The object was to solve the dynamics analysis and multi-target performance optimization of 3-P(4S) parallel mechanism. Firstly, inverse kinematics of 3-P(4S) parallel mechanism was solved, and hybrid numerical algorithm was proposed, which was composed of BP neural network and quasi-Newton method algorithm. The algorithm could solve the forward kinematics of 3-P(4S) parallel mechanism in less than 20ms by three step iterations, and the accuracy was on the level of 10－8, which can realize the high accuracy and real-time control of 3-P(4S) parallel mechanism. Then the velocity and acceleration of the moving platform and limbs were obtained by the derivative of the position solution. According to the results of kinematic analysis, the dynamic model of the mechanism was constructed by Newton Euler method, and the dynamic simulation of the numerical example was utilized to verify the dynamic model of 3-P(4S) parallel mechanism. From the verification results, both of the two were exactly the same. Finally, considering the dynamic performance, stiffness performance and speed performance of the 3-P(4S) parallel mechanism, an improved genetic algorithm was utilized to optimize the 3-P(4S) parallel mechanism. Through the multi-objective performance optimization, the dynamic performance and speed performance of the mechanism were improved by two times, and the stiffness performance was increased by three times.