Abstract: In order to overcome traditional BP neural network's problem of trapping to a local optimum in prediction of urban water quantity, the gray neutral network model (GNNM) was created by combining BP neural network with gray theory, which can realize the mutual supplement with advantages of these two methods.The PSO-GNNM (1, N) algorithm was formed by using particle swarm optimization to optimize initialized weights and threshold of GNNM.By comparison with the predicted results of traditional BP neural network and gray theory forecasting method, this algorithm had little prediction error and better generalization performance, which can provide technical support for prediction of urban water quantity.