[1]罗文海,伍毅,杜泰霖,等.基于TCN-Informer潮汐预测模型的水闸优化调度[J].中国给水排水,2026,42(9):116-122.
　LUOWenhai,WUYi,DUTailin,et al.Optimized Dispatch of Sluice Gates Based on TCN-Informer Tidal Prediction Model[J].China Water & Wastewater,2026,42(9):116-122.
点击复制

基于TCN-Informer潮汐预测模型的水闸优化调度

中国给水排水[ISSN:1000-4062/CN:12-1073/TU] 卷: 第42卷 期数: 2026年第9期 页码: 116-122 栏目: 出版日期: 2026-05-01
Title:
Optimized Dispatch of Sluice Gates Based on TCN-Informer Tidal Prediction Model
作者:
罗文海1， 伍毅2， 杜泰霖3， 高良航3， 吴青3， 桑耀翔1， 姜涛2， 陈兵1
（1.华南理工大学 环境与能源学院，广东 广州 510006；2.华南理工大学 自动化科学与工程学院，广东 广州 510641；3.广州市荔湾区水务工程建设管理中心，广东 广州 510145）
Author(s):
LUO　Wenhai1， WU　Yi2， DU　Tailin3， GAO　Lianghang3， WU　Qing3， SANG　Yaoxiang1， JIANG　Tao2， CHEN　Bing1
（1. College of Environment and Energy， South China University of Technology， Guangzhou 510006， China； 2. School of Materials Science and Engineering， South China University of Technology， Guangzhou 510641， China； 3. Liwan District Water Engineering Construction Management Center， Guangzhou 510145， China）
关键词:
TCN-Informer;  潮汐预测;  优化调度;  感潮河网;  智慧水网
Keywords:
TCN-Informer;  tidal prediction;  optimized dispatch;  tide?sensitive river network;  smart water network
摘要:
针对芳村围河网人工调度方式存在的问题，基于自主研发的蜻蜓智慧水网系统开发平台，快速、低成本建立了芳村围智慧水网系统。根据皮尔逊相关系数分析智慧水网系统采集的水位数据，选取东塱水闸闸外水位和降雨量作为输入数据，利用时域卷积网络（TCN）捕获局部特征，利用Informer建立长期依赖关系，构建TCN-Informer潮汐预测模型。该模型的平均绝对误差（MAE）为0.126 4 m、均方根误差（RMSE）为0.162 7 m、决定系数（R2）为0.919 9，展现出良好的预测性能。根据模型预测的水位曲线生成并试验了芳村围水闸优化调度方案，采用水流单向流方式，水体的氨氮浓度、溶解氧含量和透明度整体优于原调度方案，能够有效改善水环境。
Abstract:
To address the issues with the manual dispatch method of the Fangcunwei river network, a smart water network system for Fangcunwei was rapidly and cost?effectively established using the self-developed dragonfly smart water network system development platform. Based on Pearson correlation coefficient analysis of the water level data collected by the smart water network system, the water level outside the Donglang sluice gate and rainfall were selected as input data. The temporal convolutional network (TCN) was used to capture local features, and Informer was used to establish long-term dependencies, thereby constructing a TCN-Informer tidal prediction model. The model achieved a mean absolute error (MAE) of 0.126 4 m, a root mean square error (RMSE) of 0.162 7 m, and a coefficient of determination (R2) of 0.919 9, demonstrating good predictive performance. Based on the water level curve predicted by the model, an optimized dispatch scheme for the Fangcunwei sluice gates was generated and tested. Adopting a unidirectional flow method, the ammonia nitrogen concentration, dissolved oxygen content, and transparency of the water samples were overall better than those of the original dispatch scheme, effectively improving the water environment.
更新日期/Last Update: 2026-05-01