In response to the problem of the unmeasurable aerodynamic parameters in the dynamic simulation of parachute recovery systems, a six-degree-of-freedom dynamics equation and kinematic equation were established for the stable descent phase of the parachute system. The form of aerodynamics and the parameters to be identified were defined. In the foundation above, two aerodynamic parameter identification schemes based on BP neural networks were employed. These schemes involved training the neural networks with flight state data until convergence, resulting in the identification of the aerodynamic parameter model to be discerned. The effectiveness and accuracy of the two identification schemes are verified through simulation examples. The identification results of aerodynamic parameters are obtained separately, and performance evaluation metrics are calculated. The simulation results are analyzed in terms of convergence speed, identification accuracy, and other aspects, indicating that both identification schemes exhibited good agreement between predicted results and expected results. However, the double BP neural network method demonstrated superior performance. The findings of this study demonstrate the potential applicability of BP neural network methods in the identification of experimental data in future engineering applications.