基于模型-数据混合驱动的大型调水泵站优化调度方法

唐昊; 缪连杰; 汪雪韵; 金昕; 朱田生; 程海英

doi:10.16383/j.aas.c250750

基于模型-数据混合驱动的大型调水泵站优化调度方法

doi: 10.16383/j.aas.c250750 cstr: 32138.14.j.aas.c250750

唐昊^1,,
缪连杰^1,,
汪雪韵^1,,
金昕^2,,
朱田生^3,,
程海英^3,

1.
合肥工业大学电气与自动化工程学院合肥 230009
2.
安徽省水利水电勘测设计研究总院股份有限公司合肥 230088
3.
安徽省水利科学研究院合肥 230088

基金项目: 国家自然科学基金项目(62273130), 安徽省自然科学基金重点项目(2408055US008)资助

详细信息

作者简介:
唐昊：合肥工业大学电气与自动化工程学院教授. 主要研究方向为智能电网调度和大型引调水工程泵站调度. 本文通信作者. E-mail: htang@hfut.edu.cn

缪连杰：合肥工业大学电气与自动化工程学院硕士研究生. 主要研究方向为调水泵站调度优化. E-mail: lianjiemiao@mail.hfut.edu.cn

汪雪韵：合肥工业大学电气与自动化工程学院博士研究生. 主要研究方向为调水泵站调度优化. E-mail: wangxueyun@mail.hfut.edu.cn

金昕：安徽省水利水电勘测设计研究总院股份有限公司高级工程师. 主要研究方向为水利水电工程规划设计. E-mail: hjx-ai@foxmail.com

朱田生：安徽省水利科学研究院工程师. 主要研究方向为水利工程质量检测. E-mail: titia886@sina.com

程海英：安徽省水利科学研究院高级工程师. 主要研究方向为水利工程质量检测. E-mail: jczah@vip.163.com

中图分类号: Y
计量
- 文章访问数: 14
- HTML全文浏览量: 6
- 被引次数: 0
出版历程
- 收稿日期: 2025-12-30
- 录用日期: 2026-04-28
- 网络出版日期: 2026-11-03

Optimal Scheduling of Large Water Transfer Pumping Stations Based on a Model-Data Hybrid-Driven Approach

1.
School of Electrical Engineering and Automation, Hefei University of Technology, Hefei 230009
2.
Anhui Provincial Water Resources and Hydropower Survey and Design Institute Co., Ltd., Hefei 230088
3.
Anhui Province Institute of Water Conservancy Science, Hefei 230088

Funds: Supported by National Natural Science Foundation of China (62273130), and Anhui Provincial Natural Science Foundation (2408055US008)

More Information

Author Bio:
Tang Hao　Professor at the School of Electric and Automatization Engineering, Hefei University of Technology. His research interests include smart grid dispatching, large-scale water transfer project pump station scheduling. Corresponding author of this paper

Miao Lian-Jie　Master student at the School of Electric and Automatization Engineering, Hefei University of Technology. His main research interest is scheduling optimization of water transfer pumping station

Wang Xue-Yun　Ph.D. candidate at the School of Electric and Automatization Engineering, Hefei University of Technology. Her main research interest is scheduling optimization of water transfer pumping station

Jin Xin　Senior engineer in the Anhui Provincial Water Resources and Hydropower Survey and Design Institute Co., Ltd. His main research interest is planning and design of water resources and hydropower projects

Zhu Tian-Sheng　Engineer at the Anhui Province Institute of Water Conservancy Science. His main research interest is quality testing and inspection for hydraulic infrastructure

Cheng Hai-Yin　Senior engineer in the Anhui Province Institute of Water Conservancy Science. Her main research interest is quality testing and inspection for hydraulic infrastructure

摘要

摘要: 对于跨流域调水工程中由大型调水泵站、输水河道与调蓄湖泊组成的泵站输水系统, 其优化调度需同时考虑湖泊径流来水和需水的不确定性以及泵站内各机组间的相互协调, 为此, 提出一种分层协同优化机制与基于模型-数据混合驱动的调水泵站优化调度方法. 首先, 基于圣维南方程组, 建立了泵站与湖泊间输水河道的一维非恒定流模型, 以精确描述泵站流量动态变化下的水力响应过程; 在此基础上, 考虑径流来水和需水的不确定性, 构建了调水泵站双层优化调度模型, 其中上层模型以最小化能耗成本与调度期末湖泊的调蓄库容偏差为目标优化调度期内各时段泵站流量, 下层模型以泵站运行效率最大为目标优化不同流量要求下各时段泵站机组的流量最优分配方案; 最后, 构建了一种嵌入数学求解器的深度强化学习机制, 提出了基于模型-数据混合驱动的SE-Dueling DDQN优化方法, 以提升策略的适应性和鲁棒性. 算例结果表明, 所提模型与方法能有效应对径流来水和需水的不确定性, 且在能耗成本优化与调蓄库容控制精度方面均优于典型的传统优化算法.
- 大型调水泵站 /
- 优化调度 /
- 深度强化学习 /
- 数学求解器 /
- 不确定性
Abstract: For the pumping station water-conveyance system in inter-basin water transfer projects, which consist of large pumping station, canal, and regulating lake, its optimal operation must simultaneously address the uncertainties of lake runoff and water demand, as well as the coordinated operation of multiple pump units. To this end, a hierarchical collaborative optimization framework and a model-data hybrid-driven optimal scheduling method for water transfer pumping station are proposed. First, based on the Saint-Venant equations, a one-dimensional unsteady flow model is established for the conveyance canal between the pumping station and the regulating lake, so as to accurately characterize the hydraulic response under dynamic variations in the flow of pumping station. On this basis, a two-layer optimal scheduling model is developed by considering the uncertainties of runoff inflow and water demand. In the upper-layer model, the flow of pumping station in each time period is optimized with the objectives of minimizing energy consumption cost and the deviation of the lake storage from its target value at the end of the scheduling horizon. In the lower-layer model, the flow allocation among pump units in each time period is optimized under given flow requirements, with the objective of maximizing pumping station operating efficiency. Furthermore, a deep reinforcement learning framework embedded with a mathematical solver is constructed, and a model-data hybrid-driven SE-Dueling DDQN is proposed to enhance the adaptability and robustness of the scheduling strategy. Case study results demonstrate that the proposed model and method can effectively cope with the uncertainties in runoff and water demand, and outperform typical conventional optimization algorithms in both energy cost reduction and storage control accuracy.
- large water transfer pumping stations /
- optimal scheduling /
- deep reinforcement learning /
- mathematical solver /
- uncertainty

HTML全文

图 1 泵站输水系统

Fig. 1 Pumping station water conveyance system

下载: 全尺寸图片幻灯片

图 2 双层优化调度机制

Fig. 2 Two-layer optimization scheduling mechanism

下载: 全尺寸图片幻灯片

图 3 智能体序贯决策过程

Fig. 3 Sequential decision-making process of the agent

下载: 全尺寸图片幻灯片

图 4 Dueling DDQN网络结构

Fig. 4 The network structure of Dueling DDQN

下载: 全尺寸图片幻灯片

图 5 SE-Dueling DDQN学习框架

Fig. 5 The framework of SE-Dueling DDQN

下载: 全尺寸图片幻灯片

图 6 径流来水和需水预测

Fig. 6 Runoff inflow and water demand prediction

下载: 全尺寸图片幻灯片

图 7 平均奖励值变化

Fig. 7 The variation of average reward value

下载: 全尺寸图片幻灯片

图 8 泵站流量和机组运行数量

Fig. 8 Pumping station flow rate and number of operating units

下载: 全尺寸图片幻灯片

图 9 湖泊调蓄库容变化

Fig. 9 The variation of lake storage capacity

下载: 全尺寸图片幻灯片

图 10 站后水位变化

Fig. 10 The variation of water level behind the station

下载: 全尺寸图片幻灯片

图 11 泵站单位提水成本与能耗

Fig. 11 Unit water lifting cost and energy consumption of pumping station

下载: 全尺寸图片幻灯片

图 12 平均奖励值

Fig. 12 The average reward value

下载: 全尺寸图片幻灯片

表 1 各场景下瓦埠湖调蓄库容数据($ \times 10^4\,\;{\rm{m}}^3 $)

Table 1 Regulation and storage capacity data of Wabu Lake under various scenarios($ \times 10^4\,\;{\rm{m}}^3 $)

场景	初始调蓄库容	期末目标调蓄库容
场景一	4165	3580
场景二	8580	8580
场景三	2426	2456
场景四	8580	8580
场景五	2060	2576
场景六	8580	8580

下载: 导出CSV

表 2 八月(汛期)电价

Table 2 Electricity price for August (flood season)

时段	价格(元/kWh)
0:00~9:00	0.2911
9:00~16:00	0.6182
16:00~20:00; 22:00~24:00	1.0645
20:00~22:00	1.2597

下载: 导出CSV

表 3 十二月(非汛期)电价

Table 3 Electricity price for December (non-flood season)

时段	价格(元/kWh)
23:00 ~ 次日8:00	0.2682
8:00 ~ 15:00	0.5950
15:00 ~ 19:00; 21:00 ~ 23:00	1.0406
19:00 ~ 21:00	1.2355

下载: 导出CSV

表 4 各场景下算法的性能表现

Table 4 The performance of various algorithms in different scenarios

场景	算法	能耗成本(元)	调蓄库容偏差($ 10^4\,\;{\rm{m}}^3 $)
场景一	SE-Dueling DDQN	113380	+4.9
	PSO	114774	+8.8
	GA	117372	+10.4
场景二	SE-Dueling DDQN	42261	−5.8
	PSO	42471	−10.2
	GA	42587	−9.2
场景三	SE-Dueling DDQN	352402	+4.2
	PSO	371608	+5.5
	GA	364530	+4.6
场景四	SE-Dueling DDQN	38176	−10.0
	PSO	39233	−11.2
	GA	41098	−11.6
场景五	SE-Dueling DDQN	416866	+4.8
	PSO	436099	+9.9
	GA	420999	+7.7
场景六	SE-Dueling DDQN	108881	−8.9
	PSO	109759	−8.9
	GA	110826	−10.4

下载: 导出CSV

表 5 各算法训练与决策耗时

Table 5 Training and decision-making time consumption of various algorithms

算法	训练耗时(min)	决策耗时(s)
SE-Dueling DDQN	137	0.04
PSO	0	126
GA	0	141

下载: 导出CSV

表 6 不同算法的优化结果

Table 6 Optimization results under different algorithms

算法	能耗成本(元)	期末调蓄库容偏差($ 10^4\,\;{\rm{m}}^3 $)
SE-Dueling DDQN	42261	−5.8
SE-DDQN	42595	−5.7
SE-DQN	43012	−6.3
Dueling DDQN	44195	−10.7

下载: 导出CSV

表 7 不同误差水平下的优化结果

Table 7 Optimization results under different error

误差水平	能耗成本(元)	期末调蓄库容偏差($ 10^4\,\;{\rm{m}}^3 $)
$ k=1.0 $	42261	-5.8
$ k=1.5 $	42861	-7.1
$ k=2.0 $	42940	-10.3

下载: 导出CSV

参考文献(30)

[1]	刘为锋, 郭旭宁, 李云玲, 李洁玉, 刘奇, 张元教, 等. 南水北调东线一期工程补偿调度研究. 中国工程科学, 2025, 27(5): 236−248 doi: 10.15302/J-SSCAE-2025.07.006 Liu Wei-Feng, Guo Xu-Ning, Li Yun-Ling, Li Jie-Yu, Liu Qi, Zhang Yuan-Jiao, et al. Research on compensation operation of the first phase of the Eastern Route of South-to-North Water Diversion Project. Strategic Study of CAE, 2025, 27(5): 236−248 doi: 10.15302/J-SSCAE-2025.07.006
[2]	李昕阳, 郭旭宁, 李维雨, 王银堂, 董甲平, 万新宇, 等. 南水北调东线工程优化调度与区段水源组成分析. 水资源保护, 2025, 41(3): 171−178 doi: 10.3880/j.issn.1004-6933.2025.03.020 Li Xin-Yang, Guo Xu-Ning, Li Wei-Yu, Wang Yin-Tang, Dong Jia-Ping, Wan Xin-Yu, et al. Optimal operation and sectional water source composition analysis of the Eastern Route of South-to-North Water Diversion Project. Water Resources Protection, 2025, 41(3): 171−178 doi: 10.3880/j.issn.1004-6933.2025.03.020
[3]	郑和震, 张召, 吴辉明, 雷晓辉. 梯级泵站输水系统日优化调度及经济运行研究. 水利学报, 2016, 47(12): 1558−1565 doi: 10.13243/j.cnki.slxb.20151350 Zheng He-Zhen, Zhang Zhao, Wu Hui-Ming, Lei Xiao-Hui. Study on daily optimal scheduling and economic operation of cascade pumping station water conveyance system. Journal of Hydraulic Engineering, 2016, 47(12): 1558−1565 doi: 10.13243/j.cnki.slxb.20151350
[4]	刘小莲, 李贞蓉, 王雪妮, 翟宇, 张雷克, 郭维维, 等. 梯级泵站优化调度模型参数敏感性分析及其求解算法改进. 工程科学与技术, 2024, 5(6): 1−18 doi: 10.12454/j.jsuese.202400699 Liu Xiao-Lian, Li Zhen-Rong, Wang Xue-Ni, Zhai Yu, Zhang Lei-Ke, Guo Wei-Wei, et al. Parameter sensitivity analysis of optimal scheduling model for cascade pumping stations and improvement of its solution algorithm. Advanced Engineering Sciences, 2024, 5(6): 1−18 doi: 10.12454/j.jsuese.202400699
[5]	郭永灵, 张海晨, 朱兴林. 基于动态规划法的单级泵站日经济运行优化模型. 中国农村水利水电, 2020(1): 192−196 doi: 10.3969/j.issn.1007-2284.2020.01.039 Guo Yong-Ling, Zhang Hai-Chen, Zhu Xing-Lin. Optimization model for daily economic operation of single-stage pumping station based on dynamic programming. China Rural Water and Hydropower, 2020(1): 192−196 doi: 10.3969/j.issn.1007-2284.2020.01.039
[6]	胡懿, 罗竹梅, 曾云. 基于双层优化模型的不同型号混流泵经济运行. 排灌机械工程学报, 2025, 43(9): 923−928 doi: 10.3969/j.issn.1674-8530.23.0207 Hu Yi, Luo Zhu-Mei, Zeng Yun. Economic operation of mixed-flow pumps with different models based on bi-level optimization model. Journal of Drainage and Irrigation Machinery Engineering, 2025, 43(9): 923−928 doi: 10.3969/j.issn.1674-8530.23.0207
[7]	Liu X L, Tian Y, Lei X H, Wang H, Liu Z R, Wang J. An improved self-adaptive grey wolf optimizer for the daily optimal operation of cascade pumping stations. Applied Soft Computing, 2019, 75: 473−493 doi: 10.1016/j.asoc.2018.11.039
[8]	马夏敏, 张雷克, 刘小莲, 田雨, 王雪妮, 邓显羽, 等. 基于麻雀搜索算法的梯级泵站优化调度. 水力发电学报, 2024, 43(5): 43−53 doi: 10.11660/slfdxb.20240505 Ma Xia-Min, Zhang Lei-Ke, Liu Xiao-Lian, Tian Yu, Wang Xue-Ni, Deng Xian-Yu, et al. Optimal scheduling of cascade pumping stations based on sparrow search algorithm. Journal of Hydroelectric Engineering, 2024, 43(5): 43−53 doi: 10.11660/slfdxb.20240505
[9]	Kong L Z, Li Y Q, Tang H W, Yuan S Y, Yang Q, Ji Q f, et al. Predictive control for the operation of cascade pumping stations in water supply canal systems considering energy consumption and costs. Applied Energy, 2023, 341: Article No. 121103 doi: 10.1016/j.apenergy.2023.121103
[10]	Yan P R, ZHANG Z, Lei X H, Hou Q Z, Wang H. A multi-objective optimal control model of cascade pumping stations considering both cost and safety. Journal of Cleaner Production, 2022, 345: Article No. 131171 doi: 10.1016/j.jclepro.2022.131171
[11]	Mallios Z, Siarkos I, Karagiannopoulos P, Tsiarapas A. Pumping energy consumption minimization through simulation-optimization modelling. Journal of Hydrology, 2022, 612: Article No. 128062 doi: 10.1016/j.jhydrol.2022.128062
[12]	吴怡, 李智, 王京晶, 曹悦, 雷晓辉. 梯级泵站输水系统旬优化调度及经济运行研究. 人民黄河, 2018, 40(5): 144−152 doi: 10.3969/j.issn.1000-1379.2018.05.031 Wu Yi, Li Zhi, Wang Jing-Jing, Cao Yue, Lei Xiao-Hui. Study on ten-day optimal scheduling and economic operation of cascade pumping station water conveyance system. Yellow River, 2018, 40(5): 144−152 doi: 10.3969/j.issn.1000-1379.2018.05.031
[13]	桑国庆, 曹升乐, 郭瑞, 官庆朔, 白咸勇. 梯级泵站输水系统运行效率优化模型. 系统工程理论与实践, 2014, 34(8): 2179−2185 Sang Guo-Qing, Cao Sheng-Le, Guo Rui, Guan Qing-Shuo, Bai Xian-Yong. Operation efficiency optimization model for cascade pumping station water conveyance system. Systems Engineering-Theory & Practice, 2014, 34(8): 2179−2185
[14]	Olszewski P. Genetic optimization and experimental verification of complex parallel pumping station with centrifugal pumps. Applied Energy, 2016, 178: 527−539 doi: 10.1016/j.apenergy.2016.06.084
[15]	Abkenar S M, Stanley S D, Miller C J, Chase D V, Mcelmurry S P. Evaluation of genetic algorithms using discrete and continuous methods for pump optimization of water distribution systems. Sustainable Computing: Informatics and Systems, 2015, 8: 18−23 doi: 10.1016/j.suscom.2014.09.003
[16]	Chen W P, Tao T, Zhou A J, Zhang L, Liao L, Wu X M, et al. Genetic optimization toward operation of water intake-supply pump stations system. Journal of Cleaner Production, 2021, 279: Article No. 123573 doi: 10.1016/j.jclepro.2020.123573
[17]	Torregrossa D, Capitanescu F. Optimization models to save energy and enlarge the operational life of water pumping systems. Journal of Cleaner Production, 2019, 213: 89−98 doi: 10.1016/j.jclepro.2018.12.124
[18]	刘祥祥, 朱一松, 高越飞, 邓轲, 刘炳义, 陈华. 改进粒子群算法耦合水力模型的供水优化调度分析. 中国给水排水, 2025, 41(9): 59−65 doi: 10.19853/j.zgjsps.1000-4602.2025.09.009 Liu Xiang-Xiang, Zhu Yi-Song, Gao Yue-Fei, Deng Ke, Liu Bing-Yi, Chen Hua. Analysis of water supply optimal scheduling based on improved particle swarm optimization coupled with hydraulic model. China Water Wastewater, 2025, 41(9): 59−65 doi: 10.19853/j.zgjsps.1000-4602.2025.09.009
[19]	Wang X N, Ma X M, Liu X L, Zhang L k, Tian Y, Ye C. Research on optimal operation of cascade pumping stations based on an improved sparrow search algorithm. Water Science & Technology, 2023, 88(8): 1982−2001 doi: 10.2166/wst.2023.308
[20]	Guan J Y, Tang H, Wang J Y, Yao J G; Wang K; Mao W B. A GAN-based fully model-free learning method for short-term scheduling of large power system. IEEE Transactions on Power Systems, 2021, 37(4): 2655−2665 doi: 10.1109/tpwrs.2021.3121673
[21]	唐昊, 刘畅, 杨明, 汤必强, 许丹, 吕凯. 考虑电网调峰需求的工业园区主动配电系统调度学习优化. 自动化学报, 2021, 47(10): 2449−2463 doi: 10.16383/j.aas.c190079 Tang Hao, Liu Chang, Yang Ming, Tang Bi-Qiang, Xu Dan, Lv Kai. Learning optimization for scheduling of active distribution system in industrial parks considering peak shaving demand of power grid. Acta Automatica Sinica, 2021, 47(10): 2449−2463 doi: 10.16383/j.aas.c190079
[22]	Xu W, Zhang X L, Peng A B, Liang Y. Deep reinforcement learning for cascaded hydropower reservoirs considering inflow forecasts. Water Resources Management, 2020, 34(9): 3003−3018 doi: 10.1007/s11269-020-02600-w
[23]	Wu R X, Wang R, Hao J, Wu Q, Wang P. Multiobjective multihydropower reservoir operation optimization with transformer-based deep reinforcement learning. Journal of Hydrology, 2024, 632: Article No. 130904 doi: 10.1016/j.jhydrol.2024.130904
[24]	Luo W, Wang C, Zhang Y H, Zhao J S, Huang Z F, Wang J Q, et al. A deep reinforcement learning approach for joint scheduling of cascade reservoir system. Journal of Hydrology, 2025, 651: Article No. 132515 doi: 10.1016/j.jhydrol.2024.132515
[25]	Gan T, Jiang Y, Zhao H, He J Y, Duan H. Research on low-energy consumption automatic real-time regulation of cascade gates and pumps in open-canal based on reinforcement learning. Journal of Hydroinformatics, 2024, 26(7): 1673−1691 doi: 10.2166/hydro.2024.020
[26]	Zhu J, Zhang Z, Lei X H, Jing X, Wang H, Yan P R. Ecological scheduling of the middle route of South-to-North Water Diversion Project based on a reinforcement learning model. Journal of Hydrology, 2021, 596: Article No. 126107 doi: 10.1016/j.jhydrol.2021.126107
[27]	周举, 陈黎明, 胡腾飞, 陆昊, 施勇, 韦诚. 洪潮影响下潭江流域安全行洪阈值研究. 水资源保护, 2024, 40(5): 53−61 doi: 10.3880/j.issn.1004-6933.2024.05.007 Zhou Ju, Chen Li-Ming, Hu Teng-Fei, Lu Hao, Shi Yong, Wei Cheng. Study on safe flood discharge threshold of Tanjiang River Basin under the influence of flood and tide. Water Resources Protection, 2024, 40(5): 53−61 doi: 10.3880/j.issn.1004-6933.2024.05.007
[28]	贾玉博, 粟晓玲, 周伟, 武连洲, 王太山, 王芊予, 等. 水网布局下陕西省主要水系的径流丰枯转移特征及丰枯遭遇研究. 水资源与水工程学报, 2025, 36(4): 8−16 doi: 10.11705/j.issn.1672-643X.2025.04.02 Jia Yu-Bo, Su Xiao-Ling, Zhou Wei, Wu Lian-Zhou, Wang Tai-Shan, Wang Qian-Yu, et al. Study on runoff wet-dry transfer characteristics and wet-dry encounter of major river systems in Shaanxi Province under water network layout. Journal of Water Resources and Water Engineering, 2025, 36(4): 8−16 doi: 10.11705/j.issn.1672-643X.2025.04.02
[29]	张朝, 刘海星, 顾建强, 武连洲, 王太山, 王芊予, 等. 基于需水不确定性的供水管网优化运行研究. 给水排水, 2019, 55(11): 124−130 doi: 10.13789/j.cnki.wwe1964.2019.11.027 Zhang Chao, Liu Hai-Xing, Gu Jian-Qiang, Wu Lian-Zhou, Wang Tai-Shan, Wang Qian-Yu, et al. Research on optimal operation of water supply network based on water demand uncertainty. Water & Wastewater Engineering, 2019, 55(11): 124−130 doi: 10.13789/j.cnki.wwe1964.2019.11.027
[30]	谢雨祚, 郭生练, 钟斯睿, 汪肖雅, 王俊. 金沙江下游梯级水库运行期设计洪水计算方法研究. 水利学报, 2025, 56(3): 354−363 doi: 10.13243/j.cnki.slxb.20240382 Xie Yu-Zuo, Guo Sheng-Lian, Zhong Si-Rui, Wang Xiao-Ya, Wang Jun. Study on design flood calculation method for downstream cascade reservoirs of Jinsha River during operation period. Journal of Hydraulic Engineering, 2025, 56(3): 354−363 doi: 10.13243/j.cnki.slxb.20240382