Water quality and water quantity joint scheduling method based on distributed hydrological model

Abstract

The invention relates to the technical field of water environment treatment, in particular to a water quality and water quantity joint scheduling method based on a distributed hydrological model. The method comprises the steps of arranging high-precision monitoring equipment to collect multi-source data, constructing a basic data set, analyzing spatial and temporal distribution characteristics and dynamic change rules of pollutants, designing a multilayer neural network model for training optimization, and generating a water quality prediction result and a scheduling strategy. The invention solves or at least alleviates the capability limitation problem of a non-point source pollution model in the aspect of describing a pollutant migration path and mechanism, and provides a water quality and water quantity joint scheduling method based on a distributed hydrological model. According to the method, through fusion of multi-source data and a dynamic optimization regulation strategy, accurate prediction and efficient treatment of pollutants under complex hydrological conditions are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of water resource management and water environment governance, and in particular to a water quality and water quantity joint scheduling method based on a distributed hydrological model. Background Art

[0002] The coordinated regulation of water quality and quantity is a core component of river basin water environment management. Especially in the context of increasingly prominent non-point source pollution, accurately predicting and effectively addressing pollutant migration and transformation processes under complex hydrological conditions has become a critical issue that needs to be addressed. In recent years, non-point source pollution models have rapidly developed as an important tool for quantitatively characterizing pollution load characteristics and screening for optimal prevention and control measures. They mainly include empirical models and physical models. Empirical models are established through statistical analysis, such as the output coefficient method, inventory analysis method, and discharge coefficient method. They are widely used due to their advantages of few parameters, simple structure, and acceptable accuracy. However, these models have difficulty describing the specific pathways and mechanisms of pollutant migration, limiting their further application under complex hydrological conditions. Although physical models can more accurately simulate pollutant migration and transformation processes, they have high requirements for data quality and computing resources, and still face many challenges in practical application.

[0003] Currently, the application of distributed hydrological models at the basin scale is gradually increasing, providing a new technical means for the joint scheduling of water quality and quantity. However, existing methods still have shortcomings in the following aspects: First, traditional models do not accurately depict the spatiotemporal distribution characteristics of rainfall, runoff, and pollutant concentrations, making it difficult to meet the needs of high-precision predictions; second, the model training process lacks adaptive capabilities and cannot dynamically adjust parameters based on real-time monitoring data, resulting in large deviations between prediction results and actual conditions; third, existing scheduling strategies are mostly based on static rules, which are difficult to cope with dynamic changes such as sudden changes in rainfall intensity, abnormal runoff fluctuations, and accelerated pollutant diffusion, and their emergency response capabilities are insufficient. In addition, the application of multi-source data fusion technology in water quality prediction is still immature, and the integration efficiency of remote sensing imagery, meteorological data, and ground monitoring data is low, which affects the comprehensiveness and accuracy of prediction results.

[0004] Therefore, developing a method for jointly scheduling water quality and quantity based on a distributed hydrological model to accurately predict the spatiotemporal distribution of pollutant concentrations and formulate targeted dynamic scheduling strategies is of great practical significance. This invention aims to improve the accuracy and timeliness of water quality predictions by constructing a high-precision monitoring system, combining a multi-layer neural network model with an adaptive weight adjustment mechanism. Furthermore, through multi-source data fusion technology and spectral analysis methods, it identifies high-risk areas and time periods, and formulates hierarchical and classified scheduling strategies, thereby comprehensively improving the scientific nature and effectiveness of river basin water environment management. Summary of the Invention

[0005] The present invention aims to overcome the shortcomings of existing technologies, addressing or at least alleviating the limitations of non-point source pollution models in describing pollutant migration pathways and mechanisms. It provides a method for jointly scheduling water quality and quantity based on a distributed hydrological model. By integrating multi-source data and dynamically optimizing control strategies, this method enables accurate prediction and efficient control of pollutants under complex hydrological conditions.

[0006] To achieve the above object, the present invention provides the following technical solution: a water quality and water quantity joint scheduling method based on a distributed hydrological model, characterized in that it includes the following steps:

[0007] S1. Deploy high-precision monitoring equipment in the target watershed to simultaneously collect rainfall, runoff, pollutant concentrations, and meteorological data;

[0008] S2. In the initial stage, record the hydrological boundary conditions, water quality parameters and spatial distribution data in different time periods and build a basic data set;

[0009] S3. Obtain the spatiotemporal distribution characteristics of rainfall intensity on runoff process through experimental data in basic data set;

[0010] S4. Extract the dynamic variation pattern of pollutant concentration in time series based on experimental data;

[0011] S5. Comprehensively analyze the spatiotemporal distribution characteristics and dynamic change patterns, design a multi-layer neural network model, and integrate hydrological, meteorological, and pollutant concentration data for training;

[0012] S6. Iteratively optimize the multi-layer neural network model based on actual monitoring data to form a prediction model and algorithm;

[0013] S7. Apply the optimized model to water quality prediction, generate the spatiotemporal distribution map of typical inflowing tributaries, and formulate corresponding scheduling strategies.

[0014] In order to further realize the present invention, the following technical solutions may be preferably used:

[0015] Preferably, step S2 includes the following steps:

[0016] S201. Construct a set of hydrodynamic equations, using the Navier-Stokes equations in the Cartesian coordinate system as the basis, and combine them with the material transport equations to describe the migration and transformation process of pollutants;

[0017] S202, dividing the calculation area into grids, using quadrilateral grids to simulate river banks and triangular grids to simulate reservoir bays, and setting the grid sizes to suit the terrain characteristics;

[0018] S203. Set boundary conditions and initial conditions, where the boundary conditions include inflow flow, outflow flow and pollutant concentration, and the initial conditions include initial water level velocity and pollutant distribution.

[0019] Preferably, during the meshing process, the quadrilateral mesh size is set to 200m×200m, which is suitable for the main river channel and regular shoreline area; the triangular mesh size is set to 5000m 2 , which is suitable for reservoir bays and irregular shoreline areas; the two grids are seamlessly connected through shared nodes to ensure the overall continuity of the computational domain;

[0020] The setting of the boundary conditions includes hydrodynamic boundary conditions and water quality boundary conditions. The hydrodynamic boundary conditions are controlled by upstream flow and downstream water level, and the water quality boundary conditions are controlled by upstream pollutant concentration and downstream pollutant flux; the initial conditions are generated by interpolation of historical data, including the initial water level velocity field and pollutant concentration field.

[0021] Preferably, the dynamic change law decomposes the time series signal into multi-scale components through wavelet transform, and the main frequency amplitude is the amplitude of the low-frequency component in the decomposed signal, which reflects the main trend of pollutant concentration change over time.

[0022] Preferably, an adaptive weight adjustment mechanism is adopted in the model training process, including calculating the long-term mean and short-term fluctuation of the input data, extracting abnormal features, judging whether the model weight needs to be updated based on the Markov chain, and triggering the model weight adjustment when the abnormal feature coefficient is greater than the set threshold;

[0023] The adaptive weight adjustment mechanism also includes monitoring the changing trend of pollutant concentrations within a time window, and adjusting model parameters to adapt to new hydrological conditions when the changing trend deviates from the expected range.

[0024] Preferably, the step S7 is specifically as follows:

[0025] In the water quality prediction process, multi-source data fusion technology is used to integrate remote sensing image meteorological data and ground monitoring data into the model input layer; the prediction results are output in the form of spatiotemporal distribution maps, showing the changing trends of pollutant concentrations over time and space;

[0026] During the scheduling strategy formulation process, high-risk areas and time periods are identified based on the forecast results, and targeted control measures are proposed; vegetation buffer zones are added in areas with high pollution loads, and the operation of sewage interception facilities is strengthened during high-flow periods;

[0027] Analyze the deviation between the corrected pollutant concentration value and the actual monitoring value, determine the type of anomaly, and select the scheduling strategy based on the anomaly type, which includes sudden changes in rainfall intensity, abnormal runoff fluctuations, accelerated pollutant diffusion, and noise interference;

[0028] Among them, the sudden change of rainfall intensity is manifested as a rapid upward trend in pollutant concentration with changes in rainfall intensity, the abnormal fluctuation of runoff is manifested as nonlinear fluctuations in pollutant concentration with changes in runoff volume, the accelerated diffusion of pollutants is manifested as significant changes in pollutant concentration in a short period of time, and noise interference is manifested as the presence of random high-frequency components in pollutant concentration.

[0029] Preferably, step S7 includes the following steps:

[0030] S71. Identify high-risk areas and time periods

[0031] Based on the spatiotemporal distribution map, continuous areas where pollutant concentrations exceed the warning value by 20% are extracted, high-risk periods lasting more than three monitoring cycles are marked, and composite risk areas within the basin are identified where both the runoff mutation rate is greater than 15% and the pollutant concentration gradient is greater than 0.3 mg / (L·km);

[0032] S72. Develop targeted scheduling strategies

[0033] For high-risk areas: start key monitoring mode and increase data sampling frequency to 200% of the normal value;

[0034] For the continuous period: deploy mobile emergency response equipment to the predicted pollution diffusion front one hour in advance;

[0035] For complex risk areas: Simultaneously activate the three-pronged prevention and control system of pump stations, gates, and dosing devices;

[0036] S73. Execute basic scheduling

[0037] Level 1 prevention and control measures will be activated within 5km upstream of high-risk areas: recreational water intakes will be closed and navigation will be restricted;

[0038] Implement pre-dispatching 2 hours before the high-risk period: reduce the reservoir water level to 90% of the design value;

[0039] An emergency buffer zone is set up 10 km downstream of the complex risk area: adsorption barriers and online monitoring buoys are deployed.

[0040] Preferably, the step S7 further includes the following steps:

[0041] S74. Determine the type of exception

[0042] Analyze the spectrum characteristics of monitoring data:

[0043] Rainfall mutation type: 0.5-2Hz low frequency band energy accounts for more than 40%;

[0044] Abnormal runoff type: continuous oscillation waveform appears in the 2-5Hz mid-frequency band;

[0045] Pollution diffusion type: pulse cluster characteristics exist in the high frequency band of 5-10Hz;

[0046] S75. Dynamically adjust strategy parameters

[0047] When rainfall changes suddenly: for every 10mm / h increase in rainfall, the pump station power will be increased by 15%;

[0048] When the runoff is abnormal: for every 5% deviation of the flow rate from the reference value, the gate opening is adjusted by 3%;

[0049] When the pollution spreads: for every 0.1 mg / (L·h) increase in concentration gradient, shorten the dosing interval by 20%.

[0050] Preferably, after completing the parameter adjustment in step S75, if the change rate of at least one device parameter is greater than 15%, the following steps are immediately executed:

[0051] S76, Strategy Optimization Control

[0052] When rainfall and runoff anomalies occur simultaneously, gate regulation will be started 5 minutes later after the pump station power is increased to 120%;

[0053] When runoff and pollution anomalies occur simultaneously, the automatic dosing device will be activated immediately when the flow rate stabilizes within the range of ±5%; when triple anomalies occur simultaneously, the pump station will be started to drain water after 60% of the dosage of the first round of drug delivery is completed;

[0054] S77, transition control start

[0055] When the sudden rainfall is relieved and the pump station power drops below 110%, the power reduction program is triggered;

[0056] When the runoff fluctuation remains stable within the range of ±5% for 30 minutes, the gate reset procedure is triggered;

[0057] When the pollution concentration is lower than the warning value for two consecutive hours, the monitoring frequency reduction program is triggered.

[0058] Preferably, after the transition control of step S77 is completed, if the fluctuation rate of the equipment parameter is less than 5% for 1 hour, the following steps are performed:

[0059] S78, scenario-based post-disaster disposal:

[0060] After the reservoir entrance treatment is completed, the diversion channel will be kept open until the turbidity of the water flow drops by 50%;

[0061] During the disposal period at the confluence of tributaries, the dissolved oxygen content downstream is tested after each purification cycle is completed;

[0062] After the treatment of the dam front area is completed, the ultrafiltration device is maintained at 50% power for 12 hours;

[0063] S79, Effect Verification Modeling:

[0064] Rainfall disposal verification: calculate the pumping and drainage energy consumption ratio per unit rainfall and optimize the pump station combination plan;

[0065] Runoff disposal verification: Analyze the correlation between gate adjustment times and flow rate stability;

[0066] Pollution disposal verification: Establish a dose-effect model of the amount of chemical added and the reduction in concentration.

[0067] The beneficial effects of the present invention are:

[0068] The present invention achieves accurate characterization of pollutant migration paths under complex hydrological conditions by constructing a distributed hydrological model and integrating multi-source data, solves the problem that empirical models are difficult to describe the pollutant migration mechanism, and improves the applicability and prediction accuracy of the model.

[0069] At the same time, the present invention significantly enhances the model's ability to respond to abnormal conditions by introducing an adaptive weight adjustment mechanism and a dynamic optimization scheduling strategy. It can quickly adjust control measures in complex scenarios such as sudden rainfall changes and abnormal runoff, thereby ensuring the safety of the water environment in the basin.

[0070] In addition, the multi-layer neural network model proposed in the present invention combined with wavelet transform technology can effectively extract the dynamic change law of pollutant concentration, providing reliable technical support for the formulation of scientific and reasonable water quality and water quantity joint scheduling strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is the overall flow chart of the present invention.

[0072] Figure 2 Detailed flowchart of step S2 of the present invention.

[0073] Figure 3 This is a flowchart of step S7 of the present invention.

[0074] Figure 4 Schematic diagram of grid division of the present invention.

[0075] Figure 5 This is a flow chart of the adaptive weight adjustment mechanism of the present invention.

[0076] Figure 6 This is a flowchart of the abnormality type judgment of the present invention.

[0077] Figure 7 This is a control flow chart for the strategy optimization of the present invention.

[0078] Figure 8 This is a flow chart for the transition control start-up of the present invention.

[0079] Figure 9 This is a scenario-based post-processing flowchart of the present invention. DETAILED DESCRIPTION

[0080] In the description of the present invention, it should also be noted that, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to mechanical connections or electrical connections; they may refer to direct connections or indirect connections through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0081] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0082] Example 1

[0083] Currently, the application of distributed hydrological models at the basin scale is gradually increasing, providing a new technical means for the joint scheduling of water quality and quantity. However, existing methods still have shortcomings in the following aspects: First, traditional models do not accurately depict the spatiotemporal distribution characteristics of rainfall, runoff, and pollutant concentrations, making it difficult to meet the needs of high-precision predictions; second, the model training process lacks adaptive capabilities and cannot dynamically adjust parameters based on real-time monitoring data, resulting in large deviations between prediction results and actual conditions; third, existing scheduling strategies are mostly based on static rules, which are difficult to cope with dynamic changes such as sudden changes in rainfall intensity, abnormal runoff fluctuations, and accelerated pollutant diffusion, and their emergency response capabilities are insufficient. In addition, the application of multi-source data fusion technology in water quality prediction is still immature, and the integration efficiency of remote sensing imagery, meteorological data, and ground monitoring data is low, which affects the comprehensiveness and accuracy of prediction results.

[0084] Reference Figure 1 This embodiment discloses a water quality and water quantity joint scheduling method based on a distributed hydrological model, comprising the following steps:

[0085] S1. Deploy high-precision monitoring equipment in the target watershed to simultaneously collect rainfall, runoff, pollutant concentrations, and meteorological data;

[0086] S2. In the initial stage, record the hydrological boundary conditions, water quality parameters and spatial distribution data in different time periods and build a basic data set;

[0087] S3. Obtain the spatiotemporal distribution characteristics of rainfall intensity on runoff process through experimental data in basic data set;

[0088] S4. Extract the dynamic variation pattern of pollutant concentration in time series based on experimental data;

[0089] S5. Comprehensively analyze the spatiotemporal distribution characteristics and dynamic change patterns, design a multi-layer neural network model, and integrate hydrological, meteorological, and pollutant concentration data for training;

[0090] S6. Iteratively optimize the multi-layer neural network model based on actual monitoring data to form a prediction model and algorithm;

[0091] S7. Apply the optimized model to water quality prediction, generate the spatiotemporal distribution map of typical inflowing tributaries, and formulate corresponding scheduling strategies.

[0092] Reference Figure 2 , the step S2 comprises the following steps:

[0093] S201. Construct a set of hydrodynamic equations, using the Navier-Stokes equations in the Cartesian coordinate system as the basis, and combine them with the material transport equations to describe the migration and transformation process of pollutants;

[0094] S202, dividing the calculation area into grids, using quadrilateral grids to simulate river banks and triangular grids to simulate reservoir bays, and setting the grid sizes to suit the terrain characteristics;

[0095] S203. Set boundary conditions and initial conditions, where the boundary conditions include inflow flow, outflow flow and pollutant concentration, and the initial conditions include initial water level velocity and pollutant distribution.

[0096] In step S201, the two-dimensional hydrodynamic control equations are the Navier-Stokes equations in the Cartesian coordinate system, which consists of the water flow continuity equation, the momentum equation along the water flow direction (x direction), and the momentum equation perpendicular to the water flow direction (y direction).

[0097]

[0098] In the above formulas, η is the water surface elevation, h is the total water depth, g is the acceleration of gravity, ρ is the density of water, and ρ0 is the reference density of fresh water. f = 2Ωsinφ is the Coriolis force coefficient (Ω is the angular velocity of rotation, φ is the geographical latitude), P a is the atmospheric pressure, s ij is the radiation stress tensor. S and (u s ,v s) are the emission amount and velocity of the point source, respectively. and is the average value of the velocity over depth, defined as:

[0099]

[0100] (τ sx ,τ sy ) and (τ bx ,τ by ) are the water surface wind stress tensor and the riverbed surface stress tensor. It can be determined by the square law of resistance (friction resistance is proportional to the square of flow velocity):

[0101]

[0102] c f is the drag coefficient or riverbed friction, is the average velocity of water depth on the riverbed, and the friction velocity on the riverbed is The friction of the riverbed can be estimated using the Chezy number C or the Manning number M:

[0103]

[0104] The unit of Xie Cai coefficient is m 1 / 2 / s, the unit of Manning coefficient is m 1 / 3 / s. Manning coefficient and riverbed roughness height (roughness) k s The relationship is as follows:

[0105]

[0106] The Manning coefficient value is generally between 20 and 40m 1 / 3 / s.

[0107] (τ sx ,τ sy ) is the water surface wind stress tensor, wind stress It can be obtained by the following empirical formula

[0108]

[0109] Where ρ a is the air density, c d is the air resistance coefficient, The wind speed 10m above the water surface. The friction rate caused by wind stress can be expressed as

[0110]

[0111] T ijis the lateral stress, including viscous friction, turbulent friction and differential convection, which can be estimated using the eddy viscosity coefficient formula based on the depth-averaged velocity gradient:

[0112]

[0113] Where A is the horizontal eddy viscosity coefficient.

[0114] According to the Kolmogrov and Prandtle theory, the turbulent eddy viscosity coefficient v τ Proportional to the square root of the turbulent kinetic energy k and the characteristic eddy viscosity scale l. If the dissipation scale is l (dissipation rate ε=κ 3 / 2 / l), the expression of eddy viscosity coefficient expressed in κ and ε can be obtained:

[0115]

[0116] Where c μ is an empirical constant.

[0117] The logarithmic eddy viscosity coefficient can be calculated by the following formula

[0118]

[0119] Where U τ =max(U τs ,U τb ), c1 and c2 are constants. When c1 = 0.41 and c2 = -0.41, the expression is a standard parabola.

[0120] In 1996, Smagorinsky proposed a formula that relates the effective eddy viscosity coefficient to the characteristic length at the sub-grid scale:

[0121]

[0122] c s is called the "Smagorinsky constant", l represents the characteristic length, and the deformation rate is defined as:

[0123]

[0124] According to the law of conservation of mass, considering factors such as convection, diffusion and degradation during the migration of pollutants, the transport equation of pollutants can be written as

[0125]

[0126] Where c is the concentration of pollutants; (D x ,D y ) are the diffusion coefficients in the x and y directions. K dis the linear degradation coefficient of the pollutant, that is, the degradation of the pollutant conforms to the first-order reaction formula:

[0127]

[0128] Reference Figure 4 In the meshing process, the quadrilateral mesh size is set to 200m×200m, which is suitable for the main river channel and regular shoreline area; the triangular mesh size is set to 5000m 2 , which is suitable for reservoir bays and irregular shoreline areas; the two grids are seamlessly connected through shared nodes to ensure the overall continuity of the computational domain;

[0129] The setting of the boundary conditions includes hydrodynamic boundary conditions and water quality boundary conditions. The hydrodynamic boundary conditions are controlled by upstream flow and downstream water level, and the water quality boundary conditions are controlled by upstream pollutant concentration and downstream pollutant flux; the initial conditions are generated by interpolation of historical data, including the initial water level velocity field and pollutant concentration field.

[0130] The dynamic change law decomposes the time series signal into multi-scale components through wavelet transform, and the main frequency amplitude is the amplitude of the low-frequency component in the decomposed signal, which reflects the main trend of pollutant concentration change over time.

[0131] Reference Figure 5 ,The model training process adopts an adaptive weight adjustment mechanism, including calculating the long-term mean and short-term fluctuation of the input data, extracting abnormal features, and determining whether the model weight needs to be updated based on the Markov chain, and triggering the model weight adjustment when the abnormal feature coefficient is greater than the set threshold;

[0132] The adaptive weight adjustment mechanism also includes monitoring the changing trend of pollutant concentrations within a time window, and adjusting model parameters to adapt to new hydrological conditions when the changing trend deviates from the expected range.

[0133] The step S7 is specifically as follows:

[0134] In the water quality prediction process, multi-source data fusion technology is used to integrate remote sensing image meteorological data and ground monitoring data into the model input layer; the prediction results are output in the form of spatiotemporal distribution maps, showing the changing trends of pollutant concentrations over time and space;

[0135] During the scheduling strategy formulation process, high-risk areas and time periods are identified based on the forecast results, and targeted control measures are proposed; vegetation buffer zones are added in areas with high pollution loads, and the operation of sewage interception facilities is strengthened during high-flow periods;

[0136] Reference Figure 6, analyze the deviation between the corrected pollutant concentration value and the actual monitoring value, determine the anomaly type, and select the scheduling strategy according to the anomaly type, which includes sudden change in rainfall intensity, abnormal runoff fluctuation, accelerated pollutant diffusion and noise interference;

[0137] Among them, the sudden change of rainfall intensity is manifested as a rapid upward trend in pollutant concentration with changes in rainfall intensity, the abnormal fluctuation of runoff is manifested as nonlinear fluctuations in pollutant concentration with changes in runoff volume, the accelerated diffusion of pollutants is manifested as significant changes in pollutant concentration in a short period of time, and noise interference is manifested as the presence of random high-frequency components in pollutant concentration.

[0138] Reference Figure 3 、 Figure 7 、 Figure 8 and Figure 9 , the step S7 comprises the following steps:

[0139] S71. Identify high-risk areas and time periods

[0140] Based on the spatiotemporal distribution map, continuous areas where pollutant concentrations exceed the warning value by 20% are extracted, high-risk periods lasting more than three monitoring cycles are marked, and composite risk areas within the basin are identified where both the runoff mutation rate is greater than 15% and the pollutant concentration gradient is greater than 0.3 mg / (L·km);

[0141] S72. Develop targeted scheduling strategies

[0142] For high-risk areas: start key monitoring mode and increase data sampling frequency to 200% of the normal value;

[0143] For the continuous period: deploy mobile emergency response equipment to the predicted pollution diffusion front one hour in advance;

[0144] For complex risk areas: Simultaneously activate the three-pronged prevention and control system of pump stations, gates, and dosing devices;

[0145] S73. Execute basic scheduling

[0146] Level 1 prevention and control measures will be activated within 5km upstream of high-risk areas: recreational water intakes will be closed and navigation will be restricted;

[0147] Implement pre-dispatching 2 hours before the high-risk period: reduce the reservoir water level to 90% of the design value;

[0148] An emergency buffer zone is set up 10 km downstream of the complex risk area: adsorption barriers and online monitoring buoys are deployed.

[0149] The step S7 further includes the following steps:

[0150] S74. Determine the type of exception

[0151] Analyze the spectrum characteristics of monitoring data:

[0152] Rainfall mutation type: 0.5-2Hz low frequency band energy accounts for more than 40%;

[0153] Abnormal runoff type: continuous oscillation waveform appears in the 2-5Hz mid-frequency band;

[0154] Pollution diffusion type: pulse cluster characteristics exist in the high frequency band of 5-10Hz;

[0155] S75. Dynamically adjust strategy parameters

[0156] When rainfall changes suddenly: for every 10mm / h increase in rainfall, the pump station power will be increased by 15%;

[0157] When the runoff is abnormal: for every 5% deviation of the flow rate from the reference value, the gate opening is adjusted by 3%;

[0158] When the pollution spreads: for every 0.1 mg / (L·h) increase in concentration gradient, shorten the dosing interval by 20%.

[0159] After completing the parameter adjustment in step S75, if at least one device parameter change rate is greater than 15%, the following steps are immediately executed:

[0160] S76, Strategy Optimization Control

[0161] When rainfall and runoff anomalies occur simultaneously, gate regulation will be started 5 minutes later after the pump station power is increased to 120%;

[0162] When runoff and pollution anomalies occur simultaneously, the automatic dosing device will be activated immediately when the flow rate stabilizes within the range of ±5%; when triple anomalies occur simultaneously, the pump station will be started to drain water after 60% of the dosage of the first round of drug delivery is completed;

[0163] S77, transition control start

[0164] When the sudden rainfall is relieved and the pump station power drops below 110%, the power reduction program is triggered;

[0165] When the runoff fluctuation remains stable within the range of ±5% for 30 minutes, the gate reset procedure is triggered;

[0166] When the pollution concentration is lower than the warning value for two consecutive hours, the monitoring frequency reduction program is triggered.

[0167] After the transition control of step S77 is completed, if the equipment parameter fluctuation rate is less than 5% for 1 hour, the following steps are performed:

[0168] S78, scenario-based post-disaster disposal:

[0169] After the reservoir entrance treatment is completed, the diversion channel will be kept open until the turbidity of the water flow drops by 50%;

[0170] During the disposal period at the confluence of tributaries, the dissolved oxygen content downstream is tested after each purification cycle is completed;

[0171] After the treatment of the dam front area is completed, the ultrafiltration device is maintained at 50% power for 12 hours;

[0172] S79, Effect Verification Modeling:

[0173] Rainfall disposal verification: calculate the pumping and drainage energy consumption ratio per unit rainfall and optimize the pump station combination plan;

[0174] Runoff disposal verification: Analyze the correlation between gate adjustment times and flow rate stability;

[0175] Pollution disposal verification: Establish a dose-effect model of the amount of chemical added and the reduction in concentration.

[0176] The present invention achieves accurate characterization of pollutant migration paths under complex hydrological conditions by constructing a distributed hydrological model and integrating multi-source data, solves the problem that empirical models are difficult to describe the pollutant migration mechanism, and improves the applicability and prediction accuracy of the model.

[0177] At the same time, the present invention significantly enhances the model's ability to respond to abnormal conditions by introducing an adaptive weight adjustment mechanism and a dynamic optimization scheduling strategy. It can quickly adjust control measures in complex scenarios such as sudden rainfall changes and abnormal runoff, thereby ensuring the safety of the water environment in the basin.

[0178] Example 2

[0179] In order to better enable relevant personnel in this technical field to fully understand and implement the present invention, the specific implementation principle of the present invention is supplemented below with reference to a specific application scenario.

[0180] The present invention provides a water quality and quantity joint scheduling method based on a distributed hydrological model. The core of the method is to achieve accurate prediction and efficient control of pollutant migration paths and concentration changes under complex hydrological conditions through multi-source data fusion and dynamic optimization control strategies.

[0181] The first step of this method is to deploy high-precision monitoring equipment in the target watershed. These devices include rain gauges, flow meters, water quality sensors, and meteorological stations, which are used to simultaneously collect rainfall, runoff, pollutant concentrations, and meteorological data. The distribution of these devices should cover key nodes in the entire watershed, such as the main river channel, tributary confluences, reservoir bays, and key areas susceptible to pollution. For example, in a typical watershed application, a set of rain gauges and flow meters are deployed every 5 kilometers along the main river channel, and water quality sensors are added near the tributary inlets to capture changes in pollutant concentrations. All monitoring equipment transmits data in real time to the central processing system through a wireless communication network. The central system is responsible for preliminary cleaning and organization of the data to ensure the accuracy of subsequent analysis.

[0182] In the initial phase, hydrological boundary conditions, water quality parameters, and spatial distribution data for different time periods are recorded and a basic dataset is constructed. This process is based on historical data combined with experimental data obtained through field sampling. For example, rainfall, runoff, and pollutant concentration data from the past three years are used as a baseline, and initial water level, velocity, and pollutant concentration fields are generated through interpolation. Subsequently, a hydrodynamic model is established based on the Navier-Stokes equations in a Cartesian coordinate system, combining material transport equations to describe the migration and transformation of pollutants. The computational domain is meshed using a quadrilateral grid size of 200 m x 200 m, suitable for main river channels and regular shorelines; and a triangular grid size of 5000 square meters, suitable for reservoir bays and irregular shorelines. The two grids are seamlessly connected through shared nodes to ensure the overall continuity of the computational domain. Boundary conditions include hydrodynamic boundary conditions and water quality boundary conditions. The former is controlled by upstream flow and downstream water level, while the latter is controlled by upstream pollutant concentrations and downstream pollutant fluxes.

[0183] An important part of this method is to obtain the spatiotemporal distribution characteristics of rainfall intensity on the runoff process through experimental data in the basic data set. Specifically, statistical analysis tools are used to perform correlation analysis on the collected rainfall and runoff data to extract the relationship between rainfall intensity and runoff generation. For example, by drawing a scatter plot of rainfall intensity and runoff coefficient, it was found that the two showed a nonlinear growth trend. Further fitting resulted in the empirical formula Q = aP^b+c, where Q is the runoff, P is the rainfall intensity, and a, b, and c are fitting parameters. This formula can effectively reflect the influence of rainfall intensity on the runoff process and provide an important basis for subsequent model training.

[0184] One of the core techniques of this method is to extract the dynamic variation patterns of pollutant concentrations over time series based on experimental data. To improve prediction accuracy, wavelet transform technology is used to decompose the time series signal of pollutant concentration into multi-scale components, and the dominant frequency amplitude of the low-frequency component is extracted as the main trend indicator. For example, monitoring data from a certain watershed shows that pollutant concentrations fluctuate periodically during the rainy season. After wavelet transform decomposition, it was found that the amplitude of the low-frequency component was significantly higher than that of the high-frequency component, indicating that the main trend of pollutant concentration is driven by long-term rainfall patterns. Based on this, a dynamic variation model of pollutant concentration can be constructed: C(t) = A·sin(ωt+φ)+B, where C(t) is the variation of pollutant concentration over time, A is the dominant frequency amplitude, ω is the angular frequency, φ is the phase angle, and B is the baseline concentration. This model can accurately depict the dynamic variation patterns of pollutant concentration and lay the foundation for subsequent model training.

[0185] A comprehensive analysis of spatiotemporal distribution characteristics and dynamic change patterns was conducted, and a multi-layer neural network model was designed. The model was then trained by integrating hydrological, meteorological, and pollutant concentration data. Specifically, the multi-layer neural network model consists of an input layer, a hidden layer, and an output layer. The input layer receives rainfall, runoff, pollutant concentration, and meteorological data. The hidden layer uses a multi-layer perceptron structure to extract features, and the output layer predicts the spatiotemporal distribution of pollutant concentrations. An adaptive weight adjustment mechanism was introduced during model training. Abnormal features were extracted by calculating the long-term mean and short-term fluctuations of the input data, and the need to update the model weights was determined based on a Markov chain. For example, when the abnormal feature coefficient is greater than a set threshold, the model weight adjustment is triggered. At the same time, the changing trend of pollutant concentrations within the time window is monitored. When the changing trend deviates from the expected range, the model parameters are adjusted to adapt to the new hydrological conditions. In addition, cross-validation technology is used in model training to evaluate prediction performance to ensure that the model has good generalization capabilities.

[0186] One of the key steps in this method is to iteratively optimize the multi-layer neural network model based on actual monitoring data. Specifically, the model prediction results are compared with the actual monitoring values, the errors are calculated, and the model parameters are adjusted. For example, a certain prediction result shows that the pollutant concentration is significantly lower than the actual monitoring value during a certain period of time. Through analysis, it is found that the sudden change in rainfall intensity during this period caused abnormal fluctuations in runoff, which in turn affected the distribution of pollutant concentrations. For such situations, the Bayesian optimization algorithm is used to adjust the model hyperparameters, while increasing the sample size of training data to improve the model's response to abnormal conditions. After multiple iterative optimizations, a prediction model and algorithm are finally formed, with a prediction accuracy of over 90%.

[0187] The optimized model is applied to water quality prediction, generating a spatiotemporal distribution map of typical tributaries entering the reservoir, and formulating corresponding scheduling strategies. Specifically, multi-source data fusion technology is used to integrate remote sensing images, meteorological data, and ground monitoring data into the model input layer. The prediction results are output in the form of a spatiotemporal distribution map, showing the changing trends of pollutant concentrations over time and space. For example, a prediction result shows that the pollutant concentration near the inlet of a tributary exceeds the warning value by 20% and lasts for more than three monitoring cycles, identifying it as a high-risk area. For such high-risk areas, a key monitoring mode is activated, the data sampling frequency is increased to 200% of the normal value, and mobile emergency treatment equipment is deployed to the predicted pollution diffusion front one hour in advance. For complex risk areas, that is, areas where the runoff mutation rate is greater than 15% and the pollutant concentration gradient is greater than 0.3 mg / L / km, the three-in-one prevention and control system of pump stations, gates, and dosing devices is simultaneously activated.

[0188] During the scheduling strategy development process, the spectral characteristics of monitoring data are further analyzed to identify anomaly types and dynamically adjust strategy parameters. For example, a sudden rainfall change manifests as a low-frequency band of 0.5 to 2 Hz accounting for more than 40% of energy; a runoff anomaly manifests as a sustained oscillation in the mid-frequency band of 2 to 5 Hz; and a pollution diffusion anomaly manifests as pulse clusters in the high-frequency band of 5 to 10 Hz. Strategy parameters are adjusted based on the anomaly type. For every 10 mm / hour increase in rainfall, the pumping station power is increased by 15%, for every 5% deviation in flow rate from the baseline, the gate opening is adjusted by 3%, and for every 0.1 mg / L / hour increase in concentration gradient, the dosing interval is shortened by 20%. For example, if a sudden rainfall change causes the pumping station power to increase to 120%, gate adjustment is initiated after a 5-minute delay. Once the flow rate stabilizes within a range of ±5%, the automatic dosing device is activated immediately, and the pumping station begins draining after 60% of the initial dose has been delivered. When the sudden rainfall change is lifted and the pump station power drops below 110%, the power reduction program is triggered; when the runoff fluctuation is stable within the range of plus or minus 5% for 30 minutes, the gate reset program is triggered; when the pollution concentration is lower than the warning value for 2 consecutive hours, the monitoring frequency reduction program is triggered.

[0189] After completing transition control, if the fluctuation rate of equipment parameters is less than 5% for one hour, subsequent operations are executed. For example, after the reservoir entrance treatment is completed, the diversion channel remains open until the water turbidity drops by 50%; during the tributary confluence treatment, the dissolved oxygen content downstream is tested after each purification cycle; after the dam front area treatment is completed, the ultrafiltration device is maintained at 50% power for 12 hours. In addition, by calculating the unit rainfall pumping energy consumption ratio, the pump station combination scheme is optimized, the correlation between the number of gate adjustments and flow rate stability is analyzed, and a dose-effect model for the amount of chemical dosage and the concentration reduction is established. For example, the results of a certain experiment showed that the optimal unit rainfall pumping energy consumption ratio was 0.8 kWh / mm, the number of gate adjustments was negatively correlated with flow rate stability, and the dose-effect model for the amount of chemical dosage and the concentration reduction was ΔC = k·log(M), where ΔC is the concentration reduction, M is the chemical dosage, and k is the proportionality coefficient.

[0190] In summary, the present invention achieves accurate characterization of pollutant migration paths under complex hydrological conditions by constructing a distributed hydrological model to integrate multi-source data, solves the problem that empirical models are difficult to describe the mechanism of pollutant migration, and improves the applicability and prediction accuracy of the model. By introducing an adaptive weight adjustment mechanism and a dynamic optimization scheduling strategy, the model's ability to respond to abnormal conditions is significantly enhanced, and it can quickly adjust control measures in complex scenarios such as sudden rainfall changes and runoff anomalies, thereby ensuring the safety of the basin's water environment. The proposed multi-layer neural network model combined with wavelet transform technology can effectively extract the dynamic change law of pollutant concentrations, providing reliable technical support for the formulation of a scientific and reasonable joint scheduling strategy for water quality and water quantity.

[0191] The above are only preferred specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solutions and inventive concepts of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A water quality and quantity joint scheduling method based on a distributed hydrological model, characterized in that: The following steps are involved: S1. Deploy high-precision monitoring equipment in the target watershed to simultaneously collect rainfall, runoff, pollutant concentrations, and meteorological data; S2. In the initial stage, the hydrological boundary conditions, water quality parameters and spatial distribution data in different time periods are recorded and the basic data set is constructed; S3. Obtain the spatiotemporal distribution characteristics of rainfall intensity on runoff process through experimental data in basic data set; S4. Extract the dynamic variation pattern of pollutant concentration in time series based on experimental data; S5. Comprehensively analyze the spatiotemporal distribution characteristics and dynamic change patterns, design a multi-layer neural network model, and integrate hydrological, meteorological, and pollutant concentration data for training; S6. Iteratively optimize the multi-layer neural network model based on actual monitoring data to form a prediction model and algorithm; S7. Apply the optimized model to water quality prediction, generate the spatiotemporal distribution map of typical inflowing tributaries, and formulate corresponding scheduling strategies.

2. The water quality and quantity joint scheduling method based on a distributed hydrological model according to claim 1 is characterized in that: The step S2 comprises the following steps: S201. Construct a set of hydrodynamic equations, using the Navier-Stokes equations in the Cartesian coordinate system as the basis, and combine them with the material transport equations to describe the migration and transformation process of pollutants; S202, dividing the calculation area into grids, using quadrilateral grids to simulate river banks and triangular grids to simulate reservoir bays, and setting the grid sizes to suit the terrain characteristics; S203. Set boundary conditions and initial conditions, where the boundary conditions include inflow flow, outflow flow and pollutant concentration, and the initial conditions include initial water level velocity and pollutant distribution.

3. The water quality and quantity joint scheduling method based on a distributed hydrological model according to claim 2 is characterized in that: During the meshing process, the quadrilateral mesh size was set to 200 m × 200 m, suitable for river main channels and regular shoreline areas; the triangular mesh size was set to 5000 m², suitable for reservoir bays and irregular shoreline areas. The two meshes were seamlessly connected through shared nodes to ensure the overall continuity of the computational domain. The setting of the boundary conditions includes hydrodynamic boundary conditions and water quality boundary conditions. The hydrodynamic boundary conditions are controlled by upstream flow and downstream water level, and the water quality boundary conditions are controlled by upstream pollutant concentration and downstream pollutant flux; the initial conditions are generated by interpolation of historical data, including the initial water level velocity field and pollutant concentration field.

4. The water quality and quantity joint scheduling method based on a distributed hydrological model according to claim 1 is characterized in that: The dynamic change law decomposes the time series signal into multi-scale components through wavelet transform, and the main frequency amplitude is the amplitude of the low-frequency component in the decomposed signal, which reflects the main trend of pollutant concentration change over time.

5. The water quality and quantity joint scheduling method based on a distributed hydrological model according to claim 1 is characterized in that: The model training process uses an adaptive weight adjustment mechanism, which includes calculating the long-term mean and short-term fluctuations of the input data, extracting abnormal features, and determining whether the model weights need to be updated based on the Markov chain. When the abnormal feature coefficient is greater than the set threshold, the model weight adjustment is triggered; The adaptive weight adjustment mechanism also includes monitoring the changing trend of pollutant concentrations within a time window, and adjusting model parameters to adapt to new hydrological conditions when the changing trend deviates from the expected range.

6. The water quality and quantity joint scheduling method based on a distributed hydrological model according to claim 1 is characterized in that: The step S7 is specifically as follows: In the water quality prediction process, multi-source data fusion technology is used to integrate remote sensing image meteorological data and ground monitoring data into the model input layer; the prediction results are output in the form of spatiotemporal distribution maps, showing the changing trends of pollutant concentrations over time and space; During the scheduling strategy formulation process, high-risk areas and time periods are identified based on the prediction results, and targeted governance measures are proposed; Increase vegetation buffer zones in areas with high pollution loads and strengthen the operation of sewage interception facilities during high flow periods; Analyze the deviation between the corrected pollutant concentration value and the actual monitoring value, determine the type of anomaly, and select the scheduling strategy based on the anomaly type, which includes sudden changes in rainfall intensity, abnormal runoff fluctuations, accelerated pollutant diffusion, and noise interference; Among them, the sudden change of rainfall intensity is manifested as a rapid upward trend in pollutant concentration with changes in rainfall intensity, the abnormal fluctuation of runoff is manifested as nonlinear fluctuations in pollutant concentration with changes in runoff volume, the accelerated diffusion of pollutants is manifested as significant changes in pollutant concentration in a short period of time, and noise interference is manifested as the presence of random high-frequency components in pollutant concentration.

7. The water quality and quantity joint scheduling method based on a distributed hydrological model according to claim 6 is characterized in that: The step S7 comprises the following steps: S71. Identify high-risk areas and time periods Based on the spatiotemporal distribution map, continuous areas where pollutant concentrations exceeded the warning value by 20% were extracted, high-risk periods lasting more than three monitoring cycles were marked, and composite risk areas within the basin were identified that simultaneously met the conditions of runoff mutation rate > 15% and pollutant concentration gradient > 0.3 mg / (L·km); S72. Develop targeted scheduling strategies For high-risk areas: start key monitoring mode and increase data sampling frequency to 200% of the normal value; For the continuous period: deploy mobile emergency response equipment to the predicted pollution diffusion front one hour in advance; For complex risk areas: Simultaneously activate the three-pronged prevention and control system of pump stations, gates, and dosing devices; S73. Execute basic scheduling Level 1 prevention and control measures will be activated within 5km upstream of high-risk areas: recreational water intakes will be closed and navigation will be restricted; Implement pre-dispatching 2 hours before the high-risk period: reduce the reservoir water level to 90% of the design value; An emergency buffer zone is set up 10 km downstream of the complex risk area: adsorption barriers and online monitoring buoys are deployed.

8. The water quality and quantity joint scheduling method based on a distributed hydrological model according to claim 7 is characterized in that: The step S7 further includes the following steps: S74. Determine the type of exception Analyze the spectrum characteristics of monitoring data: Rainfall mutation type: 0.5-2Hz low frequency band energy accounts for more than 40%; Abnormal runoff type: continuous oscillation waveform appears in the 2-5Hz mid-frequency band; Pollution diffusion type: pulse cluster characteristics exist in the high frequency band of 5-10Hz; S75. Dynamically adjust strategy parameters When rainfall changes suddenly: for every 10mm / h increase in rainfall, the pump station power will be increased by 15%; When the runoff is abnormal: for every 5% deviation of the flow rate from the reference value, the gate opening is adjusted by 3%; When the pollution spreads: for every 0.1 mg / (L·h) increase in concentration gradient, shorten the dosing interval by 20%.

9. The water quality and quantity joint scheduling method based on a distributed hydrological model according to claim 8 is characterized in that: After completing the parameter adjustment in step S75, if at least one device parameter change rate is greater than 15%, immediately execute the following steps: S76, Strategy Optimization Control: When rainfall and runoff anomalies occur simultaneously, gate adjustment is activated 5 minutes after the pump station power increases to 120%. When runoff and pollution anomalies occur simultaneously, the automatic dosing device is activated immediately after the flow rate stabilizes within ±5%. When three anomalies occur simultaneously, the pump station is activated after 60% of the dosage of the first round of drug delivery is completed. S77, transition control starts. When the sudden rainfall change is lifted and the pump station power drops below 110%, the power reduction program is triggered; when the runoff fluctuation is stable within the range of ±5% for 30 minutes, the gate reset program is triggered; when the pollution concentration is lower than the warning value for 2 consecutive hours, the monitoring frequency reduction program is triggered.

10. The water quality and quantity joint scheduling method based on a distributed hydrological model according to claim 9 is characterized in that: After the transition control of step S77 is completed, if the equipment parameter fluctuation rate is less than 5% for 1 hour, the following steps are performed: S78. Scenario-based post-processing: After the reservoir entrance treatment is completed, the diversion channel will be kept open until the water turbidity drops by 50%. During the tributary confluence treatment, the dissolved oxygen content downstream will be tested after each purification cycle. After the dam front area treatment is completed, the ultrafiltration device will be maintained at 50% power for 12 hours. S79. Effect verification modeling: Rainfall treatment verification: calculate the energy consumption ratio of drainage per unit rainfall and optimize the pump station combination plan; runoff treatment verification: analyze the correlation between the number of gate adjustments and flow rate stability; pollution treatment verification: establish a dose-effect model of the amount of chemical dosage and the concentration reduction.

Images