The water demand forecast can supply vital basis for reginoal water resources plan, utilization and management. On the basis of analysis of the influence factors of residential domestic water consumption in Baotou City by gray correlation analysis, the paper forecasted the regional domestic water consumption in 2009 and 2010 by modeling a multiple linear regression model, gray model (1,1) and gray linear combination model, and compared their prediction accuracy. The results showed that the main factors to regional domestic water consumption were population of urban water, per capita living space and water price. Forecast of water consumption in 2009 and 2010 had highest accuracy by gray linear combination model; relative errors were 13.6% and 6.5% respectively; relative error of root mean square was 10.7%. Prediction accuracy of combination forecast model was better than that of single one and made results more accurate, reasonable and realistic.