A large doubly-fed wind turbine is taken as the research object and the dynamic load characteristics of the drive train are modeled. Moreover， the identification algorithm of low-speed shaft torsion load of the transmission chain based on the combination of torsional vibration and generator torque characteristics is proposed as well. Applying the multi-body dynamics rigid-flexible coupling model in the simulation aims to verify the vibration characteristics of the transmission chain and dynamically identify the load. The simulation results are aligned with the dynamic identification of the load in a relatively high degree. The average deviation above the rated wind speed is about 2%， and the deviation of the 1 Hz equivalent fatigue load is around 6%. Furthermore， through the prototype test， the test results are close to the dynamic identified load and the deviation of the equivalent fatigue load is below 5%， which meets the engineering requirements. The drive train dynamic load identification method， using low-pass filtering algorithm， can achieve auto-programming which only needs the wind turbine self-detection and operational data， and can also obtain the low-speed axial torsion load during the operation of the wind turbine， which provides favorable support for the wind turbine security monitoring.