<p>A multi-feature semi-parametric modeling method is proposed aiming at the inverse dynamics modeling problem of the 7-DOF heavy-duty hydraulic manipulator (HDHM), which is widely applied in construction machinery. The method can compensate for the influence of nonlinear factors on manipulator inverse dynamics modeling, and further improve the accuracy and generalization ability of the dynamics model. First, the construction process of the parametric model is de-scribed in detail. Then, based on recurrent neural network (RNN), a multi-feature semi-parametric dynamics model (MSDM) is constructed and validated under two operating conditions (high-speed and medium-load, high-speed and heavy-load) of the 7-DOF HDHM. The experimental results show that the MSDM model performs optimally compared to the other models mentioned in this paper, laying the foundation for path optimization and control of the 7-DOF HDHM.</p>