PCA-BP神经网络在流域水质评价中的应用

Application of PCA-BP Neural Network on Water Quality Evaluation in Major Drainage Basin

针对BP神经网络法在进行多污染特征的流域水质评价时面临的训练样本、验证样本的稀缺问题,提出一种基于主成分分析PCA-BP神经网络的水质评价模型。首先利用污染分担率算法筛选出能够全面反映流域整体超标情况的一组水质指标,然后利用主成分方法获取流域水质污染特征,解决训练样本过少的问题,并通过设计模型的验证条件,解决没有验证样本的问题。通过实例研究,表明主成分PCA-BP神经网络适合用于流域的水质评价,评价结果较为精确、可信。

In the application of BP neural network method on water quality evaluation in major drainage basin with multi-pollution characteristics, few training samples and validation samples can be found. An improved water quality evaluation method is introduced based on principal component analysis （PCA） - BP neural network. Pollution ratio is used to filter out a set of pollution date as indicatiors to reflect water sin. The principal component method is applied to get pollution characteristics of and to solve the problem of few training samples. of no validation sample. Case study in the paper drainage qualityy of the babasin water quality The model validation criteria is designed to solve the problem shows that the principal component PCA -BP neural network is suitable for the drainage basin of water quality evaluation, and the result is accurate and credible.