Abstract: China is a big country of water resource and in the accelerating process of urbanization,it is facing a series of major challenges such as urban population growth and water pollution,etc,so scientific and rational forecasting to water resource demand becomes a key task to protect environment and maintain sustainable development.This paper summarized various neural network algorithms in the context of water resource demand forecasting,and introduced fuzzy feedback method to improve entropy method to determine the weighting factor of combination forecasting model,to establish neural network forecasting model.The algorithm can not only automatically deduce future change trends of water resources based on historical data,but also introduce feedback and evolution mechanism,so that the users can adjust the solution accuracy to control convergence speed of algorithm.The experiment shows that,the neural network based on combination model proposed in this paper has better performance in application background when data accuracy is not high and hydrological data is incomplete.

Key words: Prediction for water resource requirement,Neural network,Combination model,Fuzzy feedback