We have developed an efficient scheme for the generation of accurate repulsive potentials for self-consistent charge density-functional-based tight-binding calculations, which involves energy-volume scans of bulk polymorphs with different coordination numbers. The scheme was used to generate an optimized parameter set for various ZnO polymorphs. The new potential was subsequently tested for ZnO bulk, surface, and nanowire systems as well as for water adsorption on the low-index wurtzite (101̅0) and (112̅0) surfaces. By comparison to results obtained at the density functional level of theory, we show that the newly generated repulsive potential is highly transferable and capable of capturing most of the relevant chemistry of ZnO and the ZnO/water interface.