Abstract: To effectively analyze the atmospheric drag characteristics in Very-low Earth orbit (VLEO) and mitigate biases of traditional empirical models, this study conducts simulation experiments on satellite samples at two typical orbital altitudes of 350 km and 450 km. Relying on 7-day predictive ephemerides in the Consultative Committee for Space Data Systems Orbit Comprehensive Message (CCSDS OCM) format and measured two-line element (TLE) data of the SpaceX Starlink constellation, a novel analytical framework for atmospheric drag characteristics of low-Earth orbit satellites is established. The main innovations of this framework lie in fully exploiting the maneuver information contained in the public predictive ephemerides, so that thrust and atmospheric drag effects can be separated and cross-validated without relying on accelerometer data. First, a high-precision transformation algorithm between osculating and mean orbital elements is adopted to reduce estimation errors of orbital altitude decay induced by osculating elements. Subsequently, a least-squares numerical fitting method optimized via strategic point selection is utilized to improve the fitting precision of atmospheric drag acceleration. Meanwhile, precise orbit determination is performed using non-maneuver segment data extracted from predictive ephemerides to solve for the satellite ballistic coefficient. Finally, historical TLE data are applied to retrieve the true atmospheric density in low Earth orbit, enabling simultaneous evaluation of satellite on-orbit status and dynamic analysis of spatial atmospheric density variations.Experimental results demonstrate that the atmospheric density of the NRLMSISE-00 model retrieved from predictive ephemerides achieves high consistency with native model outputs, which verifies the validity of the proposed ballistic coefficient estimation and atmospheric density retrieval methods. The overall deviation between the true atmospheric density retrieved from TLE data and the NRLMSISE-00 empirical model is approximately 30%, revealing altitude-dependent systematic biases within the empirical model. Atmospheric density increases by 5 to 7 times when the orbital altitude decreases by 100 km, and the atmospheric density at a fixed altitude can vary by nearly threefold over half a year. The proposed research system can provide references for atmospheric density correction and the improvement of orbital prediction accuracy.