the digital predistorter based on MP and GMP models can be used for the linearization of the 5G RoF fronthaul system, and the GMP predistorter has stronger linearization ability for the system than the MP.