A neural network model analysis of the residual chlorine in the water distribution system is reported. We use one year data of the clorine concentration to determine network parameters and predict the next year's data. We find the neural network has a good prediction ability. To improve it，we update the network parameters every day using a certain period of data.