神经网络在次级河流回水区叶绿素a浓度预测中的应用

Application of the neural network in prediction for chlorophyll-a in branch backwater region

以长江次级河流之一的临江河为研究对象,探讨神经网络应用于次级河流回水区叶绿素a浓度短期预测的可行性。利用主成分分析法(PCA)选取对叶绿素a浓度影响较大的指标,在这些指标数据的基础上建立RBF神经网络模型。网络训练和测试的结果表明,模型模拟精度较高,说明RBF神经网络模型可以用于次级河流回水区叶绿素a浓度的短期预测。通过对临江河回水区叶绿素a影响因子的分析,表明控制水体中磷含量应是临江河回水区富营养化防治的重点。

Taking Linjiang river which is a branch of Yangtze River as research object to evaluate the feasibility of neural network model for simulating chlorophyll-a trend in branch backwater region. By using the method of principal component analysis （PCA） to select the main indexes which affect the chlorophyll-a trend, the RBF neural network model was created based on the database of indexes. The training and testing results of model indicated that the simulating accuracy of model was high; it showed that the RBF neural network model could be used for simulating the chlorophyll-a short-term trend in branch backwater region. By analyzing the influencing factors of cblorophyll-a in Linjiang river backwater region, the result showed that controlling phosphorus content would be important to prevent and control Linjiang river backwater region＇ s eutrophication.