Abstract: To address the randomness and periodicity of the residential water consumption (RWC) data along with overfitting problem caused by the large dispersion of traditional gray model, a dynamic gray model group consisting of five GM(1,1) models was proposed based on gray model theory. Based on the annual RWC data of Hengshui City from 2007 to 2019, the dynamic gray model group was used to project the future changes of annual RWC in Hengshui City during 2020-2030, and meanwhile residual tests and corrections were conducted using the projected results; the dynamic gray model group was compared with five GM(1,1) models to test the model accuracy. The results showed that the projected relative error of the dynamic gray model group was smaller than that of the traditional GM(1, 1) model, implying better accuracy and applicability. The annual RWC in Hengshui approached 17.95 million m³ in 2019 and was expected to increase to 28.62 million m³ in 2030, which indicated that the future RWC in Hengshui City would be at an obvious uptrend, and this result was in line with the future population growth and socio-economic development trend. The projected results of RWC in this study was capable of providing reference for optimal water supply and water resources allocation in Hengshui City.