To get accurate parameters in both static and dynamic models, an identification method of model parameters based on the analysis of steady state error was presented. Firstly, the relationship between steady state errors and input and interference signals in a servo system was determined, and the influence of torque ripples on the parameter identification was eliminated based on the study above. Then, the friction torque was deduced by using the steady state error, and the static and dynamic parameters were identified through the genetic algorithm. Finally, the friction torque was compensated according to the identified model, and the compensated effect was analyzed. The experimental results show that the steady state error of uniform motion has decreaseed from 36 μm to ±3 μm, that of uniformly accelerated motion decreased from 34 μm to ±3 μm, and that of sinusoidal motion decreased from 35 μm to ±7 μm. It concludes that the dynamic and static parameters of LuGre can be precisely obtained by this identification method, and the tracking accuracy of the servo system can also be improved through friction compensation on the basis of the proposed model.