Water quality prediction method and system based on parameter dynamic correction

Abstract

This application provides a water quality prediction method and system based on dynamic parameter correction, which can be applied to the field of eco-hydrological dynamics. It includes: acquiring hydrological and water quality data of the target water body; determining the assimilation value corresponding to the observed values ​​based on a pre-constructed high-dimensional water quality model and observed concentrations of various nitrogen forms at multiple times; determining the relative deviation based on the observed values ​​and assimilation values ​​at the same time; generating a stimulus signal and correcting the first-order reaction rate constant in the high-dimensional water quality model in response to the absolute value of the relative deviation exceeding a preset stimulus threshold; performing water balance calculations using water level and flow rate based on a pre-constructed zero-dimensional water quality model to correct nitrogen cycle dynamic parameters and obtain parameter variation trends including the first-order reaction rate constant; determining the time-varying parameter sequence of the high-dimensional water quality model based on the parameter variation trend; and inputting the time-varying parameter sequence into the high-dimensional water quality model to output the predicted concentrations of water quality indicators for the target water body within a preset time period.

Description

Technical Field

[0001] This application relates to the field of eco-hydrological dynamics technology, specifically to a water quality prediction method and system based on dynamic parameter correction. Background Technology

[0002] Water quality models can describe the migration and transformation patterns of substances such as nitrogen, phosphorus, dissolved oxygen, and organic matter in water bodies, and are an important tool for water environment management decisions. The accuracy of model predictions is highly dependent on the accuracy of its biogeochemical parameters.

[0003] In related technologies, ensemble Kalman filtering and its variants are widely used for the synchronous updating of parameters and states. However, these methods typically passively recursively extrapolate states in the prediction step, lacking an active response mechanism to parameter deviations, resulting in a lag in the model's response to dynamic changes in water quality. Furthermore, the correction of key biochemical reaction parameters often relies on the direct stimulation of state variables by observed data. When the sensitivity between some parameters and observed variables is weak, insufficient or even inability to update parameters can easily occur. Related techniques usually perform parameter correction directly in high-dimensional models, which is computationally expensive and makes it difficult to obtain systematic parameter change trends, limiting the application of water quality models in real-time or near-real-time prediction scenarios.

[0004] Therefore, how to achieve proactive, efficient, and systematic dynamic correction of water quality model parameters has become a pressing technical problem to be solved in the field of water quality prediction. Summary of the Invention

[0005] In view of the above problems, this application provides a water quality prediction method and system based on parameter dynamic correction, which can realize active, efficient and systematic dynamic correction of water quality model parameters.

[0006] According to a first aspect of this application, a water quality prediction method based on dynamic parameter correction is provided, comprising: acquiring hydrological data and water quality data of a target water body, wherein the hydrological data includes water level and flow rate, and the water quality data includes observed values ​​of nitrogen concentrations of various forms at multiple time points; determining an assimilation value corresponding to the observed values ​​based on a pre-constructed high-dimensional water quality model and the observed values ​​of nitrogen concentrations of various forms at the multiple time points, wherein the assimilation value is the optimal estimate of the true state of the water body obtained by fusing model predictions and observed values; determining a relative deviation based on the observed values ​​and the assimilation value at the same time point; and responding to a situation where the absolute value of the relative deviation exceeds a preset value. The system identifies a stimulation threshold, generates a stimulation signal, and corrects the first-order reaction rate constant in the high-dimensional water quality model based on the stimulation signal. It then performs water balance calculations using the water level and flow rate based on a pre-constructed zero-dimensional water quality model. Under the condition that the relative deviations of the current water quality components and the upstream reactant components have the same sign, it corrects the nitrogen cycle kinetic parameters to obtain the parameter variation trend, including the first-order reaction rate constant. Based on the parameter variation trend, it determines the time-varying parameter sequence of the high-dimensional water quality model. Finally, it inputs the time-varying parameter sequence into the high-dimensional water quality model and outputs the predicted concentration of water quality indicators for the target water body within a preset time period.

[0007] According to an embodiment of this application, determining the assimilation value corresponding to the observed values ​​based on a pre-constructed high-dimensional water quality model and the observed concentrations of various nitrogen forms at multiple times includes: superimposing random perturbations following a preset probability distribution onto the model state variables to generate multiple initial state vectors; inputting each initial state vector into the high-dimensional water quality model for forward prediction to obtain multiple predicted state vectors; updating the multiple predicted state vectors using the observed values ​​to obtain multiple analytical state vectors; and calculating the arithmetic mean of the multiple analytical state vectors as the assimilation value.

[0008] According to an embodiment of this application, generating a stimulus signal and correcting the first-order reaction rate constant in the high-dimensional water quality model based on the stimulus signal includes: generating a stimulus signal based on the magnitude of the relative deviation exceeding the stimulus threshold and the simulation time step, wherein the intensity of the stimulus signal is proportional to the magnitude and inversely proportional to the time step; generating an exponential decay factor based on the stimulus signal and a preset correction coefficient; and multiplying the original rate constant by the exponential decay factor to obtain the corrected first-order reaction rate constant.

[0009] According to an embodiment of this application, the water balance calculation based on a pre-constructed zero-dimensional water quality model, using the water level and the flow rate, includes: taking the difference between the total inflow and the total outflow as the rate of change of water storage over time, and taking the product of the water level and the water surface area as the water storage; and establishing concentration change equations for recalcitrant particulate organic nitrogen, readily degradable particulate organic nitrogen, dissolved organic nitrogen, ammonia nitrogen, and nitrate nitrogen based on nitrogen cycle kinetics, wherein the concentration change equations are used to describe the changes in the concentration of each form of nitrogen over time.

[0010] According to embodiments of this application, when the relative deviations of the current water quality components and the upstream reactant components are determined to be of the same sign, the correction of nitrogen cycle kinetic parameters includes: correcting the hydrolysis rate constants of recalcitrant particulate organic nitrogen and readily degradable particulate organic nitrogen based on their own relative deviations; correcting the mineralization rate constant of dissolved organic nitrogen when the relative deviations of recalcitrant particulate organic nitrogen, readily degradable particulate organic nitrogen, and dissolved organic nitrogen are of the same sign; correcting the nitrification rate constant when the relative deviations of dissolved organic nitrogen and ammonia nitrogen are of the same sign; correcting the denitrification rate constant when the relative deviations of ammonia nitrogen and nitrate nitrogen are of the same sign; and correcting the sediment-water interface exchange flux of dissolved organic nitrogen when the relative deviations of dissolved organic nitrogen, ammonia nitrogen, and nitrate nitrogen are of the same sign.

[0011] According to an embodiment of this application, determining the time-varying parameter sequence of the high-dimensional water quality model based on the parameter change trend includes: linearly scaling the initial calibration value of the high-dimensional water quality model based on the relative position of the current value of the parameter change trend within its preset change range and a preset scaling factor, thereby generating the time-varying parameter sequence.

[0012] According to an embodiment of this application, the high-dimensional water quality model is a two-dimensional or three-dimensional water quality model, including a hydrodynamic module and a water quality module; wherein, the hydrodynamic module is used to describe the changes in water level and flow velocity over time and space, and the water quality module is used to simulate the migration and transformation of various forms of nitrogen.

[0013] According to an embodiment of this application, the predicted concentrations of water quality indicators include nitrate nitrogen concentration, ammonia nitrogen concentration, total nitrogen concentration, and dissolved oxygen concentration. The step of inputting the time-varying parameter sequence into the high-dimensional water quality model and outputting the predicted concentrations of water quality indicators of the target water body within a preset time period includes: outputting at least one of nitrate nitrogen concentration, ammonia nitrogen concentration, total nitrogen concentration, and dissolved oxygen concentration; and outputting an error evaluation index between simulated values ​​and observed values, wherein the error evaluation index includes at least one of root mean square error, mean absolute percentage error, and correlation coefficient.

[0014] The second aspect of this application provides a water quality prediction system based on dynamic parameter correction, comprising: a data acquisition module for acquiring hydrological data and water quality data of a target water body, wherein the hydrological data includes water level and flow rate, and the water quality data includes observed values ​​of nitrogen concentrations of various forms at multiple time points; a first determination module for determining an assimilation value corresponding to the observed values ​​based on a pre-constructed high-dimensional water quality model and the observed values ​​of nitrogen concentrations of various forms at the multiple time points, wherein the assimilation value is the optimal estimate of the true state of the water body obtained by fusing model predictions and observed values; a second determination module for determining a relative deviation based on the observed values ​​and the assimilation value at the same time point; and a stimulus signal generation module for responding to an absolute value exceeding the relative deviation. A stimulation signal is generated by passing a preset stimulation threshold, and the first-order reaction rate constant in the high-dimensional water quality model is corrected based on the stimulation signal; a correction module is used to perform water balance calculations based on the pre-constructed zero-dimensional water quality model using the water level and the flow rate, and correct the nitrogen cycle kinetic parameters when the relative deviations between the current water quality components and the upstream reactant components are of the same sign, to obtain the parameter change trend including the first-order reaction rate constant; a third determination module is used to determine the time-varying parameter sequence of the high-dimensional water quality model based on the parameter change trend; and a water quality output module is used to input the time-varying parameter sequence into the high-dimensional water quality model and output the predicted concentration of water quality indicators of the target water body in a preset time period.

[0015] According to a third aspect of this application, an electronic device is provided, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.

[0016] According to a fourth aspect of this application, a computer-readable storage medium is also provided, on which a computer program or instructions are stored, wherein the computer program or instructions, when executed by a processor, implement the steps of the above-described method.

[0017] According to a fifth aspect of this application, a computer program product is also provided, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method.

[0018] Compared with related technologies, this application has the following advantages:

[0019] The water quality prediction method based on parameter dynamic correction provided in this application obtains a water body estimation benchmark that is closer to the actual state than the original data by acquiring hydrological data and water quality data and determining the assimilated value that integrates model predictions and observations. Then, the relative deviation is calculated based on the observations and assimilated values ​​at the same time. When the absolute value of this relative deviation exceeds a preset stimulus threshold, a stimulus signal is generated responsively, and the first-order reaction rate constant in the high-dimensional water quality model is actively corrected. This transforms the traditional passive parameter update into an active prediction step adjustment, significantly improving the model's timeliness in responding to dynamic changes in water quality and overcoming the problem of parameter update lag. The method utilizes a zero-dimensional model for efficient water balance calculations and corrects nitrogen cycle kinetic parameters only when the relative deviations between the current water quality components and upstream reactant components have the same sign. This ensures that the parameter adjustment direction is consistent with the error propagation direction of the nitrogen cycle reaction chain, avoiding erroneous parameter adjustments caused by reverse error or irrelevant interference. This makes the parameter change trend, including the first-order reaction rate constant, have clear physical meaning and solves the inherent problem of insufficient response of low-sensitivity parameters in traditional data assimilation. Based on the changing trend of this parameter, the time-varying parameter sequence of the high-dimensional water quality model is determined. The time-varying law of the parameters efficiently identified by the low-dimensional model is mapped to the high-dimensional model. This not only preserves the physical trend of parameter changes with time and operating conditions, but also significantly reduces the computational cost of directly correcting parameters in the high-dimensional model. Finally, the time-varying parameter sequence is input into the high-dimensional model to output the predicted concentration of water quality indicators. Thus, efficient, proactive, and systematic dynamic correction of water quality model parameters is achieved while ensuring prediction accuracy. Attached Figure Description

[0020] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0021] Figure 1 A technical roadmap of a water quality prediction method based on dynamic parameter correction according to an embodiment of this application is shown.

[0022] Figure 2 A flowchart of a water quality prediction method based on parameter dynamic correction according to an embodiment of this application is shown.

[0023] Figure 3 A schematic diagram of a parameter correction framework based on ensemble Kalman filtering according to an embodiment of this application is shown.

[0024] Figure 4 A schematic diagram of a theoretical frequency curve fitted from an empirical frequency according to an embodiment of this application is shown.

[0025] Figure 5 A schematic diagram showing the performance comparison of first-order reaction parameter correction according to an embodiment of this application is provided.

[0026] Figure 6 A schematic diagram of nitrogen cycle transformation in water according to an embodiment of this application is shown.

[0027] Figure 7 A schematic diagram illustrating the correction effect of nitrogen cycle kinetic parameters under different schemes according to embodiments of this application is shown.

[0028] Figure 8 A graph showing the trend of the time-varying denitrification rate constant according to an embodiment of this application is presented.

[0029] Figure 9 A schematic diagram showing the comparison between simulated and measured nitrate nitrogen values ​​at different monitoring points according to embodiments of this application is provided.

[0030] Figure 10 An architecture diagram of a water quality prediction system based on parameter dynamic correction according to an embodiment of this application is shown. Detailed Implementation

[0031] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0032] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0033] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0034] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0035] First, let's explain the technical terms that appear in the embodiments:

[0036] Assimilation value: refers to the posterior estimate obtained by fusing model predictions and actual observations through a data assimilation algorithm (such as ensemble Kalman filtering); the assimilation value is neither the original observation value nor a simple model prediction value, but the optimal estimate of the true state of the water body obtained after data assimilation.

[0037] Relative bias: refers to the degree of relative deviation between the observed value and the assimilated value at the same time. It is used to quantify the difference between the model assimilation result and the actual observation. The relative bias is calculated in a symmetrical form to eliminate the calculation bias caused by the deviation in a single direction.

[0038] Stimulus signal: A signal generated when the absolute value of the relative deviation is greater than or equal to a preset stimulation threshold, used to drive parameter adjustment; the intensity of the stimulation signal is related to the magnitude of the relative deviation exceeding the threshold and the simulation time step, and is used to actively trigger the correction of model parameters.

[0039] Stimulation threshold: A pre-set critical value for determining whether a stimulus signal should be generated; when the absolute value of the relative deviation is less than the stimulation threshold, the model assimilation value is considered to be consistent with the observed value, and no parameter adjustment is required; when the absolute value of the relative deviation is greater than or equal to the stimulation threshold, active parameter correction is triggered.

[0040] Exponential decay response function: A function used to update the first-order response rate decay function, which is in the form of an exponential decay function. This function generates an exponential decay factor based on the stimulus signal and the modified stride adjustment coefficient. The original rate constant is multiplied by this decay factor to achieve a smooth update of the parameters over time.

[0041] First-order reaction rate constant: A parameter describing the rate of change of concentration of a substance over time in a first-order kinetic reaction, used to characterize the reaction speed of biochemical processes such as pollutant degradation, nitrification, and denitrification.

[0042] Zero-dimensional water quality model: A water quality model that does not consider the spatial distribution differences of water bodies, treats the target water body as a completely mixed homogeneous unit, and only simulates the changes in the concentration of water components over time; the zero-dimensional model is used to quickly perform water balance calculations and identify trends in nitrogen cycle kinetic parameters.

[0043] High-dimensional water quality models: Two-dimensional or three-dimensional water quality models, as opposed to zero-dimensional models, include hydrodynamic modules and water quality modules, which can simulate the distribution and changes of water level, flow velocity, and water component concentration in space (horizontal and / or vertical directions) and time.

[0044] Current water quality components: refers to the target water quality components that are directly affected by the reaction process to be corrected. For example, ammonia nitrogen is the current component in nitrification, and nitrate nitrogen is the current component in denitrification.

[0045] Upstream reactant components: These refer to the preceding components in the nitrogen cycle reaction chain that can be transformed into the current water quality components. For example, if particulate organic nitrogen is hydrolyzed to generate dissolved organic nitrogen, then particulate organic nitrogen is an upstream reactant of dissolved organic nitrogen; if dissolved organic nitrogen is mineralized to generate ammonia nitrogen, then dissolved organic nitrogen is an upstream reactant of ammonia nitrogen; if ammonia nitrogen is nitrified to generate nitrate nitrogen, then ammonia nitrogen is an upstream reactant of nitrate nitrogen.

[0046] Same sign for relative deviation: This means that at the same moment, the relative deviation of the current water quality component and the relative deviation of the upstream reactant component are both positive or both are negative. Same sign for relative deviation indicates that the model overestimates or underestimates the current component and the upstream component in the same direction. In this case, the corresponding reaction parameters are considered to have a basis for correction. Conversely, if the signs are opposite, no parameter correction is performed.

[0047] Nitrogen cycle kinetic parameters: These are parameters used to describe the rates of biochemical reactions such as hydrolysis, mineralization, nitrification, and denitrification of organic nitrogen during the nitrogen cycle. They include, but are not limited to, the rate constants for particulate organic nitrogen hydrolysis, mineralization, nitrification, and denitrification, as well as the sediment-water interface exchange flux.

[0048] Time-varying parameter sequence: refers to the sequence of water quality model parameter values ​​that change over time; the time-varying parameter sequence is generated by mapping the parameter change trend obtained from zero-dimensional model correction to a high-dimensional water quality model, and can reflect the dynamic change characteristics of parameters under different times or different hydrological conditions.

[0049] To address the aforementioned technical problems, this application provides a water quality prediction method based on dynamic parameter correction, comprising: acquiring hydrological data and water quality data of a target water body, wherein the hydrological data includes water level and flow rate, and the water quality data includes observed values ​​of nitrogen concentrations of various forms at multiple time points; determining an assimilation value corresponding to the observed values ​​based on a pre-constructed high-dimensional water quality model and the observed values ​​of nitrogen concentrations of various forms at the multiple time points, wherein the assimilation value is the optimal estimate of the true state of the water body obtained by fusing model predictions and observed values; determining a relative deviation based on the observed values ​​and assimilation values ​​at the same time point; and responding to a situation where the absolute value of the relative deviation exceeds a predetermined value. A stimulation threshold is set, a stimulation signal is generated, and the first-order reaction rate constant in the high-dimensional water quality model is corrected based on the stimulation signal; based on the pre-constructed zero-dimensional water quality model, water balance calculation is performed using the water level and the flow rate, and when it is determined that the relative deviations between the current water quality components and the upstream reactant components have the same sign, the nitrogen cycle kinetic parameters are corrected to obtain the parameter variation trend including the first-order reaction rate constant; the time-varying parameter sequence of the high-dimensional water quality model is determined based on the parameter variation trend; and the time-varying parameter sequence is input into the high-dimensional water quality model to output the predicted concentration of water quality indicators of the target water body in a preset time period.

[0050] Figure 1 A technical roadmap of a water quality prediction method based on dynamic parameter correction according to an embodiment of this application is shown. Figure 1 As shown, the water quality prediction method based on dynamic parameter correction proposed in this application forms a complete closed-loop process according to the progressive logic of data-driven—active parameter correction—low-dimensional trend identification—high-dimensional mapping prediction.

[0051] In the data acquisition and preprocessing stage, hydrological data (water level, flow rate) and water quality data (observed values ​​of nitrogen concentrations in various forms) of the target water body are collected, and statistical tests, hydrological year classification, and trend analysis are performed on the data. Data preprocessing identifies water quality change characteristics under different hydrological conditions, providing a reliable input basis for subsequent parameter correction. Then, the first-order response parameter active correction stage begins. In the prediction step of the ensemble Kalman filter, the relative deviation between the observed and assimilated values ​​is calculated. When the deviation exceeds the stimulus threshold, a stimulus signal is generated, and the first-order response rate constant is actively updated using an exponentially decaying response function. This stage introduces a stimulus-response mechanism, transforming the traditional passive parameter update into an active prediction step adjustment, thereby overcoming the model response lag problem and significantly improving the timeliness and accuracy of parameter correction.

[0052] Next, the process moves to the zero-dimensional model integration and parameter calibration stage. A zero-dimensional water quality model incorporating multiple reactions in the nitrogen cycle is constructed. Water balance calculations are performed using water level and flow rate, and an error allocation mechanism is designed based on the reverse tracing of the nitrogen cycle reaction chain. Specifically, nitrogen cycle kinetic parameters are only calibrated when the relative deviations of the current water quality components and upstream reactants have the same sign. This stage leverages the high computational efficiency of the zero-dimensional model to quickly obtain the changing trends of key parameters, such as Labile Particulate Organic Nitrogen (LPON), Refractory Particulate Organic Nitrogen (RPON), and Dissolved Organic Nitrogen (DON). A same-sign calibration rule ensures that the parameter adjustment direction aligns with the error propagation logic, effectively addressing the problem of insufficient or no response from low-sensitivity parameters in traditional data assimilation. The process then moves to the high-dimensional model parameter mapping and simulation stage. The parameter changing trends obtained from the zero-dimensional model calibration are linearly scaled and mapped to the high-dimensional water quality model, generating a time-varying parameter sequence. The time-varying parameter sequence is input into a high-dimensional model, and hourly calculations are performed based on the hydrodynamic transport equation and the water quality response equation to output the predicted concentrations of water quality indicators (including nitrate nitrogen, ammonia nitrogen, total nitrogen, dissolved oxygen, etc.) of the target water body within a preset time period. This stage, through low-dimensional to high-dimensional parameter trend mapping, preserves the physical trends of parameter changes with time and operating conditions while avoiding the huge computational cost of directly performing ensemble Kalman filter parameter correction in the high-dimensional model, thus achieving a balance between accuracy and efficiency.

[0053] like Figure 1 As shown, the above four stages form a closed-loop iteration: the output of the high-dimensional model and its assimilated value are fed back to the active parameter correction and zero-dimensional model correction stages, achieving continuous dynamic optimization of the parameters. Figure 1 The complete technical approach shown in this application enables efficient, proactive, and systematic dynamic correction of water quality model parameters under different hydrological conditions, significantly improving the accuracy and computational efficiency of water quality prediction. It is applicable to water quality prediction and integrated management of various water bodies such as lakes, reservoirs, and rivers.

[0054] Figure 2 A flowchart of a water quality prediction method based on dynamic parameter correction according to an embodiment of this application is shown. Figure 2 As shown, the water quality prediction method based on dynamic parameter correction according to the embodiments of this application may include steps S210 to S270. This application provides a water quality prediction method based on dynamic parameter correction. This method achieves active parameter correction by introducing a stimulus-response mechanism and combines a parameter mapping framework between zero-dimensional and high-dimensional models to achieve efficient and systematic dynamic parameter identification.

[0055] In step S210, hydrological data and water quality data of the target water body are acquired. The hydrological data includes water level and flow rate, and the water quality data includes observed values ​​of nitrogen concentrations of various forms at multiple times.

[0056] In one example, hydrological and water quality data for the target water body are acquired. The hydrological data includes water level and flow rate, while the water quality data includes observed concentrations of various nitrogen forms at multiple time points. This data can be obtained through on-site monitoring stations, such as water level sensors, flow sensors, and water quality analysis instruments. By acquiring this multi-source data, fundamental data support can be provided for subsequent parameter calibration and model prediction, thereby ensuring the data-driven nature of the calibration process.

[0057] In step S220, based on the pre-constructed high-dimensional water quality model and the observed values ​​of nitrogen concentrations at the multiple time points, an assimilation value corresponding to the observed value is determined. The assimilation value is the optimal estimate of the true state of the water body obtained by fusing the model prediction value and the observed value.

[0058] In one example, based on a pre-built high-dimensional water quality model and observed concentrations of various nitrogen forms at multiple time points, an assimilation value corresponding to the observed values ​​is determined. This assimilation value is the optimal estimate of the true state of the water body obtained by fusing model predictions and observed values. Specifically, the assimilation value is neither the original observed value nor simply the model prediction value, but rather a posterior estimate obtained by fusing the two through a data assimilation algorithm. By determining the assimilation value, an estimate closer to the true state of the water body can be obtained than the original observed values ​​and model predictions, thus providing a more reliable benchmark for subsequent relative deviation calculations and reducing the interference of observation noise and model errors on the parameter correction direction.

[0059] In step S230, the relative bias is determined based on the observations and assimilation values ​​at the same time.

[0060] In one example, the relative bias is determined based on the observed and assimilated values ​​at the same time. This relative bias quantifies the degree of deviation between the model's assimilation results and the actual observations. By employing a symmetrical formula for calculating the relative bias, computational bias caused by unidirectional bias can be avoided, making the bias index comparable across different concentration levels, thus providing a unified triggering benchmark for parameter correction.

[0061] In step S240, in response to the absolute value of the relative deviation exceeding a preset stimulation threshold, a stimulation signal is generated and the first-order reaction rate constant in the high-dimensional water quality model is corrected based on the stimulation signal.

[0062] In one example, in response to the absolute value of the relative deviation exceeding a preset stimulus threshold, a stimulus signal is generated, and the first-order reaction rate constant in the high-dimensional water quality model is corrected based on the stimulus signal. When the relative deviation is small, the model state is considered to be consistent with the observation, and no parameter adjustment is required; when the relative deviation exceeds the threshold, it indicates that there is a significant deviation in the model. At this time, a stimulus signal related to the deviation amplitude is generated, and this signal is used to drive the update of the first-order reaction rate constant. Through this stimulus-response mechanism, model deviations can be actively identified and parameters adjusted in advance during the prediction step. Compared with the traditional ensemble Kalman filter method that relies solely on the analysis step to passively update parameters, this significantly improves the timeliness and proactivity of parameter response, thereby reducing the predictive lag of the model.

[0063] In step S250, based on the pre-built zero-dimensional water quality model, water balance calculation is performed using the water level and the flow rate. If the relative deviations between the current water quality components and the upstream reactant components are of the same sign, the nitrogen cycle kinetic parameters are corrected to obtain the parameter change trend including the first-order reaction rate constant.

[0064] In one example, based on a pre-built zero-dimensional water quality model, water balance calculations are performed using the water level and flow rate. When the relative deviations between the current water quality components and the upstream reactant components are determined to have the same sign, nitrogen cycle kinetic parameters are corrected to obtain the parameter variation trends, including the first-order reaction rate constant. The zero-dimensional model does not consider spatial heterogeneity, has high computational efficiency, and can quickly reflect the average changes in water quality components over time. By introducing the reverse tracing relationship of the nitrogen cycle reaction chain, parameter correction is only performed when the relative deviation direction between the current water quality components and the upstream reactants is consistent, avoiding erroneous parameter adjustments caused by inconsistent error propagation directions. In this way, continuous variation trends of multiple key nitrogen cycle kinetic parameters over time can be obtained, and these trends have clear physical meaning and error propagation logic support, thus solving the problem of insufficient or no response of some parameters in traditional data assimilation.

[0065] In step S260, the time-varying parameter sequence of the high-dimensional water quality model is determined based on the parameter change trend.

[0066] In one example, the time-varying parameter sequence of the high-dimensional water quality model is determined based on the parameter variation trend. Mapping the parameter variation trend obtained from the zero-dimensional model to the high-dimensional model allows key parameters in the high-dimensional model to dynamically change with time and operating conditions. By employing trend mapping, the parameter variation patterns obtained through efficient correction of the low-dimensional model can be transferred to the high-dimensional model, avoiding the enormous computational burden of directly performing parameter correction in the high-dimensional model, while ensuring the physical rationality of the time-varying parameter characteristics.

[0067] In step S270, the time-varying parameter sequence is input into the high-dimensional water quality model, and the predicted concentration of water quality indicators of the target water body in a preset time period is output.

[0068] In one example, the time-varying parameter sequence is input into the high-dimensional water quality model, which outputs the predicted concentrations of water quality indicators for the target water body within a preset time period. After obtaining the time-varying parameter sequence, the high-dimensional water quality model calculates the predicted water quality concentrations at each monitoring point hourly based on the hydrodynamic transport equation and the water quality response equation. By outputting the predicted concentration sequence, quantitative decision-making basis can be provided for water environment management and water quality early warning, realizing a complete closed loop from data acquisition and parameter correction to prediction output.

[0069] The water quality prediction method based on parameter dynamic correction provided in this application obtains a water body estimation benchmark that is closer to the actual state than the original data by acquiring hydrological data and water quality data and determining the assimilated value that integrates model predictions and observations. Then, the relative deviation is calculated based on the observations and assimilated values ​​at the same time. When the absolute value of this relative deviation exceeds a preset stimulus threshold, a stimulus signal is generated responsively, and the first-order reaction rate constant in the high-dimensional water quality model is actively corrected. This transforms the traditional passive parameter update into an active prediction step adjustment, significantly improving the model's timeliness in responding to dynamic changes in water quality and overcoming the problem of parameter update lag. The method utilizes a zero-dimensional model for efficient water balance calculations and corrects nitrogen cycle kinetic parameters only when the relative deviations between the current water quality components and upstream reactant components have the same sign. This ensures that the parameter adjustment direction is consistent with the error propagation direction of the nitrogen cycle reaction chain, avoiding erroneous parameter adjustments caused by reverse error or irrelevant interference. This makes the parameter change trend, including the first-order reaction rate constant, have clear physical meaning and solves the inherent problem of insufficient response of low-sensitivity parameters in traditional data assimilation. Based on the changing trend of this parameter, the time-varying parameter sequence of the high-dimensional water quality model is determined. The time-varying law of the parameters efficiently identified by the low-dimensional model is mapped to the high-dimensional model. This not only preserves the physical trend of parameter changes with time and operating conditions, but also significantly reduces the computational cost of directly correcting parameters in the high-dimensional model. Finally, the time-varying parameter sequence is input into the high-dimensional model to output the predicted concentration of water quality indicators. Thus, efficient, proactive, and systematic dynamic correction of water quality model parameters is achieved while ensuring prediction accuracy.

[0070] Figure 3 A schematic diagram of a parameter correction framework based on ensemble Kalman filtering according to an embodiment of this application is shown. This embodiment further describes a specific implementation method for determining the assimilation value corresponding to the observation, which is based on the ensemble Kalman filtering framework, such as... Figure 3 As shown.

[0071] According to an embodiment of this application, determining the assimilation value corresponding to the observed values ​​based on a pre-constructed high-dimensional water quality model and the observed concentrations of various nitrogen forms at multiple times includes: superimposing random perturbations following a preset probability distribution onto the model state variables to generate multiple initial state vectors; inputting each initial state vector into the high-dimensional water quality model for forward prediction to obtain multiple predicted state vectors; updating the multiple predicted state vectors using the observed values ​​to obtain multiple analytical state vectors; and calculating the arithmetic mean of the multiple analytical state vectors as the assimilation value.

[0072] First, perform the initialization steps. Set the system's initial state vector. This vector contains pollutant concentration, first-order reaction rate constant, and auxiliary variables (such as water temperature, dissolved oxygen, and other auxiliary states that affect the reaction rate). An initial error covariance matrix is ​​set. This is used to characterize the uncertainty of the initial state. To characterize the uncertainty of the model prediction, a random perturbation following a preset probability distribution (usually Gaussian white noise) is superimposed on the initial state vector, generating multiple initial state vectors, each representing a possible true state of the system. By employing ensemble perturbation, the state estimation problem of nonlinear water quality systems can be handled without making linearization assumptions about the model.

[0073] Next, the time loop begins. For time i (i=1, 2, ...), a prediction step is executed. The analysis set obtained at the previous time (time i-1) is... Each analytical state vector is input into the high-dimensional water quality model, and under the influence of boundary conditions (including external drivers such as river flow, river water quality, reservoir temperature, reservoir dissolved oxygen concentration, and sediment release load), a time step is recursively pushed forward. This yields multiple predicted state vectors, forming the prediction set at time i. . This represents the set of posterior state estimates obtained after data assimilation at time i-1, containing historical observation information. Each predicted state vector reflects the model's prior estimate of the system state without considering the current observation information. By generating the prediction set, the uncertainty brought about by the propagation of model dynamics can be quantified, providing prior distribution information for subsequent data fusion.

[0074] Then, perform the analysis step. Obtain the observation vector at time i. (Including measured data such as water level and concentrations of various nitrogen forms) and their observation error covariance matrix Using observation vectors and observation error covariance matrix For the prediction set Each predicted state vector in the dataset is updated using Kalman filtering. The update process calculates the Kalman gain based on the deviation between the predicted state and the observation, the prediction error covariance, and the observation error covariance, and adjusts each predicted state vector accordingly to obtain the analysis set at time i. This set of analyses will serve as the input for the prediction step at the next time step (time i+1), forming a cyclical iteration.

[0075] Calculate the assimilation value at time i. For the analysis set... The arithmetic mean of all analyzed state vectors is calculated as the assimilated value at that moment. The assimilated value is neither the original observation nor a simple model prediction, but rather the optimal estimate of the true state of the water body obtained after data assimilation. Simultaneously, the statistical distribution of the analysis set can be used to estimate the uncertainty of the assimilated value (e.g., calculating the set variance as the estimation error). By averaging multiple analyzed state vectors, random fluctuations in individual estimates can be suppressed, resulting in a stable and reliable assimilated value. This provides an accurate benchmark for subsequent relative deviation calculations, avoiding parameter miscorrection due to accidental errors in a single estimate.

[0076] According to an embodiment of this application, generating a stimulus signal and correcting the first-order reaction rate constant in the high-dimensional water quality model based on the stimulus signal includes: generating a stimulus signal based on the magnitude of the relative deviation exceeding the stimulus threshold and the simulation time step, wherein the intensity of the stimulus signal is proportional to the magnitude and inversely proportional to the time step; generating an exponential decay factor based on the stimulus signal and a preset correction coefficient; and multiplying the original rate constant by the exponential decay factor to obtain the corrected first-order reaction rate constant.

[0077] In one example, the first-order reaction kinetic equation is calculated as shown in Equation (1):

[0078] (1).

[0079] Where C represents the pollutant concentration. Let t be the reaction rate constant and t be time. It represents the change in pollutant concentration per unit time.

[0080] The stimulus-response strategy is implemented through the following steps: First, the observation at time i-1... and assimilation value relative deviation The calculation is shown in formula (2):

[0081] (2).

[0082] in, and These are the observed value and the assimilated value at time i-1, respectively. This represents the relative deviation at time i-1.

[0083] when At that time, a stimulus signal is generated. ,when hour, =0. The calculation is shown in formula (3):

[0084] (3).

[0085] in, For time step, The natural logarithm is represented by the intensity of the stimulus signal, which is directly proportional to the deviation amplitude and inversely proportional to the time step ΔT, thus ensuring numerical stability. Then, based on the stimulus signal and a preset correction coefficient, an exponential decay factor is generated. The original rate constant is multiplied by this exponential decay factor to obtain the corrected first-order response rate constant.

[0086] The model uses an exponentially decaying response function to update the first-order reaction rate constant, thus completing the active parameter correction. The calculation is shown in formula (4):

[0087] (4).

[0088] in, Indicates the time interval The response value at that location, To correct the stride adjustment coefficient, The exponential decay factor, also known as the stimulus signal, gradually weakens the influence of the stimulus signal on parameter adjustment over time. It controls the decay rate of parameter adjustment. By employing the above-mentioned stimulus signal generation and exponential decay response methods, the magnitude of parameter adjustment can be reasonably matched with the degree of model deviation, while achieving smooth parameter updates, avoiding model oscillations caused by abrupt parameter changes, and improving the stability and convergence of parameter correction.

[0089] According to an embodiment of this application, the water balance calculation based on a pre-constructed zero-dimensional water quality model, using the water level and the flow rate, includes: taking the difference between the total inflow and the total outflow as the rate of change of water storage over time, and taking the product of the water level and the water surface area as the water storage; and establishing concentration change equations for recalcitrant particulate organic nitrogen, readily degradable particulate organic nitrogen, dissolved organic nitrogen, ammonia nitrogen, and nitrate nitrogen based on nitrogen cycle kinetics, wherein the concentration change equations are used to describe the changes in the concentration of each form of nitrogen over time.

[0090] In one example, the water balance calculation, based on the water balance relationship, calculates the total inflow flow. Total outflow The difference is taken as the rate of change of water storage over time. Specifically, the water quantity calculation of the zero-dimensional water quality model is shown in formula (5):

[0091] (5).

[0092] in, It represents the change in the volume of a container per unit time.

[0093] The product of water level and water surface area can be used as the water storage capacity, or the relationship between water storage capacity and water level can be obtained by fitting a water level-reservoir capacity curve, for example, by using a quadratic polynomial as shown in formula (6):

[0094] (6).

[0095] in, Let be the water storage capacity of the reservoir, and a, b, and c be the binomial coefficients obtained by fitting the water level-reservoir capacity curve. For the reservoir water level, This represents the square of the reservoir's water level.

[0096] In one example, the nitrogen cycle kinetic equations consider processes such as organic nitrogen hydrolysis, mineralization, nitrification, and denitrification, as shown in equation set (7):

[0097] (7).

[0098] in, The concentration of recalcitrant particulate matter. For the concentration of easily degradable particulate matter, This refers to the concentration of dissolved organic nitrogen. It is the water storage capacity of the reservoir. , and These are the hydrolysis rates (d) of recalcitrant and readily degradable particulate organic nitrogen, respectively. -1 ), The mineralization rate of dissolved organic nitrogen (d) -1 ), and These are the first-order nitrification rate constant and the denitrification rate constant (d). -1 ), and These are the 0th order nitrification rate and the denitrification rate (gNm). -3 d -1 ), , , , and These are the external loads (gNd) of recalcitrant particulate organic nitrogen, readily degradable particulate organic nitrogen, soluble organic nitrogen, ammonia nitrogen, and nitrate nitrogen. -1 ), and These are the average sedimentation rates (md) of recalcitrant and readily degradable particulate organic nitrogen, respectively. -1 ), , and These are the sediment-water interface exchange fluxes (gNm³) for dissolved organic nitrogen, ammonia nitrogen, and nitrate nitrogen, respectively. -2 d -1 ), This represents the average water depth.

[0099] The above equations describe the biochemical processes of various forms of nitrogen in water, such as hydrolysis, mineralization, nitrification, and denitrification, as well as physical processes such as external load input, sedimentation, and sediment-water interface exchange. By employing the above zero-dimensional model, nitrogen transformation can be rapidly simulated under average conditions, providing an efficient computational platform for parameter trend identification, thereby reducing overall computational costs.

[0100] According to embodiments of this application, when the relative deviations of the current water quality components and the upstream reactant components are determined to be of the same sign, the correction of nitrogen cycle kinetic parameters includes: correcting the hydrolysis rate constants of recalcitrant particulate organic nitrogen and readily degradable particulate organic nitrogen based on their own relative deviations; correcting the mineralization rate constant of dissolved organic nitrogen when the relative deviations of recalcitrant particulate organic nitrogen, readily degradable particulate organic nitrogen, and dissolved organic nitrogen are of the same sign; correcting the nitrification rate constant when the relative deviations of dissolved organic nitrogen and ammonia nitrogen are of the same sign; correcting the denitrification rate constant when the relative deviations of ammonia nitrogen and nitrate nitrogen are of the same sign; and correcting the sediment-water interface exchange flux of dissolved organic nitrogen when the relative deviations of dissolved organic nitrogen, ammonia nitrogen, and nitrate nitrogen are of the same sign.

[0101] In one example, based on reverse tracing of the nitrogen cycle reaction chain, parameter correction is performed only when the relative error between the current substance and the upstream reactant has the same sign. Specifically, this includes:

[0102] Hydrolysis rate constant of recalcitrant particulate organic nitrogen Hydrolysis rate constant of easily degradable particulate organic nitrogen Correction is made based on its own relative deviation; the mineralization rate constant of dissolved organic nitrogen is adjusted accordingly. Correction is applied when the relative deviations of recalcitrant particulate organic nitrogen, readily degradable particulate organic nitrogen, and soluble organic nitrogen have the same sign; this applies to the nitrification rate constant. Correction is applied when the relative deviations of dissolved organic nitrogen and ammonia nitrogen have the same sign; for the denitrification rate constant... Correction is applied when the relative deviations of ammonia nitrogen and nitrate nitrogen have the same sign; for sediment-water interface exchange flux of dissolved organic nitrogen. Correction is performed when the relative deviations of dissolved organic nitrogen, ammonia nitrogen, and nitrate nitrogen have the same sign.

[0103] By adopting the above-mentioned error allocation rule based on reverse tracing of the nitrogen cycle reaction chain, it is ensured that the parameter adjustment direction is consistent with the error propagation direction. The corresponding parameter is only triggered when the error is propagated in the same direction along the reaction chain, avoiding incorrect parameter adjustment caused by error reversal or irrelevant interference. This improves the accuracy and physical interpretability of parameter correction and effectively solves the problem of no response or erroneous response of parameters in traditional data assimilation.

[0104] According to an embodiment of this application, determining the time-varying parameter sequence of the high-dimensional water quality model based on the parameter change trend includes: linearly scaling the initial calibration value of the high-dimensional water quality model based on the relative position of the current value of the parameter change trend within its preset change range and a preset scaling factor, thereby generating the time-varying parameter sequence.

[0105] In one example, based on the relative position of the current value of the parameter change trend within its preset range of change and a preset scaling factor, the initial calibration value of the high-dimensional water quality model is linearly scaled to generate the time-varying parameter sequence. The calculation formula is shown in formula (8):

[0106] (8).

[0107] Where M is the magnification factor, For time-varying parameters, For the initial calibration value, To correct the parameters, and The maximum and minimum values ​​of the parameters are corrected separately. The time-varying parameters are then input into the high-dimensional model to complete the water quality simulation and prediction. This formula maps the normalized parameter variation trend obtained from the zero-dimensional model to the parameter value range of the high-dimensional model. By adopting this linear scaling mapping method, the time-varying patterns of parameters efficiently identified by the low-dimensional model can be transferred to the high-dimensional model. Simultaneously, the scaling factor M adjusts the magnitude of parameter variation, ensuring that the parameters of the high-dimensional model retain both the physical variation trend and adaptability to differences between the high-dimensional and low-dimensional models in terms of spatial resolution and parameter sensitivity. This significantly reduces computational costs while ensuring the physical rationality and spatiotemporal continuity of the time-varying parameters.

[0108] According to an embodiment of this application, the high-dimensional water quality model is a two-dimensional or three-dimensional water quality model, including a hydrodynamic module and a water quality module; wherein, the hydrodynamic module is used to describe the changes in water level and flow velocity over time and space, and the water quality module is used to simulate the migration and transformation of various forms of nitrogen.

[0109] In one example, the high-dimensional water quality model is a two-dimensional or three-dimensional water quality model, including a hydrodynamic module and a water quality module. The hydrodynamic module is used to describe the changes in water level and flow velocity over time and space, and its governing equations include continuity equations and momentum equations. The governing equations of the hydrodynamic module are shown in the following set of equations (9):

[0110] (9).

[0111] Where u and v are the flow velocities along the x and y directions, respectively; Let Q represent the water depth and Q represent the external source term. ρ is the vertical eddy viscosity coefficient, f is the Coriolis force coefficient, and ρ0 is the water density. and Indicates the horizontal pressure gradient; and The unbalanced term representing the horizontal Reynolds stress; and This represents external momentum sources or sinks (e.g., external forces generated by hydraulic structures, water intake, drainage, wave stress, etc.). The water quality module simulates the migration and transformation of various nitrogen forms. After obtaining the flow field provided by the hydrodynamic module, it calculates the concentration of each water quality component by combining it with biochemical reaction kinetic equations. By adopting the above-mentioned modular high-dimensional water quality model structure, it can realistically reflect the hydrodynamic processes and water quality evolution of the target water body, and has the ability to predict spatial distribution. After inputting the time-varying parameter sequence into the water quality module, the model can dynamically update the chemical reaction rate while maintaining accurate simulation of the hydrodynamic field, thereby improving the spatial distribution accuracy and temporal dynamic response capability of water quality prediction.

[0112] According to an embodiment of this application, the predicted concentrations of water quality indicators include nitrate nitrogen concentration, ammonia nitrogen concentration, total nitrogen concentration, and dissolved oxygen concentration. The step of inputting the time-varying parameter sequence into the high-dimensional water quality model and outputting the predicted concentrations of water quality indicators of the target water body within a preset time period includes: outputting at least one of nitrate nitrogen concentration, ammonia nitrogen concentration, total nitrogen concentration, and dissolved oxygen concentration; and outputting an error evaluation index between simulated values ​​and observed values, wherein the error evaluation index includes at least one of root mean square error, mean absolute percentage error, and correlation coefficient.

[0113] In one example, the predicted concentrations of water quality indicators include at least one of nitrate nitrogen concentration, ammonia nitrogen concentration, total nitrogen concentration, and dissolved oxygen concentration. After inputting the time-varying parameter sequence into the high-dimensional water quality model, the predicted concentration sequence is output. Simultaneously, to quantitatively assess the reliability of the prediction results, an error evaluation index between the simulated and observed values ​​is also output. This error evaluation index includes at least one of root mean square error, mean absolute percentage error, and correlation coefficient. By outputting these error evaluation indices, users can be provided with a quantitative basis for the model's predictive performance, facilitating comparisons between different parameter correction schemes and providing feedback information for further model optimization, thus forming a complete closed loop from parameter correction to prediction output and performance evaluation.

[0114] To further demonstrate the water quality prediction method based on dynamic parameter correction provided in this application, this embodiment takes a reservoir as the research object and specifically implements and verifies the above-mentioned water quality prediction method based on dynamic parameter correction. The reservoir is the largest valley-type reservoir in a certain region, with a maximum storage capacity of 43.75 × 10⁻⁶. 8 m 3 Water surface area 188 km² 2 It supplies approximately 60% of the city's domestic and industrial water. The study period is from January 1, 2015 to December 31, 2020, covering typical operating conditions such as water level rise and stable operation. The following section combines... Figures 4 to 9 The specific implementation steps of this embodiment will be described in detail.

[0115] During the data collection phase, the hydrological and water quality data used in this experiment were obtained from a municipal hydrological station and a reservoir management bureau. Hydrological data included daily monitoring data of the flow rates of River A, River B, and the C diversion canal, as well as the reservoir water level, from 1990 to 2020. Water quality data came from monthly monitoring data at six monitoring stations in the reservoir area and at the inlet and outlet, covering indicators such as dissolved oxygen, permanganate index, total nitrogen, ammonia nitrogen, nitrate nitrogen, and total phosphorus. Sampling and testing all complied with national hydrological and water quality monitoring standards. Meteorological data came from meteorological station 54511, including hourly wind speed and direction data, used to drive the hydrodynamic model.

[0116] During the data preprocessing stage, Figure 4 A schematic diagram of a theoretical frequency curve fitted from an empirical frequency, according to an embodiment of this application, is shown. Figure 4 As shown, the empirical frequency was calculated using the Weibull formula, combined with Pearson Type III theoretical curve fitting, with the flow rates corresponding to 25% and 75% of the theoretical frequency (4.70 m³ / s). 3 / s, 10.53 m 3The data is categorized into dry years, normal years, and wet years based on rainfall concentration; and into flood season (June-September) and non-flood season (January-May, October-December) based on the concentrated rainfall period. Shapiro-Wilk test for normality and Levene's test for homogeneity of variance were used. The Mann-Whitney U test was used to compare seasonal water quality differences, and the Kruskal-Wallis test was used to analyze annual hydrological differences. Outliers were removed at 1.5 times the interquartile range. As a fundamental statistical parameter in hydrological frequency analysis, it represents the arithmetic mean of the flow rate within the target time period, reflecting the average inflow level of the watershed over a multi-year timescale. The dispersion coefficient is defined as the ratio of the root mean square error to the multi-year average flow, reflecting the relative dispersion of annual runoff over a multi-year time series. This is the deviation coefficient, used to describe the degree of asymmetry in the distribution of annual runoff series on both sides of the mean. In actual hydrological analysis, Usually with Maintaining a certain ratio relationship ensures a good fit between the theoretical frequency curve and the empirical frequencies from measured data, for example... The inflow load is calculated based on the tributary flow and water quality concentration, as shown in formula (10):

[0117] (10).

[0118] in, and A collection consisting of months and rivers flowing into the reservoir. and These represent water quality indicators and the year, respectively. When Avoid using the values ​​{1, 2, … , 12}, {6, 7, 8}, and {1, …, 5, 10, …, 12}. They are Annual water quality indicators, flood season and non-flood season The corresponding inbound load. and These are the rivers flowing into the reservoir. exist Year Monthly water quality indicators Average concentration (mg / L) and average flow rate (m³) 3 / s).

[0119] According to the hydrological frequency analysis results, 2016 was a high-water year, 2019 was a low-water year, and the remaining related years were normal-water years. This classification is not used as a model calculation parameter, but rather to characterize the different hydrological conditions covered in this embodiment, in order to verify the applicability and stability of the method under different hydrological backgrounds. Figure 4The theoretical frequency curve fits the empirical data well, indicating that the Pearson Type III distribution can effectively describe the statistical characteristics of the reservoir's inflow.

[0120] In the active calibration phase of the first-order reaction parameters, the effect of parameter calibration was verified using denitrification as the target first-order reaction. The basic parameters were set as follows: initial nitrate nitrogen concentration 1.00 mg / L, and a baseline denitrification rate constant k at 20℃. deni20 =0038d -1 The stimulation threshold α = 0.05, and the corrected stride adjustment coefficient A = 0.5.

[0121] The formula for calculating the relative error between the observed value and the assimilated value is shown in (11):

[0122] (11).

[0123] in, and These are the observed value and the assimilated value at time i-1, respectively. Let be the relative error at time i-1.

[0124] when At that time, a stimulus signal is generated. The specific calculation process is shown in formula (12):

[0125] (12).

[0126] in, For time step; when hour, =0.

[0127] The first-order reaction rate constant is updated by the exponentially decaying response function to complete the active parameter correction. The calculation process is shown in formula (13):

[0128] (13).

[0129] in, To correct the stride adjustment coefficient.

[0130] Figure 5 A schematic diagram comparing the performance of first-order reaction parameter correction according to embodiments of this application is shown. Figure 5 As shown, where, Figure 5 (a) shows a comparison of the true value, simulated value, passive correction value of traditional ensemble Kalman filter and active correction value of stimulus-response in this application for nitrate nitrogen concentration, and shows the differences of different methods in the concentration change process through local magnification; Figure 5(b) shows a comparison of the real sequence, simulated sequence, passively corrected sequence, and actively corrected sequence of the denitrification rate constant k; Figure 5 (c) and Figure 5 (d) in the figure shows the variance of the estimation error of nitrate nitrogen concentration and the variance of the estimation error of denitrification rate constant, respectively, which are used to characterize the differences in uncertainty of state variables and parameter variables under different correction methods; Figure 5 (e) in the figure shows the external stimulus signal S(t) triggered by the relative deviation; Figure 5 Figure (f) shows the model response R(t), which characterizes the response of the parameter correction process to the stimulus signal. Experimental results show that the correlation coefficient between the parameter sequence after traditional ensemble Kalman filtering and the true value is only 0.35, with significant phase lag and amplitude attenuation. However, after active correction using the stimulus-response strategy proposed in this application, the correlation coefficient between the parameter sequence and the true value increases to 0.96, and the mean absolute percentage error decreases from 0.87% to 0.40%. Under the conditions of a monitoring frequency of once every 3 days and a relative error of ≤10%, the robustness is good. These results demonstrate that the stimulus-response strategy effectively overcomes the problems of parameter update lag and insufficient response in traditional methods.

[0131] Zero-dimensional model integration and parameter calibration stage

[0132] The water quantity calculation for the zero-dimensional water quality model is shown in formulas (14) and (15):

[0133] (14).

[0134] (15).

[0135] Where V is the water storage capacity of the reservoir (m³) 3 ), and These are the total inflow and total outflow (m³) 3 / s), WL is the reservoir water level (m), R 2 =0.9999.

[0136] Figure 6 A schematic diagram of nitrogen cycle transformation in water provided in an embodiment of this application is shown. Figure 6 As shown, recalcitrant particulate organic nitrogen and easily degradable particulate organic nitrogen At the hydrolysis rate constant and Under the action of [unclear], it is converted into dissolved organic nitrogen. Dissolved organic nitrogen In the mineralization rate constant Under the action of [unclear], it is converted into ammonia nitrogen ammonia nitrogen It is converted into nitrate nitrogen through nitration. This process is determined by the first-order nitration rate constant. and zero-order nitrification rate Characterization; Nitrate nitrogen It is converted into nitrogen gas through denitrification. And escape from the water body; this process is driven by the first-order denitrification rate constant. and zero-order denitrification rate Characterization. Recalcitrant particulate organic nitrogen and readily degradable particulate organic nitrogen can also be characterized through sedimentation processes. and Sedimentation occurs and water enters the sediment; sediment-water interface exchange flux. , and These are used to characterize the exchange of dissolved organic nitrogen, ammonia nitrogen, and nitrate nitrogen between sediments and overlying water, respectively. It should be noted that the external inflow load is represented by the component loading terms in the zero-dimensional water quality kinetic equation. Figure 6 It is mainly used to illustrate the nitrogen cycle reaction chain inside the water body and the sediment-water interface exchange relationship. The external load arrows are not drawn separately.

[0137] The nitrogen cycle kinetic equations take into account processes such as organic nitrogen hydrolysis, mineralization, nitrification, and denitrification. The equation set (16) is shown below:

[0138] (16).

[0139] The meanings of the symbols are the same as in the above embodiments, and will not be repeated here.

[0140] Figure 7 This diagram illustrates the correction effects of nitrogen cycle kinetic parameters under different schemes according to embodiments of this application. In the diagram, T represents the true parameter sequence, and S1, S2, S3, and S4 represent the parameter sequences obtained under four parameter correction schemes. Specifically, S1 represents the baseline scheme without dynamic parameter correction, S2 represents the scheme using fixed calibration parameters, S3 represents the direct parameter correction scheme without introducing the same sign constraint of the nitrogen cycle reaction chain, and S4 represents the complete scheme combining stimulus-response active correction, zero-dimensional water quality model parameter trend identification, and high-dimensional model time-varying parameter mapping as described in this application. In the diagram, ρ1, ρ2, ρ3, and ρ4 represent the correlation coefficients between S1, S2, S3, and S4 and the parameter sequence T, respectively, used to evaluate the ability of different schemes to capture the time-varying trend of parameters.

[0141] Figure 7 (a) to Figure 7(f) in the figure corresponds to the correction results of different nitrogen cycle kinetic parameters, including the hydrolysis rate constant of particulate organic nitrogen, the mineralization rate constant of dissolved organic nitrogen, the nitrification rate constant, the denitrification rate constant, and the sediment-water interface exchange flux. Figure 7 (a) shows the corrected rate constant for the hydrolysis of recalcitrant particulate organic nitrogen. Figure 7 (b) shows the corrected nitration rate constant. Figure 7 (c) shows the corrected hydrolysis rate constant of readily degradable particulate organic nitrogen. Figure 7 (d) in the figure shows the corrected result for the denitrification rate constant. Figure 7 (e) shows the corrected results for the mineralization rate constant of dissolved organic nitrogen. Figure 7 (f) shows the corrected results for the exchange flux of dissolved organic nitrogen at the sediment-water interface. Figure 7 (a) to Figure 7 As can be seen from (f), S1 and S2 exhibit insufficient or near-constant responses across multiple parameters, failing to reflect the dynamic characteristics of parameter changes over time; while S3 can produce parameter fluctuations, some parameters show low or even negative correlation with the reference sequence; in contrast, S4... Figure 7 (a) to Figure 7 The parameters shown in (f) can more stably track the changing trend of the real parameter sequence T, indicating that the same sign correction rule and parameter trend mapping method proposed in this application can improve the stability and physical consistency of nitrogen cycle kinetic parameter correction.

[0142] Figure 8 A graph illustrating the trend of time-varying denitrification rate constant according to an embodiment of this application is shown.

[0143] As attached Figure 8 As shown, the horizontal axis represents time (2016 to 2020), and the vertical axis represents the denitrification rate constant (d). -1 The denitrification rate constant exhibits significant seasonal and interannual variations: it is higher in summer and lower in winter, with the overall rate in a wet year (2016) being lower than that in a dry year (2019). This mapping method transfers the time-varying patterns of parameters efficiently identified by the low-dimensional model to the high-dimensional model, preserving the physical trends while significantly reducing computational costs.

[0144] The governing equations are shown in equation (17), and the three-dimensional momentum equations are shown in equation set (18):

[0145] (17).

[0146] (18).

[0147] Where u and v are the flow velocities along the x and y directions, respectively; Let Q represent the water depth and Q represent the external source term. ρ is the vertical eddy viscosity coefficient, f is the Coriolis force coefficient, and ρ0 is the water density. and Indicates the horizontal pressure gradient; and The unbalanced term representing the horizontal Reynolds stress; and This represents external momentum sources or sinks (e.g., external forces generated by hydraulic structures, water intake, water outflow, wave stress, etc.). The Manning roughness coefficient is 0.023, the horizontal turbulent viscosity coefficient is 5 m² / s, and the calculation step size is 5 min. The water quality model simulates the migration and transformation of various nitrogen forms. Initial parameter calibration values ​​are referenced in Table 1. The core focus is on the time-varying optimization of the denitrification rate constant RcDenWat.

[0148] Table 1: Explanation of System Response Parameters in the 3D Hydrodynamic-Water Quality Model

[0149]

[0150] k based on zero-dimensional model correction deni20 The time-varying parameters for the trend of change are calculated using the following formula (19):

[0151] (19).

[0152] Where M=6.66, k is a time-varying parameter. calb =0.085d -1 , and Correct the maximum and minimum values ​​of the parameters respectively.

[0153] Figure 9 A schematic diagram comparing simulated and measured nitrate nitrogen values ​​at different monitoring points according to embodiments of this application is shown. Figure 9 (a) to Figure 9 As shown in (h), nitrate nitrogen ( Taking the concentration as an example, the simulation results and measured values ​​of the fixed parameter calibration scheme (blue line) and the time-varying parameter mapping scheme of this application (gray dashed line) were compared at multiple monitoring points (monitoring point A, monitoring point B, monitoring point C, etc.).

[0154] Specifically, Figure 9 (a) Figure 9 (b) Figure 9 (c) Figure 9 (d) Figure 9 (e) Figure 9 (f) in Figure 9 (g) and Figure 9 The (h) in the figure corresponds to monitoring point A, monitoring point B, monitoring point C, monitoring point D, monitoring point E, monitoring point F, monitoring point G and monitoring point H respectively. Figure 9 (a) to Figure 9 In the (h) graph, the horizontal axis represents time (2016 to 2020), and the vertical axis represents nitrate nitrogen concentration (mg / L). Measured data points are used to characterize changes in monitored concentrations, while simulation curves from the fixed-parameter scheme and the time-varying-parameter scheme are used to characterize the prediction results under the two parameter settings. In each subgraph, the root mean square error (RMSE1), mean absolute percentage error (MAPE1), and correlation coefficient (ρ1) correspond to the fixed-parameter calibration scheme, while the root mean square error (RMSE2), mean absolute percentage error (MAPE2), and correlation coefficient (ρ2) correspond to the time-varying-parameter mapping scheme of this application. Figure 9 (a) to Figure 9 As shown in (h), the simulated curve (blue) of the fixed parameter scheme deviates significantly from the measured data (red dots), especially underestimating the data during the flood season and high-water years. However, after adopting the time-varying parameter scheme of this application, the agreement between the simulated curve and the measured data is significantly improved. Quantitative indicators show that the root mean square error decreased from 0.43 mg / L to 0.33 mg / L, the mean absolute percentage error decreased from 72.97% to 49.63%, the correlation coefficient increased from 0.14 to 0.51, and the nitrate nitrogen simulation error decreased by 22.63%-31.98%. Furthermore, compared to directly applying ensemble Kalman filtering (ensemble size 1000) in the two-dimensional model, the computational performance of this framework is improved by more than 58 times, and the total simulation time for 5 years of operating conditions decreased from 7681.37 core hours to 132.13 core hours.

[0155] Based on the hydrological year type classification results, the simulation verification period of the embodiment covers different hydrological conditions such as high-water years, normal-water years, and low-water years. 2016 was a high-water year, 2019 was a low-water year, and the other relevant years were normal-water years. Simultaneously, the simulation period covers both flood season and non-flood season. Therefore, the verification process of the model parameter dynamic correction method includes different inflow rates, external loads, and hydrodynamic disturbance conditions, which can be used to evaluate the applicability and stability of the method under different hydrological scenarios. Figure 9 As shown in Table 2, compared with the traditional fixed parameter scheme, the root mean square error of nitrate nitrogen simulation decreased from 0.43 to 0.33 mg / L, the mean absolute percentage error decreased from 72.97% to 49.63%, and the correlation coefficient increased from 0.14 to 0.51. Compared with directly applying ensemble Kalman filtering (ensemble size 1000) in the two-dimensional model, the framework provided by this application embodiment improves the computational performance by more than 58 times, and the total time for 5-year operating condition simulation decreased from 7681.37 core hours to 132.13 core hours.

[0156] Table 2: Comparison of Predictive Performance between Calibration and Time-Varying Parameter Scheme Models

[0157]

[0158] This embodiment verifies the effectiveness of the technical solution using actual data from Miyun Reservoir: the accuracy of first-order reaction parameter correction is significantly improved, the zero-dimensional model can effectively capture the time-varying trend of parameters, and the two-dimensional model achieves dual optimization of accuracy and computational efficiency through parameter mapping. The simulation error of nitrate nitrogen is reduced by 22.63%-31.98%, and the computational cost is reduced by more than 58 times, fully supporting the engineering application value of the data observation-parameter correction-model prediction framework.

[0159] Based on the above-described water quality prediction method based on dynamic parameter correction, embodiments of this application also provide a water quality prediction system based on dynamic parameter correction. The following will be combined with... Figure 10 The device is described in detail.

[0160] Figure 10 An architecture diagram of a water quality prediction system based on parameter dynamic correction according to an embodiment of this application is shown.

[0161] like Figure 10 As shown, the water quality prediction system 1000 based on parameter dynamic correction in this embodiment includes a data acquisition module 1010, a first determination module 1020, a second determination module 1030, a stimulus signal generation module 1040, a correction module 1050, a third determination module 1060, and a water quality output module 1070.

[0162] The data acquisition module 1010 is used to acquire hydrological data and water quality data of the target water body. The hydrological data includes water level and flow rate, and the water quality data includes observed values ​​of nitrogen concentrations at multiple time points. In one embodiment, the data acquisition module 1010 can be used to perform step S210 described above, which will not be repeated here.

[0163] The first determining module 1020, based on a pre-constructed high-dimensional water quality model and the observed concentrations of various nitrogen forms at the multiple time points, determines an assimilation value corresponding to the observed values. This assimilation value is the optimal estimate of the true state of the water body obtained by fusing the model predictions and the observed values. In one embodiment, the first determining module 1020 can be used to execute step S220 described above, which will not be repeated here.

[0164] The second determining module 1030 is used to determine the relative deviation based on the observations and assimilated values ​​at the same time. In one embodiment, the second determining module 1030 can be used to perform step S230 described above, which will not be repeated here.

[0165] The stimulus signal generation module 1040 is used to generate a stimulus signal in response to the absolute value of the relative deviation exceeding a preset stimulus threshold, and to correct the first-order reaction rate constant in the high-dimensional water quality model based on the stimulus signal. In one embodiment, the stimulus signal generation module 1040 can be used to perform step S240 described above, which will not be repeated here.

[0166] The correction module 1050 is used to perform water balance calculations based on a pre-built zero-dimensional water quality model, using the water level and the flow rate. When the relative deviations between the current water quality components and the upstream reactant components are determined to have the same sign, the module corrects the nitrogen cycle kinetic parameters to obtain the parameter variation trend, including the first-order reaction rate constant. In one embodiment, the correction module 1050 can be used to execute step S250 as described above, which will not be repeated here.

[0167] The third determining module 1060 is used to determine the time-varying parameter sequence of the high-dimensional water quality model based on the parameter change trend. In one embodiment, the third determining module 1060 can be used to execute step S260 described above, which will not be repeated here.

[0168] The water quality output module 1070 is used to input the time-varying parameter sequence into the high-dimensional water quality model and output the predicted concentration of water quality indicators of the target water body within a preset time period. In one embodiment, the water quality output module 1070 can be used to perform step S270 described above, which will not be repeated here.

[0169] According to embodiments of this application, any multiple modules among the data acquisition module 1010, the first determination module 1020, the second determination module 1030, the stimulus signal generation module 1040, the correction module 1050, the third determination module 1060, and the water quality output module 1070 can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some functions of one or more of these modules can be combined with at least some functions of other modules and implemented in one module. According to embodiments of this application, at least one of the data acquisition module 1010, the first determination module 1020, the second determination module 1030, the stimulus signal generation module 1040, the correction module 1050, the third determination module 1060, and the water quality output module 1070 can be at least partially implemented as a hardware circuit, such as a field-programmable gate array, a programmable logic array, a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit, or any other reasonable method of integrating or packaging the circuit, or implemented in software, hardware, or firmware, or in any appropriate combination of any of these three implementation methods. Alternatively, at least one of the data acquisition module 1010, the first determination module 1020, the second determination module 1030, the stimulus signal generation module 1040, the correction module 1050, the third determination module 1060, and the water quality output module 1070 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0170] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0171] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

Claims

1. A water quality prediction method based on dynamic parameter correction, characterized in that, include: Acquire hydrological and water quality data of the target water body, wherein the hydrological data includes water level and flow rate, and the water quality data includes observed values ​​of nitrogen concentrations of various forms at multiple times; Based on the pre-constructed high-dimensional water quality model and the observed values ​​of nitrogen concentrations at multiple times, an assimilation value corresponding to the observed value is determined. The assimilation value is the optimal estimate of the true state of the water body obtained by fusing the model prediction value and the observed value. The relative bias is determined based on the observations and assimilated values ​​at the same time. In response to the absolute value of the relative deviation exceeding a preset stimulus threshold, a stimulus signal is generated and the first-order reaction rate constant in the high-dimensional water quality model is corrected based on the stimulus signal. Based on a pre-built zero-dimensional water quality model, water balance calculations are performed using the water level and the flow rate. Under the condition that the relative deviations between the current water quality components and the upstream reactant components have the same sign, the nitrogen cycle kinetic parameters are corrected to obtain the parameter variation trend including the first-order reaction rate constant. The time-varying parameter sequence of the high-dimensional water quality model is determined based on the parameter variation trend; as well as The time-varying parameter sequence is input into the high-dimensional water quality model, and the predicted concentration of water quality indicators of the target water body in a preset time period is output.

2. The method according to claim 1, characterized in that, The assimilation value corresponding to the observed values, based on the pre-constructed high-dimensional water quality model and the observed concentrations of various nitrogen forms at multiple time points, includes: Multiple initial state vectors are generated by superimposing random perturbations that follow a preset probability distribution onto the model state variables. Each initial state vector is input into the high-dimensional water quality model for forward prediction, resulting in multiple predicted state vectors. The observed values ​​are used to update the multiple predicted state vectors to obtain multiple analytical state vectors; and The arithmetic mean of the plurality of analytical state vectors is calculated as the assimilation value.

3. The method according to claim 1, characterized in that, The generation of stimulus signals and the correction of the first-order reaction rate constant in the high-dimensional water quality model based on the stimulus signals include: A stimulation signal is generated based on the magnitude by which the relative deviation exceeds the stimulation threshold and the simulated time step, wherein the intensity of the stimulation signal is directly proportional to the magnitude and inversely proportional to the time step. Based on the stimulus signal and a preset correction coefficient, an exponential decay factor is generated. The original rate constant is multiplied by the exponential decay factor to obtain the corrected first-order response rate constant.

4. The method according to claim 1, characterized in that, The water balance calculation based on the pre-built zero-dimensional water quality model, using the water level and the flow rate, includes: Based on the water balance relationship, the difference between the total inflow and the total outflow is taken as the rate of change of water storage over time, and the product of water level and water surface area is taken as the water storage; and Based on the nitrogen cycle kinetics, concentration change equations for recalcitrant particulate organic nitrogen, readily degradable particulate organic nitrogen, dissolved organic nitrogen, ammonia nitrogen, and nitrate nitrogen were established. These concentration change equations are used to describe the changes in the concentration of each form of nitrogen over time.

5. The method according to claim 1, characterized in that, Assuming the relative deviations between the current water quality components and the upstream reactant components are of the same sign, the correction parameters for nitrogen cycle kinetics include: The hydrolysis rate constant of recalcitrant particulate organic nitrogen is corrected when the relative deviations between recalcitrant particulate organic nitrogen and dissolved organic nitrogen have the same sign. The hydrolysis rate constant of easily degradable particulate organic nitrogen is corrected when the relative deviation between easily degradable particulate organic nitrogen and dissolved organic nitrogen has the same sign. The mineralization rate constant of dissolved organic nitrogen is corrected when the relative deviations of dissolved organic nitrogen and ammonia nitrogen have the same sign. The nitration rate constant is corrected when the relative deviations of ammonia nitrogen and nitrate nitrogen have the same sign. The denitrification rate constant is corrected based on the relative deviation of nitrate nitrogen, and the direction of correction is constrained by the sign relationship between the relative deviations of ammonia nitrogen and nitrate nitrogen. The sediment-water interface exchange fluxes of dissolved organic nitrogen, ammonia nitrogen, and nitrate nitrogen are corrected based on the relative deviations of the corresponding water quality components, and a linkage correction is performed when the relative deviations of multiple water quality components have the same sign.

6. The method according to claim 1, characterized in that, The process of determining the time-varying parameter sequence of the high-dimensional water quality model based on the parameter variation trend includes: Based on the relative position of the current value of the parameter change trend within its preset change range and the preset scaling factor, the initial calibration value of the high-dimensional water quality model is linearly scaled to generate the time-varying parameter sequence.

7. The method according to any one of claims 1 to 6, characterized in that, The high-dimensional water quality model is a two-dimensional or three-dimensional water quality model, including a hydrodynamic module and a water quality module; wherein, the hydrodynamic module is used to describe the changes in water level and flow velocity over time and space, and the water quality module is used to simulate the migration and transformation of various forms of nitrogen.

8. The method according to claim 1, characterized in that, The predicted concentrations of the water quality indicators include nitrate nitrogen concentration, ammonia nitrogen concentration, total nitrogen concentration, and dissolved oxygen concentration. The step of inputting the time-varying parameter sequence into the high-dimensional water quality model and outputting the predicted concentrations of the water quality indicators for the target water body within a preset time period includes: Output at least one of the following: nitrate nitrogen concentration, ammonia nitrogen concentration, total nitrogen concentration, and dissolved oxygen concentration; and Output an error evaluation index between simulated and observed values, wherein the error evaluation index includes at least one of root mean square error, mean absolute percentage error, and correlation coefficient.

9. A water quality prediction system based on dynamic parameter correction, characterized in that, The system includes: The data acquisition module is used to acquire hydrological data and water quality data of the target water body. The hydrological data includes water level and flow rate, and the water quality data includes observed values ​​of nitrogen concentrations of various forms at multiple times. The first determining module determines the assimilation value corresponding to the observed value based on the pre-constructed high-dimensional water quality model and the observed values ​​of nitrogen concentrations at multiple times. The assimilation value is the optimal estimate of the true state of the water body obtained by fusing the model prediction value and the observed value. The second determination module is used to determine the relative bias based on the observations and assimilated values ​​at the same time. The stimulation signal generation module is used to generate a stimulation signal in response to the absolute value of the relative deviation exceeding a preset stimulation threshold, and to correct the first-order reaction rate constant in the high-dimensional water quality model based on the stimulation signal. The calibration module is used to perform water balance calculations based on a pre-built zero-dimensional water quality model, using the water level and the flow rate, and to correct the nitrogen cycle kinetic parameters when the relative deviations between the current water quality components and the upstream reactant components are of the same sign, thereby obtaining the parameter change trend including the first-order reaction rate constant. The third determining module is used to determine the time-varying parameter sequence of the high-dimensional water quality model based on the parameter change trend; and The water quality output module is used to input the time-varying parameter sequence into the high-dimensional water quality model and output the predicted concentration of water quality indicators of the target water body within a preset time period.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 8.

Images