基于多元回归和神经网络途径的太湖富营养化指标分析与预测

水体富营养化是目前太湖面临的一个重大环境问题,有效地预测湖泊的水质变化对防治富营养化至关重要. 现有的预测方案很多依赖于一些常见的与目标变量相关的因素,针对太湖这一具体对象,难以全面准确地分析相关因素之间的联系以及预见目标变量的变化趋势. 本文首先寻找出太湖富营养化的所有可能的影响因素,提供较为全面的备选变量以期更好地查找出目标变量的相关因素. 选择总氮这一富营养化重要评价指标作为分析和预测的对象,根据各影响因子与总氮之间的相关系数筛选出影响程度较高的因素. 分别运用多元线性回归分析方法和BP神经网路预测方法对总氮的变化进行预测研究,并将两种方法的预测性能进行比较. 结果表明,从较为全面的变量中选出的影响因子可较好地预测总氮的变化情况,多元线性回归和BP神经网络方法的实验结果都较好,从拟合优度和均方误差的角度看,BP神经网络的预测效果更好.

Taihu Lake Eutrophication Index Analysis and PredictionBased on Multiple Regression and Neural Network

Eutrophication of body of water is a major environmental problem of the Taihu lake,and the eutrophication prediction is an effective early warning method to know the change of lake water quality. Now there are a lot of prediction schemes depending on the existing known factors related to the target,and it is difficult to fully accurate analysis of the relationship between the related factors and predict the change trend of target variable. In this paper,firstly,we find out all of the possible affecting factors of the eutrophication of Taihu lake,then we can obtain a better target variable of related factors with a more comprehensive variables provided. We select the eutrophication of total nitrogen as the object of analysis and forecasting. The factors are selected by their correlation with total nitrogen. We conduct the prediction study for the change of total nitrogen with multiple linear regression analysis method and BP neural network prediction method,and compare the performances of the two methods. The results show that the variables selected can well predict the changes of total nitrogen. The experimental results obtained from multiple linear regression and BP neural network method are both accurate,from the perspectives of the goodness of fit and the mean square error,and the BP neural network is better.