Email Alerts
RSS
Water level prediction of lower Jingjiang Waterway in Yangtze River based on temporal convolution network
-
摘要:
航道水位的精准预测对保障船舶通航安全具有重要意义。以长江下荆江航道为研究区域，采用2019—2020年水文时间序列数据为训练集，2021年数据为验证集，构建基于时间卷积网络（TCN）的长江下荆江水位变化预测模型，并与基于长短时记忆神经网络（LSTM）和支持向量机（SVM）的水位预测模型进行计算精度的对比。结果表明：不同站点TCN对应的最优输入时间窗口存在一定差异，监利站、调弦口站及石首站的最优输入时间窗口分别为前2 d、前2 d和前3 d；TCN在2021年下荆江各站点水位预测结果的纳什系数（ENS）和决定系数（R2）均高于0.995，均方根误差（ERMS）小于0.21 m，整体预测效果优于LSTM，两者预测精度均较高，均显著优于SVM；但随着预测时间尺度的增加，水位预测精度整体呈下降趋势；TCN模型各站点枯水期大部分时段的水位预测绝对误差小于0.2 m，这表明TCN在航道水位预测领域具有较好的应用潜力。通过分析TCN在航道水位预测中的适用性和优越性，可为长江航道水位预测精度提升和船舶安全通航提供技术支撑。
Abstract:Accurate prediction of water level in waterway is of great significance for ensuring ships’ navigational safety. The lower Jingjiang waterway in the Yangtze River was taken as the research area, and the hydrological data from 2019 to 2020 and from 2021 were adopted as the train set and test set, respectively. A temporal convolution network (TCN) model was developed for water level prediction of the Lower Jingjiang waterway. Then the long short-term memory network (LSTM) and the support vector machine (SVM) were constructed for accuracy comparing with TCN. The results showed that there were differences of optimal input time windows of TCN in different stations. Jianli station, Tiaoxiankou station and Shishou station’s optimal input time windows were 2 days, 2 days and 3 days, respectively. In 2021, the Nash-Sutcliffe efficiency coefficient and determination coefficient of the water level prediction of TCN at each station in the lower Jingjiang River were higher than 0.995, and the RMSE was basically below 0.21 m. The overall performance of TCN was better than LSTM, and both of them could accurately predict the water level process and perform better than SVM. However, with the increase of prediction time scale, the prediction accuracy of water level showed a downward trend. In terms of different periods, the absolute error of TCN water level prediction in dry season was basically below 0.2 m, indicating that TCN has a great potential in the field of water level prediction. By analyzing the applicability and superiority of the TCN model, this study can provide technical support for improving the accuracy of water level prediction in the Yangtze River channel and the safe navigation of ships.
-
-
![]()
图 2 时间卷积网络结构和基础单元示意
Figure 2. Structure of TCN and schematic diagram of TCN basic unit
![]()
图 3 不同输入时间窗口下3个站点训练期和测试期的水位预测效果
Figure 3. Water level prediction in training and testing periods of three stations under different input time windows
![]()
图 4 TCN模型在不同站点处未来1~3 d的水位预测值与实测值对比情况
Figure 4. Comparison of predicted and measured water level of TCN model at different stations in the next three days
表 1 TCN模型的超参数设置
Table 1 Hyperparameter settings of TCN model
模型超参数 参数设置 模型超参数 参数设置 优化算法 Adam 卷积核个数 32 激活函数 ReLU 膨胀因子 [1/2/4/8] 损失函数 Mean Squared Error (MSE) 学习率 0.001 Dropout 0.2 批大小 30 卷积核尺寸 3 最大迭代轮数 200 
下载: 导出CSV
表 2 TCN、LSTM、SVM在不同站点处2021年的水位预测精度
Table 2 TCN, LSTM and SVM prediction accuracy of water level in 2021 at different stations
模型 站点 未来1 d 未来2 d 未来3 d ENS ERMS/m R2 ENS ERMS/m R2 ENS ERMS/m R2 监利 0.997 0 0.167 5 0.997 7 0.993 4 0.301 6 0.989 2 0.988 0 0.339 2 0.988 3 TCN 调弦口 0.997 1 0.165 9 0.998 2 0.992 3 0.278 4 0.992 2 0.987 8 0.341 9 0.988 6 石首 0.995 8 0.207 8 0.998 0 0.992 5 0.280 2 0.993 6 0.983 9 0.357 4 0.984 2 监利 0.996 2 0.189 9 0.996 2 0.990 9 0.318 5 0.990 5 0.982 7 0.387 6 0.982 5 LSTM 调弦口 0.996 1 0.193 3 0.996 5 0.989 7 0.334 1 0.988 7 0.979 7 0.379 3 0.980 7 石首 0.996 3 0.193 9 0.997 0 0.990 0 0.321 9 0.989 8 0.978 5 0.360 1 0.978 8 监利 0.993 3 0.260 9 0.996 9 0.980 2 0.402 3 0.987 3 0.966 2 0.523 9 0.973 0 SVM 调弦口 0.992 0 0.294 8 0.996 1 0.985 4 0.437 7 0.985 6 0.964 3 0.543 8 0.976 2 石首 0.991 7 0.291 5 0.995 8 0.977 6 0.398 1 0.984 0 0.967 6 0.520 5 0.972 4
下载: 导出CSV
表 3 3个站点未来1~3 d预测结果不同区间绝对误差在丰、枯水期出现天数
Table 3 The number of days with different absolute error in the next three days in dry season and wet season
单位：d 预测时期 绝对误差/m 监利 调弦口 石首 未来1 d 未来2 d 未来3 d 未来1 d 未来2 d 未来3 d 未来1 d 未来2 d 未来3 d 丰水期（5—10月） <0.2 108 96 80 107 94 79 98 91 78 [0.2, 0.5] 75 71 85 77 76 81 84 72 79 >0.5 1 17 19 0 14 24 2 21 27 枯水期（11月—翌年4月） <0.2 168 112 107 165 145 111 174 138 116 [0.2, 0.5] 11 63 63 14 29 55 5 36 52 >0.5 0 4 9 0 5 13 0 5 11 注：绝对误差为2019年测试期内水位模型预测值与真实值的差值。监利站未来1~3 d的水位最大绝对误差为0.518 8、0.814 4和0.837 8 m，调弦口的为0.441 7、0.809 5和0.9180 m，石首的为0.661 8、0.796 8和0.933 1 m。
下载: 导出CSV
-
[1] 交通运输部水运局. 2018年水路运输市场发展情况和2019年市场展望[J]. 中国水运,2019(4):10-11 Water Transport Bureau of the Ministry of Transport of the People’s Republic of China. Development of waterway transport market in 2018 and market outlook in 2019[J]. China Water Transport, 2019(4): 10-11. (in Chinese)
[2] ZHOU T, JIANG Z, LIU X, et al. Research on the long-term and short-term forecasts of navigable river’s water-level fluctuation based on the adaptive multilayer perceptron[J]. Journal of Hydrology, 2020, 591(2): 125285.
[3] 杨成刚, 李思璇, 董炳江, 等. 三峡水库运用后长江中下游枯水位变化成因研究[J]. 泥沙研究,2021,46(5):34-40 YANG Chenggang, LI Sixuan, DONG Bingjiang, et al. Study on the mechanism of lower water level in middle and lower reaches of the Yangtze River after impounding of the Three Gorges Reservoir[J]. Journal of Sediment Research, 2021, 46(5): 34-40. (in Chinese)
[4] 施勇, 栾震宇, 陈炼钢, 等. 长江中下游江湖关系演变趋势数值模拟[J]. 水科学进展,2010,21(6):832-839 SHI Yong, LUAN Zhenyu, CHEN Liangang, et al. Numerical study of the evolution trend in the river-lake relationship in the middle and lower reaches of the Yangtze River[J]. Advances in Water Science, 2010, 21(6): 832-839. (in Chinese)
[5] HU T F, MAO J Q, PAN S Q, et al. Water level management of lakes connected to regulated rivers: an integrated modeling and analytical methodology[J]. Journal of Hydrology, 2018, 562: 796-808. doi: 10.1016/j.jhydrol.2018.05.038
[6] ZHAO G, PANG B, XU Z X, et al. A hybrid machine learning framework for real-time water level prediction in high sediment load reaches[J]. Journal of Hydrology, 2020, 581: 124422. doi: 10.1016/j.jhydrol.2019.124422
[7] 陈炼钢, 陈黎明, 徐祎凡, 等. 基于越冬水鸟生境模拟的拟建鄱阳湖水利枢纽生态控制水位探讨[J]. 湖泊科学,2020,32(5):1519-1528 doi: 10.18307/2020.0523 CHEN Liangang, CHEN Liming, XU Yifan, et al. Ecological water level for proposed hydraulic construction locking Lake Poyang based upon wintering waterfowls habitat modelling[J]. Journal of Lake Sciences, 2020, 32(5): 1519-1528. (in Chinese) doi: 10.18307/2020.0523
[8] 戴凌全, 王煜, 汤正阳, 等. 三峡水库枯水期补水调度对洞庭湖越冬白鹤(Grus leucogeranus)摄食栖息地的影响[J]. 湖泊科学,2022,34(4):1208-1218 doi: 10.18307/2022.0414 DAI Lingquan, WANG Yu, TANG Zhengyang, et al. Impact of water supply of the Three Gorges Reservoir on the foraging habitat of Grus leucogeranus in Lake Dongting during the dry season[J]. Journal of Lake Sciences, 2022, 34(4): 1208-1218. (in Chinese) doi: 10.18307/2022.0414
[9] NOURANI V, HOSSEINI BAGHANAM A, ADAMOWSKI J, et al. Applications of hybrid wavelet-Artificial Intelligence models in hydrology: a review[J]. Journal of Hydrology, 2014, 514: 358-377. doi: 10.1016/j.jhydrol.2014.03.057
[10] MOROVATI K, NAKHAEI P, TIAN F, et al. A Machine learning framework to predict reverse flow and water level: A case study of Tonle Sap Lake[J]. Journal of Hydrology, 2021, 603: 127168. doi: 10.1016/j.jhydrol.2021.127168
[11] ZHANG Q, YE X C, WERNER A D, et al. An investigation of enhanced recessions in Poyang Lake: comparison of Yangtze River and local catchment impacts[J]. Journal of Hydrology, 2014, 517: 425-434. doi: 10.1016/j.jhydrol.2014.05.051
[12] 李云良, 张奇, 李淼, 等. 基于BP神经网络的鄱阳湖水位模拟[J]. 长江流域资源与环境,2015,24(2):233-240 doi: 10.11870/cjlyzyyhj201502008 LI Yunliang, ZHANG Qi, LI Miao, et al. Using BP neural networks for water level simulation in Poyang Lake[J]. Resources and Environment in the Yangtze Basin, 2015, 24(2): 233-240. (in Chinese) doi: 10.11870/cjlyzyyhj201502008
[13] 胡腾飞, 施勇, 毛劲乔, 等. 基于结构-参数同步优化的河湖水位模型及应用[J]. 排灌机械工程学报,2022,40(5):461-466, 481 HU Tengfei, SHI Yong, MAO Jinqiao, et al. River-lake water level model based on structure-parameter synchronized optimization and its application[J]. Journal of Drainage and Irrigation Machinery Engineering, 2022, 40(5): 461-466, 481. (in Chinese)
[14] 欧阳添, 闪锟, 周博天, 等. 基于LSTM网络的在线藻类时序数据预测研究: 以三峡水库为例[J]. 湖泊科学,2021,33(4):1031-1042 doi: 10.18307/2021.0406 OUYANG Tian, SHAN Kun, ZHOU Botian, et al. Research on the online forecasting of algal kinetics based on time-series data and LSTM neural network: taking Three Gorges Reservoir as an example[J]. Journal of Lake Sciences, 2021, 33(4): 1031-1042. (in Chinese) doi: 10.18307/2021.0406
[15] 邱俊杰, 郑红, 程云辉. 基于多尺度LSTM预测模型研究[J]. 系统仿真学报,2022,34(7):1593-1604 QIU Junjie, ZHENG Hong, CHENG Yunhui. Research on prediction of model based on multi-scale LSTM[J]. Journal of System Simulation, 2022, 34(7): 1593-1604. (in Chinese)
[16] CHANG F J, CHEN P A, LU Y R, et al. Real-time multi-step-ahead water level forecasting by recurrent neural networks for urban flood control[J]. Journal of Hydrology, 2014, 517: 836-846. doi: 10.1016/j.jhydrol.2014.06.013
[17] ZHU S L, HRNJICA B, PTAK M, et al. Forecasting of water level in multiple temperate lakes using machine learning models[J]. Journal of Hydrology, 2020, 585: 124819. doi: 10.1016/j.jhydrol.2020.124819
[18] 郭燕, 赖锡军. 基于循环神经网络的洞庭湖水位预测研究[J]. 长江流域资源与环境,2021,30(3):689-698 GUO Yan, LAI Xijun. Research on water level prediction of Dongting Lake based on recurrent neural network[J]. Resources and Environment in the Yangtze Basin, 2021, 30(3): 689-698. (in Chinese)
[19] 郭燕, 赖锡军. 基于长短时记忆神经网络的鄱阳湖水位预测[J]. 湖泊科学,2020,32(3):865-876 doi: 10.18307/2020.0325 GUO Yan, LAI Xijun. Water level prediction of Lake Poyang based on long short-term memory neural network[J]. Journal of Lake Sciences, 2020, 32(3): 865-876. (in Chinese) doi: 10.18307/2020.0325
[20] BARZEGAR R, AALAMI M T, ADAMOWSKI J. Coupling a hybrid CNN-LSTM deep learning model with a Boundary Corrected Maximal Overlap Discrete Wavelet Transform for multiscale Lake water level forecasting[J]. Journal of Hydrology, 2021, 598: 126196. doi: 10.1016/j.jhydrol.2021.126196
[21] BAI S, KOLTER J Z, KOLTUN V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling[EB/OL]. arXiv: 1803.01271. （2018-04-18）[2022-08-06] https://arxiv.org/abs/1803.01271
[22] 严冬梅, 何雯馨, 陈智. 融合情感特征的基于RoBERTa-TCN的股价预测研究[J]. 数据分析与知识发现,2022,6(12):123-133 YAN Dong, HE Wenxin, CHEN Zhi. Research on stock price prediction based on RoBERTa-TCN with emotional characteristics[J]. Data Analysis and Knowledge Discovery, 2022, 6(12): 123-133. (in Chinese)
[23] 刘鑫, 夏军强, 邓珊珊, 等. 下荆江急弯段凸冲凹淤演变过程与机理[J]. 科学通报,2022,67(22):2672-2683 doi: 10.1360/TB-2022-0177 LIU Xin, XIA Junqiang, DENG Shanshan, et al. Analysis of evolution processes and mechanisms of sharp bends in the Lower Jingjiang Reach[J]. Chinese Science Bulletin, 2022, 67(22): 2672-2683. (in Chinese) doi: 10.1360/TB-2022-0177
[24] 刘杰, 金勇杰, 田明. 基于VMD和TCN的多尺度短期电力负荷预测[J]. 电子科技大学学报,2022,51(4):550-557 LIU Jie, JIN Yongjie, TIAN Ming. Multi-scale short-term load forecasting based on VMD and TCN[J]. Journal of University of Electronic Science and Technology of China, 2022, 51(4): 550-557. (in Chinese)
[25] 唐贤伦, 陈洪旭, 熊德意, 等. 基于极端梯度提升和时间卷积网络的短期电力负荷预测[J]. 高电压技术,2022,48(8):3059-3067 doi: 10.13336/j.1003-6520.hve.20211771 TANG Xianlun, CHEN Hongxu, XIONG Deyi, et al. Short-term power load forecasting based on extreme gradient boosting and temporal convolutional network[J]. High Voltage Engineering, 2022, 48(8): 3059-3067. (in Chinese) doi: 10.13336/j.1003-6520.hve.20211771
[26] 李文武, 张鹏宇, 石强, 等. 基于聚合混合模态分解和时序卷积神经网络的综合能源系统负荷修正预测[J]. 电网技术,2022,46(9):3345-3357 LI Wenwu, ZHANG Pengyu, SHI Qiang, et al. Correction prediction of integrated energy system load based on aggregated mixed mode decomposition and TCN[J]. Power System Technology, 2022, 46(9): 3345-3357. (in Chinese)
[27] HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780. doi: 10.1162/neco.1997.9.8.1735
[28] BUYUKYILDIZ M, TEZEL G, YILMAZ V. Estimation of the change in lake water level by artificial intelligence methods[J]. Water Resources Management, 2014, 28(13): 4747-4763. doi: 10.1007/s11269-014-0773-1
[29] 李伶杰, 王银堂, 胡庆芳, 等. 基于随机森林与支持向量机的水库长期径流预报[J]. 水利水运工程学报,2020(4):33-40 LI Lingjie, WANG Yintang, HU Qingfang, et al. Long-term inflow forecast of reservoir based on Random Forest and support vector machine[J]. Hydro-Science and Engineering, 2020(4): 33-40. (in Chinese)
[30] LIM H, RYU H, RHUDY M B, et al. Deep learning-aided synthetic airspeed estimation of UAVs for analytical redundancy with a temporal convolutional network[J]. IEEE Robotics and Automation Letters, 2022, 7(1): 17-24. doi: 10.1109/LRA.2021.3117021
[31] 刘丽, 裴行智, 雷雪梅. 基于时间卷积注意力网络的剩余寿命预测方法[J]. 计算机集成制造系统,2022,28(8):2375-2386 LIU Li, PEI Xingzhi, LEI Xuemei. Temporal convolutional attention network for remaining useful life estimation[J]. Computer Integrated Manufacturing Systems, 2022, 28(8): 2375-2386. (in Chinese)
[32] GERS F A, SCHMIDHUBER J, CUMMINS F. Learning to forget: continual prediction with LSTM[J]. Neural Computation, 2000, 12(10): 2451-2471. doi: 10.1162/089976600300015015
[33] 张成成, 陈求稳, 徐强, 等. 基于支持向量机的太湖梅梁湾叶绿素a浓度预测模型[J]. 环境科学学报,2013,33(10):2856-2861 ZHANG Chengcheng, CHEN Qiuwen, XU Qiang, et al. A chlorophyll-a prediction model for Meiliang Bay of Taihu based on Support Vector Machine[J]. Acta Scientiae Circumstantiae, 2013, 33(10): 2856-2861. (in Chinese)
-
期刊类型引用(2)
1. 李同西，肖山民，王晨阳. 山区河流库区变动回水段碍航特性分析研究. 中国水运. 2024(11): 113-115 . 
百度学术
2. 李同西，肖山民，王晨阳. 山区河流库区变动回水段碍航特性分析研究. 中国水运. 2024(21): 113-115 . 百度学术
其他类型引用(1)
计量
- 文章访问数: 200
- HTML全文浏览量: 79
- PDF下载量: 61
- 被引次数: 3
