Beam pumping units are the main production modes in the oilfield oil recovery process. As the power-driven equipment is driven by larger electrical power of the pumping motor, the high energy cost problem is widespread and hard to solve. The motor working efficiency is difficult to accurately match to the complex dynamic loads, as the relationship between the motor driven torque and dynamic loads is hard to master precisely, and the accurate calculation of some important variables cannot be effectively solved by the traditional mechanism modelling method. To solve this problem, in this paper, the dynamic changes of the polished rod loads are considered as the dynamic load changes of the beam pumping units and a serial hybrid model for the motor load torque is studied. Firstly, according to the working principles of the beam pumping units and the mechanism models of the pumping motor, belt pulley-gear box and four-bar linkage mechanism are respectively built. Secondly, for the underground frictions, a rough mechanism model based on the energy conservation equation is built first and then a data-driven soft sensor model is adopted to obtain the dynamic fluid levels instead of an off-line measurement, which uses a C index based on the on-line verification method to prevent the over-fitting in data-driven modelling process. Through the research of this article, a serial hybrid model is proposed to build the dynamic relationship between the pumping motor load torque and the polished rod loads, the underground frictions can be taken as a dynamic variable related to the time and the down-hole dynamic fluid level can be obtained by an on-line calculation. The proposed hybrid modelling method is applied to four normal oil wells, and comparisons between the simulation results and the actual test results verify its effectiveness.