基于VMD-LSSVR的河水水质预测方法  

A prediction model of the river water quality based on the VMD-LSSVR

作　　者：白云[1] 李勇 BAI Yun;LI Yong(National Research Base of Intelligent Manufacturing Service,Chongqing Technology and Business University,Chongqing 400067,China;College of Earth and Environmental Sciences,Lanzhou University,Lanzhou 730000,China)

机构地区：[1]重庆工商大学国家智能制造服务国际科技合作基地,重庆400067 [2]兰州大学资源环境学院,兰州730000

出　　处：《安全与环境学报》2020年第3期1162-1168,共7页Journal of Safety and Environment

基　　金：国家自然科学基金项目(71801044);教育部人文社科研究项目(17YJC630003);重庆市自然科学基金项目(cstc2018jcyjAX0436)。

摘　　要：为了更好地掌握河水水质的变化规律,提出了一种基于变分模态分解(VMD)和最小二乘支持向量回归(LSSVR)的组合水质预测方法。通过VMD将水质指标分解成一系列有限带宽的模态分量以降低其非平稳性,然后对各分量分别建立LSSVR预测模型,并利用Pearson相关分析确定各分量输入变量,最后将各分量预测结果进行整合得到最终的水质指标预测值。以长江朱沱监测断面的高锰酸钾指数(CODMn)进行模型性能验证。结果表明,与其他现有模型相比,该方法具有更高的预测精度,为河水水质污染预控提供了有效技术支持。The paper is aimed to present a prediction method for a simulated water quality by using the variational mode decomposition (VMD) and the least square support vector regression (LSSVR) so as to better grasp the pattern of variation of the river water quality. However,as is known,it remains challenging because of the hydrological uncertainties in the single scale.Therefore,it is necessary to explore and find more internal information in data to improve the prediction accuracy. To achieve the purpose,we have,first of all,use the VMD technique to decompose the original time series of water quality index into a series of mode components with the limited bandwidth to decrease its non-stationary ones. And,secondly,we have established an appropriate LSSVR model according to its specific characteristics in each mode of the sequence. In addition,to enhance the prediction performance and accountability of this method,we have also used the Pearson correlation analysis approach to determine the input variables of each mode. And,finally,each mode’s prediction value obtained by using the LSSVR model can be integrated to achieve the final prediction result of water quality index. The experimental confirmation of the proposed method has been tested via the permanganate index (CODMn) concentration data during 2013-2016 in the Zhutuo monitoring station of the Yangtze River,China. The simulation results indicate that the above proposed method precisely predicts CODMnconcentration.The evaluation parameters of the mean absolute error (MAE),the mean absolute percentage error (MAPE) and root mean square error (RMSE) can reach 0. 131 mg/L,5. 508% and 0. 207 mg/L,respectively. Also,the method prediction result can be compared with those of the back propagation neural networks (BPNN),the LSSVR and the ensemble empirical mode decomposition (EEMD)-LSSVR model. The results show that,as compared with the other 3 methods,the proposed method can capture the details and special changing features of the time series of CODMnconcentration more ef

关 键 词：环境学 河水水质预测 变分模态分解 最小二乘支持向量回归 高锰酸钾指数 

分 类 号：X824[环境科学与工程—环境工程]