The ESA mission EnVision will address its main scientific questions through a detailed mapping of the surface and interior properties of Venus. A precise reconstruction of the spacecraft trajectory is a key requirement for the EnVision scientific investigations, including radio science. To precisely constrain the orbit evolution, refined models of the dynamical forces are included in the Precise Orbit Determination (POD) process. We developed a methodology based on a batch-sequential filter that enables a joint estimation of Venus gravity and atmospheric density. Our approach yields an accurate compensation of atmospheric mismodeling, simulated through semi-empirical predictions of the atmospheric density provided by general circulation models (GCM), e.g., Venus Climate Database (VCD). Numerical simulations of the EnVision radio science investigation were carried out by using a perturbative analysis of the dynamical forces, which accounts for atmospheric density errors ≥ 200%. By adjusting a set of atmospheric scale factors, our proposed strategy enables an estimation of the atmospheric density at the spacecraft altitudes with an accuracy of 25%. The improved dynamical model yields accuracies in the orbit reconstruction of 1-2 m, 30-40 m and 20-30 m in the radial, transverse and normal directions.