Water resource accurate scheduling method and system based on ecological system service flow

Abstract

The invention provides a water resource accurate scheduling method and system based on ecological system service flow, and the system comprises a distributed hydrological-ecological coupling modeling module, a self-adaptive multi-objective optimization decision module, and a dynamic feedback and rolling decision module. The modeling module is used for performing refined modeling on accumulated snow ablation, frozen soil hydrothermal coupling and glacier melt water based on an energy balance principle; establishing a three-dimensional river connectivity index model, a wetland hydrological suitability model and a vegetation water stress index model, and realizing dynamic quantification of an ecological system service function; an improved NSGA-III algorithm is adopted to generate a Pareto leading edge scheduling scheme solution set; rolling updating and real-time correction are realized through data assimilation, scene prediction, scheme evaluation and instruction generation; according to the technical scheme, the key hydrological process can be accurately described, dynamic quantification of ecological system service is realized, and the robustness and adaptability of a scheduling system are improved.

Description

Technical Field

[0001] This application relates to the field of water resource management and environmental engineering technology, and in particular to a method and system for precise water resource scheduling based on ecosystem service flows. Background Technology

[0002] As the core region of the global water tower, high-altitude mountainous areas not only safeguard the water supply security of hundreds of millions of people downstream, but also support the ecological patterns upon which plateau wetlands, valley vegetation, and rare species depend for survival. With intensifying climate change and increased human disturbance, hydrological processes in these regions exhibit significant non-steadiness and high uncertainty. Against this backdrop, how to simultaneously enhance ecosystem services such as water conservation, biodiversity maintenance, and water purification while ensuring socio-economic goals like water supply, flood control, and power generation has become a key issue in integrated watershed management. Although existing scheduling systems have incorporated information technology, they still reveal deep-seated structural deficiencies when dealing with the complex hydrological-ecological coupling mechanisms of high-altitude mountainous areas.

[0003] Currently, most mainstream scheduling models are based on lumped or semi-distributed hydrological simulation frameworks, which struggle to accurately depict key physical processes unique to high-altitude mountainous areas, such as snowmelt, permafrost heat and water exchange, and glacial meltwater. Snowmelt is often simplified to the temperature index method, neglecting the multiple regulatory effects of dynamic feedback from radiation, wind speed, and albedo; the freezing-thawing phase transition of the active permafrost layer and its nonlinear impact on water migration are often roughly characterized by empirical functions; and glacial meltwater generally relies on static retreat rate extrapolation, failing to consider the interactive effects of air temperature, precipitation phase, and ice surface dust cover. Such simplified modeling leads to systematic biases in runoff process simulations, especially during the spring snowmelt season and the summer glacial meltwater peak season, resulting in inaccurate predictions of runoff generation and concentration timing and magnitude, severely weakening the scientific basis for scheduling decisions.

[0004] Correspondingly, in terms of quantifying ecosystem services, traditional methods often employ single indicators or static habitat suitability curves, failing to establish a dynamic mapping relationship between hydrological conditions and ecological responses. River connectivity assessments are often limited to cross-sectional continuity judgments, neglecting the comprehensive regulation of fish migration and material transport by three-dimensional hydraulic connectivity (vertical, horizontal, and vertical). Wetland hydrological suitability models lack sensitivity analysis of key variables such as water level fluctuation frequency, inundation duration, and wet-dry cycle. Vegetation water stress assessments rely heavily on soil moisture content thresholds, failing to integrate remote sensing-derived vegetation physiological states and root zone water dynamics. Furthermore, existing optimization algorithms often employ fixed crossover and mutation probability strategies, which are prone to local convergence or uneven solution distribution when facing the inherent strong nonlinearity, high dimensionality, and dynamic constraints of scheduling problems in high-altitude and cold mountainous areas. Moreover, scheduling frameworks generally lack closed-loop interaction capabilities with the real system, failing to assimilate multimodal real-time sensing data and continuously revise schemes, resulting in a persistent disconnect between model outputs and actual system states. The aforementioned problems collectively make it difficult for the existing scheduling system to achieve the synergistic maximization of socio-economic and ecological benefits. There is an urgent need to build a new paradigm for precise water resource scheduling that deeply integrates physical mechanisms, dynamically quantifies ecological services, and has adaptive optimization and real-time feedback capabilities. Summary of the Invention

[0005] This application provides a precise water resource scheduling method and system based on ecosystem service flow, which solves the structural defects of traditional scheduling systems caused by the non-steady state and high uncertainty of hydrological processes in high-altitude and cold mountainous areas. It can accurately characterize key hydrological processes, realize dynamic quantification of ecosystem services, and improve the robustness and adaptability of the scheduling system.

[0006] To achieve the above objectives, the technical solution of this application embodiment is as follows:

[0007] In a first aspect, embodiments of this application provide a precise water resource scheduling system based on ecosystem service flows. The system includes: a distributed hydrological-ecological coupling modeling module, an adaptive multi-objective optimization decision-making module, and a dynamic feedback and rolling decision-making module; the distributed hydrological-ecological coupling modeling module is connected to the dynamic feedback and rolling decision-making module through the adaptive multi-objective optimization decision-making module.

[0008] The distributed hydrology-ecology coupled modeling module is used to construct snow ablation models, permafrost active layer hydrothermal coupled models, and glacial meltwater models in high-altitude cold mountainous areas based on the principles of energy balance and physical mechanisms. Simultaneously, it constructs three-dimensional river connectivity index models, wetland hydrological suitability models, and vegetation water stress index models, and outputs the predicted runoff process line, corresponding predicted sequences of river connectivity index, wetland suitability index, and vegetation stress index within the future speed regulation cycle, thereby realizing the dynamic quantification of ecosystem service functions.

[0009] The adaptive multi-objective optimization decision-making module is used to generate a Pareto front scheduling scheme solution set based on the predicted runoff process line and the corresponding predicted sequences of river connectivity index, wetland suitability index, and vegetation stress index within the future speed regulation cycle, using an improved NSGA-III algorithm. This is achieved by simultaneously optimizing the river connectivity index, wetland suitability index, and vegetation stress index while satisfying the objective constraints. The objective constraints are at least one of the following: maximum safe storage capacity of the reservoir group, minimum safe storage capacity of the reservoirs, downstream flood control safety flow, and minimum demand for water supply and power generation. Each solution in the Pareto front scheduling scheme solution set corresponds to a set of 3-hourly discharge flow plans for each reservoir over the next 72 hours.

[0010] The dynamic feedback and rolling decision module receives real-time monitoring data from the IoT sensing network, updates the initial field of the model through the data assimilation unit, and drives the scenario prediction unit to generate hydrological-ecological prediction sequences for future periods. The scheme evaluation and optimization unit, based on a rolling window mechanism and an objective function, determines the function value of each Pareto front scheduling scheme solution in the solution set, and selects the Pareto front scheduling scheme solution with the highest function value as the optimal solution. Based on the optimal solution, a discharge flow command is generated, which is then converted into a gate control signal by the command generation unit to control the discharge flow of each reservoir, forming a closed-loop feedback. The objective function is determined based on a weighted sum of standardized water supply, power generation, flood control, and ecological objectives, with the weight value of each objective set based on social environmental preferences.

[0011] In one possible implementation, the snow ablation model for high-altitude and cold mountainous areas uses a 10-meter resolution digital elevation model as the basic grid unit. The net radiation flux received by each unit is calculated hourly. Shortwave radiation is determined by the solar altitude angle, atmospheric transparency, and surface albedo. Longwave radiation is calculated by the Stefan-Boltzmann law combined with a cloud cover correction factor. Sensible heat and latent heat fluxes are solved by wind speed, air temperature, humidity, and surface roughness parameters based on the Moning-Obukhov similarity theory. The snow surface temperature is solved iteratively by the heat conduction equation. When the surface temperature reaches 0°C and there is a positive net energy input, a phase change process is initiated. The amount of melted water is determined by dividing the energy surplus by the latent heat constant. The surface albedo is dynamically updated with the degree of snow aging, and its decay rate is jointly driven by snow density, liquid water content, and dust deposition flux.

[0012] In one possible implementation, the hydrothermal coupling model of the active layer of permafrost is a one-dimensional vertical unsteady-state thermal-water coupling model. The model divides the soil profile into computational layers with a thickness of 0.05 meters, and each layer independently tracks the evolution of the temperature field and water content field. The latent heat of phase change is introduced into the heat conduction equation, and the ice-water phase change threshold is set in the range of -0.5℃ to 0.5℃. The effective heat capacity method is used to handle the nonlinear thermophysical property changes during the phase change process. The water migration equation adopts the Richards equation, considering the nonlinear relationship between the unfrozen water content and the matrix potential under freezing conditions. Its parameters are determined by the van Genuchten-Mualem model combined with the freezing correction function. The upper boundary condition is dynamically switched by the snow cover state. When there is no snow cover, measured meteorological forcing is used, and when there is snow cover, the heat flux and water flux are coupled and transferred through the snow bottom layer. The lower boundary is set to a zero flux condition. The model time step is 1 hour, and it is solved by an implicit finite difference scheme.

[0013] In one possible implementation, the adaptive multi-objective optimization decision module employs an improved NSGA-III algorithm with a population size of 300. Reference points are uniformly generated in the five-dimensional objective space using the Das-Dennis method. The crossover operation uses simulated binary crossover with a fixed distribution exponent of 20, and the mutation operation uses polynomial mutation with a fixed distribution exponent of 20. The crossover probability P... c With the probability of mutation P m Based on the current algebra t and the maximum algebra T max Dynamic adjustment: , In addition, a constraint violation feedback mechanism is introduced: if the optimal solution violates the maximum safe storage capacity of the reservoir group or the downstream flood control safety flow for 5 consecutive generations, the mutation probability P is temporarily increased. m Up to 0.6; when the solution set of the Pareto front scheduling scheme is unevenly distributed in the direction of the reference point, the crossover probability of individuals in the sparse direction is locally increased, and the algorithm terminates when the hypervolume index changes by less than 10 after 500 generations or for 100 consecutive generations. -4 .

[0014] In one possible implementation, the real-time monitoring data includes air temperature, precipitation, wind speed, humidity, and radiation provided by automatic weather stations; snow depth, snow water equivalent, and snow layer temperature profiles provided by snow monitoring stations; ground temperature and soil moisture content at different depths provided by permafrost monitoring wells; flow rate, water level, and water temperature provided by river hydrological stations; and high-resolution images of land cover and snow cover range acquired by regularly flying UAV remote sensing platforms.

[0015] The data assimilation unit is used to assimilate the real-time monitoring data into the state variables of the distributed hydrological-ecological coupling modeling module after timestamp alignment and quality control, using an ensemble Kalman filter algorithm. The ensemble size is 50, and the localization radius of the covariance is set to 15 kilometers.

[0016] In one possible implementation, the ensemble Kalman filter algorithm compares the real-time observation data with the simulated observations corresponding to all models during the assimilation process, calculates the Kalman gain, and then adjusts the state variables of all model members in reverse to make the model state approximate the real ecosystem state to the greatest extent possible; the covariance localization radius is set to 15 kilometers to suppress long-range spurious correlations caused by finite set samples.

[0017] In one possible implementation, the scenario prediction unit drives the model in the distributed hydrological-ecological coupled modeling module and connects to a high-resolution numerical weather prediction product as a meteorological boundary forcing for the next 72 hours. The high-resolution numerical weather prediction product has a temporal resolution of 3 hours and generates a detailed prediction sequence every 3 hours for the next 72 hours. The detailed prediction sequence includes the runoff process at the watershed outlet and each control section, the spatiotemporal changes of the river's comprehensive connectivity index, the evolution of the water level and suitability index of major wetlands, and the spatial distribution map of the vegetation stress index.

[0018] In one possible implementation, the scheme evaluation and optimization unit divides the 72-hour prediction period into 24 3-hour intervals, generates candidate discharge flow schemes for each interval, and adopts a rolling window mechanism in the evaluation process: only the schemes of the first 3 hours are finally selected, and the subsequent schemes are reserved as alternatives; and, under the condition that the target constraints are met and the downstream cross-section ecological flow is not lower than the dynamic threshold, the function value of each Pareto front scheduling scheme solution in the Pareto front scheduling scheme solution set is determined by the objective function; the dynamic threshold is dynamically adjusted based on the season and the fish migration period.

[0019] In one possible implementation, the instruction generation unit incorporates a reservoir gate hydraulic characteristic model. This model converts the discharge flow instruction into control signals for specific gates. The control signals include opening height, number of gates to be opened, and opening duration. These control signals are then transmitted to the gate monitoring systems at each reservoir via an industrial control network, thereby driving the actuators of the gate monitoring systems to precisely control the discharge flow. Furthermore, the new discharge flow acts on the real river system, and its effect is captured by the Internet of Things (IoT) sensing network as new real-time monitoring data, thus obtaining closed-loop feedback.

[0020] Secondly, embodiments of this application provide a method for precise water resource scheduling based on ecosystem service flows. The method is applied to a precise water resource scheduling system based on ecosystem service flows as described in the first aspect. The method includes:

[0021] Based on the principles of energy balance and physical mechanisms, a distributed hydro-ecological coupling modeling module is used to construct a snow ablation model, a hydrothermal coupling model of the active layer of permafrost, and a glacial meltwater model for high-altitude and cold mountainous areas. Simultaneously, a three-dimensional river connectivity index model, a wetland hydrological suitability model, and a vegetation water stress index model are constructed. The model outputs the predicted runoff process line, the corresponding predicted sequences of river connectivity index, wetland suitability index, and vegetation stress index within the future speed regulation cycle, thereby realizing the dynamic quantification of ecosystem service functions.

[0022] Based on the predicted runoff process curve and the corresponding predicted sequences of river connectivity index, wetland suitability index, and vegetation stress index within the future speed regulation cycle, an adaptive multi-objective optimization decision module employs an improved NSGA-III algorithm to collaboratively optimize the river connectivity index, wetland suitability index, and vegetation stress index while satisfying objective constraints, generating a Pareto front scheduling solution set. The objective constraints are at least one of the following: maximum safe operating capacity of the reservoir group, minimum safe operating capacity of the reservoirs, downstream flood control safety flow, and minimum demand for ensuring water supply and power generation. Each solution in the Pareto front scheduling solution set corresponds to a set of 3-hourly discharge flow schemes for each reservoir within the next 72 hours.

[0023] Through a dynamic feedback and rolling decision-making module, real-time monitoring data from the IoT sensing network is received. The data assimilation unit updates the initial field of the model, driving the scenario prediction unit to generate hydrological-ecological prediction sequences for future periods. The scheme evaluation and optimization unit determines the function value of each Pareto front scheduling scheme solution in the Pareto front scheduling scheme solution set based on a rolling window mechanism and an objective function, and selects the Pareto front scheduling scheme solution with the highest function value as the optimal solution. Based on the optimal solution, a discharge flow command is generated, and the discharge flow command is converted into a gate control signal by the command generation unit to control the discharge flow of each reservoir, forming a closed-loop feedback. The objective function is determined based on the weighted sum of standardized water supply, power generation, flood control, and ecological objectives, with the weight value of each objective set based on social environmental preferences.

[0024] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0025] (1) By constructing a refined distributed hydrological model based on the principle of energy balance, the key physical processes such as snow melting, permafrost heat exchange and glacial meltwater in high-altitude and cold mountainous areas were accurately characterized, overcoming the systematic bias in runoff prediction caused by traditional simplified modeling, and providing a solid scientific basis for scheduling decisions.

[0026] (2) By establishing a dynamic response three-dimensional river connectivity index model, wetland hydrological suitability model and vegetation water stress index model, the transformation of ecosystem service function from static assessment to dynamic quantification has been realized, and a precise mapping relationship between hydrological situation and ecological response has been established, making ecological goals quantifiable and optimizable.

[0027] (3) By introducing an adaptive crossover and mutation probability mechanism, the improved NSGA-III algorithm, and by integrating constraint violation feedback and solution set distribution equilibrium control strategy, significantly improves the global search capability and convergence efficiency of the algorithm in dealing with the inherent strong nonlinearity, high dimension and dynamic constraints of scheduling problems in high-altitude and cold mountainous areas, and avoids getting trapped in local optima.

[0028] (4) By constructing a dynamic feedback and rolling decision architecture consisting of data assimilation, scenario prediction, scheme evaluation and instruction generation units, the system achieves efficient assimilation of multimodal real-time perception data and models, accurate prediction of future scenarios, and rolling generation and closed-loop correction of scheduling instructions, ensuring real-time matching between scheduling schemes and actual system states, and greatly improving the robustness and adaptability of the scheduling system. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments of this application or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 A block diagram of a precise water resource scheduling system based on ecosystem service flows provided in this application embodiment;

[0031] Figure 2 A flowchart illustrating the workflow of a distributed hydrological-ecological coupling modeling module provided in this application embodiment;

[0032] Figure 3 A flowchart illustrating the workflow of an adaptive multi-objective optimization decision-making module provided in this application embodiment;

[0033] Figure 4 A flowchart illustrating the dynamic feedback and rolling decision-making module provided in this application embodiment;

[0034] Figure 5 A flowchart illustrating a precise water resource scheduling method based on ecosystem service flows, provided as an embodiment of this application. Detailed Implementation

[0035] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0036] In the description of the embodiments of this application, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the embodiments of this application and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. The terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, the terms "installed," "connected," and "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; a mechanical connection or an electrical connection; a direct connection or an indirect connection through an intermediate medium; or a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in the embodiments of this application according to the specific circumstances.

[0037] Figure 1 This application provides a block diagram of a precise water resource scheduling system based on ecosystem service flows. As shown in the figure, the system 100 may include: a distributed hydrological-ecological coupling modeling module 110, an adaptive multi-objective optimization decision-making module 120, and a dynamic feedback and rolling decision-making module 130; the distributed hydrological-ecological coupling modeling module 110 is connected to the dynamic feedback and rolling decision-making module 130 through the adaptive multi-objective optimization decision-making module 120.

[0038] The distributed hydrological-ecological coupling modeling module 110 is used to construct a snow ablation model, a hydrothermal coupling model of the active layer of permafrost, and a glacial meltwater model in high-altitude and cold mountainous areas based on the principles of energy balance and physical mechanisms. Simultaneously, it constructs a three-dimensional river connectivity index model, a wetland hydrological suitability model, and a vegetation water stress index model, and outputs the predicted runoff process line, the corresponding predicted sequences of river connectivity index, wetland suitability index, and vegetation stress index within the future speed regulation cycle, thereby realizing the dynamic quantification of ecosystem service functions.

[0039] The adaptive multi-objective optimization decision module 120 is used to generate a Pareto front scheduling scheme solution set based on the predicted runoff process line, the corresponding river connectivity index, wetland suitability index, and vegetation stress index within the future speed regulation cycle. It adopts an improved NSGA-III algorithm to coordinately optimize the river connectivity index, wetland suitability index, and vegetation stress index while satisfying the objective constraints. The objective constraints are at least one of the following: the maximum safe storage capacity of the reservoir group, the minimum safe storage capacity of the reservoirs, the downstream flood control safety flow, and the minimum demand for ensuring water supply and power generation. Each solution in the Pareto front scheduling scheme solution set corresponds to a set of outflow plans for each reservoir in 3-hour increments over the next 72 hours.

[0040] The dynamic feedback and rolling decision module 130 is used to receive real-time monitoring data from the Internet of Things sensing network, update the initial field of the model through the data assimilation unit, drive the scenario prediction unit to generate hydrological-ecological prediction sequences for future periods, and the scheme evaluation and optimization unit determines the function value of each Pareto front scheduling scheme solution in the Pareto front scheduling scheme solution set based on the rolling window mechanism and the objective function, and takes the Pareto front scheduling scheme solution with the highest function value as the optimal solution; and generates a discharge flow command based on the optimal solution, and converts the discharge flow command into a gate control signal through the command generation unit to control the discharge flow of each reservoir, forming a closed-loop feedback; the objective function is determined based on the weighted sum of the standardized water supply target, power generation target, flood control target and ecological target, and the weight value of each target is set based on social environmental preferences.

[0041] According to the above scheme, by constructing a variety of refined distributed hydrological models, the key physical processes in high-altitude and cold mountainous areas were accurately characterized; a dynamic response three-dimensional river connectivity index model, wetland hydrological suitability model, and vegetation water stress index model were established, realizing the transformation of ecosystem service functions from static assessment to dynamic quantification; combined with the improved NSGA-III algorithm and the integration of constraint violation feedback and solution set distribution equilibrium control strategies, the algorithm's global search capability and convergence efficiency were significantly improved; and efficient assimilation of multimodal real-time sensing data and models, accurate prediction of future scenarios, and rolling generation and closed-loop correction of scheduling instructions were achieved, ensuring real-time matching between the scheduling scheme and the actual ecosystem state.

[0042] In one possible implementation, the snow ablation model for high-altitude and cold mountainous areas uses a 10-meter resolution digital elevation model as the basic grid unit. The net radiation flux received by each unit is calculated hourly. Shortwave radiation is determined by the solar altitude angle, atmospheric transparency, and surface albedo. Longwave radiation is calculated by the Stefan-Boltzmann law combined with a cloud cover correction factor. Sensible heat and latent heat fluxes are solved by wind speed, air temperature, humidity, and surface roughness parameters based on the Moning-Obukhov similarity theory. The snow surface temperature is solved iteratively by the heat conduction equation. When the surface temperature reaches 0°C and there is a positive net energy input, the phase change process is initiated. The amount of melted water is determined by dividing the energy surplus by the latent heat constant. The surface albedo is dynamically updated with the degree of snow aging, and its decay rate is jointly driven by snow density, liquid water content, and dust deposition flux.

[0043] In one possible implementation, the hydrothermal coupling model of the active layer of permafrost is a one-dimensional vertical unsteady-state thermo-water coupling model. The model divides the soil profile into computational layers with a thickness of 0.05 meters, and each layer independently tracks the evolution of the temperature field and water content field. The latent heat of phase change is introduced into the heat conduction equation, and the ice-water phase change threshold is set in the range of -0.5℃ to 0.5℃. The effective heat capacity method is used to handle the nonlinear thermophysical property changes during the phase change process. The water migration equation adopts the Richards equation, considering the nonlinear relationship between the unfrozen water content and the matrix potential under freezing conditions. Its parameters are determined by the van Genuchten-Mualem model combined with the freezing correction function. The upper boundary condition is dynamically switched by the snow cover state. When there is no snow cover, the measured meteorological forcing is used, and when there is snow cover, the heat flux and water flux are coupled and transferred through the snow bottom layer. The lower boundary is set to a zero flux condition. The model time step is 1 hour, and it is solved by an implicit finite difference scheme.

[0044] In one possible implementation, the glacial meltwater model is a distributed model based on the coupling of energy and mass balance. The model divides the glacier surface into elevation zones, and within each zone, net radiation is calculated based on remotely sensed ice surface albedo, dust optical thickness, and particle size distribution. The turbulent flux calculation method is consistent with the snow accumulation model. Glacier surface temperature is controlled by the heat conduction equation, and the phase transition interface location is solved using the Stefan problem. After meltwater is generated, it flows into river channels via an ice surface runoff network. The runoff path is determined by the flow direction matrix extracted from the digital elevation model, the runoff velocity is calculated using the Manning formula, and the roughness coefficient is assigned regionally based on the remote sensing results of ice surface roughness. The interannual variation of glacial mass balance is driven by long-term meteorological sequences, and the glacier tongue terminus position is dynamically updated in conjunction with the glacier geometry.

[0045] In one possible implementation, the three-dimensional river connectivity index model comprehensively assesses river continuity from three dimensions: longitudinal, lateral, and vertical. Longitudinal connectivity is assessed by calculating the resistance to fish swimming upstream based on riverbed slope, flow velocity, and obstacle location, using a resistance function based on swimming ability thresholds. Lateral connectivity is assessed by evaluating the riverbank material exchange capacity through floodplain inundation frequency and duration, determined using water level-area curves combined with historical flood event statistics. Vertical connectivity is assessed by calculating the vertical exchange flux through riverbed permeability coefficient and groundwater level difference, with the permeability coefficient obtained through joint inversion of field borehole tests and geophysical exploration. The three-dimensional indices are weighted and fused to generate a comprehensive connectivity index, with the weights determined by expert scoring combined with principal component analysis.

[0046] In one possible implementation, the wetland hydrological suitability model uses the core area of ​​the wetland as the calculation unit, extracts key characteristic variables of the water level time series, including annual average water level, water level fluctuation, number of days with high water level, number of days with low water level, wet-dry cycle, and frequency of extreme low water level occurrence; establishes response surfaces of each characteristic variable with typical wetland plant cover, biomass, and seed germination rate through long-term ecological monitoring data, and uses a Gaussian kernel function to fit the nonlinear relationship; after normalizing each response surface, a weighted composite wetland hydrological suitability index is synthesized, with the weights determined according to the species importance value.

[0047] In one possible implementation, the vegetation water stress index model integrates remote sensing and ground observation data; it uses a 16-day MODIS NDVI product as a basis and combines Sentinel-2 high-resolution imagery to correct spatial details; it simultaneously acquires root zone soil moisture content profiles, which are monitored in real time by TDR sensors installed at depths of 0.1 meters, 0.3 meters, and 0.6 meters; by establishing a hysteresis regression relationship between NDVI and effective root zone water content, it identifies the water threshold of vegetation physiological response; when the measured water content is lower than the threshold and NDVI shows a decreasing trend, it is determined to be a water stress state; the stress intensity is quantified by the degree of deviation from the threshold and the duration, generating a spatially distributed vegetation water stress index map.

[0048] In one possible implementation, the adaptive multi-objective optimization decision-making module employs an improved NSGA-III algorithm with a population size of 300. Reference points are uniformly generated in the five-dimensional objective space using the Das-Dennis method. The crossover operation uses simulated binary crossover with a fixed distribution exponent of 20, and the mutation operation uses multinomial mutation with a fixed distribution exponent of 20. The crossover probability P... c With the probability of mutation P m Based on the current algebra t and the maximum algebra T max Dynamic adjustment: , In addition, a constraint violation feedback mechanism is introduced: if the optimal solution violates the maximum safe storage capacity of the reservoir group or the downstream flood control safety flow for 5 consecutive generations, the mutation probability P is temporarily increased. m Up to 0.6; when the solution set of the Pareto front scheduling scheme is unevenly distributed in the direction of the reference point, the crossover probability of individuals in the sparse direction is locally increased. The algorithm terminates when the hypervolume index changes by less than 10 after 500 generations or for 100 consecutive generations. -4 .

[0049] In one possible implementation, the real-time monitoring data includes air temperature, precipitation, wind speed, humidity, and radiation provided by automatic weather stations; snow depth, snow water equivalent, and snow layer temperature profiles provided by snow cover monitoring stations; ground temperature and soil moisture content at different depths provided by permafrost monitoring wells; flow rate, water level, and water temperature provided by river hydrological stations; and high-resolution images of land cover and snow cover extent acquired by regularly flying UAV remote sensing platforms. This data assimilation unit, after timestamp alignment and quality control, uses an ensemble Kalman filter algorithm to assimilate the real-time monitoring data into the state variables of the distributed hydro-ecological coupling modeling module. The ensemble size is 50, and the localization radius of the covariance is set to 15 kilometers.

[0050] In one possible implementation, the ensemble Kalman filter algorithm compares the real-time observation data with the simulated observations corresponding to all models during the assimilation process, calculates the Kalman gain, and then adjusts the state variables of all model members in reverse to make the model state approximate the real ecosystem state as closely as possible; the covariance localization radius is set to 15 kilometers to suppress long-range spurious correlations caused by finite set samples.

[0051] In one possible implementation, the scenario prediction unit drives the model in the distributed hydrological-ecological coupled modeling module and connects to a high-resolution numerical weather prediction product as a meteorological boundary forcing for the next 72 hours. The high-resolution numerical weather prediction product has a temporal resolution of 3 hours and generates a detailed prediction sequence every 3 hours for the next 72 hours. The detailed prediction sequence includes the runoff process at the watershed outlet and each control section, the spatiotemporal changes of the river's comprehensive connectivity index, the evolution of the water level and suitability index of major wetlands, and the spatial distribution map of the vegetation stress index.

[0052] In one possible implementation, the scheme evaluation and optimization unit divides the 72-hour forecast period into 24 3-hour intervals, generates candidate discharge flow schemes for each interval, and adopts a rolling window mechanism in the evaluation process: only the schemes of the first 3 hours are finally selected, and the subsequent schemes are reserved as alternatives; and, under the condition that the objective constraint is met and the downstream cross-section ecological flow is not lower than the dynamic threshold, the function value of each Pareto front scheduling scheme solution in the Pareto front scheduling scheme solution set is determined by the objective function; the dynamic threshold is dynamically adjusted based on the season and the fish migration period.

[0053] In one possible implementation, the instruction generation unit incorporates a reservoir gate hydraulic characteristic model. This model converts the discharge flow instruction into a control signal for a specific gate. The control signal includes the opening height, number of gates to be opened, and opening duration. This control signal is then transmitted to the gate monitoring system at each reservoir via an industrial control network, thereby driving the actuator of the gate monitoring system to precisely control the discharge flow. Furthermore, the new discharge flow acts on the real river system, and its effect is captured by the Internet of Things (IoT) sensing network as new real-time monitoring data, thus obtaining closed-loop feedback.

[0054] The workflow of the technical solution of this application will be described below with reference to an embodiment. The water resource precision scheduling system based on ecosystem service flow proposed in this application is deployed in the joint scheduling center of a large watershed in a high-altitude cold mountainous area. As the water source guarantee for hundreds of millions of people and important ecological functional areas downstream, the hydrological process of this watershed is jointly dominated by snowmelt, permafrost water and heat exchange and glacial meltwater, exhibiting high spatiotemporal heterogeneity and dynamic uncertainty.

[0055] Reference Figure 1 The system 100 consists of three core functional modules: a distributed hydrological-ecological coupled modeling module 110, an adaptive multi-objective optimization decision-making module 120, and a dynamic feedback and rolling decision-making module 130. These three modules are connected sequentially to form a complete technical chain from physical process simulation and multi-objective scheme optimization to real-time closed-loop control. The data flow begins with a detailed description of the watershed underlying surface and atmospheric forcing, which is then transformed into hydrological and ecological state variables through the coupled model. These variables then serve as inputs for optimization decisions, and finally, an executable gate control command is generated through a rolling mechanism. The system continuously receives feedback from the Internet of Things (IoT) sensing network to achieve self-correction.

[0056] The distributed hydrological-ecological coupling modeling module is the core of this application. Its task is to perform millimeter-level precision distributed simulations of key hydrophysical processes in high-altitude and cold mountainous areas, and simultaneously realize the dynamic quantification of ecosystem service functions. The workflow of this distributed hydrological-ecological coupling modeling module can be referred to... Figure 2As shown. First, based on a 10-meter resolution digital elevation model of the watershed, the entire study area was divided into millions of regular grid cells, each of which independently performed surface energy and moisture balance calculations. For the snowmelt process, the model calculated the net radiation flux cell by cell with a time step of 1 hour. The shortwave radiation incident amount was jointly determined by the solar altitude angle, atmospheric transparency parameters, and the surface albedo of the cell. The surface albedo is a dynamic variable, with its initial value obtained from remote sensing inversion, and then decayed according to the snow aging model. The decay rate was jointly driven by the simulated snow density, liquid water content, and dust deposition flux estimated from aerosol monitoring data for that cell. The calculation of longwave radiation strictly followed the Stefan-Boltzmann law and introduced an emissivity coefficient corrected by cloud cover observation data. The solutions for sensible heat flux and latent heat flux were based on the Moning-Obukhov similarity theory. The input variables included wind speed, air temperature, specific humidity, and surface dynamic roughness determined by looking up a table based on the surface cover type. At each time step, the model iteratively calculates the snow surface temperature by solving the heat conduction equation. When the surface temperature reaches 0 degrees Celsius and there is a positive net energy input to the cell, the phase change process is considered to have started, and the amount of meltwater is determined by dividing the energy surplus in that time step by the latent heat of melting of water. Meltwater generated by all cells flows into a distributed runoff network according to the flow direction extracted by the digital elevation model.

[0057] A one-dimensional vertical unsteady-state thermo-water coupled model was constructed to investigate the hydrothermal transport process in the active layer of permafrost. This model divides the soil profile of each grid cell into several computational layers with a thickness of 0.05 meters from top to bottom, and each layer independently tracks the spatiotemporal evolution of its temperature and volumetric water content. A latent heat source term for the ice-water phase transition was introduced into the heat conduction governing equations. The phase transition threshold was set in the temperature range of -0.5°C to 0.5°C, and the effective heat capacity method was used to handle the nonlinear specific heat capacity changes caused by the phase transition within this range. The water migration process was described by the Richards equation, which underwent a key modification under freezing conditions. Specifically, the relationship between soil matric potential and water content was described by the van Genuchten-Mualem model, but its parameters are functions of temperature. A freezing correction function was used to characterize the characteristic that the unfrozen water content decreases sharply with decreasing temperature. The upper boundary condition of the one-dimensional vertical unsteady-state thermo-hydro coupled model is dynamically switched according to the output of the snow cover module: when the element is covered by snow, the upper boundary receives heat and moisture fluxes from the snow cover layer; when there is no snow cover, the near-surface air temperature and precipitation provided by meteorological station observations or reanalysis data are directly used forcing. The lower boundary of the one-dimensional vertical unsteady-state thermo-hydro coupled model is set to a zero flux condition. The coupled equations are discretized using a fully implicit finite difference scheme with a time step of 1 hour, ensuring the stability and accuracy of the numerical solution and simulating the complex hydro-thermal coupling effects during the freezing and thawing of the active permafrost layer.

[0058] For the meltwater process in glaciers, a distributed model based on the coupling of energy and mass balance was adopted. The distributed model first divides each glacier surface into several elevation zones according to a digital elevation model. Within each elevation zone, the net radiation received by the glacier surface is calculated based on high-resolution ice surface albedo products, dust optical thickness, and particle size distribution data retrieved from remote sensing. The calculation method for turbulent flux is consistent with the aforementioned snow accumulation model. The temperature field inside the glacier is controlled by the heat conduction equation, while the location of the phase transition interface (i.e., the depth at which the ice temperature reaches 0 degrees Celsius) is determined by solving the Stefan problem. The calculated meltwater is considered as the source term forming on the glacier surface, and then enters the ice surface runoff network according to the ice surface flow direction matrix extracted from the high-precision digital elevation model. The runoff velocity is calculated using the Manning formula, where the roughness coefficient is not a fixed value but is spatially assigned based on the ice surface roughness (such as ice fissure density and glacial till coverage) retrieved from remote sensing. In addition, the distributed model updates the glacier's mass balance year by year by driving long-term meteorological sequences, and combines glacier geometric models to dynamically simulate the advance and retreat of the glacier tongue's terminus, thereby reflecting the long-term impact of glacier resource changes on runoff on an interdecadal scale.

[0059] While simulating the aforementioned refined hydrological processes, the hydro-ecological coupled modeling module runs a dynamic ecosystem service quantification model in parallel. The three-dimensional river connectivity index model comprehensively assesses the continuity of the river system from three dimensions: longitudinal, lateral, and vertical. Longitudinal connectivity focuses on the migration routes of aquatic organisms (especially fish). By calculating the slope and velocity distribution along the river channel and combining this with the location of artificial obstacles such as dams and weirs, a resistance function based on the swimming ability threshold of species is constructed to quantify the difficulty of upstream migration. Lateral connectivity aims to assess the material exchange capacity between the river and the floodplain. Its core is to establish a water level-inundation area relationship curve and use historical flood observation data to statistically analyze the inundation frequency and duration at different water levels, thereby characterizing the availability and stability of riparian habitats. Vertical connectivity focuses on the exchange between surface water and groundwater. The spatial distribution of permeability coefficients is obtained through joint inversion of in-situ borehole permeability tests of riverbed sediments and geophysical exploration data. This is then combined with the simulated difference between groundwater level and river level to calculate the vertical exchange flux. Finally, after standardization, the indices of these three dimensions are weighted and fused using a weighting method determined by expert scoring and principal component analysis to generate a comprehensive river connectivity index, which can dynamically respond to changes in river flow.

[0060] The wetland hydrological suitability model uses the core area of ​​each independent wetland patch as the computational unit. The model extracts the simulated or observed water level time series for this unit and calculates six key hydrological characteristic variables: annual average water level, interannual water level variation, number of days with high water levels (above a certain threshold), number of days with low water levels (below a certain threshold), number of wet-dry cycles, and frequency of extreme low water events. By analyzing long-term ecological monitoring data of the wetland (including the cover of typical wetland plants, aboveground biomass, and seed bank germination rate), a nonlinear response relationship between each ecological indicator and the above six hydrological characteristic variables is established. This relationship is fitted using a Gaussian kernel function to form a multidimensional response surface. Subsequently, the response surfaces of each ecological indicator are normalized, and weights are assigned according to the importance of each species in the wetland ecosystem (such as dominance and rarity), and a weighted composite wetland hydrological suitability index is synthesized. This index directly reflects the degree to which the current and predicted hydrological conditions support the integrity of the wetland ecosystem.

[0061] The vegetation water stress index model integrates remote sensing observations and in-situ ground monitoring data. MODIS NDVI products with a 16-day cycle serve as the basic data source for vegetation growth status, while Sentinel-2 imagery with higher spatial resolution is used for spatial detail correction and fusion to improve monitoring accuracy on heterogeneous surfaces. Simultaneously, soil moisture monitoring stations deployed in typical vegetation distribution areas acquire real-time root zone soil moisture profiles using time-domain reflectometer sensors installed at depths of 0.1 meters, 0.3 meters, and 0.6 meters. By analyzing long-term series data, a hysteresis regression relationship is established between NDVI changes and effective root zone water content, thereby identifying the water threshold that triggers vegetation physiological decline. When the simulated or measured root zone water content falls below this threshold, and the synchronous NDVI trend shows a downward trend, the system determines that the vegetation in that area is under water stress. The intensity of stress is quantified by two factors: the magnitude of soil moisture content falling below the threshold and the cumulative duration of falling below the threshold, ultimately generating a vegetation water stress index distribution map with a spatial resolution of 100 meters.

[0062] See Figure 3The adaptive multi-objective optimization decision-making module receives output from the distributed hydrological-ecological coupled modeling module, including the runoff forecast process line for the future scheduling cycle, and the corresponding predicted sequences of river connectivity index, wetland suitability index, and vegetation stress index. The core task of this module is to collaboratively optimize multiple often conflicting objectives while meeting constraints such as the maximum safe storage capacity of the reservoir group, the minimum safe storage capacity of the reservoirs, the downstream flood control safety flow, and the minimum demand for water supply and power generation. These objectives typically include: maximizing water supply benefits, maximizing power generation, minimizing flood control risk, maximizing the comprehensive river connectivity index, maximizing the comprehensive wetland suitability index, and minimizing the vegetation stress index.

[0063] This adaptive multi-objective optimization decision-making module employs an improved NSGA-III algorithm to solve this problem. The algorithm sets the population size to 300. In the five-dimensional objective space, a set of reference points is uniformly generated using the Das-Dennis method to guide the population towards a uniform distribution along the Pareto front. The algorithm's evolutionary operations include crossover and mutation. The crossover operation uses simulated binary crossover with a distribution exponent set to 20 to control the similarity between offspring and parents. The mutation operation uses multinomial mutation, also with a distribution exponent set to 20. The core improvement of this application lies in the introduction of an adaptive probability mechanism and a feedback control strategy. The crossover probability P... c With the probability of mutation P m It is not fixed, but rather depends on the current evolutionary generation t and the preset maximum evolutionary generation T. max Dynamic adjustment, the adjustment formula is: P c = 0.9 - 0.4 × (t / T max ), P m = 0.1 + 0.4 ×(t / T maxThis means that in the early stages of evolution, the algorithm tends to have a higher crossover probability to promote global exploration; while in the later stages, it increases the mutation probability to enhance local fine-grained search. Furthermore, the algorithm incorporates a constraint violation feedback mechanism: the system continuously monitors whether the optimal solution in the population violates key hard constraints in the target constraints (such as the maximum safe storage capacity of a reservoir group and the lower limit of downstream flood control safe flow). If the optimal solution violates the same constraint for five consecutive generations, it is determined that the algorithm may be stuck in a local stagnation near the infeasibility region. At this time, the mutation probability Pm is temporarily increased to 0.6 to generate greater perturbation and help the population escape the current region. Simultaneously, the algorithm also monitors the uniformity of the solution set distribution along the reference point direction. If it finds that the solutions in certain target directions are too sparse, it temporarily increases the crossover probability of individuals located near that sparse direction to encourage searching in that direction, thereby ensuring the diversity and breadth of the final Pareto front scheduling scheme solution set. The algorithm's termination condition is set at 500 generations of evolution, or the change in the hypervolume index of the Pareto front corresponding to 100 consecutive generations of the population is less than 10. -4 After optimization, the module outputs a Pareto front solution set containing hundreds of non-dominated scheduling schemes. Each solution corresponds to a set of 3-hourly discharge flow schemes for each reservoir over the next 72 hours.

[0064] Dynamic feedback and rolling decision-making modules are key to achieving "precision" and "adaptability," forming a closed loop from perception to decision-making to execution. (See also...) Figure 4 First, the data assimilation unit receives real-time monitoring data from the IoT sensing network within the watershed. This real-time monitoring data can include: temperature, precipitation, wind speed, humidity, and radiation provided by automatic weather stations; snow depth, snow water equivalent, and snow layer temperature profiles provided by snow cover monitoring stations; ground temperature and soil moisture content at different depths provided by permafrost monitoring wells; flow rate, water level, and water temperature provided by river hydrological stations; and high-resolution images of land cover and snow cover extent acquired by regularly flying UAV remote sensing platforms. All real-time monitoring data undergoes immediate timestamp alignment, unit standardization, and strict quality control upon access, removing obvious outliers and filling in reasonable missing values. The processed real-time monitoring data is then fed into an ensemble Kalman filter assimilation system. This system uses the aforementioned distributed hydro-ecological coupling model as the forecasting model, maintaining a model state set containing 50 members. Each member operates independently after a small perturbation is added to the initial field. At the assimilation time, the ensemble Kalman filter assimilation algorithm compares the real-time observed data with the simulated observed values ​​corresponding to all model members, calculates the Kalman gain, and then adjusts the state variables of all model members (such as snow water equivalent, soil moisture content, river flow, etc. in each grid) in reverse to make the model state approximate the real system state as closely as possible. The covariance localization radius is set to 15 kilometers to suppress long-range spurious correlations caused by finite set samples.

[0065] After assimilation to obtain an updated initial model field, the scenario prediction unit is activated. This unit drives the model in the hydro-ecological coupled modeling module and simultaneously incorporates high-resolution numerical weather prediction products as meteorological boundary forcing for the next 72 hours. These high-resolution numerical weather prediction products have a temporal resolution of 3 hours, generating detailed prediction sequences every 3 hours for the next 72 hours. These detailed prediction sequences include: runoff processes at the watershed outlet and various control sections, spatiotemporal variations in the river's comprehensive connectivity index, evolution of water levels and suitability indices in major wetlands, and spatial distribution maps of vegetation stress indices. These prediction results constitute a "scenario library" for the next decision-making cycle.

[0066] The scheme evaluation and optimization unit then makes rolling decisions based on this scenario. This unit divides the 72-hour forecast period into 24 decision steps with 3-hour intervals. First, it calls the Pareto scheduling scheme solution set generated by the adaptive multi-objective optimization decision module. The evaluation process employs a rolling window mechanism: the system does not select all scheduling instructions for the next 72 hours at once, but only makes a final decision for the upcoming first 3-hour period (i.e., the first step of the rolling window). For this period, the evaluation unit iterates through the first-step instructions corresponding to all schemes in the Pareto solution set, and performs rapid simulation verification in conjunction with the predicted state for this period provided by the scenario prediction unit. Verification must ensure that after the instructions are executed, the reservoir state still meets all objective constraints, and the flow at the downstream key section is not lower than the dynamic ecological flow threshold (this threshold may vary with seasons, fish migration periods, etc.). Among the candidate instructions that pass verification, the evaluation unit calculates an objective function value. This function is a standardized weighted sum of various objectives such as water supply, power generation, flood control, and ecology. The weights can be set by the dispatcher based on social environmental preferences. Finally, the solution that maximizes the overall objective function value is selected as the optimal solution for the current time period, and the downstream flow is generated. The instructions in steps 2 through 24 are reserved as alternatives, and their final selection will be repeated in the next rolling decision cycle (after 3 hours). This mechanism allows the system to reassess and adjust subsequent scheduling plans every 3 hours using the latest observational data assimilation results and updated weather forecasts, effectively addressing forecast uncertainties and achieving dynamic correction.

[0067] The instruction generation unit receives the discharge flow instruction from the scheme evaluation and optimization unit. This unit incorporates a reservoir gate hydraulic characteristic model, converting the target flow value in the discharge flow instruction into control parameters such as the opening height, number of gates, and opening duration. These parameters, after being verified by safety logic, are distributed to the gate monitoring systems at each reservoir via the industrial control network, driving the actuators to precisely control the discharge flow. Simultaneously, the new discharge flow acts on the real river system, and its effects are captured by the IoT sensing network, forming new observation data and initiating the next cycle of "data assimilation—scenario prediction—scheme evaluation—instruction generation." Thus, the entire system constitutes a closed-loop control system of perception, decision-making, execution, feedback, and re-perception, ensuring that scheduling decisions are always synchronized and matched with the complex and ever-changing hydrological and ecological realities of the high-altitude mountainous region.

[0068] The workflow of this application will be further described below with reference to Example 2. When dealing with sudden extreme weather events (such as accelerated snowmelt due to extreme high temperatures in spring), the snow storage in the watershed is significantly higher than usual. The snowmelt model of the distributed hydrological-ecological coupling modeling module detects that the net radiation flux is abnormally high due to the continuous high temperature and clear sky. The simulation shows that the snowmelt runoff intensity in the next 24 hours may reach more than twice the historical average for the same period. This raises two urgent issues: first, the flood control pressure brought about by the surge in runoff; and second, the scouring and damage that short-term high flow may cause to the habitat structure of downstream river channels.

[0069] The data assimilation unit of the dynamic feedback and rolling decision module was the first to detect the anomaly. Temperature data transmitted from automatic weather stations throughout the basin continued to break historical extremes, and snow depth sensors at snow monitoring stations showed that the snow layer was rapidly thinning. The Kalman filter assimilation system quickly assimilated these real-time observations into the model, significantly correcting the initial snow water equivalent field and energy balance state of each grid cell, making the model forecast closer to the rapidly changing reality. Based on the assimilated state and the updated extreme high-temperature weather forecast, the scenario prediction unit generated a high-resolution forecast for the next 72 hours. The forecast scenario clearly showed that the main river sections would experience flood peaks approaching warning levels in 12 hours. At the same time, the three-dimensional river connectivity index model predicted that due to the excessively fast flow velocity and the rapid rise and fall of water levels, the lateral connectivity index of the river (characterizing floodplain inundation) would improve briefly and then deteriorate sharply, while vertical connectivity might suffer long-term negative impacts due to riverbed erosion.

[0070] The adaptive multi-objective optimization decision module was urgently invoked, but its optimization objective weights and constraints were dynamically adjusted. In the optimization model, the risk weight of the flood control objective was temporarily increased, while the ecological objectives placed greater emphasis on indicators related to maintaining river geomorphological stability. Simultaneously, the downstream ecological flow lower limit constraint was replaced with a stricter "flow change rate" constraint to mitigate flood peaks and prevent severe river erosion. The improved NSGA-III algorithm initiated optimization under an adaptive probabilistic mechanism. Due to the more stringent constraints, the algorithm's built-in constraint violation feedback mechanism was quickly triggered. In the early stages of evolution, a large number of individuals in the population were detected violating the new flow change rate constraint. The algorithm then temporarily increased the mutation probability to 0.6, prompting the population to generate greater mutations and rapidly explore feasible solution spaces that satisfy the new constraints. After accelerated evolution, the algorithm converged in a short time, generating a new set of Pareto scheduling solutions. These schemes share the common feature of pre-releasing reservoir capacity and finely controlling the staggered discharge of multiple reservoirs. While ensuring that the total discharge meets flood control requirements, they greatly smooth the discharge process line, sacrificing some short-term power generation benefits, but effectively reducing the downstream flood peak level and protecting the river channel habitat structure.

[0071] The scheme evaluation and optimization unit employs a rolling window mechanism to evaluate these new schemes. In the first 3-hour decision window, it selects an instruction to "slightly increase discharge to begin storing flood peak capacity." The instruction generation unit converts this instruction into a gate control signal and executes it. Three hours later, new observational data confirms that the snowmelt rate is higher than predicted in the previous period. The data assimilation unit immediately updates the model state, and the scenario prediction unit subsequently generates an updated prediction of a higher and earlier flood peak. Based on this new scenario, the scheme evaluation and optimization unit re-evaluates the remaining schemes in the Pareto solution set and may combine the latest forecasts to fine-tune and optimize the model parameters, performing rapid optimization again to select an instruction to "further increase discharge to free up more storage capacity" for the next 3-hour window. Through this rapid closed-loop iteration every 3 hours, the system can dynamically track the development of extreme events, like "autopilot," and continuously adjust the scheduling strategy, ultimately achieving the precise scheduling goal of ensuring flood control safety while minimizing ecological impact. This fully demonstrates the robustness and adaptability of the scheduling system provided in this application in the face of high uncertainty.

[0072] Figure 5 This document provides a flowchart of a precise water resource scheduling method based on ecosystem service flows, as illustrated in an embodiment of this application. This method can be applied to a precise water resource scheduling system based on ecosystem service flows, and may include the following steps.

[0073] S501. Based on the principles of energy balance and physical mechanisms, a distributed hydrological-ecological coupling modeling module is used to construct a snow ablation model, a hydrothermal coupling model of the active layer of permafrost, and a glacial meltwater model for high-altitude and cold mountainous areas. Simultaneously, a three-dimensional river connectivity index model, a wetland hydrological suitability model, and a vegetation water stress index model are constructed. The model outputs the predicted runoff process line, the corresponding predicted sequences of river connectivity index, wetland suitability index, and vegetation stress index within the future speed regulation cycle, thereby realizing the dynamic quantification of ecosystem service functions.

[0074] S502. Based on the predicted runoff process line and the corresponding predicted sequences of river connectivity index, wetland suitability index and vegetation stress index within the future speed regulation cycle, the improved NSGA-III algorithm is adopted through the adaptive multi-objective optimization decision module to coordinately optimize the river connectivity index, wetland suitability index and vegetation stress index while satisfying the objective constraints, and generate the Pareto front scheduling scheme solution set.

[0075] Among them, the objective constraint is at least one of the following: the maximum safe storage capacity of the reservoir group, the minimum safe storage capacity of the reservoir, the downstream flood control safety flow, and the minimum demand for ensuring water supply and power generation; each solution in the Pareto frontier scheduling scheme solution set corresponds to a set of outflow schemes for each reservoir in 3 hours over the next 72 hours.

[0076] S503. Through the dynamic feedback and rolling decision module, real-time monitoring data from the Internet of Things sensing network is received. The initial field of the model is updated by the data assimilation unit, which drives the scenario prediction unit to generate hydrological-ecological prediction sequences for future periods. The scheme evaluation and optimization unit determines the function value of each Pareto front scheduling scheme solution in the solution set based on the rolling window mechanism and the objective function, and selects the Pareto front scheduling scheme solution with the highest function value as the optimal solution. Based on the optimal solution, a discharge flow command is generated. The discharge flow command is converted into a gate control signal by the command generation unit to control the discharge flow of each reservoir, forming a closed-loop feedback.

[0077] The objective function is determined by a weighted sum of standardized water supply, power generation, flood control, and ecological objectives, with the weight of each objective set based on social environmental preferences.

[0078] According to the above scheme, by constructing a variety of refined distributed hydrological models, the key physical processes in high-altitude and cold mountainous areas were accurately characterized; a dynamic response three-dimensional river connectivity index model, wetland hydrological suitability model, and vegetation water stress index model were established, realizing the transformation of ecosystem service functions from static assessment to dynamic quantification; combined with the improved NSGA-III algorithm and the integration of constraint violation feedback and solution set distribution equilibrium control strategies, the algorithm's global search capability and convergence efficiency were significantly improved; and efficient assimilation of multimodal real-time sensing data and models, accurate prediction of future scenarios, and rolling generation and closed-loop correction of scheduling instructions were achieved, ensuring real-time matching between the scheduling scheme and the actual ecosystem state.

[0079] The various embodiments in this specification are described in a progressive manner. For the same or similar parts between the various embodiments, please refer to each other. Each embodiment focuses on describing the differences from other embodiments.

[0080] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of this application.

Claims

1. A precise water resource scheduling system based on ecosystem service flows, characterized in that, The system includes: a distributed hydro-ecological coupling modeling module, an adaptive multi-objective optimization decision-making module, and a dynamic feedback and rolling decision-making module; the distributed hydro-ecological coupling modeling module is connected to the dynamic feedback and rolling decision-making module through the adaptive multi-objective optimization decision-making module. The distributed hydrology-ecology coupled modeling module is used to construct snow ablation models, permafrost active layer hydrothermal coupled models, and glacial meltwater models in high-altitude cold mountainous areas based on the principles of energy balance and physical mechanisms. Simultaneously, it constructs three-dimensional river connectivity index models, wetland hydrological suitability models, and vegetation water stress index models, and outputs the predicted runoff process line, corresponding predicted sequences of river connectivity index, wetland suitability index, and vegetation stress index within the future speed regulation cycle, thereby realizing the dynamic quantification of ecosystem service functions. The adaptive multi-objective optimization decision-making module is used to generate a Pareto front scheduling scheme solution set based on the predicted runoff process line and the corresponding predicted sequences of river connectivity index, wetland suitability index, and vegetation stress index within the future speed regulation cycle, using an improved NSGA-III algorithm. This is achieved by simultaneously optimizing the river connectivity index, wetland suitability index, and vegetation stress index while satisfying the objective constraints. The objective constraints are at least one of the following: maximum safe storage capacity of the reservoir group, minimum safe storage capacity of the reservoirs, downstream flood control safety flow, and minimum demand for water supply and power generation. Each solution in the Pareto front scheduling scheme solution set corresponds to a set of 3-hourly discharge flow plans for each reservoir over the next 72 hours. The dynamic feedback and rolling decision module receives real-time monitoring data from the IoT sensing network, updates the initial field of the model through the data assimilation unit, and drives the scenario prediction unit to generate hydrological-ecological prediction sequences for future periods. The scheme evaluation and optimization unit, based on a rolling window mechanism and an objective function, determines the function value of each Pareto front scheduling scheme solution in the solution set, and selects the Pareto front scheduling scheme solution with the highest function value as the optimal solution. Based on the optimal solution, a discharge flow command is generated, which is then converted into a gate control signal by the command generation unit to control the discharge flow of each reservoir, forming a closed-loop feedback. The objective function is determined based on a weighted sum of standardized water supply, power generation, flood control, and ecological objectives, with the weight value of each objective set based on social environmental preferences.

2. The system according to claim 1, characterized in that, The snow ablation model for high-altitude and cold mountainous areas uses a 10-meter resolution digital elevation model as the basic grid unit. The net radiation flux received by each unit is calculated hourly. Shortwave radiation is determined by the solar altitude angle, atmospheric transparency, and surface albedo. Longwave radiation is calculated by the Stefan-Boltzmann law combined with a cloud cover correction factor. Sensible and latent heat fluxes are solved by wind speed, air temperature, humidity, and surface roughness parameters based on the Moning-Obukhov similarity theory. The snow surface temperature is solved iteratively by the heat conduction equation. When the surface temperature reaches 0°C and there is a positive net energy input, the phase change process is initiated. The amount of melted water is determined by dividing the energy surplus by the latent heat constant. The surface albedo is dynamically updated with the degree of snow aging, and its decay rate is jointly driven by snow density, liquid water content, and dust deposition flux.

3. The system according to claim 1, characterized in that, The hydrothermal coupling model of the active layer of permafrost is a one-dimensional vertical unsteady-state thermo-water coupling model. The model divides the soil profile into computational layers with a thickness of 0.05 meters. Each layer independently tracks the evolution of the temperature field and water content field. The latent heat of phase change is introduced into the heat conduction equation. The ice-water phase change threshold is set in the range of -0.5℃ to 0.5℃. The effective heat capacity method is used to handle the nonlinear thermophysical property changes during the phase change process. The water migration equation adopts the Richards equation, considering the nonlinear relationship between the unfrozen water content and the matrix potential under freezing conditions. Its parameters are determined by the van Genuchten-Mualem model combined with the freezing correction function. The upper boundary condition is dynamically switched by the snow cover state. When there is no snow cover, the measured meteorological forcing is used. When there is snow cover, the heat flux and water flux are coupled and transferred through the snow bottom layer. The lower boundary is set to a zero flux condition. The model time step is 1 hour and is solved by implicit finite difference scheme.

4. The system according to claim 1, characterized in that, The adaptive multi-objective optimization decision-making module employs an improved NSGA-III algorithm with a population size of 300. Reference points are uniformly generated in the five-dimensional objective space using the Das-Dennis method. The crossover operation uses simulated binary crossover with a fixed distribution exponent of 20, and the mutation operation uses multinomial mutation with a fixed distribution exponent of 20. The crossover probability P... c With the probability of mutation P m Based on the current algebra t and the maximum algebra T max Dynamic adjustment: , In addition, a constraint violation feedback mechanism is introduced: if the optimal solution violates the maximum safe storage capacity of the reservoir group or the downstream flood control safety flow for 5 consecutive generations, the mutation probability P is temporarily increased. m Up to 0.6; when the solution set of the Pareto front scheduling scheme is unevenly distributed in the direction of the reference point, the crossover probability of individuals in the sparse direction is locally increased, and the algorithm terminates when the hypervolume index changes by less than 10 after 500 generations or for 100 consecutive generations. -4 .

5. The system according to claim 1, characterized in that, The real-time monitoring data includes air temperature, precipitation, wind speed, humidity, and radiation provided by automatic weather stations; snow depth, snow water equivalent, and snow layer temperature profiles provided by snow monitoring stations; ground temperature and soil moisture content at different depths provided by permafrost monitoring wells; flow rate, water level, and water temperature provided by river hydrological stations; and high-resolution images of land cover and snow cover range acquired by regularly flying UAV remote sensing platforms. The data assimilation unit is used to assimilate the real-time monitoring data into the state variables of the distributed hydrological-ecological coupling modeling module after timestamp alignment and quality control, using an ensemble Kalman filter algorithm. The ensemble size is 50, and the localization radius of the covariance is set to 15 kilometers.

6. The system according to claim 5, characterized in that, During the assimilation process, the ensemble Kalman filter algorithm compares the real-time observation data with the simulated observations corresponding to all models, calculates the Kalman gain, and then adjusts the state variables of all model members in reverse to make the model state approximate the real ecosystem state as closely as possible. The localization radius of the covariance is set to 15 kilometers to suppress long-range spurious correlations caused by finite set samples.

7. The system according to claim 6, characterized in that, The scenario prediction unit drives the model in the distributed hydrological-ecological coupled modeling module and connects to high-resolution numerical weather prediction products as meteorological boundary forcing for the next 72 hours. The high-resolution numerical weather prediction products have a temporal resolution of 3 hours and generate detailed prediction sequences every 3 hours for the next 72 hours. The detailed prediction sequences include the runoff process at the watershed outlet and each control section, the spatiotemporal changes of the river's comprehensive connectivity index, the evolution of water levels and suitability indices of major wetlands, and the spatial distribution map of vegetation stress index.

8. The system according to claim 7, characterized in that, The scheme evaluation and optimization unit divides the 72-hour prediction period into 24 3-hour intervals, generates candidate discharge flow schemes for each interval, and adopts a rolling window mechanism in the evaluation process: only the schemes of the first 3 hours are finally selected, and the subsequent schemes are reserved as alternatives; and, under the condition that the target constraints are met and the downstream cross-section ecological flow is not lower than the dynamic threshold, the function value of each Pareto front scheduling scheme solution in the Pareto front scheduling scheme solution set is determined by the objective function; the dynamic threshold is dynamically adjusted based on the season and the fish migration period.

9. The system according to claim 8, characterized in that, The instruction generation unit incorporates a reservoir gate hydraulic characteristic model, which is used to convert the discharge flow instruction into a control signal for a specific gate. The control signal includes the opening height, the number of gates to be opened, and the opening duration. The control signal is then transmitted to the gate monitoring system at each reservoir via an industrial control network, thereby driving the actuator of the gate monitoring system to precisely control the discharge flow. Furthermore, the new discharge flow acts on the real river system, and its effect is captured by the Internet of Things sensing network as new real-time monitoring data, thus obtaining closed-loop feedback.

10. A method for precise water resource allocation based on ecosystem service flows, characterized in that, The method is applied to a precise water resource scheduling system based on ecosystem service flows according to any one of claims 1 to 9, the method comprising: Based on the principles of energy balance and physical mechanisms, a distributed hydro-ecological coupling modeling module is used to construct a snow ablation model, a hydrothermal coupling model of the active layer of permafrost, and a glacial meltwater model for high-altitude and cold mountainous areas. Simultaneously, a three-dimensional river connectivity index model, a wetland hydrological suitability model, and a vegetation water stress index model are constructed. The model outputs the predicted runoff process line, the corresponding predicted sequences of river connectivity index, wetland suitability index, and vegetation stress index within the future speed regulation cycle, thereby realizing the dynamic quantification of ecosystem service functions. Based on the predicted runoff process curve and the corresponding predicted sequences of river connectivity index, wetland suitability index, and vegetation stress index within the future speed regulation cycle, an adaptive multi-objective optimization decision module employs an improved NSGA-III algorithm to collaboratively optimize the river connectivity index, wetland suitability index, and vegetation stress index while satisfying objective constraints, generating a Pareto front scheduling solution set. The objective constraints are at least one of the following: maximum safe operating capacity of the reservoir group, minimum safe operating capacity of the reservoirs, downstream flood control safety flow, and minimum demand for ensuring water supply and power generation. Each solution in the Pareto front scheduling solution set corresponds to a set of 3-hourly discharge flow schemes for each reservoir within the next 72 hours. Through a dynamic feedback and rolling decision-making module, real-time monitoring data from the IoT sensing network is received. The data assimilation unit updates the initial field of the model, driving the scenario prediction unit to generate hydrological-ecological prediction sequences for future periods. The scheme evaluation and optimization unit determines the function value of each Pareto front scheduling scheme solution in the Pareto front scheduling scheme solution set based on a rolling window mechanism and an objective function, and selects the Pareto front scheduling scheme solution with the highest function value as the optimal solution. Based on the optimal solution, a discharge flow command is generated, and the discharge flow command is converted into a gate control signal by the command generation unit to control the discharge flow of each reservoir, forming a closed-loop feedback. The objective function is determined based on the weighted sum of standardized water supply, power generation, flood control, and ecological objectives, with the weight value of each objective set based on social environmental preferences.

Images