基于三线性成分分解-神经网络算法组合的水质COD高维荧光光谱模型研究

Three-dimensional EEMs fluorescence spectra have the benefit of high sensitivity, less sample and simple operation, and hence can be widely applied to the analysis of environmental organic pollutants. Thus, this proposal will conduct researches on the model of the water quality COD in the Zhuzhou region of Xiangjiang Rive via the EEMs fluorescence spectra. However, fluorescence intensities ars highly nonlinear with values of COD due to the diversity and complexity of components in the aquatic environment, and traditional quantitative methods in the linear principle can't obtain a accurate result when modelling the water quality COD. Since the integration of several algorithms in building a model can obtain more advantages, this proposal proposes a strategy of combining the trilinear component decomposition in multi-way calibration and the neural network. It can realize the characterization of organic pollutants and then build a prediction model of the water quality COD even in the presence of unknown interference and nonlinearity. This proposal is expected to provide a effective analytical method and a technical support in on-line monitoring in water pollution.

三维EEMs荧光光谱具有灵敏度高、需样量少和方法简便快速等优点，在水质有机污染物分析方面具有广泛的应用前景。本项目拟选取湘江株洲段流域作为研究区域，开展基于EEMs荧光光谱分析的水质COD模型研究。由于水质成分的多样性和复杂性，荧光强度与水质COD之间存在高度的非线性，应用常规的线性定量校正方法对水质COD进行建模分析不能得到准确的预测结果。根据多个算法组合建模的互补性原理，本项目将化学计量学多维校正方法中的三线性成分分解与神经网络进行组合建模，在未知干扰和非线性因素存在的情况下，对水环境中有机污染物进行荧光光谱表征，并建立有机污染物综合指标COD的预测模型。本项目的完成有望为在线绿色水质污染监测研究提供有效的分析手段和技术支持。

三维EEMs荧光光谱具有灵敏度高、需样量少和方法简便快速等优点，在水质有机污染物分析方面具有广泛的应用前景。本项目选取湘江株洲段流域作为研究区域，开展基于EEMs荧光光谱分析的水质COD模型研究。根据多个算法组合建模的互补性原理，本项目将化学计量学多维校正方法中的三线性成分分解与神经网络进行组合建模，在未知干扰和非线性因素存在的情况下，对水环境中有机污染物进行荧光光谱表征，并建立有机污染物综合指标COD的预测模型。研究结果显示，模型相关系数大于0.85，表明模型具有良好的预测性能。本项目的完成为在线绿色水质污染监测研究提供有效的分析手段和技术支持。