Abstract: The kinematic precision of the Delta parallel manipulator is reduced during movement because of static errors. In this study, a new method for identifying the kinematic parameters is proposed. Firstly, an error model is established that involves nine main sources of errors. A modified particle swarm optimization algorithm based on particle classification and dynamically changing weight by dimensions (PC-DCWD) is then produced. This algorithm introduces an average extreme value to classify populations, and it adopts asynchronous evolutionary strategies for different particles while enhancing collaboration between populations. This introduces the concept of a distance factor to dynamically change the inertia weight by dimensions, while considering that different dimensions of the same particle have differences when searching the best location. By comparing the PC-DCWD algorithm with the linearly decreasing weight particle swarm optimization algorithm, and testing five classical functions, the simulation result shows that the PC-DCWD falls into the local optimum effectively, increases the accuracy and convergence rate of the algorithm, and has a better performance in optimization. Finally, by identifying nine main errors in the error model, the simulation result shows that the identification value is almost equal to the ideal value. Therefore, the proposed algorithm is considered effective and practical.