Integrated numerical–physical modeling for optimizing riverbed dredging in Mixed Pumped-Storage Power Stations: a case study of Lianghekou project Authors: Rui Bai, Jiaye Li, Yingjun Song, Dezhi Fu, Wei Jin, Jun Chen, Jianbing Li Published: 2026 ABSTRACT The Mixed Pumped-Storage Power Station (MPSPS) constitutes an emerging strategy for the development of integrated clean-energy bases that combine hydropower, wind, and solar resources. However, a critical literature gap exists regarding effective engineering measures to mitigate the “pumping funnel” phenomenon in the lower reservoir, which remains a key operational challenge for MPSPS. Taking the Lianghekou MPSPS as a case study, this study proposes riverbed dredging as a mitigation measure and develops an integrated numerical-physical modeling approach for its optimization. The optimal scheme was selected based on a combined criterion of pumping-effect improvement and construction cost. The most promising scheme was further verified through large-scale physical model experiments of the entire lower reservoir. The results show that the selected riverbed dredging strategy substantially improves the pumping reliability of the Lianghekou MPSPS system, expanding the river’s cross-sectional area by 15–23%, reducing reverse flow by 30–42%, suppressing pumping funnel depth to 65% of its original value, and increasing the continuous pumping duration by 61%. Within the context of the Lianghekou project, these findings provide a quantitative basis for engineering design, construction, and operational scheduling, while also offering potential reference for similar MPSPS applications in large-scale renewable energy integration. Introduction The Mixed Pumped-Storage Power Station (MPSPS), formed through expanding cascade hydropower systems, constitutes a specialized pumped storage facility recognized as an efficient “high-quality long-duration energy storage” solution1–3. Within the context of modern power systems with large-scale renewable energy integration, the MPSPS embraces rapid development and significantly facilitates the development of basin-scale clean energy bases integrating hydropower, wind, and solar energy and power storage4. Many MPSPS have been put into operation or under construction worldwide, and they have broad application prospects in the future. The pumping operation in the MPSPS lower reservoir creates distinctive hydraulic conditions that can lead to a “pumping funnel” phenomenon5–7. Particularly in cases where the lower reservoir is situated in a canyon-type river channel, the pumping funnel effect intensifies, which can readily undermine the pumping reliability of the MPSPS. Li5, verified this phenomenon using dual hydrodynamic models and summarized the rule of pumping reliability. Consequently, the pump-turbine units of the MPSPS may operate inefficiently and unsteadily, and full pumping power and storage capacity may not be guaranteed. This potentially results in additional energy losses that fail to meet the design specifications of the power station. These factors collectively reduce the energy storage efficiency of the MPSPS project and diminish its economic viability. To improve the pumping reliability in the lower reservoir of the MPSPS, engineering measures, including riverbed dredging, should be implemented to optimize hydraulic conditions under pumping operations. Research has found that the most direct solution is to reduce the pumping power of the MPSPS or decrease the number of operating pump units, thereby reducing the pumping discharge, extending the sustainable pumping time, and effectively improving the pumping reliability. Under the premise of maintaining engineering installed capacity and design specifications, feasible engineering measures include optimizing inlet/outlet locations, adjusting inlet/outlet angles, implementing riverbed dredging, and locally widening the water surface around the inlet/outlet. It should be noted that other potential measures, including optimization of the inlet/outlet layout and local widening of the water surface, have been quantitatively evaluated in our previous study5. That study showed that while inlet/outlet optimization improves pumping reliability to some extent, local widening alone is insufficient because pumping reliability is constrained by the narrowest cross-section along the river channel. In contrast, riverbed dredging directly enlarges the critical cross-sectional area and thus provides a more effective and reliable improvement. Therefore, the physical model experiments in the present study adopt the optimized inlet/outlet layout from our previous work and focus on riverbed dredging as the primary measure for enhancing conveyance capacity. Scholars have carried out some research on riverbed dredging8–10. As an essential component of water conservancy and hydropower engineering construction, riverbed dredging can effectively improve the hydraulic conveyance capacity of river channels, enhance navigational efficiency, strengthen flood control capabilities, mitigate flood risks, and improve water environmental quality11–13. Nowadays, there are various approaches and techniques for riverbed dredging, including underwater dredging and dry riverbed dredging14,15. Due to the increase of pumping power, the phenomenon of pumping funnel has been discovered, thereby triggering research on pumping reliability and riverbed dredging. For example, in the design stage of recent projects such as the Guangma MPSPS and Ankang MPSPS, Fan et al.16, and Yang et al.17, put forward a concept regarding riverbed dredging in the lower reservoir of MPSPSs to improve the pumping hydraulic conveyance capacity. Overall, research on the pumping reliability and riverbed dredging of MPSPSs is still in the early stages. Furthermore, studies that use methods such as hydrodynamic numerical simulations and physical model experiments to investigate the pumping reliability of MPSPSs are even less common. Therefore, based on the existing literature survey, there is a lack of theoretical studies and summaries regarding the riverbed dredging practice for MPSPSs. In summary, to improve the pumping reliability of MPSPSs and support the design and optimization of riverbed dredging measures, it is necessary to analyze the hydrodynamic simulation results of different riverbed dredging plans, evaluate the pumping reliability of each dredging plan, to recommend an appropriate dredging plan based on a comprehensive consideration of dredging effects and costs. And then the physical model experiment for the entire lower reservoir of the MPSPS should be conducted for the recommended dredging plan to verify the corresponding pumping reliability. Methods This paper proposes a riverbed dredging plan analysis method for the pumping reliability of a MPSPS, combining hydrodynamic simulation with physical model experiments. Hydrodynamic numerical simulations are conducted for the proposed riverbed dredging plans to evaluate the simulation results and their pumping reliability, thereby selecting an appropriate dredging solution. Based on this, a physical model experiment for the recommended dredging plan is performed to compare and support the accuracy of the numerical simulation result, and to verify the pumping reliability of the MPSP after dredging. Hydrodynamic numerical simulation Model selection MIKE 21, developed by the Danish Institute of Hydraulics, is a two-dimensional free-surface flow numerical simulation system for free-surface flows. Its core fluid dynamics module is based on the Saint-Venant equations, derived from the three-dimensional incompressible Navier-Stokes equations18–21. The accuracy of the model is highly dependent on the riverbed roughness parameter. The Manning’s coefficient is influenced by multiple factors, such as riverbed morphology, vegetation cover, and river sinuosity22,23. After continuous improvement and engineering validation, MIKE 21 has been successfully applied in simulation analyses of major domestic water conservancy projects, such as the Yangtze River Estuary Management and the South-to-North Water Diversion Project24. Model construction The hydrodynamic modelling process for MPSPS lower reservoirs under pumping operations involves four critical phases: riverbed topographic processing and meshing, boundary condition configuration, temporal discretization optimization, and roughness parameter calibration14. Firstly, detailed measured topographic data of the entire reservoir is collected, organized, and converted into high-precision raster topographic data, generating two-dimensional triangular or quadrilateral grids25. To balance the relationship between computational accuracy, grid density, and simulation time, an uneven grid division method can be employed. The grid size is appropriately increased in relatively less important areas, while local river sections near the inlet/outlet are refined with higher grid resolution26. Secondly, the flow boundary conditions of the model are the time series of the three flow inlets/outlets of the lower reservoir: the discharge flow from the upper reservoir power station (flowing into the lower reservoir), the discharge flow from the lower reservoir power station (flowing out to the downstream), and the pumping flow of the MPSPS (pumping from the lower reservoir to the upper reservoir). Next, the total calculation time of the hydrodynamic model generally corresponds to the longest continuous full-power pumping condition designed for the MPSPS, which typically meets the energy storage needs of the power system for 8 h. To balance computational accuracy and simulation time, a segmented simulation strategy is employed27. A larger time step is used when the pumping reliability is high, and a smaller time step is used when the pumping reliability is low, ensuring the accuracy of the simulation results during critical moments. Finally, based on the flow data and corresponding water level relationships at various measured cross-sections of the reservoir, the riverbed roughness of the hydrodynamic model is calibrated. Considering that the reservoir is narrow and riverine, there may be significant differences in factors such as channel slope, bed material composition, and cross-sectional geometry along different river segments. To ensure that the calibration results accurately reflect the real terrain features and flow resistance of the reservoir, the riverbed roughness is divided into segments, and a segment-by-segment calibration approach is used to improve calibration accuracy. Optimization of riverbed dredging plan The selection of the riverbed dredging plans is primarily to evaluate the pumping hydrodynamic effect and the construction cost of these dredging plans, and then make a comprehensive techno-economic comparison. Based on this, the pumping effect and the construction cost of each dredging plan are integrated to calculate the comprehensive evaluation index \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{E}_{summary}$$\end{document}. The formula is as follows:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{l}{E}_{summary}={V}_{pump}-{C}_{con}\:\end{array}$$\end{document} Where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{pump}$$\end{document} represents the pumping reliability of the MPSPS and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{C}_{con}$$\end{document} represents the construction cost of the riverbed dredging plan. Considering that both \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{pump}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{C}_{con}$$\end{document} are dimensionless indicators with comparable value ranges (0–2), the subtraction form is adopted rather than the ratio form for two main reasons. First, subtraction provides an intuitive range (0.13–0.94 in this case), where values near 0 reflect low cost-effectiveness and values near 1 reflect high cost-effectiveness. Second, unlike the ratio form, subtraction is unaffected by the magnitude of the denominator, avoiding the bias that arises when a small denominator inflates the ratio of a mediocre scheme. Consequently, the subtraction form yields more reasonable and engineering-consistent rankings. Pumping effect evaluation To evaluate the pumping hydrodynamic effect of riverbed dredging plans, it is necessary to analyze the hydrodynamic simulation results of different dredging plans. The specific steps are as follows: (1) Based on the proposed riverbed dredging plans, construct the riverbed topography for each plan. (2) Perform hydrodynamic modeling for the different topographies and conduct numerical simulations under various pumping conditions. (3) Analyze the hydrodynamic simulation results corresponding to each dredging plan, extract quantifiable characteristic values, and calculate the weighted pumping effect evaluation index. According to the calculation method for the pumping reliability index of MPSPS proposed by Li et al.5,, the pumping effect evaluation index \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{pump}$$\end{document} is considered as follows:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{l}{V}_{pump}={V}_{reH}+{V}_{reQ}\:\:\end{array}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{reH}$$\end{document} is the pumping duration guarantee rate index, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{reQ}$$\end{document} is the pumping flow guarantee rate index, both are dimensionless quantities ranging from 0 to 1. Thus, the value range of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{pump}$$\end{document} is from 0 to 2. Construction cost evaluation To evaluate the construction cost of riverbed dredging plans, it is necessary to analyze from two perspectives: total cost and construction risk. On one hand, the dredging methods and construction organization of each plan are analyzed, considering the auxiliary costs and quantities of dredging construction, the total construction cost is calculated. On the other hand, the dredging construction difficulty and safety risks under construction are assessed. The weighted construction cost evaluation indexes of different dredging plans \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{C}_{con}$$\end{document} are calculated using the following formula:3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{l}{C}_{con}={K}_{1}\left(\frac{{C}_{cost}}{{C}_{budget}}\right)+{K}_{2}{R}_{diff}\:\:\end{array}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{C}_{con}$$\end{document} includes two components: the acceptance level of the construction cost and technology risk for dredging construction. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{C}_{budget}$$\end{document} represents a reasonable and acceptable cost budget for dredging, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{C}_{cost}$$\end{document} denotes the calculated construction cost for implementing each dredging Plan. Considering that investors may abandon the project if the construction cost exceeds the budget by 100%, the value of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\frac{{C}_{cost}}{{C}_{budget}}$$\end{document} is expected to be below 2. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{R}_{diff}$$\end{document} is the technical risk coefficient associated with different construction methods adopted by each dredging Plan, determined by factors of construction difficulty and safety risk, its value is artificially set and introduced in the section "Construction methods", with a value range of 0 to 0.8 in this case. Finally, the construction cost and technology risk are comprehensively considered, with their relative importance determined by weight coefficients satisfying \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{K}_{1}+{K}_{2}=1$$\end{document}. In the present case study, the investor places greater emphasis on controlling construction cost overruns, so the weight for direct construction cost is set to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{K}_{1}=0.7$$\end{document} and given the relatively high level of construction technology and the investor’s construction management capability, the safety risk is considered manageable; therefore, the weight of technology risk is set to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{K}_{2}=0.3$$\end{document}. It should be noted that these weights may be adjusted for different regions or different investors. Physical experiment A physical model experiment of the entire lower reservoir is conducted for the selected dredging plan to verify the pumping reliability of the MPSPS. Model design and calibration Unlike a conventional partial river-reach model, a complete reservoir model should be established in this model design. The model should cover the entire river section of the lower reservoir, including the reach upstream to the upper reservoir dam and downstream to the lower reservoir dam. The elevation range must extend from the reservoir bottom to the normal water level, and a certain high level should be considered to facilitate the experiment. Considering that the length of the riverine reservoir can be pretty long (up to several tens of kilometers), the similarity ratio should be determined in such a way that it minimizes the overall size of the model, while prioritizing the accuracy of the physical model experiment observations. This ensures that the model remains manageable while providing reliable data for the experiment. Since the riverbed topography is artificially smoothed, the roughness of the model is typically lower than that of a natural river channel. After the model terrain is completed and the reservoir capacity adjustment is verified, it is necessary to calibrate the roughness. The calibration process is based on the measured flow and corresponding water level relationships at different cross-sections. The measured flow is converted to the model flow according to the similarity criterion. The water levels at corresponding cross-sections in the model are measured. After conversion using the similarity criterion, the model water levels are compared with the measured river channel water levels to assess the error. Analyze the trend of roughness increase or decrease at each cross-section. Adjust the model roughness by applying thicker or smoother cement coatings, or adding or removing small stones locally. Repeat the above steps until the model-measured water levels are in close agreement with the measured river channel water levels. Model case design and experimentation To conduct the physical model experiment, the initial water level is set to the normal water level of the lower reservoir, and flow conditions of different cases are drawn up, including discharge from the upper reservoir, discharge released into downstream river, and discharge pumped by the MPSPS. During the experiment, observe the changes in water levels at various cross-sections of the physical model in real-time, with particular attention to the pumping funnel phenomenon near the inlet/outlet of the lower reservoir. Convert the actual hourly changes of water levels based on the similarity criterion, and analyze key indicators such as the occurrence time and maximum depth of the pumping funnel, and continuous pumping duration. Summarize the results under different operating conditions and analyze the pumping reliability under the recommended dredging terrain of the riverine reservoir. Additionally, compare the water level variation process, pumping reliability, and other indicators from the physical model experiment with the corresponding results in the hydrodynamic numerical simulation under the same conditions. This will help assess the matching accuracy between the numerical and physical model results. Case study Project information This study focuses on the Lianghekou (LHK) MPSPS, an ongoing project on Sichuan’s Yalong River configured with the LHK Reservoir as its upper reservoir and an expanded Yagenyiji (YGYJ) Reservoir as its lower reservoir. Figure 1 depicts the resulting “two reservoirs, three stations” joint system, which is built upon this mixed pumped-storage scheme. Fig. 1Structure Schematic Diagram of the LHK MPSPS System. Structure Schematic Diagram of the LHK MPSPS System. The lower reservoir of the LHK MPSPS is a canyon-type riverine reservoir, a significant pumping funnel phenomenon has occurred near the inlet/outlet of the YGYJ reservoir, severely affecting its pumping reliability. Riverbed dredging plans Construction methods Based on existing construction methods, riverbed dredging can generally be classified into three categories: water surface construction, shoreline construction, and diversion tunnel construction. Different dredging plans should be proposed by analyzing the characteristics and effects of construction methods. The first method is water surface construction, which includes using dredging vessels and building surface working platforms to directly perform underwater dredging. This method is suitable for situations where the water flow is relatively calm and the river channel is relatively wide. The YGYJ reservoir of the LHK MPSPS is located in a narrow canyon terrain with rapid water flow, making it difficult to anchor dredging vessels and working platforms. Due to the high safety risks, this method is not recommended. The second method is shoreline construction. The simple approach is shoreline trenching, where construction machinery is used to excavate the riverbed soil and rocks along both riverbanks in accessible sections for dredging operations. This method, compared to surface construction, ensures the safety of construction machinery and personnel. Considering that the construction operates close to the water surface, the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{R}_{diff}$$\end{document} is assigned as 0.4. However, due to the limited boom length of the machinery, it is difficult to extend into the middle of the river, resulting in a less significant dredging effect. The improved approach is alternating blanketing and retreating dredging on both banks. Specifically, first fill material like stones and gravels is blanketed from the shore into the river to create a temporary working platform. Then, the machinery retreats while simultaneously excavating the riverbed soil and artificial fill material. After completing dredging on one bank, the process continues on the other bank. Because the blanketed material blocks the river channel and may cause localized backwater rise, there is a certain risk of inundation, so the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{R}_{diff}$$\end{document} is assigned as 0.8. The third method is diversion tunnel construction, which involves building cofferdams upstream and downstream of the construction section and creating a diversion tunnel for water conveyance, forming a closed area to provide dry conditions for dredging operations in the riverbed. This method can be combined with the diversion tunnel construction of the inlet/outlet of the lower reservoir, which only requires a relatively small cost to expand the scope of the closed dry area. Therefore, this \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{R}_{diff}$$\end{document} is assigned as 0.1. Dredging plans design First, using no riverbed dredging as a control plan, designated as Plan 0. This plan involves only the diversion construction for excavating and leveling the inlet/outlet, without conducting any riverbed dredging. Second, two riverbed dredging plans (Plans 1 and 2) are proposed using the shore-based construction method. Plan 1 involves simple riverbed dredging using the shore trenching method. Considering the limited boom length of the machinery, Plan 2 further develops the approach by implementing alternating blanketing and retreating dredging on both banks. Third, four riverbed dredging plans (Plans 3 to 6) are proposed using diversion tunnel construction. These four plans combine the necessary inlet/outlet diversion construction in Plan 0 with the riverbed dredging diversion construction. The only difference lies in the varying riverbed dredging range, the locations of the upstream and downstream cofferdams, and the arrangement and length of the diversion tunnels. Table 1Dredging Plans for the YGYJ Reservoir of the LHK MPSPS System.Dredging Plan No.Dredging Plan NameConstruction MethodInlet/Outlet Construction Diversion Integrated with Riverbed DredgingPlan 0No Dredging (Comparison)—Non-integratedPlan 1Shoreline TrenchingShoreline ConstructionNon-integratedPlan 2Shoreline Trenching & Blanketing and Retreating DredgingaShoreline ConstructionNon-integratedPlan 3Dry Excavation(Long Reach)Diversion Tunnel ConstructionIntegratedPlan 4Dry Excavation(Short Reach)Diversion Tunnel ConstructionIntegratedPlan 5Dry Excavation (Short Reach) & Tunnel Flow BypassbDiversion Tunnel ConstructionIntegratedPlan 6Dry Excavation(Extended Reach)Diversion Tunnel ConstructionIntegratedaBlanketing and Retreating Dredging means creating a temporary working platform from the shore forward to the river, and retreating the machinery while simultaneously excavating the shoreline riverbed soil and artificial fill material; bTunnel Flow Bypass means retaining the diversion tunnel as a permanent structure and keeping the flow-through function. Table 1Dredging Plans for the YGYJ Reservoir of the LHK MPSPS System.Dredging Plan No.Dredging Plan NameConstruction MethodInlet/Outlet Construction Diversion Integrated with Riverbed DredgingPlan 0No Dredging (Comparison)—Non-integratedPlan 1Shoreline TrenchingShoreline ConstructionNon-integratedPlan 2Shoreline Trenching & Blanketing and Retreating DredgingaShoreline ConstructionNon-integratedPlan 3Dry Excavation(Long Reach)Diversion Tunnel ConstructionIntegratedPlan 4Dry Excavation(Short Reach)Diversion Tunnel ConstructionIntegratedPlan 5Dry Excavation (Short Reach) & Tunnel Flow BypassbDiversion Tunnel ConstructionIntegratedPlan 6Dry Excavation(Extended Reach)Diversion Tunnel ConstructionIntegratedaBlanketing and Retreating Dredging means creating a temporary working platform from the shore forward to the river, and retreating the machinery while simultaneously excavating the shoreline riverbed soil and artificial fill material; bTunnel Flow Bypass means retaining the diversion tunnel as a permanent structure and keeping the flow-through function. Dredging Plans for the YGYJ Reservoir of the LHK MPSPS System. Dry Excavation (Long Reach) Dry Excavation (Short Reach) Dry Excavation (Extended Reach) aBlanketing and Retreating Dredging means creating a temporary working platform from the shore forward to the river, and retreating the machinery while simultaneously excavating the shoreline riverbed soil and artificial fill material; bTunnel Flow Bypass means retaining the diversion tunnel as a permanent structure and keeping the flow-through function. Among these four diversion tunnel dredging plans, Plan 3 considers a dredging range of 1300 m, including 200 m upstream and 1100 m downstream from the inlet/outlet. It requires the excavation of a diversion tunnel with a length of 1173 m. Plan 4 reduces the dredging range compared to Plan 3 by removing the 200 m upstream from the inlet/outlet and leaving only the 1100 m downstream. It requires the excavation of a diversion tunnel with a length of 922 m. Plan 5, based on Plan 4, retains the flow-through function of the diversion tunnel under pumping operations, to analyze the role of the permanent retention of the diversion tunnel in improving the pumping reliability. Plan 6 further extends the dredging range compared to Plan 4, reaching 2000 m downstream from the inlet/outlet. It requires the excavation of a diversion tunnel with a length of 1645 m. All dredging plans are shown in Fig. 2. Fig. 2Schematic Diagram of Riverbed Dredging Topography and Cross-Section Construction for the YGYJ Reservoir of the LHK MPSPS System. Schematic Diagram of Riverbed Dredging Topography and Cross-Section Construction for the YGYJ Reservoir of the LHK MPSPS System. Hydrodynamic simulation Modeling and parameter calibration The hydrodynamic modeling of the lower reservoir of the LHK MPSPS is focused on the entire YGYJ reservoir area, covering a river segment with a total length of 19.6 km. Based on the proposed riverbed dredging plans, the elevation data of riverbed topography in each plan has been adjusted and processed through spatial interpolation, serving as the topographical conditions for hydrodynamic modeling. An uneven grid division method was used, and different grid densities were experimented. A grid density that balances computational accuracy and simulation time was chosen. The river segment near the inlet/outlet (from 1 km upstream to 3 km downstream) is the key focus area for the pumping funnel phenomenon. A refined grid with a resolution of 3 m×3 m was used in this region to ensure the accuracy of the hydrodynamic calculations. A grid size of 8 m×8 m was used for the remaining river segments. The total number of grids for the entire reservoir area is approximately 161,000, and the simulation time for each pumping condition is around 3 to 5 h. A time-step strategy selecting different time steps based on pumping reliability is adopted to balance computational efficiency, accuracy, and convergence. Before the occurrence of unreliable pumping or a sudden drop in pumping flow (typically happening between 6 and 7 h, and in unfavorable conditions between 4 and 6 h), a time step of 1–5 s is used for numerical simulation. After the occurrence of unreliable pumping or a sudden drop in pumping flow, a time step of 0.1 s is used for numerical simulation to ensure the convergence of the model iteration. The YGYJ reservoir was divided into 21 river segments for the riverbed roughness calibration. The measured flow rates of 237 m3/s and 657 m3/s, along with the corresponding measured water levels at each cross-section, were used as calibration references. The roughness values were repeatedly adjusted until the simulated water levels at each cross-section closely matched the observed values. Through error statistics, the calibrated average water level errors for the 21 river segments ranged from 1.9% to 4.1%, which met the accuracy specifications. For the river segments undergoing riverbed dredging, the terrain was relatively regular, and the flow conditions were better, so a lower riverbed roughness value was chosen. As shown in Fig. 3. Fig. 3Hydrodynamic Modeling, Meshing, and Roughness Calibration for the YGYJ Reservoir of the LHK MPSPS System. Hydrodynamic Modeling, Meshing, and Roughness Calibration for the YGYJ Reservoir of the LHK MPSPS System. As shown in Fig. 3(a), the hydraulic layout comprises three flow boundaries: Boundary 1 (inflow from the LHK Reservoir to the YGYJ Reservoir), Boundary 2 (pumping flow withdrawn from the YGYJ Reservoir for the MPSPS), and Boundary 3 (outflow from the YGYJ Reservoir released to the downstream river). Simulation cases setup Considering the coordinated operational modes of LHK Hydropower Station (HPS), YGYJ HPS, and LHK MPSPS, various pumping cases that may occur at the lower reservoir are analyzed. The least favorable pumping conditions are selected for analyzing and comparing the pumping effects of the riverbed dredging plans. The simulated pumping conditions mainly include the outflow of the upper reservoir (Boundary 1), the pumping flow of the LHK MPSPS (Boundary 2), and the outflow of the lower reservoir (Boundary 3). The minimum discharge corresponding to the single-unit output of the LHK HPS (Boundary 1 flow rate of 129 ~ 146 m³/s) and the YGYJ HPS (Boundary 3 flow rate of 150 ~ 210 m3/s), and the full-power continuous pumping discharge of the MPSPS (Boundary 2 flow rate of 373–616 m3/s) are used as a typical operating case (Case B). Based on Case B, a less favorable case (Case A) is formed by increasing the Boundary 3 flow rate to 230 ~ 300 m3/s. Conversely, based on Case B, the flow rate of Boundary 1 is increased to 215 m3/s, forming a more favorable case (Case C). Due to the significant fluctuation of the water level between 2785 m and 2865 m of the LHK reservoir, the generating water head of the LHK HPS units also changes accordingly, causing the corresponding generating flow for the minimum single-unit output to fluctuate between 129 and 146 m3/s. Similarly, the pumping water head of the LHK MPSPS also changes accordingly, causing the full-power pumping flow to fluctuate between 349 and 600 m3/s. The pumping flow is a crucial and sensitive factor for pumping reliability. Therefore, according to different water levels of the LHK reservoir, the Case B group is further subdivided into six subcases. Among them, Cases B-1, B-2, and B-3 have larger pumping flow, leading to a more significant pumping funnel phenomenon. By analogy, Case A and C groups are corresponding sets to A-1, A-2, A-3, and C-1, C-2, C-3, respectively. Due to the fluctuation of the water level between 2598 m and 2605 m of the YGYJ reservoir, the generating water head of the YGYJ HPS units also changes accordingly, causing the corresponding generating flow for the minimum single-unit output to fluctuate between 150 and 210 m3/s. Similarly, the pumping water head of the LHK MPSPS also changes accordingly, causing the fluctuation of full-power pumping flow, e.g. full-power pumping flow in Case B-1 fluctuates between 600 and 572 m3/s. In summary, by analyzing the conditions, a total of 12 simulation cases have been formulated, as shown in Table 2. The initial water level of the simulated YGYJ reservoir for all cases is set to the normal water level of 2605 m. Table 2Pumping Cases for Hydrodynamic Simulation of the YGYJ Reservoir of the LHK MPSPS System.Case No.LHK ReservoirWater Level (m)Boundary 2LHK MPSPS Pumping Flow (m3/s)Boundary 1LHK ReservoirOutput Flow (m3/s)Boundary 3YGYJ ReservoirOutput Flow (m3/s)A-12785600 ~ 572146230 ~ 300A-22795571 ~ 546139A-32806526 ~ 505136B-12785600 ~ 572146150 ~ 210B-22795571 ~ 546139B-32806526 ~ 505136B-42825467 ~ 450133B-52847403 ~ 389129B-62865358 ~ 349129C-12785600 ~ 572215150 ~ 210C-22795571 ~ 546C-32806526 ~ 505 Table 2Pumping Cases for Hydrodynamic Simulation of the YGYJ Reservoir of the LHK MPSPS System.Case No.LHK ReservoirWater Level (m)Boundary 2LHK MPSPS Pumping Flow (m3/s)Boundary 1LHK ReservoirOutput Flow (m3/s)Boundary 3YGYJ ReservoirOutput Flow (m3/s)A-12785600 ~ 572146230 ~ 300A-22795571 ~ 546139A-32806526 ~ 505136B-12785600 ~ 572146150 ~ 210B-22795571 ~ 546139B-32806526 ~ 505136B-42825467 ~ 450133B-52847403 ~ 389129B-62865358 ~ 349129C-12785600 ~ 572215150 ~ 210C-22795571 ~ 546C-32806526 ~ 505 Pumping Cases for Hydrodynamic Simulation of the YGYJ Reservoir of the LHK MPSPS System. Numerical simulation results For the hydrodynamic modeling of the 7 riverbed dredging plans and the flow boundary conditions of the 12 simulation cases, a combined hydrodynamic simulation was conducted, resulting in 84 sets of numerical simulation results. Results of typical case for different dredging plans By fixing the typical case and comparing the pumping simulation results for different dredging plans, an analysis was conducted. Among the 12 cases, a typical unfavorable case (B-2) was selected as a representative, then a less severe case (C-3) was considered, and a more severe case (A-1) was chosen. The pumping simulation results for different dredging plans are shown in Fig. 4. Fig. 4Hydrodynamic Simulation Results of Typical Operational Cases under Different Channel Dredging Plans for the LHK MPSPS System. Hydrodynamic Simulation Results of Typical Operational Cases under Different Channel Dredging Plans for the LHK MPSPS System. Firstly, from the comparison of pumping flow rates, it can be observed that in all cases, whether it is Case A-1 (Fig. 4g), Case B-2 (Fig. 4d), or Case C-3 (Fig. 4a), there is a noticeable deviation between the actual pumping flow rate and the expected flow rate. Moreover, the pattern of this deviation is generally similar across all cases. Among the tested dredging plans, Plan 0 (the reference plan) deviates from the expected flow rate the earliest, showing the least favorable pumping performance. In contrast, Dredging Plans 3 to 6, which involve the construction of diversion tunnels, experience deviations from the expected flow rate later, exhibiting the most favorable pumping performance among all plans. The dredging plans involving shoreline construction, namely Plan 1 and Plan 2, experience deviations from the expected flow rate at a time between the reference Plan 0 and the diversion tunnel-based Plans 3 to 6; their pumping performance is intermediate relative to the other plans. Secondly, from the comparison of water levels at the same time, it can be observed that in all cases, whether it is Case A-1 (Fig. 4h and i), Case B-2 (Fig. 4e and f), or Case C-3 (Fig. 4b and c), a pumping funnel phenomenon occurs, and the pattern of this pumping funnel is generally similar across all cases. When continuous pumping reaches the same duration, dredging Plan 0 exhibits the deepest pumping funnel and the steepest water level drop. Dredging Plans 3 to 6, which involve the construction of diversion tunnels, have the shallowest pumping funnel and the most gradual water level decrease. Dredging Plans 1 and 2, which involve shoreline construction, show a pumping funnel performance between the two aforementioned cases. Further comparison of dredging Plans 3 to 6, which involve diversion tunnel construction, shows that in any case, based on the deviation in flow time and the pumping funnel degree at the same time, the performance generally follows the pattern: Plan 6 (light green) > Plan 3 (orange) ≈ Plan 5 (light purple) > Plan 4 (dark red). Thirdly, from the comparison of velocity, an analysis was conducted by fixing Case B-2 and comparing the pumping simulation velocities along the streamline for different dredging plans, as shown in Fig. 5. Observe the reverse flow velocity of river sections within the range of 0 to 1 km from the MPSPS inlet/outlet section. After 3 h of pumping (Fig. 5a), Plan 0 has the maximum reverse flow velocity, followed by Plans 1 and 2, while Plans 3 to 6 have the minimum reverse flow velocity. It is indicated that the method of diversion tunnel construction has the best effect in improving the reliability of water pumping. After 6 h of pumping (Fig. 5b), the magnitudes of the reverse flow velocities show the order: Plan 6 (light green) < Plan 3 (orange) ≈ Plan 5 (light purple) < Plan 4 (dark red), which is consistent with the results in Fig. 4. Fig. 5Projected Velocity along Streamline of Case B-2 under Different Channel Dredging Plans for the LHK MPSPS System. Projected Velocity along Streamline of Case B-2 under Different Channel Dredging Plans for the LHK MPSPS System. Results of the recommended plan under different dredging cases Based on the river dredging plan selection conclusion in the section "Optimal dredging plan selection", Dredging Plan 4 is fixed as the recommended plan. The pumping simulation effects of this plan under different cases are analyzed, as shown in Fig. 6. Fig. 6Hydrodynamic Simulation Results of the Recommended Channel Dredging Plan under Different Operational Cases for the LHK MPSPS System. Hydrodynamic Simulation Results of the Recommended Channel Dredging Plan under Different Operational Cases for the LHK MPSPS System. Figure 6a and c represent the six cases in Group B, used to analyze the impact of constant outflow from the LHK and YGYJ reservoirs, with gradually decreasing pumping flow, on the pumping reliability. The results show that a decrease in pumping flow can significantly extend the time during which the pumping flow deviates from the expected flow, reduce the depth of the shallowest pumping funnel, and slow down the slope of the water level drop. This effectively improves the pumping reliability. Figure 6d and f select three cases, A-1, B-1, and C-1, to analyze the changes in pumping reliability when the pumping flow is significant and constant, and the outflow conditions from the LHK and YGYJ reservoirs are either relaxed or tightened. The results reveal that when the pumping flow is large, increasing the outflow from the YGYJ reservoir or decreasing the outflow from the YGYJ reservoir can extend the time during which the pumping flow deviates from the expected flow, slow down the degree of the pumping funnel phenomenon, and improve the pumping reliability. Figure 6g and i select three cases, A-3, B-3, and C-3, and similar analysis shows that increasing the outflow from the LHK reservoir or decreasing the outflow from the YGYJ reservoir can improve pumping reliability. Additionally, comparing Fig. 6d and f, it can also be concluded that a decrease in pumping flow enhances pumping reliability. Optimal dredging plan selection According to the optimization method in the section "Optimization of riverbed dredging plan", the pumping effectiveness and cost of the riverbed dredging plans are evaluated. The calculation results of the evaluation indicators are shown in Table 3. Firstly, a comparison is made between the two construction methods: shoreline construction and diversion tunnel construction. In terms of construction method costs, shoreline construction has poor operability. Current construction machinery is not yet capable of achieving large-scale and deep riverbed dredging. Especially when the river cannot consistently maintain low water levels and low flow rates, the shoreline construction platform is prone to flooding, which increases construction risks and makes it difficult. Diversion tunnel construction carries lower risks but has a higher construction cost. The cost indicators for shoreline construction and diversion tunnel construction range from 0.93 to 1.12 and 0.82 to 1.39, respectively. Table 3Comprehensive Comparative Metrics of Channel Dredging Plans for the LHK MPSPS System.Plan No. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{C}_{cost}$$\end{document} (million CNY) \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\frac{{C}_{cost}}{{C}_{budget}}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{R}_{diff}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{C}_{con}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{H}_{pump}$$\end{document} (hour) \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{reH}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{reQ}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{pump}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{E}_{summary}$$\end{document} Plan 01710.570.10.434.10.340.340.670.25Plan 13491.160.40.935.10.580.571.150.21Plan 23781.260.81.125.30.620.631.260.13Plan 34061.350.10.986.70.900.881.780.80Plan 43401.130.10.826.60.880.891.760.94Plan 53541.180.10.866.70.890.911.800.94Plan 65821.940.11.398.01.000.981.980.59 Table 3Comprehensive Comparative Metrics of Channel Dredging Plans for the LHK MPSPS System.Plan No. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{C}_{cost}$$\end{document} (million CNY) \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\frac{{C}_{cost}}{{C}_{budget}}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{R}_{diff}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{C}_{con}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{H}_{pump}$$\end{document} (hour) \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{reH}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{reQ}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{V}_{pump}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{E}_{summary}$$\end{document} Plan 01710.570.10.434.10.340.340.670.25Plan 13491.160.40.935.10.580.571.150.21Plan 23781.260.81.125.30.620.631.260.13Plan 34061.350.10.986.70.900.881.780.80Plan 43401.130.10.826.60.880.891.760.94Plan 53541.180.10.866.70.890.911.800.94Plan 65821.940.11.398.01.000.981.980.59 Comprehensive Comparative Metrics of Channel Dredging Plans for the LHK MPSPS System. Regarding dredging effectiveness, the dredging effectiveness indicator for diversion tunnel construction ranges from 1.76 to 1.98, while for shoreline construction, the dredging effectiveness indicator only reaches 1.15 (Plan 1). Even with sub-base dredging, the effectiveness evaluation can only reach 1.26 (Plan 2). This level of effectiveness cannot ensure that the pumping reliability in the lower reservoir reaches the engineering design level, which represents the maximum dredging capability of shoreline construction. In contrast, diversion tunnel construction can ensure that the pumping reliability in the lower reservoir meets the engineering design level, making it more effective in improving pumping reliability. From a comprehensive evaluation perspective, the shoreline construction method has a comprehensive evaluation indicator of only 0.21(Plan 1) and 0.13 (Plan 2), which reflects poor performance and high costs, making it not a recommended solution. Although the diversion tunnel construction method has higher construction costs, it offers low construction risks and better construction effectiveness. In the short river reach (Plan 4 and Plan 5) and long river reach (Plan 3) cases, the pumping reliability for most working conditions can meet the engineering design level. In the extended river reach (Plan 6), despite the increased construction cost, it can even ensure that most working conditions meet a higher pumping standard of 8 h. The comprehensive evaluation indicator for the diversion tunnel construction method ranges from 0.59 to 0.94, higher than the shoreline construction method. Therefore, the diversion tunnel construction method is recommended for riverbed dredging. Secondly, further optimization should be conducted among the four dredging plans (Plan 3 to Plan 6) for the diversion tunnel construction. The dredging effectiveness indicator for the short river reach (Plan 4 at 1.76) is slightly lower than that for the long river reach (Plan 3 at 1.78). However, the dredging effectiveness indicator for the short river reach with diversion tunnel flow (Plan 5 at 1.80) is even higher. Given the higher construction costs for the long river reach, the comprehensive evaluation indicators for the short river reach (both Plan 4 and Plan 5 at 0.94) are better than those for the long river reach (Plan 3 at 0.80). This indicates that Plan 3, which involves dredging the upstream 200 m of the inlet/outlet river section, is not cost-effective. The dredging effectiveness for the extended river reach (Plan 6 at 1.98) is significantly higher compared to the short river reach plans. Still, the project investment increases substantially, resulting in lower cost-effectiveness. The comprehensive evaluation indicator for Plan 6 (0.59) is even worse than that for the short river reach plans. This indicates that the marginal cost-effectiveness of extending the dredging range in the downstream river section of the inlet/outlet reservoir is decreasing. Compared to Plan 4, Plan 5 shows a certain improvement in dredging effectiveness (from 1.76 to 1.80), but the increase is not significant. This suggests that retaining the diversion tunnel’s flow capacity under pumping conditions does not notably improve the pumping reliability. Overall, both plans have optimal evaluation levels, with Plan 4 incurring lower construction costs. In summary, Plan 4 is recommended as the optimal plan for riverbed dredging. It can increase the continuous pumping duration from the original 4.1 h to 6.6 h with a growth rate of 61%. Physical model validation This study designed a physical model of the lower reservoir of the LHK MPSPS based on the principle of similarity, simulating the hydraulic characteristics of the river channel. The model was calibrated for accuracy through filling experiments and roughness adjustments. Seven operating cases were set up for the experiment, and the water pumping process was simulated. Comparisons between the physical model results and numerical simulations show good agreement in terms of pumping funnel depth evolution, continuous pumping duration, and water-level decline patterns under various operating cases. These quantitative agreements validate the reliability of both the physical model setup and the numerical model predictions. The research shows that riverbed dredging significantly enhances the reliability of the pumping process, providing technical support for the continuous operation of the MPSPS under extreme conditions. Model design, fabrication, and calibration According to the principle of similarity, the physical model of the full reservoir of the YGYJ reservoir at the lower site of the LHK MPSPS was designed based on the gravitational similarity criteria. A standard model was used, with a geometric scale of 1:80, a length scale of 80, a velocity scale of 8.944, a flow rate scale of 57,243, a time scale of 8.944, a sediment particle size scale of 80, and a roughness scale of 2.076. Fig. 7Design Drawing and Actual Production Rendering of Hydraulic Physical Modeling for the LHK MPSPS System. Design Drawing and Actual Production Rendering of Hydraulic Physical Modeling for the LHK MPSPS System. The model simulation covers the entire river channel from the dam of the LHK MPSPS to the dam of the YGYJ HPS, with a total length of approximately 19.6 km. The river channel is simulated up to an elevation of 2608 m, which is 3 m above the normal water level of 2605 m. The physical model was constructed based on the measured underwater topographic profiles. In the experiment, the potential pumping-funnel impacted reach (i.e., the river section from 1 km upstream to 1.5 km downstream of the inlet/outlet) of the YGYJ reservoir is designated as having a completely similar reach. Due to limitations in the experimental site, the other river sections are aligned while ensuring that the cross-sectional terrain of the river reach is similar. The plan layout diagram and the top view of the physical model of the YGYJ reservoir can be found in Fig. 7. Since the regulating storage volume of the lower reservoir is closely related to the pumping reliability, the corresponding verification and adjustment of the storage capacity was established to ensure the accuracy and rationality of model experiment. With a constant inflow rate, the water level in the model was filled from the dead water level to the normal storage level. The filling time was recorded, and according to similarity criteria, it was converted into the actual total storage volume. This was then compared with the required storage capacity for the river channel of the YGYJ reservoir—the error after calibration was found to be 3.57%, within the acceptable range. The physical model employed techniques such as roughening or smoothing the cement surface and locally increasing or decreasing small stones to adjust the model roughness. Using measured water levels and flow rates, the channel segment roughness was calibrated. The errors in the average water levels for each segment ranged between 2.4% and 4.1%, which is within the acceptable range. The model fabrication and verification realistically demonstrate the longitudinal and lateral cross-sectional morphological characteristics, storage capacity, and hydraulic features of the YGYJ reservoir over a river channel length of 19.6 km. Creating a large-scale (1:80) physical model of the entire YGYJ reservoir river channel for the analysis and verification of pumping reliability, this represents an innovative initiative in MPSPS It serves as a pioneering guide for the design of future large-scale pumped storage power stations. Experimental flow cases and model results Since frequent adjustment of the experiment flow rate may result in unstable flow regimes and accumulated errors, the pumping flow of the physical model was adjusted to maintain stable and change every 2 h, ensuring that the total pumping volume remained consistent with actual operating condition. The physical model experiment of the YGYJ reservoir at the LHK MPSPS has verified a total of seven cases, corresponding seven hydrodynamic numerical simulations have been conducted, as shown in Table 4. The initial water level of the reservoir for the experiment is at the normal storage level of 2605 m. This study employs electromagnetic flowmeters to monitor and control the inlet and outlet rates as the water withdrawal rates in real-time. This ensures that the water withdrawal process, with its continuous variations, can be accurately simulated according to the predefined working conditions. The model is equipped with water gauge cameras along its course, enabling precise monitoring and recording of the dynamic water levels during the model’s operation. Therefore, this model can realistically reproduce and record the hydraulic characteristics of a river-type MPSPS during the water withdrawal process. Table 4Setup of Physical Model Test Conditions and Comparison of Simulation Results.Case No.Pumping flowper 2 h(m3/s)LHK power generation flow (m3/s)Downstream YGYJ outflow (m3/s)The time when the water level at Gauge 3 reaches 2594 m (h)Numerical modelPhysical modelError1616→ 603→ 591→ 5791462106.086.21−0.132600→ 593→ 586→ 5801392106.626.76−0.153542→ 531→ 520→ 5101362107.267.47−0.214484→ 474→ 464→ 4551332108.038.27−0.245462→ 603→ 591→ 4341462108.378.56−0.196616→ 603→ 295→ 5911462106.877.01−0.147616→ 603→ 591→ 5791463005.525.70−0.18 Table 4Setup of Physical Model Test Conditions and Comparison of Simulation Results.Case No.Pumping flowper 2 h(m3/s)LHK power generation flow (m3/s)Downstream YGYJ outflow (m3/s)The time when the water level at Gauge 3 reaches 2594 m (h)Numerical modelPhysical modelError1616→ 603→ 591→ 5791462106.086.21−0.132600→ 593→ 586→ 5801392106.626.76−0.153542→ 531→ 520→ 5101362107.267.47−0.214484→ 474→ 464→ 4551332108.038.27−0.245462→ 603→ 591→ 4341462108.378.56−0.196616→ 603→ 295→ 5911462106.877.01−0.147616→ 603→ 591→ 5791463005.525.70−0.18 Setup of Physical Model Test Conditions and Comparison of Simulation Results. By comparing the results of physical model tests and numerical simulations under the same working conditions, and judging according to the standard that the water level near the pumped-storage inlet and outlet basically reaches the limit, it is found that the pumping duration of the physical model test is slightly longer than that of the numerical simulation results. The error ranges from 0.13 to 0.24 h, with the relative error between 2.0% and 3.2%. The difference between the two is insignificant, and the pumping reliability result of the physical model test is slightly more optimistic. This discrepancy may arise from two potential limitations. First, scale effects: the 1:80 physical model satisfies Froude similarity but cannot simultaneously achieve Reynolds and Weber similarity, amplifying the relative influences of viscous forces and surface tension. This tends to suppress pumping funnel development and produces overly smooth flow patterns, leading to more optimistic outcomes. Second, incomplete similarity: due to laboratory space constraints, only the inlet/outlet reach was fully similar; the remaining reaches were constructed with topographic similarity only, with bent alignments. This may cause additional energy losses and reduced water-surface drawdown, further weakening the observed pumping funnel. These factors are potential limitations rather than definitive causes. Future studies should consider larger physical model scales and fully similar construction to improve accuracy. Selecting Case 2 as a representative example, Fig. 8 shows the photos of the physical model experiment and the flow fields of the numerical simulation. From the results at 2 h of pumping (Fig. 8a and d), 4 h of pumping (Fig. 8b and e), and 6 h of pumping (Fig. 8e and f), there are similarities in the flow field pattern and flow velocity magnitude near the MPSPS inlet/outlet. When the flow velocity is low, the blue tracer in the test photos is relatively faint, and the flow field arrows in the numerical simulation are also small; when the flow velocity is high, the aggregation of the blue tracer in the test photos is more obvious, and the flow field arrows in the numerical simulation are also larger. Fig. 8Photos of Physical Model Experiment and Flow Fields of Numerical Simulation under Case 2 for the LHK MPSPS System. Photos of Physical Model Experiment and Flow Fields of Numerical Simulation under Case 2 for the LHK MPSPS System. By comparing the water levels of the hydrodynamic numerical simulation under operating Case 2 with the observational results of the physical model, as shown in Fig. 9. It indicates that the depth and range of the pumping funnel during the actual water withdrawal process of the physical model experiment are generally consistent with the numerical simulation results, with the physical model experiment water levels being slightly higher than the numerical simulation results. In addition, the physical model experiments showed a hydraulic gradient downstream of the YGYJ reservoir dam, and this hydraulic phenomenon was more pronounced than in the numerical simulations. Fig. 9Result Comparison of Physical Model Experiment and Numerical Simulation under Case 2 for the LHK MPSPS System. Result Comparison of Physical Model Experiment and Numerical Simulation under Case 2 for the LHK MPSPS System. Discussion Hydrodynamic simulation Firstly, the MIKE hydrodynamic numerical model was used to investigate the pumping reliability of the MPSPS, revealing the spatiotemporal evolution of the pumping funnel phenomenon: when the outflow from the LHK reservoir is small, the pumping flow and the outflow from the YGYJ reservoir are both relatively large. Under sustained pumping, insufficient water replenishment occurs. As a result, an asymmetric water level drop zone forms near the inlet/outlet of the YGYJ reservoir (Boundary 2), and reverse flow appears downstream of the inlet/outlet, leading to the formation of the pumping funnel phenomenon. This phenomenon further reduces the cross-sectional flow area, exacerbates the continuous decline in hydraulic replenishment capacity, until reaching the limit point of pumping conditions, ultimately resulting in an unreliable pumping situation. According to the simulation results, the maximum depth of the pumping funnel reaches 5.5–6.0 m, with a maximum horizontal extent of 1.3–1.9 km. However, as a depth-averaged 2D model based on the shallow water equations, MIKE 21 cannot resolve the vertical acceleration, jet entrainment, vertical velocity distribution, or anisotropic turbulent structures near the inlet/outlet during pumping funnel development. Therefore, while computationally efficient for long-duration (8 h) and large-domain (20 km) simulations, its predictions of local funnel characteristics should be interpreted with caution. Capturing the full 3D nature of the funnel would require a 3D hydrodynamic model to resolve vertical momentum transfer and turbulent eddies, albeit at a much higher computational cost. This limitation is considered a valuable direction for further exploration in future research. Secondly, the hydrodynamic numerical simulation results indicate that, under the tested conditions, the riverbed dredging plan using diversion tunnel construction can significantly expand the cross-sectional flow area of the river channel (with an increase rate of 15% ~ 23%) and reduce reverse flow and water level gradient under pumping conditions (with a reduction rate of 30% ~ 42%). These changes effectively suppress the development of the pumping funnel (with the maximum depth reduction to 65% of the simulated value). Consequently, the continuous full-power pumping duration exceeds 8 h under most conditions, and under typical unfavorable conditions, the continuous full-power pumping duration is more than 6.5 h, meeting the design specification standards. Lastly, through the calculation of comprehensive evaluation indicators, Plan 4 is recommended as the appropriate riverbed dredging plan, with its pumping effect evaluation indicator increasing to 1.76 and the comprehensive evaluation indicator reaching 0.94, the highest among all the plans. The evaluation comparison between Plan 4 and Plan 3 shows that the dredging range expansion with an extra 200-meter river reach upstream from the inlet/outlet can slightly improve the pumping hydrodynamic effect, but not significantly. Conversely, it presents relatively low cost-effectiveness because of the higher construction cost. The evaluation comparison between Plan 4 and Plan 6 shows that the dredging range expansion downstream from the inlet/outlet can effectively improve the hydrodynamic effect. However, the construction cost of the dredging range expansion increases significantly, and its marginal cost-effectiveness quickly decreases. The evaluation comparison between Plan 4 and Plan 5 shows that permanent retention of the flow-through function of the diversion tunnel can only slightly improve the pumping hydrodynamic effect, still not significant. The improvement effect is offset by the increased construction cost, making it less meaningful to retain the flow-through function. Physical model validation Through the comparative analysis of the physical model experiment results and numerical simulation results, the following conclusions are summarized. Through the 1:80 scale full-reservoir physical model experiment, the pumping funnel phenomenon under pumping conditions was reproduced, validating the reliability of the numerical model in terms of the time process of pumping flow and the distribution of water levels along the route. The results of the physical model and numerical simulation show consistent patterns, with the hourly-averaged water-level relative error being less than 8.5% across all tested cases. The maximum relative error of 8.5% occurred in Case 7.The physical model experiments confirmed that diversion tunnel dredging effectively improves pumping reliability, increasing continuous pumping duration by 61%. The model experiment results are more optimistic compared to the numerical simulation, providing stronger support for adopting riverbed dredging measures to enhance pumping reliability. The preliminary analysis indicates that discrepancy is attributed to the model scale limitations, which lead to incomplete topographic similarity and lower fluid viscosity.In the typical unfavorable condition where the canyon-type riverine lower reservoir experiences a small outflow from the upper reservoir and a large MPSPS pumping flow, the pumping process of the MPSPS is influenced by the insufficient hydraulic replenishment capacity of the local river. This leads to the formation of a localized falling funnel near the inlet/outlet of the lower reservoir, commonly referred to the pumping funnel phenomenon.By riverbed dredging near the inlet/outlet section, the hydraulic replenishment conditions can be effectively improved, significantly reducing the depth and extent of the pumping funnel. This helps ensure that the MPSPS can still achieve continuous and long-duration pumping capacity even under unfavorable conditions, thereby greatly enhancing the reliability of the MPSPS for continuous long-period pumping. Through the 1:80 scale full-reservoir physical model experiment, the pumping funnel phenomenon under pumping conditions was reproduced, validating the reliability of the numerical model in terms of the time process of pumping flow and the distribution of water levels along the route. The results of the physical model and numerical simulation show consistent patterns, with the hourly-averaged water-level relative error being less than 8.5% across all tested cases. The maximum relative error of 8.5% occurred in Case 7. Through the 1:80 scale full-reservoir physical model experiment, the pumping funnel phenomenon under pumping conditions was reproduced, validating the reliability of the numerical model in terms of the time process of pumping flow and the distribution of water levels along the route. The results of the physical model and numerical simulation show consistent patterns, with the hourly-averaged water-level relative error being less than 8.5% across all tested cases. The maximum relative error of 8.5% occurred in Case 7. The physical model experiments confirmed that diversion tunnel dredging effectively improves pumping reliability, increasing continuous pumping duration by 61%. The model experiment results are more optimistic compared to the numerical simulation, providing stronger support for adopting riverbed dredging measures to enhance pumping reliability. The preliminary analysis indicates that discrepancy is attributed to the model scale limitations, which lead to incomplete topographic similarity and lower fluid viscosity. The physical model experiments confirmed that diversion tunnel dredging effectively improves pumping reliability, increasing continuous pumping duration by 61%. The model experiment results are more optimistic compared to the numerical simulation, providing stronger support for adopting riverbed dredging measures to enhance pumping reliability. The preliminary analysis indicates that discrepancy is attributed to the model scale limitations, which lead to incomplete topographic similarity and lower fluid viscosity. In the typical unfavorable condition where the canyon-type riverine lower reservoir experiences a small outflow from the upper reservoir and a large MPSPS pumping flow, the pumping process of the MPSPS is influenced by the insufficient hydraulic replenishment capacity of the local river. This leads to the formation of a localized falling funnel near the inlet/outlet of the lower reservoir, commonly referred to the pumping funnel phenomenon. In the typical unfavorable condition where the canyon-type riverine lower reservoir experiences a small outflow from the upper reservoir and a large MPSPS pumping flow, the pumping process of the MPSPS is influenced by the insufficient hydraulic replenishment capacity of the local river. This leads to the formation of a localized falling funnel near the inlet/outlet of the lower reservoir, commonly referred to the pumping funnel phenomenon. By riverbed dredging near the inlet/outlet section, the hydraulic replenishment conditions can be effectively improved, significantly reducing the depth and extent of the pumping funnel. This helps ensure that the MPSPS can still achieve continuous and long-duration pumping capacity even under unfavorable conditions, thereby greatly enhancing the reliability of the MPSPS for continuous long-period pumping. By riverbed dredging near the inlet/outlet section, the hydraulic replenishment conditions can be effectively improved, significantly reducing the depth and extent of the pumping funnel. This helps ensure that the MPSPS can still achieve continuous and long-duration pumping capacity even under unfavorable conditions, thereby greatly enhancing the reliability of the MPSPS for continuous long-period pumping. Conclusions This study addresses the hydrodynamic phenomenon termed the “pumping funnel” and the associated pumping reliability issue of the MPSPS. Taking the LHK MPSPS on the Yalong River as a case study, the research systematically investigates the mechanism on the pumping reliability of riverbed dredging measures, through the collaborative analysis of hydrodynamic numerical simulations and full-reservoir physical model experiments. Additionally, a dredging plan optimization method grounded in hydrodynamic analysis is developed and evaluated. This study has established a comprehensive “simulation–experiment–verification” analysis framework for assessing the pumping reliability of MPSPS, addressing the hydrodynamic complexities associated with pumping operations in canyon-type riverine lower reservoirs. The research outcomes have been directly integrated into the engineering design for riverbed dredging at the LHK MPSPS. Both hydrodynamic simulations and physical model experiments confirm that the recommended dredging plan satisfies the design specifications for pumping reliability, with the maximum relative error between the two methods being less than 8.5% across all tested cases. Within the context of the Lianghekou project, this study offers practical insights and methodological references for the design and research of future large-scale MPSPS projects. However, it should be noted that the findings are derived from a specific case study; generalization to other projects with different hydraulic and topographic conditions should be made with caution. Future work could extend the proposed framework to other MPSPS sites and further explore the three-dimensional nature of the pumping funnel using 3D hydrodynamic models to capture vertical momentum transfer and turbulent structures near the inlet/outlet.