一种紫外-可见光谱法水质COD检测的预测模型研究

A prediction model for the determination of water COD by using UV-Visible spectroscopy

针对紫外-可见光谱法水质COD检测在线、实时性的测量要求,研究了一种基于主成分分析(PCA)联合粒子群优化极限学习机(PSO-ELM)预测模型,借以预测水质COD检测数据。通过采用PCA对水质COD检测的光谱数据进行降维处理,提取其特征信息,消除向量相关性并送入PSO-ELM模型中进行建模及预测。研究结果表明,基于PCA联合PSO-ELM进行紫外-可见光谱法水质COD检测的预测模型研究,不仅预测精度较单纯的ELM模型提高了近10倍,而且相对于PSO-ELM模型的预测效率(运行时间)提升了一个数量级,这为紫外-可见光谱法水质COD在线、实时性检测创造了条件。

According to the determination of water COD by using UV-Visible spectroscopy＇s requirements of online and real-time,this paper studied a prediction model based on principal component analysis（ PCA） combined particle swarm optimization extreme learning machine（ PSO-ELM） to predict water COD detection data. The model uses PCA to reduce the dimension of water＇s spectral data,this process could extract the data＇s feature information and eliminate vector correlation.After that,the data would be used for modeling and prediction by the PSO-ELM prediction model.The research results show that the prediction model based on PCA combined PSO-ELM＇s prediction accuracy has been proved almost 10 times than the prediction model based on ELM,and the prediction effect（ run time） has been proved by one order of magnitude than the prediction model based on PSO-ELM,so these create the conditions for the determination of water COD by using UV-Visible spectroscopy online and real-time.