摘要
Abstract
Based on the eutrophication evaluation criteria for Chinese lakes and reservoirs, and RBF, GRNN, BP, and Elman neural network algorithm theories, four neural network models were constructed to evaluate the eutrophication of lakes and reservoirs. The interpolation method was used to construct network training samples. The threshold levels for eutrophication evaluation of Chinese lakes and reservoirs were considered the evaluation samples and were used for prediction. The predicted results, which were regarded as the criteria for division of eutrophication levels, were used to evaluate the eutrophication of 24 major lakes and reservoirs in China. The results show the following: The eutrophication evaluation results of the 24 major lakes and reservoirs using the RBF, GRNN, BP, and Elman neural network models were basically the same and had high precision, indicating that the four neural network models and evaluation methods are reasonable and feasible, and can provide a new way for evaluation of eutrophication of lakes and reservoirs. Compared with the BP and Elman neural network models, the RBF and GRNN neural network models not only had identical evaluation results, but had advantages of fast convergence, high prediction accuracy, less parameters to be adjusted ( only the SPREAD parameter) , and unlikely occurrence of a local minimum, and could perform quicker prediction and evaluation of the network with greater computational advantages.关键词
湖库/富营养化评价/RBF神经网络/GRNN神经网络/BP神经网络/Elman神经网络Key words
lake and reservoir/ eutrophication evaluation/ RBF neural network/ GRNN neural network/ BP neural network/ Elman neural network
分类
资源环境
