Grey wolf optimization (GWO) is a kind of swarm intelligence algorithm which simulates the rank system and predatory behavior of wolves. It has some shortcomings such as low convergence accuracy, easily falling into local optimal solution and so on. In order to improve the performance of the GWO algorithm, this paper proposes an improved gray wolf optimization(CGWO) algorithm based on the Tent mapping and normal cloud model. In the initial population stage, the algorithm employs the Tent mapping to make the population evenly distributed in the search space to improve the optimization efficiency. In the stage of attacking prey, the normal cloud model is used to update the location of the wolves, so that the algorithm has better randomness and fuzziness in the early stage, which improves the ability of global exploration and local optimal solution avoidance. In the later stage, the entropy of the normal cloud model is decreased with the increase of the number of iterations, hence, the randomness and fuzziness are reduced, which effectively improves the local exploitation ability and the convergence accuracy. 20 international standard test functions are selected to benchmark the performance of the CGWO algorithm, and the optimization results of unimodal, multi-modal and composite function are compared with various optimization algorithms. The results show that the CGWO algorithm is improved in convergence rate and accuracy, and has better balance between global exploration ability and local exploitation ability.