The dockless bike-sharing system has received widespread attention as a green mode of travel. To improve operational efficiency and promote orderly parking behavior, this article proposes a new two-stage location and allocation model based on the hexagon grids. In the first stage, we build a location model integrating the set-covering part and the flow-capturing part. The model comprehensively considers the walking endurance limit, the point-demand in the grid and the path-demand among grids to determine the optimal candidate grids. In the second stage, we establish two allocation models to calculate the renting and parking demand in each optimal candidate grid and obtain the final optimal grids. In addition, we verify the feasibility of cycling demand transfer by Moran’s index and design the effective spatial distribution of bike-sharing system by grid density function. Empirical results show that the number of the selected grids (8% are newly added grids) is only 53% of the initial grids without cycling demand lost. It means the new framework in this paper can effectively alleviate traffic congestion in crowded urban and ensure that travelers’ demand will not decrease at the same time.