China’s BeiDou satellite navigation system (BDS) has completed its full constellation in orbit since June 2020. Services have been evolved from regional (BDS-2) to global (BDS-3). This contribution evaluates the impact of solar radiation pressure (SRP) modeling on satellite orbits and geodetic parameters. To that end, we process 2 years of BDS observations (2019-2021), collected by a network of 100 ground stations. A physical a priori box-wing (bw) model based on the estimated optical properties is introduced. Various physical effects, such as yaw bias, self-shadowing, radiator emission and thermal radiation of solar panels are considered. The ECOM (Empirical CODE orbit Model, 5 parameters), ECOM+along-track and ECOM2 (both 7 and 9 parameters) models are employed on top of the a priori box-wing model in the experiment. We show that without the use of the a priori box-wing model, the ECOM+along-track model shows clear better orbit solutions during eclipse seasons for BDS-3 satellites. This is proven to be mainly due to the thermal radiation of solar panels. However, the along-track acceleration is highly correlated with LOD (length of day) and ECOM parameters. LOD estimates in this case are contaminated. The STD (standard deviation) of daily LOD estimates with respect to IERS-C04-14 series increases from 40 us (ECOM) to 85 us (ECOM+along-track). After the consideration of the a priori boxwing model, satellite orbital errors are greatly reduced for all the ECOM models. For instance, orbit misclosures of BDS-3 CAST (China Academy of Space Technology) satellites improve by a factor of two for the ECOM model during eclipse seasons; dependencies of SLR (satellite laser ranging) residuals on the sun elongation angle almost vanish for BDS-3 satellites. Furthermore, the use of the a priori box-wing model mitigates a great majority of the spurious signals in the geodetic parameters. In particular, the total amplitude of the 1, 3, 5, 7 cpy signals for the geocenter Z component has been reduced by a factor of 4.5 for the ECOM model. In general, the combination of the introduced physical a priori box-wing model and the ECOM model is preferred for BDS satellites.