Severe haze occurred in Beijing frequently during December 2015, which had launched two red alerts for atmospheric heavy pollution. In this paper, according to the lidar stereo data at the tower of the institute of atmospheric physics of the Chinese academy of sciences, the atomspheric boundary layer height of Beijing city is retrieved by gradient method, and the simulation results of the mesoscale numerical WRF model are evaluated. The results show that the two methods have good consistency, but the extreme values are not very consistent. The correlation between lidar inversion of daily variation of boundary layer height and WRF simulation results reaches 0.76, the root mean square error is 163 m, and the average deviation is -61 m. Meanwhile, the accuracy of WRF simulation in clean weather is higher than that in polluted weather. In addition, the correlation between the observed PM2.5 mass concentration and the ABLH of lidar inversion is -0.85.