A global competitive harmony search algorithm(GCHS) is proposed. In this algorithm, the conceptions of stochastic local mean and global mean are given. The competition search mechanism is built to realize two harmony vectors are competition selection, and the two harmony vectors are both generated in the each iteration. The adaptive global pitch adjustment and local learning strategy are designed to balance the global search and local search. The effects that the parameter HMS, HMCR and PAR have on the performance of the GCHS algorithm are also analyzed in detail. The numerical results show the superiority of the proposed GCHS algorithm in terms of accuracy, convergence speed, and robustness when compared with the harmony search algorithm and other seven state-of-the-art harmony search algorithms.