In order to establish the relationship model between grinding parameters and deformation error of workpiece caused by grinding force, force error modeling and realtime compensation method based on neural network and finite element simulation were proposed. A BP neural network optimized by genetic algorithm was established to characterize the relationship between grinding parameters and grinding force. Then, the mechanical analysis of the parts was carried out by using the finite element method, and the relationship model between grinding force and force deformation was established. Finally, the error mapping model of the machining parameters and cutting force was established to predict the amount of error compensation and make realtime compensation. Experimental results show that the error model of the cutting force is accurate and effective, and has high application value.