Aimed at the recognition of nutritional status for lake and reservoir not considering the indicators of nutritional status and weights index classification being less, based on entropy and probabilistic neural network (PNN) basic principles,the paper proposed entropy PNN recognition model of nutritional status of lake and reservoir, and took 24 lakes and reservoirs in the country for example analysis. By use of entropy method to determine recognition index weight, according to evaluation criteria of eutrophication of lakes and reservoirs,the paper classified lakes and reservoirs as 11 levels from extremely poor nutritional status nutritional to very severe eutrophication, proposed the grade standards of identification based on improved nutritional status of lakes and reservoirs. In grade standard field values,the paper interpolated between the use of methods to generate a random sample of PNN model for training and testing,and used correct recognition rate and duration to evaluate the performance PNN model. Finally,based on two programs ,it identified the nutritional status of 24 nationwide lakes and reservoirs. The results showed that ① PNN model for randomly generated training and testing samples, correct recognition rate reaches 98.9% and 98.6% (5 times the average), the model used to identify the nutritional status of lakes and reservoirs is reasonable and feasible. ② recognition results of lakes and reservoirs exist difference. In comparison, the recognition result considering index weights more scientifically and objectively reflect the nutritional status of lakes and reservoirs.