Aiming at the water quality pollution and euttophication problems for drinking water source of lake type, the synthesized throphicstate index (STSI) model was established based on throphic state index (TSI) and entropy weights to evaluating lake eutrophication statusthrough calculating STSI. Based on neural network simulation theory and Matlab software, BP algorithm model was improved through additional momentum method and the learning rate self-adjustment. Improved BP model of 5-3-1 type was established to simulate and predict lakeeutrophication. The synthesized throphic state index (STSI) model and improved BP model were applied to Gucheng Lake to evaluate and pre -dict throphic state, and the model results were verified. The results show that the improved BP model can determine the water body ’s state,which provides a new method to evaluate and predict eutrophication.