Abstract & Overview
A model is set up to evaluate the lake trophic status with the BP neural network.It is in accordance with the Chinese lake trophic evaluation standards and the principle and theory of the BP neural network.To overcome the shortcomings of the standardized BP neural network learning,that is the slow convergence and susceptible to partial peak value,the model is set up to get the training and test samples at different range of grading by random inter-plug.When both samples meet the required accuracy,they are used for evaluating the trophic level of Chenghai Lake and Lugu Lake from 2008 to 2012.RBF neural network is established for making comparison.The results show that the average relative error and the maximum relative error for the training and test samples of LM-BP model is respectively 0.31%,0.47% and 2.68%,5.71%,which is better than RBF model.Therefore,LM-BP model has a good ability of generalization and evaluation accuracy,and it could provide a new way for the lake trophic evaluation.The evaluation results of LM-BP model for the two lakes tell us that Chenghai Lake varies between mesotrophic and eutrophic,and Lugu Lake is between oligotrophic and mesotrophic.But the trophic level of the two lakes is descending.