Abstract: An artificial neural network method to distinguish reservoir water temperature layering model was developed. This method describes indirectly physical mechanism and physical model of water temperature layering,by means of setting feedforward network and using example training,making network to simulate better mapping relation between water temperature layering model influence factor and water temperature layering model. Comparing with common method,this method has notable characteristics as follows:good classification accuracy,strong adaption ability,applying to linear and nonlinear classification problem.