Abstract: Thermospheric neutral temperature and atomic oxygen density are key parameters for investigating thermosphere–ionosphere coupling and the orbital evolution of low-Earth-orbit spacecraft. Traditional energy-balance inversion methods based on incoherent scatter radar (ISR) rely on empirical models to provide boundary and shape parameters, and it is difficult to obtain the global optimum after least-squares fitting. To address these limitations, this study proposes a new thermospheric parameter inversion method based on a particle swarm optimization–sequential quadratic programming (PSO-SQP) hybrid optimization algorithm. Based on the ion energy balance equation, the proposed method introduces a Pseudo-Huber robust loss function and a dynamic height-screening mechanism, enabling the inversion of neutral temperature and atomic oxygen density within the effective dynamic height range of the thermosphere. Inversion results using observations from the Arecibo and Qujing incoherent scatter radars show that the hybrid algorithm enables independent retrieval of key thermospheric parameters and eliminates the dependence on empirical model parameters. Physical closed-loop validation indicates that the reconstructed ion temperature is highly self-consistent with radar measurements, with a correlation coefficient of r = 0.9380. In addition, Monte Carlo error analysis further demonstrates the strong robustness of the proposed method under Gaussian observational noise. The results of this study can support fundamental research on the thermospheric atmosphere and the development of orbital models for low-Earth-orbit spacecraft, indicating important scientific significance and practical value.