In order to solve the problem of multiple impact factors of urban water demand, slow convergence speed of BP neural network, low precision and easily falling into local optimum, this paper proposes an improved forecasting model based on the combination of grey relational analysis, mind evolutionary algorithm (MEA) and back propagation neural network (BPNN). grey relational analysis was adopted to select the main factors that influence the water requirement, and the weights and threshold values of BP neural network was optimized by using mind evolutionary algorithm (MEA) which has a strong global search ability to build the GRA-MEA-BP water demand forecast coupling model. And BP neural network model was established for comparison as well. The application results show that the GRA-MEA-BP coupling model has higher prediction accuracy and prediction speed. Thus, it can be used as an effective model for water demand forecasting.