PREDICTION OF TOTAL PHOSPHORUS IN RIVERS BASED ON ATTENTION MECHANISM OF TEMPORAL CONVOLUTIONAL NETWORKS

LI Yuanyuan; LIU Hailong

doi:10.13205/j.hjgc.202305022

Volume 41 Issue 5

May 2023

Turn off MathJax

Article Contents

Article Navigation > ENVIRONMENTAL ENGINEERING > 2023 > 41(5): 163-171

LI Yuanyuan, LIU Hailong. PREDICTION OF TOTAL PHOSPHORUS IN RIVERS BASED ON ATTENTION MECHANISM OF TEMPORAL CONVOLUTIONAL NETWORKS[J]. ENVIRONMENTAL ENGINEERING , 2023, 41(5): 163-171. doi: 10.13205/j.hjgc.202305022

Citation:

LI Yuanyuan, LIU Hailong. PREDICTION OF TOTAL PHOSPHORUS IN RIVERS BASED ON ATTENTION MECHANISM OF TEMPORAL CONVOLUTIONAL NETWORKS[J]. ENVIRONMENTAL ENGINEERING , 2023, 41(5): 163-171. doi: 10.13205/j.hjgc.202305022

Citation:

PDF( 5542 KB)

PREDICTION OF TOTAL PHOSPHORUS IN RIVERS BASED ON ATTENTION MECHANISM OF TEMPORAL CONVOLUTIONAL NETWORKS

doi: 10.13205/j.hjgc.202305022

LI Yuanyuan,
LIU Hailong^,

School of Resources and Environment, University of Electronic Science and Technology of China, Chengdu 611731, China

Received Date: 2022-07-26

Abstract

Abstract

The application of deep learning in water pollution monitoring has become a current hot topic. Taking the total phosphorus prediction of Min River water system in Pengshan District, Meishan City, Sichuan Province as the research object, a temporal convolutional network (TCN) prediction model based on temporal pattern attention (TPA) mechanism was proposed to solve the problem that the concentration of water pollutants was related to different time steps between upstream and downstream. Firstly, the Pearson correlation coefficient was applied to analyze the spatiotemporal relationship of total phosphorus concentration at upstream and downstream sites. Then, the dilated convolution and causal convolution of TCN were used to extract the dependencies of the time series data. And the TPA mechanism was used to learn the complex spatial relationship between the time series of total phosphorus concentration at different stations from TCN, which can obtain the station weights at different time steps in the time series data. Finally, this model was applied to predict the total phosphorus concentration of rivers. The research results showed that the multi-site input TPA-TCN model reduced RMSE by 36.29%, MAE by 28.18% and MAPE by 25.26%, compared with that of the single-site input TPA-TCN model. Compared with TCN model integrated with the traditional attention mechanism, the three evaluation metrics of TCN model integrated with TPA mechanism reduced by 10.24%, 10.78% and 9.94%, indicating that the multi-site input TPA-TCN model has certain advantages in predicting the total phosphorus concentration in rivers. The TPA-TCN model can be effectively applied to water quality monitoring, which has important reference significance for the prediction of water pollutants.
- water quality prediction,
- temporal convolutional network,
- temporal pattern attention mechanism,
- Pearson correlation coefficient

FullText(HTML)

References(33)

References

[1]	HORN A L, RUEDA F, HORMANN G, et al. Implementing river water quality modelling issues in mesoscale watershed models for water policy demands:an overview on current concepts, deficits, and future tasks[J]. Physics and Chemistry of the Earth, 2004, 29(11/12): 725-737.
[2]	刘东君, 邹志红. 最优加权组合预测法在水质预测中的应用研究[J]. 环境科学学报, 2012, 32(12): 3128-3132.
[3]	李娜, 李勇, 冯家成, 等. 太湖水体Chl-a预测模型ARIMA的构建及应用优化[J]. 环境科学, 2021, 42(5): 2223-2231.
[4]	曾一川, 曾会国, 袁伟皓, 等. 长江口入海通道水质综合分析与模型预测[J]. 环境工程, 2022, 40(5): 95-102 ,108.
[5]	GUO T, HE W, JIANG Z L, et al. An improved LSSVM model for intelligent prediction of the daily water level[J]. Energies, 2019, 12(1): 112.
[6]	刘世存, 杨薇, 田凯, 等. 基于多层全连接神经网络的白洋淀水质预测[J]. 农业环境科学学报, 2020, 39(6): 1283-1292.
[7]	刘攀, 郑雅莲, 谢康, 等. 水文水资源领域深度学习研究进展综述[J]. 人民长江, 2021, 52(10): 76-83.
[8]	HUAN J, CHEN B, XU X G, et al. River dissolved oxygen prediction based on Random Forest and LSTM[J]. Applied Engineering in Agriculture, 2021, 37(5): 901-910.
[9]	董泉汐. 基于深度学习的水环境时间序列预测方法研究[D]. 北京: 北京工业大学, 2020.
[10]	张贻婷, 李天宏. 基于长短时记忆神经网络的河流水质预测研究[J]. 环境科学与技术, 2021, 44(8): 163-169.
[11]	BARZEGAR R, AALAMI M T, ADAMOWSKI J. Short-term water quality variable prediction using a hybrid CNN-LSTM deep learning model[J]. Stochastic Environmental Research and Risk Assessment, 2020, 34(2): 415-433.
[12]	YANG Y R, XIONG Q Y, WU C, et al. A study on water quality prediction by a hybrid CNN-LSTM model with attention mechanism[J]. Environmental Science and Pollution Research, 2021, 28(39): 55129-55139.
[13]	BAI S, KOLTER J Z, KOLTUN V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling[C]. arXiv:1803.01271, 2018.
[14]	李荆, 刘钰, 邹磊. 基于时空建模的动态图卷积神经网络[J]. 北京大学学报(自然科学版), 2021, 57(4): 605-613.
[15]	ZHANG Y F, THORBURN P J, FITCH P. Multi-task temporal convolutional network for predicting water quality sensor data[C]//International Conference on Neural Information Processing, Springer, 2019: 122-130.
[16]	FU Y X, HU Z H, ZHAO Y C, et al. A long-term water quality prediction method based on the temporal convolutional network in smart mariculture[J]. Water, 2021, 13(20): 2907.
[17]	LI W S, WEI Y G, AN D, et al. LSTM-TCN: dissolved oxygen prediction in aquaculture, based on combined model of long short-term memory network and temporal convolutional network[J].Environmental Science and Pollution Research, 2022, 29(26): 39545-39556.
[18]	SHIH S Y, SUN F K, LEE H Y. Temporal pattern attention for multivariate time series forecasting[J]. Machine Learning, 2019, 108(8/9): 1421-1441.
[19]	崔鸿雁, 徐帅, 张利锋, 等. 机器学习中的特征选择方法研究及展望[J]. 北京邮电大学学报, 2018, 41(1): 1-12.
[20]	张阳, 冼慧婷, 赵志杰. 基于空间相关性和神经网络模型的实时河流水质预测模型[J].北京大学学报(自然科学版), 2022, 58(2): 337-344.
[21]	SCHOBER P, BOER C, SCHWARTE L A. Correlation coefficients: appropriate use and interpretation[J]. Anesthesia and Analgesia, 2018, 126(5): 1763-1768.
[22]	PANTISKAS L, VERSTOEP K, BAL H. Interpretable multivariate time series forecasting with temporal attention convolutional neural networks[C]//IEEE Symposium Series on Computational Intelligence, 2020: 1687-1694.
[23]	ZHAI N J, ZHOU X F. Temperature prediction of heating furnace based on deep transfer learning[J]. Sensors, 2020, 20(17): 4676-4702.
[24]	WAN R Z, MEI S P, WANG J, et al. Multivariate temporal convolutional network: a deep neural networks approach for multivariate time series forecasting[J]. Electronics, 2019, 8(8): 876-893.
[25]	ZHOU X H, WANG J P, CAO X K, et al. Simulation of future dissolved oxygen distribution in pond culture based on sliding window-temporal convolutional network and trend surface analysis[J]. Aquacultural Engineering, 2021, 95: 102200.
[26]	王渝红, 史云翔, 周旭, 等. 基于时间模式注意力机制的BiLSTM多风电机组超短期功率预测[J]. 高电压技术, 2022, 48(5): 1884-1892.
[27]	SONG C G, YAO L H, HUA C Y, et al. A novel hybrid model for water quality prediction based on synchrosqueezed wavelet transform technique and improved long short-term memory[J]. Journal of Hydrology, 2021, 603(A): 126879.
[28]	李光, 吴祈宗. 基于结论一致的综合评价数据标准化研究[J]. 数学的实践与认识, 2011, 41(3): 72-77.
[29]	LIU Y Q, ZHANG Q, SONG L H, et al. Attention-based recurrent neural networks for accurate short-term and long-term dissolved oxygen prediction[J]. Computers and Electronics in Agriculture, 2019, 165: 104964.
[30]	ZHOU J, WANG Y Y, XIAO F, et al. Water quality prediction method based on IGRA and LSTM[J]. Water, 2018, 10(9): 1148.
[31]	许佳辉, 王敬昌, 陈岭, 等. 基于图神经网络的地表水水质预测模型[J]. 浙江大学学报(工学版), 2021, 55(4): 601-607.
[32]	陈海涵, 吴国栋, 李景霞, 等. 基于注意力机制的深度学习推荐研究进展[J]. 计算机工程与科学, 2021, 43(2): 370-380.
[33]	王竹荣, 薛伟, 牛亚邦, 等. 基于注意力机制的泊位占有率预测模型研究[J]. 通信学报, 2020, 41(12): 182-192.