基于季节性分解与长短期记忆网络的水质动态预警

doi:10.11959/j.issn.2096-6652.202145

智能科学与技术学报 ›› 2021, Vol. 3 ›› Issue (4): 456-465.doi: 10.11959/j.issn.2096-6652.202145

基于季节性分解与长短期记忆网络的水质动态预警

许博文¹, 毕敬¹, 苑海涛², 王功明¹, 乔俊飞¹

¹ 北京工业大学信息学部，北京100124
² 北京航空航天大学自动化科学与电气工程学院，北京100191

修回日期:2021-03-12 出版日期:2021-12-15 发布日期:2021-12-01
作者简介:许博文（1996- ），男，北京工业大学信息学部硕士生，主要研究方向为异常检测、水质预警等
毕敬（1979- ），女，博士，北京工业大学信息学部副教授、博士生导师，主要研究方向为时空大数据特征建模与分析、计算智能、深度学习、数据中心节能等
苑海涛（1986- ），男，博士，北京航空航天大学自动化科学与电气工程学院副教授，IEEE高级会员，现任中国体视学学会理事。主要研究方向为云计算、边缘计算、数据中心、深度学习、智能优化、时间序列特征建模与预测等
王功明（1987- ），男，博士，北京工业大学信息学部助理研究员，主要研究方向为污水处理过程特征建模与水质预测、神经自组织结构设计与优化、离散事件动态系统安全策略优化等
乔俊飞（1968- ），男，博士，北京工业大学信息学部教授、博士生导师，主要研究方向为智能计算、自组织控制以及污水处理过程优化运行控制等
基金资助:
水体污染控制与治理科技重大专项(2018ZX07111005);国家自然科学基金资助项目(61802015);国家自然科学基金资助项目(62073005);国家自然科学基金资助项目(62173013)

Dynamic water quality warning with seasonal decomposition and long short-term memory network

Bowen XU¹, Jing BI¹, Haitao YUAN², Gongming WANG¹, Junfei QIAO¹

¹ Faculty of Information Technology, Beijing University of Technology, Beijing 100124, China
² School of Automation Science and Electrical Engineering, Beihang University, Beijing 100191, China

Revised:2021-03-12 Online:2021-12-15 Published:2021-12-01
Supported by:
The Major Science and Technology Program for Water Pollution Control and Treatment of China(2018ZX07111005);The National Natural Science Foundation of China(61802015);The National Natural Science Foundation of China(62073005);The National Natural Science Foundation of China(62173013)

摘要/Abstract

摘要：

随着地表水水质恶化日益严重，有效的水质预警预测技术对于水资源的可持续发展与应急响应机制实施至关重要。长短期记忆网络在水质时间序列预测问题中被广泛使用，但是仅使用长短期记忆网络进行水质预测并不能解决各种复杂因素造成的水质序列的不规则波动问题。为了解决该问题，提出一种数据驱动的水质预测混合模型，该模型将基于局部加权回归散点平滑（Loess）的季节与趋势分解（STL）算法与基于编解码的长短期记忆网络（LSTM-ED）结合。首先通过STL的加法模型将水质时间序列分解为3个子序列，然后利用多元LSTM-ED神经网络对子序列进行预测，通过叠加将数据恢复为实际值，最后通过季节性分段的拉依达准则进一步判断水质是否存在异常并做出预警。实验结果表明，与单一的LSTM、LSTM-ED以及基于序列分解的LSTM-ED模型相比，所提出的模型能显著地提高水质时间序列预测的精度和可靠性，并为水质动态预警提供有效的数据支持。

关键词: 季节性分解, 长短期记忆网络, 水质动态预警, 异常检测

Abstract:

Surface water quality is increasingly deteriorated in recent years, and therefore, high-quality early warning and prediction of water quality are essential for sustainability of water resources and emergency response mechanisms.Long short-term memory (LSTM) network is widely applied in the existing literature on the prediction of water quality time series.However, only applying LSTM for the prediction of water quality time series cannot well address irregular fluctuations in the water quality series caused by multiple complex factors.To solve this problem, a data-driven prediction model for the water quality time series was proposed, named STL-LSTM-ED, which was composed of seasonal-trend decomposition using locally weighted scatterplot smoothing (STL) and LSTM based on encoder-decoder (LSTM-ED).Compared with several typical models of LSTM, LSTM-ED, and a sequence decomposition method based on LSTM, the proposed STL-LSTM-ED can significantly improve the prediction accuracy and reliability of the water quality time series, and also provide the effective data support for dynamic warning of water quality.

Key words: seasonal decomposition, long short-term memory network, dynamicwarning of water quality, anomaly detection

中图分类号:

TP399

许博文, 毕敬, 苑海涛, 等. 基于季节性分解与长短期记忆网络的水质动态预警[J]. 智能科学与技术学报, 2021, 3(4): 456-465.

Bowen XU, Jing BI, Haitao YUAN, et al. Dynamic water quality warning with seasonal decomposition and long short-term memory network[J]. Chinese Journal of Intelligent Science and Technology, 2021, 3(4): 456-465.

图/表 15

图1

图2

图3

图4

表1

表2

LSTM与LSTM-ED模型的多步预测精度"

步长	LSTM			LSTM-ED
步长	MAPE	MSE	RMSE	MAPE	MSE	RMSE
1	10.71	0.91	1.15	6.92	0.56	0.79
2	10.96	0.94	1.18	7.07	0.57	0.8
3	11.34	0.97	1.22	7.21	0.59	0.81
4	11.77	1.01	1.27	7.28	0.62	0.83
5	12.2	1.05	1.32	7.35	0.64	0.83
6	12.63	1.09	1.37	7.55	0.69	0.86
7	13.02	1.13	1.42	7.62	0.71	0.87
8	13.37	1.16	1.46	7.69	0.73	0.89
9	13.69	1.19	1.5	8.17	0.76	0.92
10	13.97	1.22	1.53	8.49	0.78	0.93
11	14.2	1.24	1.56	8.91	0.83	0.95
12	14.41	1.26	1.59	9.23	0.87	0.99
13	14.58	1.28	1.61	9.49	0.9	1.02
14	14.73	1.29	1.63	9.88	0.92	1.05
15	14.85	1.3	1.65	10.32	0.94	1.08
16	14.95	1.31	1.66	10.77	0.97	1.1
17	15.04	1.32	1.67	11.51	0.99	1.13
18	15.1	1.33	1.68	11.89	1.04	1.16
19	15.15	1.33	1.69	12.21	1.13	1.19
20	15.19	1.34	1.69	12.65	1.19	1.21
平均	13.59	1.18	1.49	$9 . 11$	$0 . 82$	$0 . 97$

表2

图5

图6

图8

图7

表3

基于序列分解的LSTM-ED模型的多步预测精度"

步长	EMD-LSTM-ED			STL-LSTM-ED
步长	MAPE	MSE	RMSE	MAPE	MSE	RMSE
1	3.21	0.28	0.46	3.73	0.33	0.43
2	3.4	0.3	0.47	3.97	0.35	0.45
3	3.69	0.32	0.5	4.22	0.37	0.48
4	4.05	0.35	0.53	4.48	0.39	0.51
5	4.44	0.39	0.57	4.75	0.41	0.54
6	4.86	0.43	0.62	5.01	0.44	0.57
7	5.29	0.47	0.66	5.27	0.46	0.6
8	5.73	0.5	0.71	5.53	0.48	0.63
9	6.16	0.54	0.76	5.79	0.5	0.65
10	6.59	0.58	0.8	6.04	0.52	0.68
11	7.02	0.62	0.85	6.28	0.55	0.71
12	7.43	0.66	0.9	6.52	0.57	0.74
13	7.83	0.7	0.94	6.74	0.59	0.76
14	8.21	0.73	0.99	6.96	0.6	0.79
15	8.58	0.76	1.03	7.16	0.62	0.81
16	8.93	0.8	1.07	7.36	0.64	0.84
17	9.26	0.83	1.11	3.73	0.33	0.43
18	9.57	0.86	1.15	3.97	0.35	0.45
19	9.86	0.88	1.18	4.22	0.37	0.48
20	10.13	0.91	1.22	4.48	0.39	0.51
平均	6.71	0.60	0.83	$5 . 31$	$0 . 46$	$0 . 60$

表3

图9

图10

表4

图11

参考文献 25

[1]	JIN T , CAI S B , JIANG D X ,et al. A data-driven model for real-time water quality prediction and early warning by an integration method[J]. Environmental Science and Pollution Research International, 2019,26(29): 30374-30385.
[2]	权利敏, 杨翠丽, 乔俊飞 . 基于CFNN的污水处理过程溶解氧浓度在线控制[J]. 智能科学与技术学报, 2020,2(3): 261-267.
	QUAN L M , YANG C L , QIAO J F . CFNN-based online control for dissolved oxygen concentration of wastewater treatment processes[J]. Chinese Journal of Intelligent Science and Technology, 2020,2(3): 261-267.
[3]	OLIVEIRA P J , STEFFEN J L , CHEUNG P . Parameter estimation of seasonal ARIMA models for water demand forecasting using the harmony search algorithm[J]. Procedia Engineering, 2017,186: 177-185.
[4]	PARMAR K S , BHARDWAJ R . Water quality management using statistical analysis and time-series prediction model[J]. Applied Water Science, 2014,4(4): 425-434.
[5]	CAO Q , LIU B X , CHENG X . Notice of retraction:forecasting model based on LS-SVM and Markov chain for the water quality of Poyang Lake[C]// Proceedings of 2011 7th International Conference on Natural Computation. Piscataway:IEEE Press, 2011: 2070-2073.
[6]	LE X H , HO H V , LEE G ,et al. Application of long short-term memory (LSTM) neural network for flood forecasting[J]. Water, 2019,11(7): 1387.
[7]	AYADI A , GHORBEL O , OBEID A M ,et al. Outlier detection approaches for wireless sensor networks:a survey[J]. Computer Networks, 2017,129: 319-333.
[8]	张建晋, 王韫博, 龙明盛 ,等. 面向季节性时空数据的预测式循环网络及其在城市计算中的应用[J]. 计算机学报, 2020,43(2): 286-302.
	ZHANG J J , WANG Y B , LONG M S ,et al. Predictive recurrent net-works for seasonal spatiotemporal data with applications to urban computing[J]. Chinese Journal of Computers, 2020,43(2): 286-302.
[9]	TSENG F M , YU H C , TZENG G H . Combining neural network model with seasonal time series ARIMA model[J]. Technological Forecasting and Social Change, 2002,69(1): 71-87.
[10]	ZHOU Z J , HU C H . An effective hybrid approach based on grey and ARMA for forecasting gyro drift[J]. Chaos,Solitons ＆ Fractals, 2008,35(3): 525-529.
[11]	KHASHEI M , BIJARI M . A novel hybridization of artificial neural networks and ARIMA models for time series forecasting[J]. Applied Soft Computing, 2011,11(2): 2664-2675.
[12]	ZHANG L , ZOU Z H , SHAN W . Development of a method for com prehensive water quality forecasting and its application in Miyun reservoir of Beijing,China[J]. Journal of Environmental Sciences, 2017,56: 240-246.
[13]	SHI B , WANG P , JIANG J P ,et al. Applying high-frequency surrogate measurements and a wavelet-ANN model to provide early warnings of rapid surface water quality anomalies[J]. Science of the Total Environment, 2018,610/611: 1390-1399.
[14]	SUN Y X , LENG B , GUAN W . A novel wavelet-SVM short-time passenger flow prediction in Beijing subway system[J]. Neurocomputing, 2015,166: 109-121.
[15]	CHEN L , CHI Y G , GUAN Y Y ,et al. A hybrid attention-based EMD-LSTM model for financial time series prediction[C]// Proceedings of 2019 2nd International Conference on Artificial Intelligence and Big Data (ICAIBD). Piscataway:IEEE Press, 2019: 113-118.
[16]	CONEJO A J , PLAZAS M A , ESPINOLA R ,et al. Day-ahead electricity price forecasting using the wavelet transform and ARIMA models[J]. IEEE Transactions on Power Systems, 2005,20(2): 1035-1042.
[17]	WU D , JIANG Z K , XIE X F ,et al. LSTM learning with Bayesian and Gaussian processing for anomaly detection in industrial IoT[J]. IEEE Transactions on Industrial Informatics, 2020,16(8): 5244-5253.
[18]	NIU H L , XU K L , WANG W Q . A hybrid stock price index forecasting model based on variational mode decomposition and LSTM network[J]. Applied Intelligence, 2020,50(12): 4296-4309.
[19]	CHEN D W , ZHANG J H , JIANG S X . Forecasting the short-term metro ridership with seasonal and trend decomposition using loess and LSTM neural networks[J]. IEEE Access, 2020,8: 91181-91187.
[20]	CLEVELAND R B , CLEVELAND W S , MCRAE J E ,et al. STL:a seasonal-trend decomposition[J]. Journal of official statistics, 1990,6(1): 3-73.
[21]	HOCHREITER S , SCHMIDHUBER J . Long short-term memory[J]. Neural Computation, 1997,9(8): 1735-1780.
[22]	HABLER E , SHABTAI A . Using LSTM encoder-decoder algorithm for detecting anomalous ADS-B messages[J]. Computers ＆ Security, 2018,78: 155-173.
[23]	BAHDANAU D , CHO K , BENGIO Y . Neural machine translation by jointly learning to align and translate[J]. arXiv preprint,2014,arXiv:1409.0473.
[24]	倪维健, 孙宇健, 刘彤 ,等. 基于注意力双向循环神经网络的业务流程剩余时间预测方法[J]. 计算机集成制造系统, 2020,26(6): 1564-1572.
	NI W J , SUN Y J , LIU T ,et al. Business process remaining time pre-diction using bidirectional recurrent neural networks with attention[J]. Computer Integrated Manufacturing Systems, 2020,26(6): 1564-1572.
[25]	HUANG N E , SHEN Z , LONG S R ,et al. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis[J]. Proceedings of the Royal Society of London Series A:Mathematical,Physical and Engineering Sciences, 1998,454(1971): 903-995.

真实情况	预测为正例	预测为反例
正例	TP	FN
反例	FP	TN

k值	F1-Score	AUC
3	0.494	0.62
4	0.516	0.67
5	0.501	0.64
6	0.472	0.60
7	0.463	0.57
8	0.455	0.56

基于季节性分解与长短期记忆网络的水质动态预警

Dynamic water quality warning with seasonal decomposition and long short-term memory network

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 15

参考文献 25

相关文章 3

Metrics

推荐阅读 0

[1]	罗欣欢, 王奕璇, 李炜, 陈曦. 基于显著性信息的Fit CutMix数据增强算法在医学影像上的应用[J]. 智能科学与技术学报, 2023, 5(1): 58-68.
[2]	郁洲, 毕敬, 苑海涛. 基于改进DQN算法的复杂海战场路径规划方法[J]. 智能科学与技术学报, 2022, 4(3): 418-425.
[3]	陈苑文, 王晓, 李灵犀, 王飞跃. 基于社会媒体数据增强的交通态势感知研究及进展[J]. 智能科学与技术学报, 2022, 4(1): 1-13.