| 本文已被：浏览 109次 下载 0次 |
|
|
|
| 基于时间卷积网络的长江下荆江航道水位预测研究 |
|
李 港1,2, 李有为1,2, 舒章康3, 张 宇3, 王 江1,2, 查 伟3
|
|
1.长江航道勘察设计院武汉有限公司;2.国家内河航道整治工程技术研究中心;3.南京水利科学研究院水文水资源与水利工程学科国家重点实验室
|
|
| 摘要: |
| 航道水位的精准预测对于保障船舶通航安全具有重要意义。本研究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作为训练集，2021年数据作为验证集，构建了基于时间卷积网络（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了计算精度的对比，分析了TCN在水位预测中的适用性和优越性。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、调弦口站以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NSE和R2均高于0.995，RMSE位于0.21 m以下，整体预测效果优于LSTM，但两者预测精度均较高，均显著优于SVM，能够合理准确的预测下荆江航道的水位变化过程；但随着预测时间尺度的增加，水位预测精度整体呈现出降低的趋势；同时，TCN模型各站点枯水期的水位预测结果在大部分时段的绝对误差位于0.2m以下，表明TCN在航道水位预测领域具有较好的应用潜力，能够有效保证航道枯水期的船舶通航安全。 |
| 关键词: 水位预测 航道水位 下荆江航道 时间卷积网络 长短时记忆神经网络 |
| DOI： |
| 分类号:TV124 |
| 基金项目:江苏自然科学基金(BK20200160) |
|
| Water level prediction of Lower Jingjiang Waterway in Yangtze River based on temporal convolution network |
|
LI Gang1,2, LI Youwei1,2, SHU Zhangkang3, ZHANG Yu3, WANG Jiang1,2, ZHA Wei3
|
|
1.Changjiang Waterway Survey Design and Research Institute (Wuhan);2.National Engineering Research Center for Inland Waterway Regulation;3.State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering Science, Nanjing Hydraulic Research Institute
|
| Abstract: |
| Accurate prediction of water level in waterway is of great significance for ensuring ships’ navigational safety. The Lower Jingjiang waterway under the Yangtze River was took as the research area, 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 Lower Jingjiang waterway, Then long short-term memory network(LSTM) and support vector machine(SVM) were constructed for accuracy comparing with TCN, which aims to verify the applicability and superiority of TCN in water level prediction. The results showed that there are 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 NSE and R2 of the water level prediction of TCN at each station in Lower Jingjiang River were higher than 0.99, and the RMSE was basically below 0.21m. The overall performance of TCN was better than LSTM, both of them can accurately predict the water level process and performance 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 period, the absolute error of TCN water level prediction in dry season was basically below 0.2m, indicating that TCN has a great potential in the field of water level prediction, which can effectively ensure the navigable safety of ships in dry season. |
| Key words: Water level prediction Water level in waterway Lower Jingjiang Waterway Temporal convolutional network Long short-term memory network |
