基于两层分解和极限学习机的河流水质预测
Prediction of River Water Quality Based on Two-Layer Decomposition Method And Extreme Learning Machine
陈一帆 1李泽 1周成龙 1胡悦1
作者信息
- 1. 苏州科技大学电子与信息工程学院,江苏 苏州 215009
- 折叠
摘要
针对城市河流水质中的化学需氧量和高锰酸盐指数无法有效预测的问题,提出一种基于两层分解方法和极限学习机的河流水质预测方法.引入改进的自适应噪声完备集合经验模态分解(improved complete ensemble EMD,ICEEMDAN)分解原始水质时间序列,分频处理子分量为高频分量和低频分量,加入变分模态分解(variational mode decomposition,VMD)对高频分量进行二次分解,建立关于所有分量的极限学习机(extreme learning machine,ELM)和最小二乘支持向量机(least square support vector machine,LSSVM)预测模型,将各分量预测值线性相加得到最终的预测结果.以苏州市水道的水质数据集作为样本进行实验,上述方法得出化学需氧量和高锰酸盐指数的平均绝对误差(mean absolute error,MAE)、均方根误差(root mean square error,RMSE)和平均绝对百分比误差(mean absolute percentage error,MAPE)分别为 0.076、0.158、2.674%和0.182、0.193、0.673%,在识别和分析关键水质参数方面的稳健性优于其它对比模型,克服了集合经验模态分解(ensemble empirical mode decomposition,EEMD)重构过程中出现的模态混叠问题,特别是在高频分量情况下,能够做到预测更加有效.
Abstract
Aiming at the problem that chemical oxygen demand and permanganate index in urban river water qual-ity can not be effectively predicted,a river water quality prediction method based on a two-layer decomposition meth-od and extreme learning machine is proposed.Improved complete ensemble EMD(ICEEMDAN)was introduced to decompose the original water quality time series.The sub-components were divided into high-frequency components and low-frequency components.Variational mode decomposition(VMD)was added to decompose the high-frequency components twice,and extreme learning machine(ELM)and least squares support vector machine(LSSVM)prediction models for all components were established.The final prediction result was obtained by adding the predicted values of each component linearly.Taking the water quality data set of Suzhou watercourses as a sample,this method obtains the mean absolute error(MAE),the root mean square error(RMSE)and the mean absolute per-centage error(MAPE)of chemical oxygen demand and permanganate index are 0.076,0.158,2.674%and 0.182,0.193,0.673%respectively.The robustness of this method in identifying and analyzing key water quality parameters is better than other comparison models,which overcomes the problem of mode aliasing in the reconstruction process of ensemble empirical mode decomposition(EEMD).Especially in the case of high-frequency components,it can be predicted more effectively.
关键词
河流水质预测/改进的自适应噪声完备集合经验模态分解/变分模态分解/极限学习机/分频处理Key words
River water quality prediction/Improved complete ensemble EMD/Variational mode decomposition/Extreme Learning machine/Frequency division processing引用本文复制引用
基金项目
国家自然科学基金资助项目(61703059)
苏州市科技发展计划项目(SS202024)
出版年
2024
NETL
NSTL
万方数据