Previous work has shown that changing the velocity equation in Particle Swarm Optimisation (PSO) has an impact on the performance of the PSO. These changes have been traditionally done manually. This study investigates automating this process by evolving the velocity equation using grammatical evolution (GE). A grammar is defined to specify the syntax of the equation and options for the components making up the equation. The proposed approach was evaluated on a set of 16 continuous optimisation functions for 30 and 100 dimensions. PSO with the equations evolved by GE was found to outperform PSO with the standard velocity equation for all 16 functions. These results were found to be statistically significant at a 99% level of confidence. The reusability of the velocity equations, i.e., the ability of the velocity equation evolved for a subset of functions to produce good results for all functions, is also assessed. The study revealed that PSO with the reusable equation performed better than canonical PSO and comparably with the disposable velocity equations, i.e. equations evolved for a particular function. Furthermore, the results produced by PSO with the evolved equations performed better or the same on 15 of the 16 functions as state of the art PSO approaches applied to the same functions.