FY-3E/HIRAS-II is the first early mooring orbiting infrared hyperspectral instrument in the world. Evaluating the quality of its observation data plays a very important role in improving the data assimilation and the accuracy of global numerical weather prediction. Based on the 35 days of HIRAS-II observations from December 2021 to January 2022 and March 2022, this paper uses the innovation vector method to assess the quality of the on-orbit observation data. The distribution characteristics of O-B deviation and standard deviation are calculated by land and ocean respectively. Further matching MetOp-B/IASI observation data in the same time period and in the same region, the double-difference method is used to analyze the quality of HIRAS-II observation data, which can eliminate the influence of radiation transfer mode simulation deviation. The results show that the O-B average deviation of long wave and medium wave in most channels is less than 0.5 K, and the standard deviation is within 1 K. The standard deviation on land is larger than that on ocean (especially for window channels). Due to the deviation of ERA5 reanalysis data, the radiation value simulated by RTTOV has a systematic error in the 664-665 cm^-1 CO₂ absorption band and the 1 300-1 680 cm^-1 water vapor absorption band, which makes the deviation larger, and the double O-B bias in these bands compared with MetOp-B/IASI is close to 0 K, indicating that the O-B bias is mainly caused by the simulation error of the radiation transfer mode, rather than the low quality of the instrument observation. The large deviation near the 980-1 080 cm^-1 O₃ absorption band and the 1 300 cm^-1 CH₄ absorption band is caused by the use of fixed climate profile values in RTTOV. The O-B average deviation of short wave in most channels is between -2 K and 2 K, and the standard deviation is within 2 K. The channels near 1 920 cm^-1 are the junction of medium wave and short wave of the instrument, so different detectors will cause large O-B deviation. The large deviation of 2 267-2 380 cm^-1 due to the fact of the NLTE effect (Non-local Thermodynamic Equilibrium) is not taken into account when RTTOV simulates the brightness temperature. The deviation and standard deviation of channels greater than 2 400 cm^-1 increase gradually due to solar pollution. HIRAS-II O-B deviation is asymmetric with the scanning angle, so it is necessary to correct the scanning angle deviation when using HIRAS-II data.