Based on the annual water consumption data from 2005 to 2014 in Guangxi, the gray GM (1,1) model was used for prediction. In order to improve the prediction precision, the traditional grey GM (1,1) model was improved in different ways. It is found that the prediction results of the four grey models are ideal, and the average precision reached 99.5%. The precision of the traditional grey GM(1,1) model is 99.0%, the function transformation improved grey model is 99.4%, the revised residual grey model is 99.7%, and the weakening operator treated grey model is 99.9%, also the reliability of water consumption in Guangxi in grey model are fully verified.