A Method and System for Predicting Sediment Concentration Based on Freeze-Thaw Coupling Physical Constraint Model

Abstract

This disclosure provides a method and system for predicting sediment concentration based on a freeze-thaw coupled physical constraint model, relating to the field of sediment concentration prediction technology. The method includes: acquiring and preprocessing various raw data such as hydrological data, meteorological data, and soil and vegetation data; constructing a baseline sediment concentration physical model based on the preprocessed raw data, and constructing an input feature vector; inputting the input feature vector into a sediment concentration prediction ANN model, and outputting the sediment concentration; wherein, during the training process of the sediment concentration prediction ANN model, the baseline sediment prediction of the physical model and the sediment mass conservation constraint are introduced into the ANN model loss function, and a freeze-thaw weighted strategy is used to enhance the loss function, resulting in a sediment concentration prediction ANN model that deeply integrates the physical model and the ANN model. This disclosure improves the stability and accuracy of predictions under extreme rainfall, runoff surges, or frequent freeze-thaw cycles.

Description

Technical Field

[0001] This disclosure relates to the field of sediment concentration prediction technology, specifically to a method and system for sediment concentration prediction based on a freeze-thaw coupling physical constraint model. Background Technology

[0002] The statements in this section are merely background information relating to this disclosure and do not necessarily constitute prior art.

[0003] The spatiotemporal variations of sediment not only affect the stability of river channel morphology but also directly relate to the safety of water conservancy operations, the health of aquatic ecosystems, and the risk of downstream flooding. Therefore, accurately predicting sediment concentration variations is of great significance for watershed water and sediment management and ecological restoration. In recent years, with global warming and the increasing frequency of extreme weather events, the freeze-thaw cycle in mid-to-high latitude regions has become increasingly intense. The freeze-thaw process significantly alters soil structure and permeability, leading to complex nonlinear changes in surface runoff and sediment production mechanisms. During spring and early winter, repeated freeze-thaw cycles loosen soil structure and reduce erosion resistance, making it susceptible to significant sediment transport surges from rainfall or snowmelt. Traditional sediment prediction models struggle to characterize the impact of freeze-thaw erosion on sediment concentration. Furthermore, the development of remote sensing, automatic hydrological monitoring, and soil sensing technologies has enabled the continuous acquisition of multi-source data (such as air temperature, rainfall, flow rate, soil moisture content, snowmelt, and soil temperature), providing an accurate and comprehensive data foundation for establishing high-precision, dynamically responsive sediment concentration prediction models.

[0004] Although there are various existing methods for predicting sediment concentration, these methods still have many shortcomings:

[0005] (a) Traditional empirical models and multiple linear regression models, such as The method relies on fixed parameters for fitting, which makes it difficult to reflect the characteristics of the watershed hydrological response under freeze-thaw conditions, and its applicability is poor in different seasons and years.

[0006] (ii) Secondly, physical models can describe the dynamics of water flow and the basic transport mechanism of sediment well, but when characterizing the characteristics such as the frequency of freeze-thaw cycles and changes in soil erosivity during the freeze-thaw period, they lack model parameters related to the freeze-thaw process, and the initial conditions are difficult to obtain, resulting in significant deviations between the model and actual observations.

[0007] (iii) In addition, although pure data-driven artificial neural networks can fit sediment concentration using historical data, they lack physical constraints, are prone to overfitting, and have insufficient extrapolation ability for physical conservation and extreme conditions. The prediction results may show unreasonable negative values ​​or not conform to mass conservation. Summary of the Invention

[0008] To address the aforementioned issues, this disclosure proposes a method and system for predicting sediment concentration based on a freeze-thaw coupled physical constraint model. The freeze-thaw physical process is explicitly modeled and integrated into an ANN (physical model and artificial neural network) model. The baseline sediment prediction from the physical model and the sediment mass conservation constraint are introduced into the ANN loss function. The output is constrained by the physical constraint loss function, ensuring that the model data fits to physical laws. A freeze-thaw weighted strategy is employed to enhance the loss function, improving the model's stability and accuracy under extreme rainfall, runoff surges, or frequent freeze-thaw cycles.

[0009] According to some embodiments, the present disclosure adopts the following technical solutions:

[0010] Sediment concentration prediction methods based on freeze-thaw coupled physical constraint models include:

[0011] Acquire and preprocess various raw data, including hydrological data, meteorological data, and soil and vegetation data;

[0012] Construct input feature vectors based on various preprocessed raw data;

[0013] The input feature vector is fed into the sediment concentration prediction ANN model, and the output is the sediment concentration.

[0014] In the training process of the sediment concentration prediction ANN model, a benchmark sediment concentration physical model is constructed. The benchmark sediment prediction of the physical model and the sediment mass conservation constraint are introduced into the loss function of the ANN model. The loss function is enhanced by a freeze-thaw weighting strategy, resulting in a sediment concentration prediction ANN model that deeply integrates the physical model and the ANN model.

[0015] As one embodiment, the acquisition of various raw data, including hydrological data, meteorological data, and soil and vegetation data, includes:

[0016] Data is time-aligned to a unified time step to acquire various raw data, including hydrological data, meteorological data, and soil and vegetation data.

[0017] The hydrological data includes runoff, runoff surge, observed sediment concentration, and snowmelt; the meteorological data includes temperature, rainfall, snowfall, and extreme rainfall; and the soil and vegetation data includes soil temperature, soil volumetric water, and freeze-thaw frequency.

[0018] As one embodiment, the preprocessing process includes time-series alignment, monthly extraction, outlier removal, and normalization.

[0019] As one embodiment, the construction of the benchmark sediment concentration physical model includes:

[0020] Calculate runoff surge based on runoff volume;

[0021] Construct an indicator function, and calculate the freeze-thaw frequency based on the indicator function and soil temperature;

[0022] A rainfall indicator is constructed based on rainfall data, and a soil temperature gradient is calculated based on soil temperature.

[0023] A physical model of baseline sediment concentration was constructed based on runoff, runoff surge, observed sediment concentration, and snowmelt.

[0024] As one embodiment, the construction of the input feature vector includes: constructing the input feature vector based on runoff, snowmelt, air temperature, rainfall, snowfall, soil temperature, soil volumetric water, as well as the calculated runoff surge, freeze-thaw frequency, calculated rainfall indicator, and soil temperature gradient.

[0025] As one embodiment, the total loss function of the sediment concentration prediction ANN model includes data fitting loss, physical boundary constraints, sediment mass conservation constraints, and freeze-thaw weighted constraints. Through physical constraint loss and sediment mass conservation constraints, the prediction results of the pure data-driven model are prevented from having negative values ​​or unreasonable extreme values. Through the freeze-thaw weighted strategy, the model's response capability to freezing-thawing transition periods and extreme runoff events is enhanced.

[0026] According to some embodiments, the present disclosure adopts the following technical solutions:

[0027] A sediment concentration prediction system based on a freeze-thaw coupling physical constraint model is characterized by comprising:

[0028] The data acquisition module is used to acquire and preprocess various raw data, including hydrological data, meteorological data, and soil and vegetation data.

[0029] The feature construction module is used to construct input feature vectors based on various preprocessed raw data.

[0030] The sediment concentration prediction module is used to input the input feature vector into the sediment concentration prediction ANN model and output the sediment concentration.

[0031] In the training process of the sediment concentration prediction ANN model, a benchmark sediment concentration physical model is constructed. The benchmark sediment prediction of the physical model and the sediment mass conservation constraint are introduced into the loss function of the ANN model. The loss function is enhanced by a freeze-thaw weighting strategy, resulting in a sediment concentration prediction ANN model that deeply integrates the physical model and the ANN model.

[0032] According to some embodiments, the present disclosure adopts the following technical solutions:

[0033] A computer program product includes a computer program that, when executed by a processor, implements the sediment concentration prediction method based on a freeze-thaw coupling physical constraint model.

[0034] According to some embodiments, the present disclosure adopts the following technical solutions:

[0035] A non-transitory computer-readable storage medium is provided for storing computer instructions, which, when executed by a processor, implement the sediment concentration prediction method based on the freeze-thaw coupling physical constraint model.

[0036] According to some embodiments, the present disclosure adopts the following technical solutions:

[0037] An electronic device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the sediment concentration prediction method based on the freeze-thaw coupling physical constraint model.

[0038] Compared with the prior art, the beneficial effects of this disclosure are as follows:

[0039] The sediment concentration prediction method based on the freeze-thaw coupling physical constraint model disclosed herein is used to predict the sediment concentration of rivers in high-altitude freeze-thaw areas. It enables the model to have high prediction accuracy, good physical consistency, and assessable uncertainty under the influence of the freeze-thaw process. In particular, it has good extrapolation ability for extreme or transitional freeze-thaw periods.

[0040] This disclosure presents a sediment concentration prediction method based on a freeze-thaw coupled physical constraint model. It constructs a baseline sediment concentration physical model, explicitly models the freeze-thaw physical process, and integrates it into the ANN model to improve the prediction accuracy of sediment concentration during freeze-thaw cycles. The baseline sediment prediction from the physical model and the sediment mass conservation constraint are introduced into the ANN model's loss function to avoid negative predictions or exceeding physical upper limits. While maintaining the ANN model's ability to fit complex nonlinear relationships, the output is constrained through the physical constraint loss function, ensuring that the model data fits to physical laws. Finally, a freeze-thaw weighted strategy is used to enhance the loss function, improving the model's stability and accuracy under extreme rainfall, runoff surges, or frequent freeze-thaw cycles.

[0041] This disclosed method for predicting sediment concentration based on a freeze-thaw coupled physical constraint model significantly improves prediction accuracy while ensuring the physical consistency of sediment transport. By deeply integrating the physical model with an artificial neural network (ANN) and incorporating key features such as freeze-thaw frequency and soil temperature gradient, it avoids negative values ​​or unreasonable extreme values ​​in the prediction results of purely data-driven models through physical constraint loss and sediment mass conservation constraints. The freeze-thaw weighting strategy enhances the model's response to freezing-thawing periods and extreme runoff events, improving the robustness and generalization ability of the prediction. This invention can accurately fit historical observations while satisfying the mass conservation and physical boundary requirements of sediment transport, balancing interpretability and engineering applicability. It can be widely applied in practical scenarios such as watershed sediment early warning, soil and water conservation, and reservoir operation. Attached Figure Description

[0042] The accompanying drawings, which form part of this disclosure, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure.

[0043] Figure 1 This is a flowchart of a sediment concentration prediction method based on a freeze-thaw coupling physical constraint model according to an embodiment of this disclosure. Detailed Implementation

[0044] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0045] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains.

[0046] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this disclosure. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0047] Example 1

[0048] One embodiment of this disclosure provides a method for predicting sediment concentration based on a freeze-thaw coupling physical constraint model. The method includes the following steps:

[0049] Step 1: Acquire and preprocess various raw data, including hydrological data, meteorological data, and soil and vegetation data;

[0050] Step 2: Construct input feature vectors based on the preprocessed raw data;

[0051] Step 3: Input the input feature vector into the sediment concentration prediction ANN model, and output the sediment concentration;

[0052] In the training process of the sediment concentration prediction ANN model, a benchmark sediment concentration physical model is constructed. The benchmark sediment prediction of the physical model and the sediment mass conservation constraint are introduced into the loss function of the ANN model. The loss function is enhanced by a freeze-thaw weighting strategy, resulting in a sediment concentration prediction ANN model that deeply integrates the physical model and the ANN model.

[0053] As one embodiment, the sediment concentration prediction method based on a freeze-thaw coupled physical constraint model disclosed herein deeply integrates the physical model with an artificial neural network (ANN) and introduces key features such as freeze-thaw frequency and soil temperature gradient, which can significantly improve prediction accuracy while ensuring the physical consistency of sediment transport. Through physical constraint loss and sediment mass conservation constraints, it avoids negative values ​​or unreasonable extreme values ​​in the prediction results of purely data-driven models; through a freeze-thaw weighted strategy, it enhances the model's response to freezing-thawing periods and extreme runoff events, improving the robustness and generalization ability of the prediction. The specific training process of this disclosure is as follows:

[0054] Step 1: Acquire and preprocess various raw data, including hydrological data, meteorological data, and soil and vegetation data;

[0055] Specifically, firstly, the data is time-aligned to a unified time step T, and various raw data are collected, including hydrological data, meteorological data, and soil and vegetation data, as follows:

[0056] Hydrological data, including: runoff (m³ / s), runoff surge (m³ / s), observed sediment concentration (kg / m³), snowmelt (mm / month);

[0057] Meteorological data, including: temperature (°C), Rainfall (mm / month), snowfall (mm / month), extreme rainfall , (mm / month);

[0058] Soil and vegetation data, including: soil temperature (°C), soil volumetric water Freeze-thaw frequency (Second-rate).

[0059] Secondly, the acquired raw data undergoes preprocessing, including time-series alignment, monthly extraction, outlier removal, and normalization. The specific steps are as follows:

[0060] (1) Time series processing: For data from different sources: runoff, sediment concentration (SSC), runoff surge, snowmelt, air temperature, rainfall, snowfall, extreme rainfall, soil temperature, soil volumetric water content, and freeze-thaw frequency, the data were aligned to the daily scale; for missing time points, linear interpolation was used to fill in the missing data to ensure that all variables were within the same study time range.

[0061] (2) Monthly Alignment: Extract daily-scale data to monthly-scale. Specific requirements: Calculate monthly averages for runoff, runoff surge, sediment concentration (SSC), air temperature, soil temperature, and soil volumetric water content; calculate monthly cumulatives for snowmelt, rainfall, snowfall, and extreme rainfall; calculate freeze-thaw frequency: calculate temperatures meeting freeze-thaw conditions from hourly-scale soil temperatures and accumulate them on the daily scale, then accumulate them on the monthly scale. Align monthly data to the same monthly scale.

[0062] (3) Remove outliers: Search the data, find data with large values, and verify whether they are caused by extreme events or provide reliable data support.

[0063] (4) Normalization: Before inputting data into the model, Z-score standardization is performed:

[0064]

[0065] in, X i This is the original data. X min The minimum value of the original data. X max The maximum value of the original data is used to map the data to the range [0,1).

[0066] Furthermore, the runoff surge is calculated based on runoff volume, including:

[0067]

[0068] In the formula, For runoff, for t Flow rate at time -1.

[0069] Furthermore, an indicator function is constructed, and the freeze-thaw frequency is calculated based on the indicator function and soil temperature;

[0070]

[0071] In the formula, This is an indicator function, where 1 is recorded when the soil temperature crosses 0°C; For the first t Soil temperature at any given time, in °C; N This is for the statistical window length.

[0072] Furthermore, a rainfall indicator is constructed based on rainfall data, including:

[0073]

[0074] In the formula, For rainfall, P represents a quantile of historical rainfall; in this study, the 95th quantile is used. 95 .

[0075] Furthermore, the soil temperature gradient is calculated based on soil temperature, including:

[0076]

[0077] In the formula, for t Soil temperature at any given time for t Soil temperature at -1.

[0078] Step 2: Construct a benchmark sediment concentration physical model and input feature vector based on the preprocessed raw data;

[0079] (1) First, a physical model of baseline sediment concentration is constructed based on runoff, runoff surge, observed sediment concentration, and snowmelt amount:

[0080]

[0081] In the formula, .

[0082] (2) Freeze-thaw correction based on the benchmark sediment concentration physical model is performed as follows:

[0083]

[0084] In the formula, , φ It is the freeze-thaw effect correction factor; k It is the freeze-thaw sensitivity coefficient.

[0085] The purpose of this disclosure of freeze-thaw correction is because SSC Prediction and runoff Q and rainfall PThese mechanisms are related to various indicators, but in seasonally frozen soil regions or high-altitude watersheds, they can change significantly depending on the freeze-thaw cycle. When freeze-thaw cycles do not occur, the soil is in a frozen state and has low erosiveness. SSC Output will decrease because the soil is loose and easily eroded during freeze-thaw cycles. SSC Output will increase; during freeze-thaw cycles, soil structure is damaged, and soil erosion resistance is weakened. SSC Increased output means that in seasonally frozen soil basins or high-altitude cold basins, with the same runoff and precipitation, an increase in freeze-thaw frequency will result in higher output. SSC It will also increase the interpretability of physics by correcting and compensating for the problems of traditional models.

[0086] (3) Construct the input feature vector;

[0087] Based on runoff, snowmelt, air temperature, rainfall, snowfall, soil temperature, soil volumetric water, and calculated runoff surge, freeze-thaw frequency, calculated rainfall indicator, and soil temperature gradient, an input feature vector is constructed:

[0088]

[0089] Step 3: Construct an ANN model for predicting sediment concentration;

[0090] (1) ANN prediction of sediment concentration output:

[0091]

[0092] In the formula, W All combinations of weights and biases in the model.

[0093]

[0094] in: This is the weight matrix for layer 1; This is the bias vector for the first layer; This represents the total number of layers in the neural network.

[0095] (2) Construct the loss function:

[0096] 1) Data fitting loss:

[0097]

[0098] In the formula, To predict sediment concentration output using ANN, To observe sediment concentration.

[0099] 2) Physical boundary constraints:

[0100]

[0101] In the formula, The historical maximum observation value or the upper limit of the project can be used.

[0102] 3) Sediment mass conservation constraint:

[0103]

[0104] 4) Freeze-thaw weighted constraint:

[0105]

[0106]

[0107] In the formula, It is a sample set from the freeze-thaw period. It is the freeze-thaw sensitivity weighting coefficient.

[0108] Furthermore, the weighting coefficients include and ,in, Calculated using the validation set:

[0109]

[0110] Calculated based on a mass conservation deviation of <5%:

[0111] In summary, the total loss function of this disclosure is:

[0112]

[0113] Step 4: Model training and prediction;

[0114] (1) Using the Adam optimizer, with a learning rate of 1e-3, and a sliding window training method, train the optimal weights:

[0115]

[0116] (2) The prediction formula is:

[0117]

[0118] (3) Residual analysis:

[0119]

[0120] This disclosure incorporates the baseline sediment prediction and sediment mass conservation constraints of the physical model into the loss function of the artificial neural network (ANN) model. An ANN model is constructed with 11 input nodes as the input layer, two hidden layers, and one output layer to capture the nonlinear sediment transport characteristics under freeze-thaw conditions. Taking into account the mean square error, physical constraints, and mass conservation constraints, a total loss function is constructed. The backpropagation algorithm is then used to optimize the neural network parameters and physical weights. α , β , λ 1. λ 2. Synchronous optimization forms an adaptive parameter update loop until the model converges. A freeze-thaw weighted strategy is used to enhance the loss function. Through iterative training, the set number of iterations or optimization conditions are reached to predict sediment concentration, obtaining the prediction results. The results can be dynamically adjusted according to freeze-thaw conditions, achieving physical consistency and data-driven fusion prediction for complex surface sediment transport processes. This disclosure uses a freeze-thaw weighted strategy to enhance the loss function. Through iterative training, the set number of iterations or optimization conditions are reached to obtain a trained sediment concentration prediction ANN model for sediment concentration prediction.

[0121] Example 2

[0122] One embodiment of this disclosure provides a sediment concentration prediction system based on a freeze-thaw coupling physical constraint model, comprising:

[0123] The data acquisition module is used to acquire and preprocess various raw data, including hydrological data, meteorological data, and soil and vegetation data.

[0124] The feature construction module is used to build a benchmark sediment concentration physical model based on various preprocessed raw data and construct the input feature vector.

[0125] The sediment concentration prediction module is used to input the input feature vector into the sediment concentration prediction ANN model and output the sediment concentration.

[0126] In the training process of the sediment concentration prediction ANN model, the baseline sediment prediction of the physical model and the sediment mass conservation constraint are introduced into the loss function of the ANN model. The loss function is enhanced by a freeze-thaw weighting strategy, resulting in a sediment concentration prediction ANN model that deeply integrates the physical model and the ANN model.

[0127] Example 3

[0128] One embodiment of this disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the sediment concentration prediction method based on a freeze-thaw coupling physical constraint model.

[0129] Example 4

[0130] One embodiment of this disclosure provides a non-transitory computer-readable storage medium for storing computer instructions. When these computer instructions are executed by a processor, they implement the sediment concentration prediction method based on the freeze-thaw coupling physical constraint model.

[0131] Example 5

[0132] One embodiment of this disclosure provides an electronic device, including a processor, a memory, and a computer program; wherein the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the sediment concentration prediction method based on the freeze-thaw coupling physical constraint model.

[0133] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0134] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0135] While the specific embodiments of this disclosure have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of this disclosure. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of this disclosure are still within the scope of protection of this disclosure.

Claims

1. A method for predicting sediment concentration based on a freeze-thaw coupled physical constraint model, characterized in that, include: Acquire and preprocess various raw data, including hydrological data, meteorological data, and soil and vegetation data; Construct input feature vectors based on various preprocessed raw data; The input feature vector is fed into the sediment concentration prediction ANN model, and the output is the sediment concentration. In the training process of the sediment concentration prediction ANN model, a benchmark sediment concentration physical model is constructed, and the benchmark sediment prediction and sediment mass conservation constraint of the physical model are introduced into the loss function of the ANN model. The loss function is enhanced by the freeze-thaw weighting strategy, and a sediment concentration prediction ANN model with deep integration of the physical model and the ANN model is obtained. The total loss function of the sediment concentration prediction ANN model includes data fitting loss, physical boundary constraints, sediment mass conservation constraints, and freeze-thaw weighted constraints, specifically: 1) Data fitting loss: In the formula, To predict sediment concentration output using ANN, To observe sediment concentration; 2) Physical boundary constraints: In the formula, Take the historical maximum observation value or the upper limit of the project; As a benchmark physical model for sediment concentration; For ANN prediction of sediment concentration output; 3) Sediment mass conservation constraint: in, To predict sediment concentration output using ANN, For runoff, As a benchmark physical model for sediment concentration; 4) Freeze-thaw weighted constraint: In the formula, It is a sample set from the freeze-thaw period. t Indicates time, It is a freeze-thaw sensitivity weighting factor. For freeze-thaw frequency, As a benchmark physical model for sediment concentration. To predict sediment concentration output using ANN, For physical boundary constraints.

2. The sediment concentration prediction method based on the freeze-thaw coupling physical constraint model as described in claim 1, characterized in that, The acquisition of various raw data, including hydrological data, meteorological data, and soil and vegetation data, includes: Data is time-aligned to a unified time step to acquire various raw data, including hydrological data, meteorological data, and soil and vegetation data. The hydrological data includes runoff, runoff surge, observed sediment concentration, and snowmelt; the meteorological data includes temperature, rainfall, snowfall, and extreme rainfall; and the soil and vegetation data includes soil temperature, soil volumetric water, and freeze-thaw frequency.

3. The sediment concentration prediction method based on the freeze-thaw coupling physical constraint model as described in claim 1, characterized in that, The preprocessing process includes time-series alignment, monthly extraction, outlier removal, and normalization.

4. The sediment concentration prediction method based on the freeze-thaw coupling physical constraint model as described in claim 1, characterized in that, The construction of the benchmark sediment concentration physical model includes: Calculate runoff surge based on runoff volume; Construct an indicator function, and calculate the freeze-thaw frequency based on the indicator function and soil temperature; A rainfall indicator is constructed based on rainfall data, and a soil temperature gradient is calculated based on soil temperature. A physical model of baseline sediment concentration was constructed based on runoff, runoff surge, observed sediment concentration, and snowmelt.

5. The sediment concentration prediction method based on the freeze-thaw coupling physical constraint model as described in claim 4, characterized in that, An input feature vector is constructed based on runoff, snowmelt, air temperature, rainfall, snowfall, soil temperature, soil volumetric water, as well as the calculated runoff surge, freeze-thaw frequency, calculated rainfall indicator, and soil temperature gradient.

6. A sediment concentration prediction system based on a freeze-thaw coupled physical constraint model, specifically implementing the sediment concentration prediction method based on a freeze-thaw coupled physical constraint model as described in any one of claims 1-5, characterized in that, include: The data acquisition module is used to acquire and preprocess various raw data, including hydrological data, meteorological data, and soil and vegetation data. The feature construction module is used to construct input feature vectors based on various preprocessed raw data. The sediment concentration prediction module is used to input the input feature vector into the sediment concentration prediction ANN model and output the sediment concentration. In the training process of the sediment concentration prediction ANN model, a benchmark sediment concentration physical model is constructed. The benchmark sediment prediction of the physical model and the sediment mass conservation constraint are introduced into the loss function of the ANN model. The loss function is enhanced by a freeze-thaw weighting strategy, resulting in a sediment concentration prediction ANN model that deeply integrates the physical model and the ANN model.

7. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the sediment concentration prediction method based on the freeze-thaw coupling physical constraint model as described in any one of claims 1-5.

8. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the sediment concentration prediction method based on the freeze-thaw coupling physical constraint model as described in any one of claims 1-5.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to perform the sediment concentration prediction method based on the freeze-thaw coupling physical constraint model as described in any one of claims 1-5.

Images