The prediction of hydrodynamic parameters of offshore structures is one of the important problems in the field of offshore platform design and production, and how to predict and analyze it is always difficult. Aiming at the identification and analysis of hydrodynamic parameters of offshore platform, a structure motion model based on physical mechanism neural network is established by combining the platform motion data, and the hydrodynamic parameters are inversed. The research shows that the neural network method can accurately fit and invert the numerical results, and the inversion of the experimental results shows that the added mass parameters obtained are accurate, and the damping components of the first-order without potential flow damping at different frequencies are obtained, which provides a new method for identification and analysis of hydrodynamic parameters.