Page 184 - 《环境工程技术学报》2023年第1期
P. 184
Vol.13,No.1 环 境 工 程 技 术 学 报 第 13 卷,第 1 期
Jan.,2023 Journal of Environmental Engineering Technology 2023 年 1 月
陈伟,金柱成,俞真元,等.基于经验小波变换的鄱阳 湖 COD n 预 测 [J].环境工程技术学报,2023,13(1):180-187.
M
CHEN W,KIM J S,YU J W,et al.Forecasting COD of Poyang Lake based on empirical wavelet transform[J].Journal of Environmental Engineering
Mn
Technology,2023,13(1):180-187.
基于经验小波变换的鄱阳 湖 COD n 预测
M
1,2
1
陈伟 ,金柱成 ,俞真元 ,王晓丽 ,彭士涛 ,魏燕杰 3
1,2
1,3
1*
1.天津理工大学环境科学与安全工程学院
2.朝鲜理科大学数学系
3.交通运输部天津水运工程科学研究院
摘要 高锰酸盐指 数 (COD ) 是衡量水质状况的最重要参数之一,能反映水体受还原性物质污染的程度。结合经验小波变换
Mn
(EWT 和双向长短期记 忆 (BLSTM 神经网络,提出了一种先利 用 EW 将原始 的 COD M n 时间序列分解成若干成分,然后利用
)
T
)
BLST M 神经网络对分解出来的每个成分进行预测,最后将所有成分的预测结果重建获得最 终 COD n 预测值的新的混合模型
M
EWT-BLSTM;并 以 201 年 8 月—202 年 4 月鄱阳 湖 COD n 监测数据为研究对象,进行模型性能验证。结果表明:EWT-
0
7
M
d
BLST M 模型具有良好的预测性能,预测未 来 1 以后 的 COD M n 时,EWT-BLST M 模型的平均绝对百分比误差 为 2.25%,与单
一 BLST M 神经网络模型相比降低 了 10.53%;预测未 来 7 以后 的 COD M n 时,EWT-BLST M 模型的平均绝对百分比误差为
d
8.36%,与单 一 BLST M 神经网络模型相比降低 了 16.16%。 在 COD n 峰值处,该模型依然保持较高稳定的预测性能,说明在数
M
据相对复杂、极端的情况下,该模型依然适用。
关键词 水质预测;COD ;经验小波变 换 (EWT);双向长短期记 忆 (BLSTM);机器学习;数学模拟;鄱阳湖
Mn
中图分类号:X524 文章编号:1674-991X(2023)01-0180-08 doi:10.12153/j.issn.1674-991X.20210592
Forecasting COD of Poyang Lake based on empirical wavelet transform
Mn
1
1,3
1,2
1,2
1*
CHEN Wei , KIM Jusong , YU Jinwon , WANG Xiaoli , PENG Shitao , WEI Yanjie 3
1.School of Environmental Science and Safety Engineering, Tianjin University of Technology
2.Department of Mathematics, University of Science, DPR Korea
3.Tianjin Research Institute for Water Transport Engineering, Ministry of Transport
Abstract Permanganate index (COD ) is one of the most important parameters to measure water quality and
Mn
could reflect the degree of water pollution by reducing substances. A novel COD forecast model (EWT-BLSTM)
Mn
by combining empirical wavelet transform (EWT) and bidirectional long short-term memory (BLSTM) neural
network was proposed. First, the original COD time series was decomposed into several components by EWT.
Mn
Next, BLSTM neural network was employed to predict each decomposed component. Finally, the predictions of all
components were reconstructed to obtain the new hybrid model EWT-BLSTM for the final COD predictions.
Mn
COD data of Poyang Lake was used to evaluate the proposed forecast model. The results showed that EWT-
Mn
BLSTM model had a powerful forecast capacity. For 1, 7-day ahead forecasting, the mean absolute percentage error
(MAPE) of the forecast by EWT-BLSTM was 2.25% and 8.36%, respectively. The MAPE reduced by EWT-
BLSTM over BLSTM was 10.53% for 1-day ahead forecasting and 16.16% for 7-day ahead forecasting.
Furthermore, the proposed model showed highly stable forecasting performance for COD peak points, indicating
Mn
that the proposed method was still applicable in the case of relatively complex data with extreme points.
Key words water quality forecast; COD ; empirical wavelet transform (EWT); bidirectional long short-term
Mn
memory (BLSTM); machine learning; mathematical modelling; Poyang Lake
近年来,经济社会发展和人类活动加剧了水资 来浮游植物大量增殖、水体溶解氧浓度降低、水体
源的消耗,工业废水、生活污水的排放及面源污染等 生境受损等一系列问题 。鄱阳湖是中国第一大淡
[1]
直接对水体水质造成影响。水体污染和富营养化已 水湖,其生态系统的变化受到研究人员的关注 [2-3] 。
成为湖泊水体生态环境主要问题之一。富营养化带 目前关于鄱阳湖水质水量、富营养化和植被分布的
收稿日期:2021-10-19
基金项目:中央级公益性科研院所基本科研业务费专 项 (TKS190202,TKS20200405);天津市科技计划项 目 (20JCQNJC00100)
作者简介:陈伟(1997—),男,硕士研究生,主要研究方向为生态修复,tjutcw1997@163.com
* 责任作者:王晓丽(1972—),女,教授,主要研究方向为污染修复技术,tjutwxl@163.com