River ecological management method, device, equipment and medium

Abstract

This application provides a method, apparatus, equipment, and medium for river ecological management. The method acquires an ecological dataset, which includes four-dimensional data formed by the comprehensive ecological health index of a target river segment during a target historical period, and runoff, sediment concentration, and water temperature from river monitoring data. A water, sediment, and temperature threshold prediction model is trained based on multiple ecological data sets. After training, an implicit function describing the model is obtained. This implicit function includes a continuous function of the comprehensive ecological health index with respect to runoff, sediment concentration, and water temperature. The continuous function is obtained by learning and fitting a continuous manifold to multiple discrete ecological data sets through a neural differential equation in the water, sediment, and temperature threshold prediction model. Solving the implicit function yields the runoff threshold range, sediment concentration threshold range, and water temperature threshold range. This application provides a method for quantitatively evaluating river health and determining thresholds, enabling targeted river ecological management.

Description

Technical Field

[0001] This application relates to the field of ecological water conservancy technology, and more specifically, to a method, device, equipment, and medium for river ecological management. Background Technology

[0002] With the advancement of ecological civilization construction, river management has shifted from simply controlling water quality or flood control to comprehensively considering the dynamic coordination and macro-system regulation of multiple factors such as hydrology, sediment, water temperature, and biological habitats. The siltation and shrinkage of sedimentary rivers or river channels, as well as ecological degradation, have a significant negative impact on ecological civilization construction. How to improve these phenomena is a key research direction and technical challenge in the field of ecology. Summary of the Invention

[0003] In view of this, this application provides a method, apparatus, equipment and storage medium for river ecological management, in order to at least solve the problems existing in the related technologies.

[0004] Specifically, this application is implemented through the following technical solution:

[0005] This application provides a method for river ecological management, including:

[0006] An ecological dataset is obtained, which includes multiple ecological data, including a four-dimensional dataset consisting of the comprehensive ecological health index of the target river segment during the target historical period and runoff, sediment concentration, and water temperature from river monitoring data; the comprehensive ecological health index is determined based on the river monitoring data.

[0007] The water, sediment, and temperature threshold prediction model is trained based on the aforementioned multiple ecological data. After training, an implicit function describing the water, sediment, and temperature threshold prediction model is obtained. The implicit function includes a continuous function of the comprehensive ecological health index with respect to runoff, sediment concentration, and water temperature. The continuous function is obtained by learning and fitting the discrete multiple ecological data through the neural differential equation in the water, sediment, and temperature threshold prediction model.

[0008] Based on the implicit function, the threshold ranges for runoff, sediment concentration, and water temperature are determined respectively; these threshold ranges are used for ecological management of the target river section.

[0009] This application also provides a method and apparatus for river ecological management, comprising:

[0010] The dataset acquisition module is used to acquire an ecological dataset, which includes multiple ecological data, including a four-dimensional dataset formed by the comprehensive ecological health index of the target river segment during the target historical period and the runoff, sediment concentration, and water temperature from the river monitoring data; the comprehensive ecological health index is determined based on the river monitoring data.

[0011] The model training module is used to train the water, sediment, temperature, and threshold prediction model based on the multiple ecological data. After training, an implicit function describing the water, sediment, temperature, and threshold prediction model is obtained. The implicit function includes a continuous function of the comprehensive ecological health index with respect to runoff, sediment concentration, and water temperature. The continuous function is obtained by learning and fitting the discrete multiple ecological data through the neural differential equation in the water, sediment, temperature, and threshold prediction model.

[0012] The threshold calculation module is used to determine the runoff threshold range, sediment concentration threshold range, and water temperature threshold range based on the implicit function; the runoff threshold range, sediment concentration threshold range, and water temperature threshold range are used for ecological management of the target river section.

[0013] This application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the river ecological management methods described in the foregoing embodiments.

[0014] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the river ecological management methods described in the foregoing embodiments.

[0015] This application also provides a computer program product, including a computer program that, when run by a processor, performs the steps of any of the possible river ecological management methods described above.

[0016] The technical solutions provided by the embodiments of this application may include the following beneficial effects:

[0017] In this embodiment, an ecological dataset is constructed based on the comprehensive ecological health index of the target historical period and the corresponding runoff, sediment concentration, and water temperature data. A water, sediment, and temperature threshold prediction model is then trained based on this dataset. After training, an implicit function describing the water, sediment, and temperature threshold prediction model is obtained. This implicit function is the final product of the model training. It learns the complex, non-linear intrinsic laws between water, sediment, and temperature conditions and comprehensive ecological health indicators learned from the discrete ecological dataset. Thus, based on this implicit function, accurate, continuous, and ecologically consistent runoff threshold ranges, sediment concentration threshold ranges, and water temperature threshold ranges can be provided for the target river section. In other words, by using the above-mentioned method to quantitatively evaluate river health and determine thresholds, the accuracy of ecological management can be improved.

[0018] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this specification. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a river ecological management method according to an exemplary embodiment of this application;

[0020] Figure 2 This is a flowchart illustrating the determination of a comprehensive ecological health index according to an exemplary embodiment of this application;

[0021] Figure 3 This is a flowchart illustrating a local optimization as shown in an exemplary embodiment of this application;

[0022] Figure 4 This is a schematic diagram of the structure of a river ecological management method apparatus shown in an exemplary embodiment of this application;

[0023] Figure 5 This is a hardware structure diagram of a computer device illustrated in an exemplary embodiment of this application. Detailed Implementation

[0024] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0025] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0026] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0027] With the advancement of ecological civilization construction, river management has shifted from simply controlling water quality or flood control to comprehensively considering the dynamic coordination and macro-system regulation of multiple factors such as hydrology, sediment, water temperature, and biological habitats. The siltation and shrinkage of sedimentary rivers or river channels, as well as ecological degradation, have a significant negative impact on ecological civilization construction. How to improve these phenomena is a key research direction and technical challenge in the field of ecology.

[0028] To address the aforementioned phenomena, relevant technologies typically employ river management and control threshold methods. However, most methods primarily focus on quantifying ecological runoff thresholds. Early methods, such as the Tennant method and the 7Q10 method, mainly focus on a single static value of minimum ecological flow, determining the "bottom line" for river survival through historical flow statistics. These methods are computationally simple but fail to consider the intra-annual dynamic changes in hydrological processes and the influence of key co-factors such as sediment and water temperature. The later developed IFIM method attempts to combine hydrological data (such as flow velocity and water depth) with habitat preference information for specific aquatic organisms (such as fish), recommending flow schemes conducive to biodiversity conservation by simulating the available habitat area under different flow scenarios.

[0029] Secondly, technological explorations regarding sediment thresholds are continuously deepening. For example, the analysis of scouring and deposition patterns in the desert valley section of the upper Yellow River has yielded a series of water and sediment thresholds for different river sections and different sediment concentrations. Based on maintaining the water and sediment balance of the entire Yellow River basin, a three-level index threshold system for coordinated regulation of water and sediment in the Loess Plateau, main channel, and estuary has been constructed. Based on the inflow and sediment data of floods during the flood season in the Ningxia-Inner Mongolia section from 1973 to 2005, the correlation between the scouring and deposition characteristics and efficiency of floods and the water and sediment content at the downstream cross-section has been established, resulting in sediment concentration scouring and deposition thresholds and optimal scouring flow rates.

[0030] Furthermore, although some studies have recognized the importance of seasonal water temperature changes in sediment initiation and bed morphology evolution, the practice of embedding it as a core decision variable in the quantitative model of ecological water demand threshold is still a cutting-edge exploration.

[0031] Some ecological runoff methods (such as the Tennant method) are based on observations of rivers with low sediment loads and do not fully consider the specific ecological needs under high sediment load conditions. Current water and sediment regulation threshold indicators are mostly designed based on river morphology maintenance targets, lacking quantification of ecological process response mechanisms. Moreover, most existing studies remain at the level of qualitative analysis or isolated simulations of single-factor influences, lacking a unified and operational framework capable of simultaneously quantifying the synergistic thresholds of water, sediment, and temperature, and dynamically correlating and verifying them with explicit river health goals. In other words, there is still a significant technological gap in systematically integrating the three key elements of water, sediment, and temperature, and accurately quantifying their critical thresholds and suitable processes for overall river health.

[0032] Based on the above research, this disclosure provides a river ecological management method. The method first acquires an ecological dataset, which includes multiple ecological data points. These ecological data points include four-dimensional data formed by the comprehensive ecological health index of the target river section during a target historical period, and runoff, sediment concentration, and water temperature from river monitoring data. The comprehensive ecological health index is determined based on the river monitoring data. Then, a water, sediment, and temperature threshold prediction model is trained based on the multiple ecological data points. After training, an implicit function describing the water, sediment, and temperature threshold prediction model is obtained. This implicit function includes a continuous function of the comprehensive ecological health index with respect to runoff, sediment concentration, and water temperature. This continuous function is obtained by learning and fitting a continuous manifold to the discrete multiple ecological data points through the neural differential equation in the water, sediment, and temperature threshold prediction model. Finally, based on the implicit function, the runoff threshold range, sediment concentration threshold range, and water temperature threshold range are determined respectively. These threshold ranges are used for ecological management of the target river section.

[0033] In this embodiment, an ecological dataset is constructed based on the comprehensive ecological health index of the target historical period and the corresponding runoff, sediment concentration, and water temperature data. A water, sediment, and temperature threshold prediction model is then trained based on this dataset. After training, an implicit function describing the water, sediment, and temperature threshold prediction model is obtained. This implicit function is the final product of the model training. It learns the complex, non-linear intrinsic laws between water, sediment, and temperature conditions and comprehensive ecological health indicators learned from the discrete ecological dataset. Thus, based on this implicit function, accurate, continuous, and ecologically consistent runoff threshold ranges, sediment concentration threshold ranges, and water temperature threshold ranges can be provided for the target river section. In other words, by using the above-mentioned method to quantitatively evaluate river health and determine thresholds, the accuracy of ecological management can be improved.

[0034] To facilitate understanding of this embodiment, a detailed description of the river ecological management method disclosed in this disclosure is provided first. The executing entity of the river ecological management method provided in this disclosure is generally a computer device. This computer device can be a server, which can be an independent physical server, a server cluster composed of multiple physical servers, or a distributed system. It can also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud storage, big data, and artificial intelligence platforms. In other embodiments, the computer device can also be a terminal device, which can be a mobile device, terminal, handheld device, computing device, etc.

[0035] In other embodiments, the method can also be applied to an implementation environment consisting of computer equipment and servers, or an implementation environment consisting of terminal equipment and servers. Furthermore, this river ecological management method can also be implemented by a processor calling computer-readable instructions stored in memory.

[0036] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0037] Please see the appendix Figure 1 The above is a flowchart illustrating a river ecological management method as an exemplary embodiment of this application. Figure 1 As shown, the river ecological management method in this embodiment may include the following steps S101~S103:

[0038] S101: Obtain an ecological dataset, which includes multiple ecological data, including four-dimensional data formed by the comprehensive ecological health index of the target river section during the target historical period and the runoff, sediment concentration and water temperature in the river monitoring data; the comprehensive ecological health index is determined based on the river monitoring data.

[0039] Here, river monitoring data includes topographic data and hydrological data. Topographic data refers to multi-period, large-section topographic data reflecting the evolution of river channel morphology, while hydrological data refers to long-term hydrological yearbook data recording water and sediment changes. Regarding water temperature data in the river monitoring data, it is obtained from meteorological station measurement records. It should be noted that due to the difficulty in obtaining water temperature observation data at key sections due to their data characteristics, the inlet water temperature is calculated equivalently from air temperature data.

[0040] In this embodiment, river monitoring data may include runoff, sediment content, water temperature, water flow velocity, and water flow depth.

[0041] For example, ecological data can be expressed by the following formula express, For the first i One ecological data point, For runoff, For sand content, Water temperature It is a comprehensive ecological health index.

[0042] The following is combined Figure 2 The process of determining the comprehensive ecological health index is explained in detail.

[0043] Figure 2 A flowchart illustrating the determination of a comprehensive ecological health index, provided as an exemplary embodiment of this application. Figure 2 As shown, steps S1011 to S1012 are included:

[0044] S1011: Based on the river monitoring data, determine the weighted available area of ​​habitat, water quality index, and biological integrity index respectively.

[0045] It is understandable that, since the comprehensive ecological health index needs to be considered from multiple perspectives, this application conducts a comprehensive assessment from three perspectives: determining the weighted usable area of ​​habitat, water quality index, and biological integrity index.

[0046] Specifically, when determining the weighted usable area of ​​habitat, water quality index, and biological integrity index based on the river monitoring data, the following steps (1) to (2) may be included:

[0047] (1) For each river section, a habitat suitability index is determined based on the water flow velocity, the water flow depth, the sediment content and the water temperature, and the weighted usable area of ​​the habitat is determined based on the area of ​​each river section and the corresponding habitat suitability index.

[0048] First, it should be noted that since the target river section is usually quite long and the species’ preference for habitat varies throughout the river section, this embodiment divides the target river section into multiple river section areas.

[0049] For each river section, a habitat suitability model was pre-constructed when determining its habitat suitability index. This model integrates hydrodynamic factors (flow velocity and flow depth) with aquatic environmental factors (temperature, dissolved oxygen, etc.) to construct a multi-habitat parameter evaluation system to quantify the species' preference for habitat. The higher the habitat suitability index, the higher the probability of the species appearing.

[0050] The habitat suitability model uses a standardization method to convert each habitat parameter into a suitability index (SI) in the range of 0 to 1. An extreme value of 0 indicates that the habitat conditions are completely unsuitable for the species to survive, while an extreme value of 1 indicates that the habitat conditions are most suitable for the species to survive.

[0051] In this embodiment, based on a comprehensive trade-off of ecological priority, management cost, data accessibility, and ecological resilience, four habitat parameters—water flow depth, water flow velocity, sediment concentration, and water temperature—are selected. A suitability index is then calculated using suitability curves of typical species (usually fish). That is, each habitat parameter corresponds to a suitability curve. For each river section, the suitability index corresponding to each habitat parameter is first determined based on the water flow velocity, water flow depth, sediment content and water temperature, as well as the corresponding suitability curves. Then, the habitat suitability index (HIS) is determined based on each suitability index. In this way, the habitat suitability index corresponding to each river section can be obtained. Then, the area of ​​each river section and the corresponding habitat suitability index are used to determine the weighted usable area of ​​the habitat.

[0052] As shown in formula (1), the expression for the weighted available area of ​​habitat is:

[0053] (1)

[0054] Where WUA represents the weighted area of ​​habitat available, and HIS represents the habitat suitability index. i For the river section area, River section area i The area.

[0055] Thus, the response relationship of habitat weighted available area (WUA) to runoff Q, sediment concentration Cs, and water temperature T can be obtained through the above formula.

[0056] In some implementations, the flow velocity and flow depth of each river segment are determined by the following steps: obtaining cross-sectional topographic data of the river segment, and determining the flow velocity and flow depth using a cross-sectional morphology correction model based on the runoff and cross-sectional topographic data of the river segment.

[0057] In this embodiment, given the known runoff Q of a certain cross section of a rectangular river channel and the measured topography of that cross section, assuming that the runoff Q remains unchanged when the rectangular river channel is transformed into the measured topography, and that the vertical velocity distribution of the measured topography cross section satisfies the logarithmic law, the distribution of the water flow velocity v and the water flow depth h in the measured topography cross section can be obtained.

[0058] Please refer to formula (2), which is the expression for the runoff of a rectangular river channel:

[0059] (2)

[0060] Where v is the water flow velocity. b For the river width, h This refers to the depth of the water flow.

[0061] Please refer to formula (3), which is the expression for the runoff of a river channel of arbitrary shape:

[0062] (3)

[0063] in, , H is the cross-sectional depth of the water flow, and H is the cross-sectional height. The height of the bed surface at the transverse position x of the cross-section. v ( x , y () represents the cross-sectional velocity. , , g It is the acceleration due to gravity. Let y be the resistance gradient and y be the vertical distance from the bed surface. D 50 The median particle size, .

[0064] Based on the above formulas (1) and (3), the WUA of the target river segment can be calculated using the cross-sectional morphology correction model, as shown in formula (4):

[0065] (4)

[0066] Thus, based on formula (4) and combined with the location of the control stations, a second interpolation is performed to determine the region of each river segment along the course. to divide the various river sections Summing these values ​​yields the weighted usable area (WUA) of habitat for the target river segment.

[0067] (2) Based on the runoff, the sediment content and the water temperature, the water quality index is obtained by using a water environment prediction model to predict water quality, and the biological integrity index is obtained by using a planktonic animal and plant prediction model to predict ecology.

[0068] As is well known, the Index of Biotic Integrity (IBI) is a comprehensive biological indicator for evaluating the ecological environment of rivers. It includes the Phytoplanktonic Index of Biotic Integrity (P-IBI) based on phytoplankton and the Zoolankton Index of Biotic Integrity (Z-IBI) based on zooplankton.

[0069] To reveal the driving mechanism of water quality index (WQI) and phytoplankton community integrity index (IBI) on water, sediment, and temperature (T), this application employs the C5.0 decision tree algorithm to construct classification prediction models between WQI, P-IBI, Z-IBI, and runoff Q, sediment concentration Cs, and water temperature T. C5.0 is an improved version of the ID3 and C4.5 algorithms. It evaluates the purity of attribute splitting through information entropy calculation and uses the gain ratio to correct for information gain bias, making it particularly suitable for scenarios with continuous independent variables such as runoff and sediment concentration.

[0070] In the implementation, the "C50" package in R was used to build the model. Due to the limited data volume in this study, no cross-validation dataset was set. During model training, the Adaptive Boosting option was set to generate multiple weak classifiers and ensemble them to improve overall performance. At the same time, pruning was enabled to control the tree depth through cost complexity parameters to avoid overfitting.

[0071] Among them, the prediction model for WQI uses the Water Quality Identification Index (WQI) to reflect the river water quality status, as shown in formula (5):

[0072] (5)

[0073] Where n represents the number of water quality indicators. Let be the normalized value of the i-th water quality indicator.

[0074] Thus, by substituting the runoff, sediment content, and water temperature into formula (2), the water quality index can be obtained.

[0075] Similarly, by substituting runoff, sediment concentration, and water temperature into the P-IBI prediction model, the phytoplankton integrity index can be obtained. Similarly, by substituting runoff, sediment concentration, and water temperature into the Z-IBI prediction model, the zooplankton integrity index can be obtained. Thus, based on the phytoplankton integrity index and the zooplankton integrity index, the biological integrity index can be obtained.

[0076] In practice, WQI, P-IBI, and Z-IBI are discretized according to surface water environmental quality standards and ecological assessment standards, and are divided into several levels.

[0077] To eliminate dimensional differences and improve model convergence efficiency, this embodiment also pre-standardizes runoff, sediment concentration, and water temperature using Z-score, and performs subsequent processing based on the standardized runoff, sediment concentration, and water temperature.

[0078] Based on the above analysis, it can be seen that this application has constructed functions for water quality index with runoff, sediment concentration, and water temperature, respectively. Zooplankton integrity index as a function of runoff, sediment concentration, and water temperature Phytoplankton integrity index as a function of runoff, sediment content, and water temperature and the function of habitat-weighted available area with respect to runoff, sediment concentration, and water temperature. As shown in formulas (6) to (9):

[0079] (6)

[0080] (7)

[0081] (8)

[0082] (9)

[0083] in, Q For runoff, For sand content, T This refers to the water temperature.

[0084] S1012: Determine the comprehensive ecological health index based on the weighted available area of ​​the habitat, the water quality index, and the biological integrity index.

[0085] As shown in formula (10), this is the expression for the comprehensive ecological health index.

[0086] (10)

[0087] in, E To achieve a comprehensive ecological health index, It is a function of the comprehensive ecological health index, runoff, sediment content, and water temperature.

[0088] Thus, the comprehensive ecological health index can be determined by substituting the habitat weighted available area, water quality index and biological integrity index determined in the aforementioned steps into formula (10).

[0089] In this embodiment of the application, to determine the threshold values ​​for runoff, sediment concentration, and water temperature, it is first necessary to establish the relationship between the comprehensive ecological health index and runoff, sediment concentration, and water temperature. Therefore, this application will... This three-dimensional function is projected onto three coordinate planes to obtain the projection functions shown in formulas (11) to (13).

[0090] (11)

[0091] (12)

[0092] (13)

[0093] That is, formulas (11) to (13) respectively represent the independent mapping relationship between the decoupled comprehensive ecological health index and runoff, sediment content and water temperature.

[0094] Then, the calculation results of the long-term series of the model are grouped according to the demand time scale (such as year, month, quarter, flood season, dry season, etc.), and the instantaneous corresponding data points of the comprehensive ecological health index, runoff, sediment concentration, and water temperature within the time scale are extracted. In this way, the discrete relationships between E and Q, E and Cs, and E and T can be depicted, and thus ecological data can be obtained. .

[0095] S102: The water, sediment, and temperature threshold prediction model is trained based on the multiple ecological data. After training, an implicit function describing the water, sediment, and temperature threshold prediction model is obtained. The implicit function includes a continuous function of the comprehensive ecological health index with respect to runoff, sediment concentration, and water temperature. The continuous function is obtained by learning and fitting the discrete multiple ecological data through the neural differential equation in the water, sediment, and temperature threshold prediction model.

[0096] The water, sediment, and temperature threshold prediction model adopts a network architecture of "encoder-differential equation-decoder". As mentioned above, the obtained ecological data are all discrete points. There is usually noise between these discrete points and it is difficult to define a clear functional relationship. To solve this problem, this application uses a continuous manifold learning method based on neural differential equations to fit the above discrete ecological data and obtain a continuous function of the comprehensive ecological health index with respect to runoff, sediment concentration and water temperature.

[0097] Among them, neural differential equations are used to define feature evolution rules that evolve discrete features into continuous features. These rules guide the model to evolve the features of the input discrete ecological data samples into continuous features that cover the high-dimensional data manifold. This continuous modeling approach incorporates discrete ecological data into a unified differential equation framework, achieving a leap from discrete data to a model.

[0098] Furthermore, the above methods support irregularly sampled data and operate in the full four-dimensional space, avoiding information loss caused by dimensionality reduction.

[0099] In this application, a water, sediment, and temperature threshold prediction model is trained based on multiple ecological data to obtain an implicit function representing the model. Since the implicit function is obtained by training through the fusion of neural differential equations and continuous manifold learning, it can transform discrete ecological data into continuous data. Thus, it can learn the complex and nonlinear intrinsic laws between water, sediment, and temperature conditions and comprehensive ecological health indicators from the ecological dataset. In other words, the essence of the implicit function is a "continuous mapping relationship".

[0100] The detailed process of training the above model will be introduced later.

[0101] S103: Based on the implicit function, determine the runoff threshold range, sediment concentration threshold range, and water temperature threshold range respectively; the runoff threshold range, sediment concentration threshold range, and water temperature threshold range are used for ecological management of the target river section.

[0102] After fitting the implicit function, the function can be solved to obtain the threshold ranges for runoff, sediment concentration, and water temperature, so that ecological management of the target river section can be carried out based on each threshold.

[0103] In this embodiment, a continuous manifold learning method based on neural differential equations is used to fit discrete ecological data samples into a smooth implicit function, avoiding the interpolation distortion problem in the sample gap region of traditional discrete modeling. Simultaneously, this function integrates the nonlinear correlation between runoff, sediment concentration, water temperature, and the comprehensive ecological health index, capturing the synergistic effects of these three factors (such as the dynamic change in the sediment concentration threshold range when water temperature rises). The output threshold more closely matches the true tolerance patterns of the river ecosystem, thus reducing prediction errors.

[0104] In some implementations, after obtaining the implicit function, this application determines the water-sand temperature threshold based on robust inverse problem solving, i.e., solving the equation. ,in, The target variable is related to water, sand, and temperature.

[0105] In this embodiment, the problem to be solved is transformed into the following three optimization problems: , , That is, during the solution process, runoff, sediment concentration, and water temperature are calculated separately. Here, since the entire model is end-to-end differentiable, in some implementations, automatic differentiation techniques can be used to determine the variables to be solved (e.g., q or cs or t The precise gradient of the gradient is obtained, and then the gradient descent method is used to solve it efficiently. In this way, the mutually decoupled threshold ranges of runoff, sediment concentration, and water temperature can be obtained.

[0106] To address the issue of multiple solutions, a multi-starting-point parallel strategy can be adopted, where optimization is started simultaneously from multiple initial points randomly sampled within the domain of the variable to be solved.

[0107] In this way, the water, sediment and temperature thresholds for different ecological suitability levels at different time scales (such as year, quarter, month) can be obtained, and the water, sediment and temperature ranges corresponding to different ecological suitability levels at different time scales can be given, providing a reference for the dynamic management of rivers.

[0108] The following is combined Figure 3 The training process of the water and sediment temperature threshold prediction model described in step S102 is explained in detail.

[0109] Figure 3 A flowchart illustrating the training process of a water-sand temperature threshold prediction model provided as an exemplary embodiment of this application. Figure 3 As shown, the training of the water, sediment, and temperature threshold prediction model based on the aforementioned ecological data includes the following steps S1021~S1024:

[0110] S1021: The encoder performs feature encoding processing on the multiple ecological data respectively to obtain multiple ecological data features.

[0111] In this embodiment, the encoder uses a multi-channel feature extraction method to extract features, obtaining multi-scale features and geometric features, and then performs feature fusion on the multi-scale features and geometric features based on a cross-scale attention fusion mechanism to obtain fused features.

[0112] Specifically, step S1021 includes the following steps (A) to (B):

[0113] (A) For each ecological data, multi-scale feature extraction is performed on the ecological data to obtain multi-scale features, and geometric feature extraction is performed on the ecological data through the encoder to obtain geometric features.

[0114] In this step, when extracting multi-scale features from the ecological data, the multi-scale feature pyramid is used to process the ecological data in parallel at multiple resolutions (such as 1, half, and quarter) through downsampling operations at different ratios, thereby obtaining multi-scale features. In this way, the model can simultaneously obtain the global macro features (low-resolution channel features) and local fine features (high-resolution channel features) of the ecological data.

[0115] Geometric feature extraction includes the following steps:

[0116] (A1) Determine the feature distance between the multiple ecological data, and construct a k-nearest neighbor graph based on the feature distance between the multiple ecological data; the edges in the k-nearest neighbor graph are used to characterize the neighborhood association between ecological data.

[0117] The characteristic distance may include Euclidean distance, and in other embodiments, it may also be Manhattan distance or cosine distance, etc., which are not limited here.

[0118] (A2) Using a graph neural network, the feature information of the ecological data in the k nearest neighbor graph is fused with the feature information of the neighboring ecological data to obtain the geometric features.

[0119] Here, the k-nearest neighbor graph defines the path for the flow of feature information in the graph neural network (GNN). This allows the model to explicitly learn the local neighborhood topological relationships between ecological data, thereby learning the intrinsic geometric relationships of the data manifold.

[0120] (B) A cross-scale attention fusion mechanism is adopted to fuse the multi-scale features with the geometric features to obtain fused features, and the fused features are used as the ecological data features.

[0121] It is understandable that after obtaining multi-scale features and geometric features, a cross-scale attention fusion mechanism can be used for dynamic weighted fusion. This mechanism can intelligently determine whether to focus more on global distribution or local geometry based on the currently obtained multi-scale features and geometric features, thereby generating a fused feature with highly integrated feature information, and using this fused feature as the ecological data feature.

[0122] S1022: Input the multiple ecological data features into the neural differential equation to generate a trajectory equation; the continuous trajectory points contained in the trajectory equation are used to represent continuous ecological data.

[0123] The following section will first introduce the theoretical basis for converting discrete ecological data into continuous streams.

[0124] Assuming all four-dimensional data points (i.e., ecological data) All lie in a potentially smooth low-dimensional manifold In this embodiment, the manifold is not directly parameterized, but rather through a vector field parameterized by a deep neural network. To characterize its geometric structure.

[0125] This vector field forms the right-hand side of an ordinary differential equation (ODE) such that the manifold The following dynamic systems can be used along the virtual "evolution time" s The integral trajectory of " is generated, where the expression of the ordinary differential equation ODE is shown in formula (14):

[0126] (14)

[0127] in, It is a high-dimensional feature vector after continuous evolution.

[0128] Thus, from any initial point Starting from the initial point, which can refer to various ecological data, the trajectory of the next point can be obtained by integrating along the vector field.

[0129] Specifically, multiple ecological data features can be input into the neural differential equation as initial evolutionary states. Based on the feature evolution rules defined by the vector field function, the vector field function can be integrated using an adaptive step-size numerical integration method to generate the trajectory equation.

[0130] The trajectory equation is shown in formula (15):

[0131] (15)

[0132] Thus, all points on the trajectory equation lie on the target manifold.

[0133] S1023: Input the continuous trajectory points in the trajectory equation into the decoder to obtain the prediction result of the comprehensive ecological health index.

[0134] Here, the continuous trajectory points in the trajectory equation can be considered as continuous eigenvectors. , continuous feature vector The data is input into the decoder to obtain the predicted results of the comprehensive ecological health index.

[0135] S1024: Determine the model loss based on the prediction results, and adjust the parameters of the encoder, the neural differential equation, and the decoder based on the model loss until the preset training requirements are met.

[0136] The preset training requirements include model loss convergence or the number of training iterations reaching a preset threshold, which are not specified here.

[0137] It is understandable that after obtaining the prediction results, the model loss can be determined based on the prediction results, and the parameters of the encoder, neural differential equation and decoder can be adjusted based on the model loss until the preset training requirements are met, so as to obtain a trained water and sand temperature threshold prediction model. At the same time, the target implicit continuous function describing the trained water and sand temperature threshold prediction model can be obtained.

[0138] In other words, the target implicit continuous function is the latent mapping relationship from input to output learned and represented by the entire water and sand temperature threshold prediction model (encoder-neural differential equation-decoder).

[0139] In some implementations, when determining the model loss based on the prediction results, the data fitting loss, physical constraint loss, and geometric regularization loss can be determined separately based on the prediction results, and the model loss can be determined based on the data fitting loss, physical constraint loss, and geometric regularization loss.

[0140] In this embodiment, in order to balance the data fitting quality and physical consistency, a composite loss function (i.e., the model loss function) based on data fitting loss, physical constraint loss and geometric regularization loss is constructed as shown in formula (16):

[0141] (16)

[0142] in, For model loss, For data fitting loss, For geometric regularization loss, For physical constraint losses, the weighting coefficients of each loss are set according to actual needs.

[0143] In this implementation, the training process employs a three-stage strategy: initially focusing on global trend learning (primary optimization data fitting loss), then introducing geometric constraints to refine local structures (primary optimization geometric regularization loss), and finally strengthening physical constraints to improve the rationality and interpretability of the solution (primary optimization physical constraint loss). This strategy is achieved by adjusting the weight coefficients of each loss term; for example, increasing the weight coefficient of the primary optimization data fitting loss. The value of .

[0144] The data fitting loss measures the deviation between the model's predicted values ​​and the true ecological health index (ground value). Optionally, the data fitting loss can be enhanced by introducing adaptive weights based on local density and uncertainty to focus on sparse or highly uncertain regions.

[0145] Geometric regularization loss is based on the continuous feature vector corresponding to the prediction result. The geometric constraints of the trajectory equation (such as manifold smoothness, continuity, and curvature constraints) are determined, aiming to enhance the global smoothness of the implicit function by penalizing abnormal curvature of the data manifold. This loss term helps to suppress overfitting during model training and effectively smooths abnormal fluctuations in prediction results, ensuring that the function's behavior is continuous and stable throughout the feature space.

[0146] Specifically, for each ecological data point including runoff, sediment concentration, and water temperature, a Jacobian matrix can be calculated, and the geometric regularization loss can be determined based on the Jacobian matrix. For example, for... First, calculate the Jacobian matrix. Then penalize the Jacobian matrix (gradient vector). ,in, .

[0147] The physical constraint loss is determined based on the prediction results and the pre-set river ecological physical constraints. These constraints are quantifiable and differentiable pre-defined conditions based on the inherent laws of the river ecosystem, hydrological and hydraulic principles, and prior knowledge in the field. The core is to ensure that the model output conforms to objective ecological and physical common sense. For example, prior knowledge in the field (such as "the comprehensive ecological health index E shows a single-peak curve relationship of first increasing and then decreasing with the increase of runoff Q") can be transformed into differentiable mathematical constraints and integrated into the model optimization process.

[0148] For example, regarding the physical law that "E decreases monotonically with increasing sediment concentration Cs", it is expected that this law will be satisfied throughout the entire prediction domain of the model. Therefore, a penalty can be imposed on regions that do not meet this condition, and the calculation method is as follows: The ReLU function ensures that only when A non-zero penalty value is generated only when the time is right, thus realizing a one-sided constraint.

[0149] Similarly, other physical constraints, such as specific boundary conditions and the concavity / convexity of functions, are also implemented by constructing corresponding differentiable expressions and incorporating physical constraint losses.

[0150] Corresponding to the aforementioned embodiments of the river ecological management method, this application also provides embodiments of the river ecological management method apparatus.

[0151] Please refer to Figure 4 This is a schematic diagram illustrating the structure of a river ecological management method apparatus, as shown in an exemplary embodiment of this application. Figure 4 As shown, the river ecological management method device 400 includes:

[0152] The dataset acquisition module 410 is used to acquire an ecological dataset, which includes multiple ecological data, including a four-dimensional dataset formed by the comprehensive ecological health index of the target river section during the target historical period and the runoff, sediment concentration, and water temperature in the river monitoring data; the comprehensive ecological health index is determined based on the river monitoring data.

[0153] The model training module 420 is used to train the water, sediment, and temperature threshold prediction model based on the multiple ecological data. After training, an implicit function describing the water, sediment, and temperature threshold prediction model is obtained. The implicit function includes a continuous function of the comprehensive ecological health index with respect to runoff, sediment concentration, and water temperature. The continuous function is obtained by learning and fitting the discrete multiple ecological data through the neural differential equation in the water, sediment, and temperature threshold prediction model.

[0154] The threshold calculation module 430 is used to determine the runoff threshold range, sediment concentration threshold range, and water temperature threshold range based on the implicit function; the runoff threshold range, sediment concentration threshold range, and water temperature threshold range are used for ecological management of the target river section.

[0155] In some embodiments, the water-sand temperature threshold prediction model further includes an encoder and a decoder; the model training module 420 is specifically used for:

[0156] The encoder performs feature encoding on the multiple ecological data to obtain multiple ecological data features;

[0157] The multiple ecological data features are input into the neural differential equation to generate a trajectory equation; the continuous trajectory points contained in the trajectory equation are used to represent continuous ecological data.

[0158] The continuous trajectory points in the trajectory equation are input into the decoder to obtain the prediction result of the comprehensive ecological health index;

[0159] The model loss is determined based on the prediction results, and the parameters of the encoder, the neural differential equation, and the decoder are adjusted based on the model loss until the preset training requirements are met.

[0160] In some implementations, the model training module 420 is specifically used for:

[0161] For each piece of ecological data, multi-scale feature extraction is performed on the ecological data to obtain multi-scale features, and geometric feature extraction is performed on the ecological data through the encoder to obtain geometric features;

[0162] A cross-scale attention fusion mechanism is adopted to fuse the multi-scale features and the geometric features to obtain fused features, and the fused features are used as the ecological data features.

[0163] In some implementations, the model training module 420 is specifically used for:

[0164] Determine the Euclidean distance between the multiple ecological data points, and construct a k-nearest neighbor graph based on the Euclidean distance between the multiple ecological data points; the edges in the k-nearest neighbor graph are used to represent the neighborhood associations between the ecological data points;

[0165] Using a graph neural network, the feature information of the ecological data in the k-nearest neighbor graph is fused with the feature information of its neighboring ecological data to obtain the geometric features.

[0166] In some implementations, one term of the neural differential equation is composed of a vector field used to define the feature evolution rules;

[0167] In some implementations, the model training module 420 is specifically used for:

[0168] The multiple ecological data features are respectively input into the neural differential equation as initial evolution states. Based on the feature evolution rules defined by the vector field function, the vector field function is integrated using an adaptive step-size numerical integration method to generate the trajectory equation.

[0169] In some implementations, the model training module 420 is specifically used for:

[0170] Based on the prediction results, the data fitting loss, physical constraint loss, and geometric regularization loss are determined respectively, and the total loss is determined based on the data fitting loss, the physical constraint loss, and the geometric regularization loss.

[0171] In some implementations, the model training module 420 is specifically used for:

[0172] Based on the prediction results and the corresponding true values, the data fitting loss is determined;

[0173] Based on the prediction results and the preset river ecological physical constraints, the physical constraint loss is determined;

[0174] The geometric regularization loss is determined based on the continuous feature vectors corresponding to the prediction results and the geometric constraints of the trajectory equation.

[0175] In some implementations, the dataset acquisition module 410 is specifically used for:

[0176] Based on the river monitoring data, the weighted usable area of ​​habitat, water quality index, and biological integrity index were determined respectively.

[0177] The comprehensive ecological health index is determined based on the weighted available area of ​​the habitat, the water quality index, and the biological integrity index.

[0178] In some embodiments, the target river segment includes multiple river segment areas, and the river monitoring data also includes water flow velocity and water flow depth; the dataset acquisition module 410 is specifically used for:

[0179] For each river section, a habitat suitability index is determined based on the water flow velocity, water flow depth, sediment content, and water temperature. The weighted usable area of ​​the habitat is determined based on the area of ​​each river section and the corresponding habitat suitability index.

[0180] Based on the runoff, sediment content, and water temperature, the water quality index is obtained by using a water environment prediction model to predict water quality, and the biological integrity index is obtained by using a planktonic flora and fauna prediction model to predict ecology.

[0181] In some implementations, the dataset acquisition module 410 determines the water flow velocity and water flow depth for each river section through the following steps:

[0182] Obtain cross-sectional topographic data of the river section area;

[0183] Based on the runoff of the river section and the cross-sectional topographic data, the water flow velocity and the water flow depth are determined using a cross-sectional morphology correction model.

[0184] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0185] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0186] Corresponding to the above-described river ecological management method, this disclosure also provides a computer device, such as... Figure 5The diagram shown is a structural schematic of a computer device provided in an embodiment of this disclosure. Figure 5 As shown, the computer device 500 includes a processor 510, an internal bus 520, memory 530, a network interface 540, and non-volatile memory 550, and may also include other hardware required for its functions. One or more embodiments of this specification can be implemented in software, for example, the processor 510 reads the corresponding computer program from the non-volatile memory 550 into the memory 530 and then runs it. Of course, besides software implementation, one or more embodiments of this specification do not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution entity of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0187] The memory 530, also known as internal memory, is used to temporarily store the computational data in the processor 510, as well as the data exchanged with non-volatile memory 550 such as hard disk. The processor 510 exchanges data with non-volatile memory 550 through the memory 530.

[0188] In this embodiment, memory 530 is specifically used to store application code that executes the solution of this application, and its execution is controlled by processor 510. That is, when the computer device is running, processor 510 communicates with network interface 540, memory 530 and non-volatile memory 550 through internal bus 520, so that processor 510 executes the application code stored in memory 530 and non-volatile memory 550, thereby executing the river ecological management method described in the above method embodiment.

[0189] Processor 510 may be an integrated circuit chip with signal processing capabilities. The aforementioned processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware microservices. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor.

[0190] It is understood that the structures illustrated in the embodiments of this application do not constitute a specific limitation on the computer device 500. In other embodiments of this application, the computer device 500 may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0191] This disclosure also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the river ecological management method described in the above-described method embodiments. The storage medium may be a volatile or non-volatile computer-readable storage medium.

[0192] This disclosure also provides a computer program product carrying program code. The program code includes instructions that can be used to execute the steps of the river ecological management method in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.

[0193] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0194] The embodiments of the subject matter and functional operation described in this specification can be implemented in the following ways: digital electronic circuits, tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or combinations thereof. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible, non-transitory program carrier for execution by a data processing apparatus or for controlling the operation of a data processing apparatus. Alternatively or additionally, the program instructions may be encoded on artificially generated propagation signals, such as machine-generated electrical, optical, or electromagnetic signals, which are generated to encode information and transmit it to a suitable receiving device for execution by the data processing apparatus. The computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or combinations thereof.

[0195] The processing and logic flow described in this specification can be executed by one or more programmable computers that execute one or more computer programs to perform corresponding functions by operating on input data and generating output. The processing and logic flow can also be executed by dedicated logic circuitry—such as FPGAs (Field-Programmable Gate Arrays) or ASICs (Application-Specific Integrated Circuits), and the device can also be implemented as dedicated logic circuitry.

[0196] Computers suitable for executing computer programs include, for example, general-purpose and / or special-purpose microprocessors, or any other type of central processing unit. Typically, the central processing unit receives instructions and data from read-only memory and / or random access memory. Basic computer microservices include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices for storing data, such as disks, magneto-optical disks, or optical disks, or the computer will be operatively coupled to such mass storage devices to receive data from or transfer data to them, or both. However, a computer is not required to have such devices. Furthermore, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name a few.

[0197] Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, such as semiconductor memory devices (e.g., EPROM, EEPROM, and flash memory devices), magnetic disks (e.g., internal hard disks or removable disks), magneto-optical disks, and CD-ROM and DVD-ROM disks. Processors and memory may be supplemented by or incorporated into dedicated logic circuitry.

[0198] While this specification contains numerous specific implementation details, these should not be construed as limiting the scope of any invention or the scope of the claims, but rather are primarily intended to describe features of specific embodiments of a particular invention. Certain features described in the various embodiments herein may also be implemented in combination in a single embodiment. Conversely, various features described in a single embodiment may also be implemented separately in various embodiments or in any suitable sub-combination. Furthermore, while features may function in certain combinations as described above and even initially claimed in this way, one or more features from a claimed combination may be removed from that combination in some cases, and a claimed combination may refer to a sub-combination or a variation thereof.

[0199] Similarly, although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and microservices in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program microservices and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0200] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims may be performed in a different order and still achieve the desired result. Furthermore, the processes depicted in the drawings are not necessarily shown in a specific order or sequence to achieve the desired result. In some implementations, multitasking and parallel processing may be advantageous.

[0201] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method of river ecological management, characterized in that, include: Obtain an ecological dataset, which includes multiple ecological data, including four-dimensional data formed by the comprehensive ecological health index of the target river section during the target historical period and the runoff, sediment content and water temperature in the river monitoring data; The comprehensive ecological health index is determined based on the river monitoring data; The water, sediment, and temperature threshold prediction model is trained based on the aforementioned multiple ecological data. After training, an implicit function describing the water, sediment, and temperature threshold prediction model is obtained. The implicit function includes a continuous function of the comprehensive ecological health index with respect to runoff, sediment concentration, and water temperature. The continuous function is obtained by learning and fitting the discrete multiple ecological data through the neural differential equation in the water, sediment, and temperature threshold prediction model. Based on the implicit function, the threshold ranges for runoff, sediment concentration, and water temperature are determined respectively; these threshold ranges are used for ecological management of the target river section. The water and sediment temperature threshold prediction model further includes an encoder and a decoder; the training of the water and sediment temperature threshold prediction model based on the multiple ecological data includes: The encoder performs feature encoding on the multiple ecological data to obtain multiple ecological data features; The multiple ecological data features are input into the neural differential equation to generate a trajectory equation; the continuous trajectory points contained in the trajectory equation are used to represent continuous ecological data; the continuous trajectory points in the trajectory equation are input into the decoder to obtain the prediction result of the comprehensive ecological health index; the model loss is determined based on the prediction result, and the parameters of the encoder, the neural differential equation and the decoder are adjusted based on the model loss until the preset training requirements are met.

2. The method according to claim 1, characterized in that, The encoder performs feature encoding on the multiple ecological data to obtain multiple ecological data features, including: For each piece of ecological data, multi-scale feature extraction is performed on the ecological data to obtain multi-scale features, and geometric feature extraction is performed on the ecological data through the encoder to obtain geometric features; A cross-scale attention fusion mechanism is adopted to fuse the multi-scale features and the geometric features to obtain fused features, and the fused features are used as the ecological data features.

3. The method of claim 2, wherein, The geometric feature extraction of the ecological data includes: Determine the Euclidean distance between the multiple ecological data points, and construct a k-nearest neighbor graph based on the Euclidean distance between the multiple ecological data points; the edges in the k-nearest neighbor graph are used to represent the neighborhood associations between the ecological data points; Using a graph neural network, the feature information of the ecological data in the k-nearest neighbor graph is fused with the feature information of its neighboring ecological data to obtain the geometric features.

4. The method of claim 1, wherein, One term of the neural differential equation is composed of a vector field function, which is used to define feature evolution rules; the step of inputting the multiple ecological data features into the neural differential equation to generate a trajectory equation containing multiple continuous trajectory points includes: The multiple ecological data features are respectively used as initial evolutionary states and input into the neural differential equation. Based on the feature evolution rules defined by the vector field function, the vector field function is integrated using an adaptive step-size numerical integration method to generate the trajectory equation.

5. The method of claim 1, wherein, Determining the model loss based on the prediction results includes: Based on the prediction results, the data fitting loss, physical constraint loss, and geometric regularization loss are determined respectively, and the total loss is determined based on the data fitting loss, the physical constraint loss, and the geometric regularization loss.

6. The method of claim 5, wherein, The step of determining the data fitting loss, physical constraint loss, and geometric regularization loss based on the prediction results includes: Based on the prediction results and the corresponding true values, the data fitting loss is determined; Based on the prediction results and the preset river ecological physical constraints, the physical constraint loss is determined; The geometric regularization loss is determined based on the continuous feature vectors corresponding to the prediction results and the geometric constraints of the trajectory equation.

7. The method of claim 1, wherein, The comprehensive ecological health index is determined through the following steps: Based on the river monitoring data, the weighted usable area of ​​habitat, water quality index, and biological integrity index were determined respectively. The comprehensive ecological health index is determined based on the weighted available area of ​​the habitat, the water quality index, and the biological integrity index.

8. The method of claim 7, wherein, The target river segment includes multiple river segments, and the river monitoring data also includes water flow velocity and water flow depth; based on the river monitoring data, the weighted usable area of ​​habitat, water quality index, and biological integrity index are determined, including: For each river section, a habitat suitability index is determined based on the water flow velocity, water flow depth, sediment content, and water temperature. The weighted usable area of ​​the habitat is determined based on the area of ​​each river section and the corresponding habitat suitability index. Based on the runoff, sediment content, and water temperature, the water quality index is obtained by using a water environment prediction model to predict water quality, and the biological integrity index is obtained by using a planktonic flora and fauna prediction model to predict ecology.

9. The method of claim 8, wherein, The flow velocity and depth for each river section are determined through the following steps: Obtain cross-sectional topographic data of the river section area; Based on the runoff of the river section and the cross-sectional topographic data, the water flow velocity and the water flow depth are determined using a cross-sectional morphology correction model.

10. A river ecology management device, characterized by, include: The dataset acquisition module is used to acquire an ecological dataset, which includes multiple ecological data, including four-dimensional data formed by the comprehensive ecological health index of the target river section during the target historical period and the runoff, sediment content and water temperature in the river monitoring data. The comprehensive ecological health index is determined based on the river monitoring data; The model training module is used to train the water, sediment, temperature, and threshold prediction model based on the multiple ecological data. After training, an implicit function describing the water, sediment, temperature, and threshold prediction model is obtained. The implicit function includes a continuous function of the comprehensive ecological health index with respect to runoff, sediment concentration, and water temperature. The continuous function is obtained by learning and fitting the discrete multiple ecological data through the neural differential equation in the water, sediment, temperature, and threshold prediction model. The threshold calculation module is used to determine the runoff threshold range, sediment concentration threshold range, and water temperature threshold range based on the implicit function; the runoff threshold range, sediment concentration threshold range, and water temperature threshold range are used for ecological management of the target river section; The water-sand temperature threshold prediction model further includes an encoder and a decoder; the model training module is specifically used for: The encoder performs feature encoding on the multiple ecological data to obtain multiple ecological data features; the multiple ecological data features are input into the neural differential equation to generate a trajectory equation; the continuous trajectory points contained in the trajectory equation are used to represent continuous ecological data; the continuous trajectory points in the trajectory equation are input into the decoder to obtain the prediction result of the comprehensive ecological health index; the model loss is determined based on the prediction result, and the parameters of the encoder, the neural differential equation and the decoder are adjusted based on the model loss until the preset training requirements are met.

11. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the river ecological management method according to any one of claims 1-9.

12. A computer readable storage medium having stored thereon a computer program, characterized in that, When the program is executed by the processor, it implements the steps of the river ecological management method according to any one of claims 1-9.

Images