Method and system for predicting water and salt migration of full-film double-furrow saline-alkali soil

Abstract

The invention relates to a method and a system for predicting water and salt migration of full-film double furrows of saline-alkali soil. The method comprises the following steps: dividing a to-be-predicted region into a plurality of consistent coordinate regions based on water and salt migration key factors; training a water-salt migration first-order predictor for each coordinate area; constructing a graph neural network taking a coordinate region as a node as a second-order predictor, and integrating first-order prediction results to predict water and salt distribution of the whole region; and embedding the trained first-order predictor into a second-order predictor framework to form a water-salt migration predictor, and executing saline-alkali land full-film double-furrow water-salt migration prediction of the to-be-predicted area. According to the method, the technical problems that in the prior art, saline-alkali soil water-salt migration prediction precision is insufficient or the calculation power demand is large, and scene mobility is poor are solved, and the technical effects that the prediction precision is guaranteed, meanwhile, the calculation complexity is reduced, and the model generalization ability and the scene adaptability are improved are achieved.

Description

Technical Field

[0001] The invention relates to the field of water-salt migration analysis, and in particular to a prediction method and system for water-salt migration in double-ridge ditch with full-film film on saline-alkali land. Background Art

[0002] Saline-alkali land is a type of land with severe soil salinization. The study of its water-salt migration law is of great significance for improving saline-alkali land, improving land utilization and agricultural productivity. Accurately predicting the water-salt migration process in saline-alkali land can provide a scientific basis for the efficient use of water resources and provide technical support for the comprehensive development and utilization of saline-alkali land.

[0003] At present, there are two main types of prediction methods for water and salt migration in saline-alkali land. One is a method based on physical models, which realizes prediction by constructing linear equations for water and salt migration. This method is simple in principle and has low computational complexity, but because linear equations are difficult to fully describe the complex nonlinear water and salt migration process in the actual environment, the prediction accuracy is insufficient and it is difficult to meet the needs of actual applications. The other is a numerical simulation method based on grid segmentation, which divides the area to be analyzed into grids and calculates water and salt migration in units of grids. If accurate prediction results are to be obtained, the grid accuracy needs to be increased, resulting in a significant increase in computing power requirements. At the same time, since the soil structure, chemical composition and environmental conditions of saline-alkali land in various places are different, this method is difficult to adapt to different saline-alkali land environments, and the scene portability is poor. Summary of the invention

[0004] The present invention aims to solve the technical problems in the prior art of insufficient prediction accuracy of water and salt migration in saline-alkali land, large computing power requirements and poor scene portability, and provides a prediction method and system for water and salt migration in full-film double-ridge ditch in saline-alkali land to solve the problems.

[0005] The technical solution of the present invention to solve the above technical problems is as follows: In the first aspect, the present invention provides a method for predicting water and salt migration in double-ridge ditch with full-film film on saline-alkali land, comprising: aggregating and segmenting the area to be predicted based on the key factors of water and salt migration to obtain the segmentation results of the area to be predicted, wherein the segmentation results of the area to be predicted include a plurality of coordinate areas, and the key factors of water and salt migration belonging to the same coordinate area are consistent; traversing the plurality of coordinate areas, placing the key factors of water and salt migration in the input layer and placing the water and salt distribution labels of a plurality of sub-areas in the output layer based on a plurality of water and salt initial distribution information, and training a plurality of first-order water and salt migration models. Predictor; based on graph neural network, taking the several coordinate areas as neuron nodes, performing topological simulation on the segmentation results of the area to be predicted, building a second-order water-salt migration predictor architecture, setting the outputs of the several first-order water-salt migration predictors as the outputs of the neuron nodes, placing the overall regional water-salt distribution identifier in the output layer, and training the second-order water-salt migration predictor; using the several first-order water-salt migration predictors to replace the neuron nodes of the second-order water-salt migration predictor, and obtaining a water-salt migration predictor to perform water-salt migration prediction in the full-film double-ridge ditch of the saline-alkali land in the area to be predicted.

[0006] Optionally, based on the key factors of water and salt migration, the area to be predicted is aggregated and segmented to obtain the segmentation result of the area to be predicted, which includes: based on the water and salt migration influencing factors preset by the user end, collecting the water and salt migration monitoring logs of the first fixed area, wherein any one of the water and salt migration monitoring logs includes the water and salt migration influencing factor record value and the water and salt distribution topology record value; performing pairwise same-attribute deviation calculations on the water and salt migration monitoring logs to obtain a water and salt migration influencing factor record deviation module value set and a water and salt distribution topology record value deviation set; performing water and salt migration correlation sorting on the water and salt migration influencing factors according to the water and salt distribution topology record value deviation set and the water and salt migration influencing factor record deviation module value set to obtain the water and salt migration key factors.

[0007] Optionally, the water-salt migration influencing factors are sorted according to the water-salt distribution topology record value deviation set and the water-salt migration influencing factor record deviation modulus value set to obtain the water-salt migration key factors, including: building a benchmark data sequence according to the water-salt distribution topology record value deviation set; building a first comparison sequence according to the first attribute water-salt migration influencing factor record deviation modulus value set of the water-salt migration influencing factor record deviation modulus value set, until building an Nth comparison sequence according to the Nth attribute water-salt migration influencing factor record deviation modulus value set; constructing a grey correlation matrix from the first comparison sequence to the Nth comparison sequence according to the benchmark data sequence, performing grey correlation analysis, and extracting the water-salt migration key factors having a correlation greater than or equal to a correlation threshold.

[0008] Optionally, the water-salt transport influencing factors are sorted by water-salt transport correlation according to the water-salt distribution topology record value deviation set and the water-salt transport influencing factor record deviation modulus set to obtain the water-salt transport key factor, including: sorting the water-salt transport influencing factors by water-salt transport correlation according to the water-salt distribution topology record value deviation set and the water-salt transport influencing factor record deviation modulus set to obtain the first fixed area water-salt transport key factor; updating the first fixed area cyclic analysis until the Mth fixed area water-salt transport key factor is obtained, M is an integer, M≥100; counting the first fixed area water-salt transport key factor until the trigger frequency of the Mth fixed area water-salt transport key factor is greater than or equal to the trigger frequency threshold of the water-salt transport key factor, wherein the trigger frequency threshold is greater than or equal to The floor value of .

[0009] Optionally, based on the key factors of water and salt migration, the area to be predicted is aggregated and segmented to obtain the segmentation result of the area to be predicted, including: extracting the key factors of water and salt migration from the first fixed area water and salt migration key factors to the first key factor association set to the Mth fixed area water and salt migration key factors to the Qth key factor association set; adding the first key factor association set to the Qth key factor association set to obtain the association sum value; traversing the first key factor association set to the Qth key factor association set and summing them respectively to obtain the first key factor association sum to the Qth key factor association sum, and comparing them with the association sum values ​​respectively to obtain the weight distribution of the key factors of water and salt migration; constructing a weighted Euclidean distance evaluation function for the key factors of water and salt migration according to the weight distribution of the key factors of water and salt migration, and aggregating and segmenting the area to be predicted in combination with the Euclidean distance threshold to obtain the segmentation result of the area to be predicted.

[0010] Optionally, the plurality of coordinate regions are traversed, and based on a plurality of water-salt initial distribution information, the water-salt migration key factors are placed in an input layer, and a plurality of sub-region water-salt distribution identifiers are placed in an output layer, and a plurality of first-order predictors of water-salt migration are trained, including: extracting a first coordinate region according to the plurality of coordinate regions, and constructing a spatial coordinate domain; extracting the first water-salt initial distribution information from the plurality of water-salt initial distribution information, distributing it in the spatial coordinate domain, and obtaining a water-salt initial distribution vector; taking the spatial coordinate domain and the water-salt initial distribution vector as constraints, and collecting water-salt migration monitoring data based on the water-salt migration key factors, wherein the water-salt migration monitoring data includes time series information of key factor eigenvalue records and time series information of water-salt migration position records; based on a key factor deviation threshold , segment the key factor characteristic value record time series information in the time domain, obtain the key factor characteristic value record values ​​of the first time zone to the key factor characteristic value record values ​​of the Kth time zone; according to the key factor characteristic value record values ​​of the first time zone to the key factor characteristic value record values ​​of the Kth time zone, extract the water-salt distribution vector at the end time of the first time zone to the water-salt distribution vector at the end time of the Kth time zone from the water-salt migration position record time series information, and build a sub-region water-salt distribution mark; take the key factor characteristic value record values ​​of the first time zone to the key factor characteristic value record values ​​of the Kth time zone as input, take the sub-region water-salt distribution mark as supervision, train the first coordinate region water-salt migration first-order predictor based on the long short-term memory neural network, and add the several water-salt migration first-order predictors.

[0011] Optionally, the first time zone key factor characteristic value record values ​​up to the Kth time zone key factor characteristic value record values ​​are used as input, and the sub-region water-salt distribution identifier is used as supervision to train the first coordinate region water-salt migration first-order predictor, and add the several water-salt migration first-order predictors, including: using K value as the number of neurons, building a first-level neural network architecture based on the long short-term memory neural network; using the first time zone key factor characteristic value record values ​​up to the Kth time zone key factor characteristic value record values ​​are used as input, and the sub-region water-salt distribution identifier is used as supervision to train the first-level neural network architecture to obtain a first-level first coordinate region water-salt migration first-order predictor; when the first-level first coordinate region water-salt migration first-order predictor does not meet the convergence condition, adding The number of neurons is determined by building a two-level neural network architecture based on the long short-term memory neural network to perform cyclic training until the first-order predictor of water and salt migration in the first coordinate area of ​​level Y meets the convergence condition, and the first-order predictor of water and salt migration in the first coordinate area of ​​level Y is set as the first-order predictor of water and salt migration in the first coordinate area.

[0012] In the second aspect, the present invention provides a prediction system for water and salt migration in double-ridge ditch with full film on saline-alkali land, including: a regional segmentation module, which is used to aggregate and segment the area to be predicted based on the key factors of water and salt migration, and obtain the segmentation results of the area to be predicted, wherein the segmentation results of the area to be predicted include several coordinate areas, and the key factors of water and salt migration belonging to the same coordinate area are consistent; a first-order training module, which is used to traverse the several coordinate areas, and based on some water and salt initial distribution information, put the key factors of water and salt migration in the input layer, put some sub-area water and salt distribution labels in the output layer, and train some first-order water and salt migration models. Predictor; a second-order training module, which is used to perform topological simulation on the segmentation results of the area to be predicted based on a graph neural network, with the several coordinate areas as neuron nodes, to build a second-order water-salt migration predictor architecture, set the outputs of the several first-order water-salt migration predictors as the outputs of the neuron nodes, place the overall regional water-salt distribution identifier in the output layer, and train the second-order water-salt migration predictor; a prediction execution module, which is used to use the several first-order water-salt migration predictors to replace the neuron nodes of the second-order water-salt migration predictor, and obtain a water-salt migration predictor to execute the water-salt migration prediction of the full-film double-ridge ditch in the saline-alkali land in the predicted area.

[0013] The beneficial effects of the present invention are: Based on the key factors of water and salt migration, the predicted area is aggregated and segmented to obtain the segmentation results of the predicted area. The complex saline-alkali land area is intelligently segmented according to the key factors of water and salt migration to form several coordinate areas with internal consistency, laying the foundation for subsequent accurate prediction and avoiding the calculation pressure caused by over-refined grids in traditional methods. Traversing several coordinate areas, based on some initial water and salt distribution information, the key factors of water and salt migration are placed in the input layer, and the water and salt distribution labels of several sub-areas are placed in the output layer. Several first-order predictors of water and salt migration are trained so that each predictor can accurately capture the law of water and salt migration in a specific area and improve the local prediction accuracy. Based on the graph neural network, several coordinate regions are used as neuron nodes, and topological simulation is performed on the segmentation results of the predicted area. The second-order water-salt migration predictor architecture is built. The outputs of several first-order water-salt migration predictors are set as the outputs of neuron nodes, and the water-salt distribution identifier of the entire region is placed in the output layer. The second-order water-salt migration predictor is trained, and the outputs of each first-order predictor are integrated through the graph neural network to build a second-order predictor architecture to simulate the mutual influence and overall flow law of water-salt migration between regions, effectively solving the water-salt interaction problem between regions and improving the overall prediction accuracy. Several first-order water-salt migration predictors are used to replace the neuron nodes of the second-order water-salt migration predictor, and the water-salt migration predictor is obtained to perform the water-salt migration prediction of the saline-alkali land with full-film double ridges in the predicted area. By integrating the trained first-order predictor into the second-order predictor architecture, a complete water-salt migration predictor is formed, which realizes the high-precision prediction of the water-salt migration of the entire saline-alkali land with full-film double ridges, while maintaining the adaptability of the model to different scenarios and low computational complexity.

[0014] By combining hierarchical prediction with graph neural networks, we not only avoid the problem of insufficient prediction accuracy of traditional linear equations, but also overcome the shortcomings of grid segmentation methods, such as large amount of calculation and poor scene transferability. We achieve a balance between prediction accuracy and computational efficiency, and improve the adaptability of the model in different saline-alkali land environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of a flow chart of a method for predicting water and salt migration in double-ridge ditch with full-film film on saline-alkali land provided by the present invention; Figure 2 A schematic diagram of the structure of a prediction system for water and salt migration in double-ridge ditch with full-film film on saline-alkali land provided by the present invention.

[0016] In the accompanying drawings, the components represented by the reference numerals are as follows: A region segmentation module 11, a first-order training module 12, a second-order training module 13, and a prediction execution module 14. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0018] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0019] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in the present invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present invention.

[0020] Embodiment 1, as Figure 1 As shown, an embodiment of the present invention provides a method for predicting water and salt migration in double-ridge ditch with full-film coating on saline-alkali land, comprising: S100: Aggregate and segment the area to be predicted based on the key factors of water and salt migration to obtain a segmentation result of the area to be predicted, wherein the segmentation result of the area to be predicted includes a plurality of coordinate areas, and the key factors of water and salt migration belonging to the same coordinate area are consistent.

[0021] Specifically, first, the predicted saline-alkali land area is segmented based on key factors. By analyzing the key factors of water-salt migration (such as soil texture, terrain slope, initial salt content distribution, etc.), the entire predicted area is segmented and aggregated, and areas with similar characteristics are divided into the same coordinate area. After the segmentation is completed, the entire predicted area is divided into several clearly defined coordinate areas, and the key factors of water-salt migration in each coordinate area show consistency.

[0022] Compared with the traditional uniform grid division, the zoning based on the key factors of water and salt migration can better reflect the actual physical characteristics of water and salt migration in saline-alkali land and provide a more reasonable spatial basic unit for the establishment of subsequent predictions.

[0023] S200: traversing the plurality of coordinate regions, placing the key factors of water and salt migration in an input layer based on a plurality of water and salt initial distribution information, placing a plurality of sub-region water and salt distribution identifiers in an output layer, and training a plurality of first-order predictors of water and salt migration.

[0024] Specifically, several obtained coordinate regions are traversed and a unique first-order predictor of water and salt migration is constructed for each coordinate region.

[0025] For each coordinate region, a first-order water-salt migration predictor is constructed based on the collected initial water-salt distribution information. In the first-order water-salt migration predictor, the key factors of water-salt migration in the coordinate region (such as soil characteristics, topographic factors, meteorological conditions, etc.) are set as the input layer parameters of the first-order water-salt migration predictor, and the water-salt distribution identifiers of each sub-region in the coordinate region (i.e., the quantitative representation of the water-salt distribution state) are set as the output layer targets of the first-order water-salt migration predictor.

[0026] By using appropriate machine learning algorithms and training data, corresponding first-order predictors of water and salt transport are trained for each coordinate region. These first-order predictors of water and salt transport can predict the distribution changes of water and salt in the corresponding coordinate region based on the key factor characteristics of water and salt transport in the respective coordinate region. This method of training predictors by region can establish specialized prediction models for regions with different characteristics, improving the accuracy and adaptability of the overall prediction.

[0027] S300: Based on a graph neural network, the plurality of coordinate regions are used as neuron nodes, topological simulation is performed on the segmentation results of the area to be predicted, a second-order water-salt migration predictor architecture is constructed, the outputs of the plurality of first-order water-salt migration predictors are set as the outputs of the neuron nodes, the overall regional water-salt distribution identifier is placed in the output layer, and the second-order water-salt migration predictor is trained.

[0028] Specifically, the coordinate regions obtained in the previous steps are regarded as neuron nodes in the graph network. By performing topological simulation on the segmentation results of the predicted region, a graph structure reflecting the spatial correlation between the coordinate regions is constructed. In this architecture, the output of each neuron node (i.e., coordinate region) is provided by the output result of the trained first-order predictor of water and salt migration. These outputs reflect the prediction of water and salt migration in each independent coordinate region. The output layer of the entire graph neural network is set as the water and salt distribution mark of the entire region, that is, the representation of the global water and salt distribution state.

[0029] Through this hierarchical design, local prediction and global prediction are organically combined, so that the second-order predictor of water and salt migration can fully consider the water-salt interaction between each coordinate area, thereby realizing the comprehensive prediction of the water and salt migration process in the whole area. The training process of this second-order predictor of water and salt migration uses the actual observed water and salt distribution data of the whole area, and minimizes the error between the predicted results and the actual observed results by adjusting the parameters of the graph neural network, thereby obtaining a second-order predictor of water and salt migration with high prediction accuracy.

[0030] S400: Using the plurality of first-order water-salt migration predictors to replace the neuron nodes of the second-order water-salt migration predictor, a water-salt migration predictor is obtained to perform water-salt migration prediction for double-ridges in full-film saline-alkali land in the area to be predicted.

[0031] Specifically, several trained first-order predictors of water and salt transport are integrated into the established second-order predictor framework of water and salt transport to form a complete water and salt transport predictor.

[0032] Specifically, the trained first-order predictors of water and salt migration in each coordinate area are actually replaced into the corresponding neuron nodes of the graph neural network, so that each node is a first-order predictor of water and salt migration with the ability to predict water and salt migration in a specific area. Through the replacement integration operation, a hierarchical and modular water and salt migration predictor is constructed. This water and salt migration predictor can not only accurately characterize the water and salt migration characteristics of each local area, but also capture the water and salt interaction relationship between regions through the topological structure of the graph neural network, thereby realizing a comprehensive and accurate prediction of water and salt migration in the double-ridge ditch of the saline-alkali land with full film in the entire predicted area.

[0033] Through regional aggregation and segmentation based on the key factors of water and salt migration and the construction of a dedicated first-order predictor for water and salt migration, the prediction accuracy is improved and the water and salt migration prediction results are closer to the actual situation; by adopting a hierarchical design of integrating regional predictors with a graph neural network, the computing power requirements are reduced while ensuring accuracy, avoiding the computational burden brought by fine-grained grid calculations; in addition, the zoning prediction method based on the key factors of water and salt migration has a strong scenario migration capability, and can flexibly adapt to saline-alkali land environments with different characteristics, providing more universal technical support for the efficient utilization of water resources and the comprehensive development of saline-alkali land.

[0034] Furthermore, based on the key factors of water and salt migration, the predicted area is aggregated and segmented to obtain the segmentation results of the predicted area, which previously included: S510: Based on the water-salt migration influencing factor preset by the user end, collecting the water-salt migration monitoring log of the first fixed area, wherein any one of the water-salt migration monitoring logs includes a water-salt migration influencing factor record value and a water-salt distribution topology record value; S520: performing pairwise same-attribute deviation calculations on the water-salt migration monitoring logs to obtain a water-salt migration influencing factor record deviation modulus value set and a water-salt distribution topology record value deviation set; S530: sorting the water-salt migration influencing factors according to the water-salt distribution topology record value deviation set and the water-salt migration influencing factor record deviation modulus value set to obtain the water-salt migration key factors.

[0035] In an optional implementation, in order to obtain the key factors of water and salt migration, first, based on the water and salt migration influencing factors pre-set by the user, the water and salt migration monitoring logs are collected in the first fixed area. Each monitoring log in the collected water and salt migration monitoring log contains two types of key information: one is the specific record values ​​of various water and salt migration influencing factors, such as soil texture parameters, groundwater levels, meteorological conditions and other values; the other is the water and salt distribution topology record values ​​at the corresponding moment, that is, the spatial distribution status data of water and salt. These water and salt migration monitoring logs provide basic data support for subsequent analysis.

[0036] Then, the deviation of the same attribute is calculated for each monitoring log in the collected water-salt transport monitoring log. Specifically, the absolute value of the difference between the influencing factor record values ​​of the same attribute in any two monitoring logs is calculated to form a water-salt transport influencing factor record deviation modulus set; at the same time, the difference measure of the water-salt distribution topology record values ​​in the two logs is calculated to form a water-salt distribution topology record value deviation set. By obtaining the water-salt transport influencing factor record deviation modulus set and the water-salt distribution topology record value deviation set, the corresponding relationship between the change of the influencing factor and the change of the water-salt distribution under different times or conditions is quantified.

[0037] Subsequently, based on the obtained water-salt transport influencing factor record deviation modulus value set and water-salt distribution topology record value deviation set, the correlation between each water-salt transport influencing factor and water-salt distribution change is analyzed. Through correlation analysis, several factors with the closest relationship with water-salt transport are selected from the numerous water-salt transport influencing factors, namely, the key factors of water-salt transport. These key factors of water-salt transport are the basic parameters for subsequent regional segmentation and the construction of various predictors.

[0038] Through the above steps, the key factors that have a substantial impact on water and salt transport can be objectively identified, avoiding the subjectivity and uncertainty that may be caused by artificial selection of factors, and providing a basis for subsequent regional segmentation and the establishment of a predictor.

[0039] Furthermore, the water-salt migration influencing factors are sorted by water-salt migration correlation according to the water-salt distribution topology record value deviation set and the water-salt migration influencing factor record deviation modulus value set to obtain the water-salt migration key factors, including: S531: Building a benchmark data sequence according to the water-salt distribution topology record value deviation set; S532: constructing a first comparison sequence according to the first attribute of the water-salt migration influencing factor record deviation modulus value set, until constructing an Nth comparison sequence according to the Nth attribute of the water-salt migration influencing factor record deviation modulus value set; S533: According to the benchmark data sequence, the first comparison sequence to the Nth comparison sequence, a grey correlation matrix is ​​constructed, grey correlation analysis is performed, and the key factors of water-salt migration whose correlation is greater than or equal to a correlation threshold are extracted.

[0040] In a preferred embodiment, first, the obtained water-salt distribution topology record value deviation set is organized into a standardized benchmark data sequence. The benchmark data sequence reflects the degree of change of the water-salt distribution state in the saline-alkali land area under different conditions or time periods, and serves as a reference standard for evaluating the correlation of various influencing factors.

[0041] Then, the set of recorded deviation modulus values ​​of water-salt transport influencing factors is subjected to attribute classification processing. Specifically, the set of recorded deviation modulus values ​​of the first attribute water-salt transport influencing factors (such as soil texture) is organized into the first comparison sequence, and the set of recorded deviation modulus values ​​of the second attribute water-salt transport influencing factors (such as groundwater level) is organized into the second comparison sequence, and so on, until the set of recorded deviation modulus values ​​of the Nth attribute water-salt transport influencing factors is organized into the Nth comparison sequence. Each comparison sequence reflects the changes of a specific attribute water-salt transport influencing factor under different conditions or times.

[0042] Next, based on the benchmark data sequence and N comparison sequences, a grey correlation matrix is ​​constructed, and grey correlation analysis is performed. Specifically, first, the benchmark data sequence (the set of water-salt distribution topology record value deviations) and each comparison sequence (the set of record deviation modulus values ​​of each attribute water-salt migration influencing factor) are preprocessed, including dimensionless and standardized processing, so that data of different dimensions are comparable. Secondly, the correlation coefficient between the benchmark sequence and each comparison sequence is calculated. For each sampling point, the difference between the benchmark sequence value and the corresponding comparison sequence value is calculated, and then the correlation coefficient of the point is calculated by combining the maximum difference and the minimum difference through the grey correlation formula, so as to obtain the degree of correlation between each influencing factor and the change in water-salt distribution at each sampling point. Then, the correlation coefficients of all sampling points of each influencing factor are averaged to obtain the overall grey correlation between the influencing factor and the change in water-salt distribution, which indicates the degree of influence of the influencing factor on water-salt migration. The larger the value, the more significant the influence. After that, the influencing factors with a correlation greater than or equal to the threshold are screened according to the preset correlation threshold, and the influencing factors are determined as key factors of water-salt migration.

[0043] Through grey correlation analysis, we can objectively and quantitatively evaluate the influence of various influencing factors on water and salt migration, so as to scientifically screen out the key factors that really have a significant impact, and provide accurate parameter basis for subsequent regional segmentation and prediction model construction.

[0044] Furthermore, the water-salt migration influencing factors are sorted by water-salt migration correlation according to the water-salt distribution topology record value deviation set and the water-salt migration influencing factor record deviation modulus value set to obtain the water-salt migration key factors, including: S534: sorting the water-salt migration influencing factors according to the water-salt distribution topology record value deviation set and the water-salt migration influencing factor record deviation modulus value set to obtain the water-salt migration key factors of the first fixed area; S535: updating the first fixed area cyclic analysis until the key factor of water-salt migration in the Mth fixed area is obtained, where M is an integer and M≥100; S536: Count the key factors of water and salt migration in the first fixed area until the trigger frequency of the key factors of water and salt migration in the Mth fixed area is greater than or equal to the key factors of water and salt migration whose trigger frequency threshold is greater than or equal to The floor value of .

[0045] Specifically, the reliability of key factor identification is improved by repeating correlation analysis in multiple fixed areas and screening based on trigger frequency statistics.

[0046] First, according to the deviation set of water-salt distribution topology record values ​​and the deviation modulus set of water-salt migration influencing factors, the water-salt migration influencing factors are sorted by correlation to obtain the key factors of water-salt migration in the first fixed area. Here, the key factor identification is performed in a single fixed area, which is the same as the aforementioned correlation analysis method. Then, the analysis object is updated from the first fixed area to other fixed areas, and the correlation analysis process is repeated. This cyclic analysis continues until the analysis of M fixed areas is completed, and the key factors of water-salt migration in each fixed area are obtained. Among them, M is an integer and not less than 100, indicating that verification is required on a sufficient number of regional samples to ensure the statistical significance of the results.

[0047] Then, the analysis results of the M fixed areas are statistically processed. Specifically, the frequency of each water-salt migration influencing factor being identified as a key factor in the M fixed areas is calculated, that is, the trigger frequency. At the same time, a trigger frequency threshold is set, which is not less than For example, when M=100, the trigger frequency threshold is at least 15. Then, the water-salt migration influencing factors with a trigger frequency greater than or equal to the threshold are screened out as the universal key factors of water-salt migration.

[0048] Through multi-regional verification and frequency statistics, key factors that have significant impacts in a variety of saline-alkali land environments can be identified, avoiding the particularity or contingency that may be brought about by single-region analysis, improving the reliability and universality of key factor identification, and providing a more robust parameter basis for subsequent regional segmentation and prediction model construction.

[0049] Furthermore, based on the key factors of water and salt migration, the predicted area is aggregated and segmented to obtain the segmentation results of the predicted area, including: S110: extracting the water-salt migration key factor from the first fixed area water-salt migration key factor to the first key factor association degree set of the Mth fixed area water-salt migration key factor to the Qth key factor association degree set; S120: summing up the first key factor association degree set until the Qth key factor association degree set to obtain an association degree sum value; S130: traverse the first key factor association degree set until the Qth key factor association degree set, add them up respectively, obtain the sum of the first key factor association degrees until the Qth key factor association degree sum, compare them with the association degree sum value respectively, and obtain the weight distribution of water-salt migration key factors; S140: constructing a weighted Euclidean distance evaluation function for the key factors of water and salt migration according to the weight distribution of the key factors of water and salt migration, and aggregating and segmenting the area to be predicted in combination with the Euclidean distance threshold to obtain a segmentation result of the area to be predicted.

[0050] Specifically, when the predicted area is aggregated and segmented based on the key factors of water and salt migration, first, the correlation values ​​of the first key factor to the Qth key factor in the first fixed area to the Mth fixed area are extracted from the key factors of water and salt migration, forming Q key factor correlation sets, that is, the first key factor correlation set to the Qth key factor correlation set. Each key factor correlation set contains the correlation values ​​of a specific key factor in M ​​fixed areas, which reflect the degree of influence of the key factor on water and salt migration. Then, all the correlation values ​​from the first key factor correlation set to the Qth key factor correlation set are added and calculated to obtain the overall correlation sum value. The correlation sum value represents the overall impact intensity of all key factors and serves as the benchmark value for subsequent calculations.

[0051] Subsequently, the sum of the association degree set of each key factor is calculated respectively, that is, the sum of the association degrees of the first key factor to the sum of the association degrees of the Qth key factor. Then, the sum of the association degrees of each key factor is calculated by ratio calculation with the sum of the overall association degrees to obtain the weight distribution of each key factor as the weight distribution of the key factors of water and salt migration. The weight distribution of the key factors of water and salt migration reflects the relative importance of each key factor in the process of water and salt migration. The larger the weight, the more significant the influence of the factor. Afterwards, based on the calculated weight distribution of the key factors of water and salt migration, a weighted Euclidean distance evaluation function is constructed. This function is used to measure the distance between different location points in the key factor feature space in the area to be predicted, and weighted processing is performed on each factor according to its weight. Combined with the preset Euclidean distance threshold, the clustering algorithm is used to aggregate and segment the area to be predicted, and the areas with Euclidean distance less than the threshold are aggregated into the same coordinate area, and finally the segmentation result of the area to be predicted is obtained.

[0052] Through weighted aggregation segmentation based on the weights of key factors, a more reasonable functional zoning of the prediction area can be carried out according to the actual influence of each key factor, ensuring the consistency of water and salt migration characteristics in each coordinate area, laying the foundation for the subsequent construction of the prediction model.

[0053] Furthermore, the plurality of coordinate regions are traversed, and based on the plurality of water-salt initial distribution information, the water-salt migration key factors are placed in the input layer, and the plurality of sub-region water-salt distribution identifiers are placed in the output layer, and a plurality of first-order water-salt migration predictors are trained, including: S210: extracting a first coordinate region according to the plurality of coordinate regions, and constructing a spatial coordinate domain; S220: extracting first water-salt initial distribution information from the plurality of water-salt initial distribution information, distributing it in the spatial coordinate domain, and obtaining a water-salt initial distribution vector; S230: taking the spatial coordinate domain and the water-salt initial distribution vector as constraints, collecting water-salt migration monitoring data based on the water-salt migration key factors, wherein the water-salt migration monitoring data includes time series information of key factor characteristic value records and time series information of water-salt migration position records; S240: Based on the key factor deviation threshold, segment the key factor characteristic value record time series information in the time domain to obtain the key factor characteristic value record values ​​of the first time zone to the key factor characteristic value record values ​​of the Kth time zone; S250: According to the key factor characteristic value record values ​​of the first time zone to the key factor characteristic value record values ​​of the Kth time zone, the water-salt migration position record time series information is extracted, the water-salt distribution vector at the end time of the first time zone to the water-salt distribution vector at the end time of the Kth time zone is extracted, and the sub-region water-salt distribution mark is constructed; S260: Taking the recorded values ​​of the key factor characteristic values ​​of the first time zone to the recorded values ​​of the key factor characteristic values ​​of the Kth time zone as input, taking the water-salt distribution identifier of the sub-region as supervision, training the first-order predictor of water-salt migration in the first coordinate region based on the long short-term memory neural network, and adding the several first-order predictors of water-salt migration.

[0054] Specifically, first, traverse several coordinate regions, and determine one coordinate region each time as the first coordinate region. Secondly, construct a mathematical spatial coordinate domain based on the spatial range of the first coordinate region. The spatial coordinate domain defines the geometric boundaries of a specific coordinate region and the coordinate representation of internal spatial points, providing a spatial reference framework for the subsequent description of water and salt distribution. Then, the initial distribution information corresponding to the first coordinate region is extracted from the existing water and salt initial distribution information as the first water and salt initial distribution information, and mapped to the constructed spatial coordinate domain to form a water and salt initial distribution vector. This vector describes the spatial distribution state of water and salt in the region at the initial moment of the study.

[0055] Next, with the spatial coordinate domain and the initial water-salt distribution vector as constraints, monitoring data collection is carried out based on the key factors of water-salt migration. The collected data includes two parts: one is the time series information of the key factor characteristic value records, which reflects the changes of the key factors affecting water-salt migration over time; the other is the time series information of the water-salt migration position records, which reflects the trajectory of the migration and change of water and salt in space. Subsequently, based on the preset key factor deviation threshold, the time series information of the key factor characteristic value records is segmented in the time domain. When the change of the key factor characteristic value exceeds the key factor deviation threshold, it is considered that the water-salt migration environment has changed significantly and needs to be divided into different time zones. In this way, the key factor characteristic value record time series information is divided, and the key factor characteristic value record value from the first time zone to the Kth time zone is obtained. The change of the key factor characteristic value in each time zone is relatively stable.

[0056] After that, according to the key factor eigenvalue record values ​​of each time zone, the water-salt distribution vector of the corresponding time zone end time is extracted from the water-salt migration position record time series information, and the water-salt distribution vector of the first time zone end time is obtained until the water-salt distribution vector of the Kth time zone end time is obtained. These vectors together constitute the sub-region water-salt distribution mark, which serves as the output target of the first-order predictor of water-salt migration. After that, the key factor eigenvalue record values ​​of the first time zone until the Kth time zone eigenvalue record values ​​are used as input data, and the constructed sub-region water-salt distribution mark is used as the supervision target, and the long short-term memory neural network is used to train the first-order predictor of water-salt migration in the first coordinate area. The long short-term memory neural network has the advantage of processing time series data and can effectively capture the time dependence in the water-salt migration process. After the training is completed, the first-order predictor of water-salt migration in the first coordinate area is added to the first-order predictor set of water-salt migration for the subsequent construction of the overall predictor.

[0057] Further, taking the first time zone key factor characteristic value record value to the Kth time zone key factor characteristic value record value as input, taking the sub-region water-salt distribution mark as supervision, training the first coordinate region water-salt migration first-order predictor, adding the several water-salt migration first-order predictors, including: S261: Using K value as the number of neurons, build a first-level neural network architecture based on long short-term memory neural network; S262: using the first time zone key factor characteristic value record value to the Kth time zone key factor characteristic value record value as input, using the sub-region water-salt distribution identifier as supervision, training a first-level neural network architecture, and obtaining a first-order predictor of water-salt migration in the first coordinate region of the first level; S263: When the first-order predictor of water-salt migration in the first-level first coordinate area does not meet the convergence condition, increase The number of neurons is determined by building a two-level neural network architecture based on the long short-term memory neural network to perform cyclic training until the first-order predictor of water and salt migration in the first coordinate area of ​​level Y meets the convergence condition, and the first-order predictor of water and salt migration in the first coordinate area of ​​level Y is set as the first-order predictor of water and salt migration in the first coordinate area.

[0058] In a preferred embodiment, the K value (i.e., the number of time zones) obtained by time domain segmentation is used as a parameter for the number of neurons, and a first-level neural network architecture is constructed based on a long short-term memory neural network, so that the network structure matches the time segmentation characteristics of the input data. Each time zone corresponds to a neuron, which helps to capture the characteristic patterns of water and salt migration in different time periods. Then, the recorded values ​​of the characteristic values ​​of the key factors from the first time zone to the Kth time zone are used as network input data, and the water and salt distribution identifiers of the sub-regions are used as supervision targets to train the constructed first-level neural network architecture. Through machine learning algorithms such as back propagation, the network parameters are adjusted to minimize the error between the predicted output and the actual water and salt distribution, thereby obtaining a first-order predictor of water and salt migration in the first-level first coordinate area. Subsequently, it is evaluated whether the first-order predictor of water and salt migration in the first-level first coordinate area meets the preset convergence conditions (such as whether the prediction error is lower than a specific threshold). If not, the number of neurons is automatically increased to , a more complex secondary neural network architecture is built based on the long short-term memory neural network, and the training process is re-executed. This process of increasing the number of neurons and iterative optimization of the network structure continues until the first-order predictor of water and salt migration in the first coordinate area of ​​level Y meets the convergence condition. Finally, the first-order predictor of water and salt migration in the first coordinate area of ​​level Y that meets the convergence condition is determined as the first-order predictor of water and salt migration in the coordinate area.

[0059] Through adaptive network structure optimization, the complexity of the neural network can be automatically adjusted according to the differences in the complexity of water and salt migration in different coordinate areas. While ensuring the prediction accuracy, it avoids overfitting problems and waste of computing resources caused by overly complex network structures, thereby improving the adaptability and efficiency of the prediction model.

[0060] Embodiment 2, as Figure 2 As shown, based on the same inventive concept as the method for predicting water and salt migration in double furrows with full-film membranes in saline-alkali land provided in Example 1, the embodiment of the present invention further provides a prediction system for water and salt migration in double furrows with full-film membranes in saline-alkali land, comprising: The region segmentation module 11 is used to aggregate and segment the region to be predicted based on the key factors of water and salt migration to obtain the segmentation result of the region to be predicted, wherein the segmentation result of the region to be predicted includes a plurality of coordinate regions, and the key factors of water and salt migration belonging to the same coordinate region are consistent; The first-order training module 12 is used to traverse the plurality of coordinate regions, place the key factors of water-salt migration in the input layer, place the water-salt distribution identifiers of the plurality of sub-regions in the output layer based on the plurality of water-salt initial distribution information, and train a plurality of first-order predictors of water-salt migration; The second-order training module 13 is used to perform topological simulation on the segmentation results of the area to be predicted based on a graph neural network, taking the several coordinate areas as neuron nodes, building a second-order water-salt migration predictor architecture, setting the outputs of the several first-order water-salt migration predictors as the outputs of the neuron nodes, placing the overall regional water-salt distribution identifier in the output layer, and training the second-order water-salt migration predictor; The prediction execution module 14 is used to use the several first-order water-salt migration predictors to replace the neuron nodes of the second-order water-salt migration predictor, and obtain the water-salt migration predictor to execute the water-salt migration prediction of the saline-alkali land with full film double ridges in the predicted area.

[0061] Furthermore, the embodiment of the present application also includes a key factor acquisition module, which includes the following execution steps: Based on the water-salt migration influencing factor preset by the user end, collecting the water-salt migration monitoring log of the first fixed area, wherein any one of the water-salt migration monitoring logs includes a water-salt migration influencing factor record value and a water-salt distribution topology record value; Performing pairwise same-attribute deviation calculations on the water-salt migration monitoring logs to obtain a water-salt migration influencing factor record deviation modulus value set and a water-salt distribution topology record value deviation set; The water-salt migration influencing factors are sorted by water-salt migration correlation according to the water-salt distribution topology record value deviation set and the water-salt migration influencing factor record deviation modulus value set to obtain the water-salt migration key factors.

[0062] Furthermore, the key factor acquisition module also includes the following execution steps: Building a benchmark data sequence based on the water-salt distribution topology record value deviation set; A first comparison sequence is constructed according to the first attribute of the water-salt migration influencing factor record deviation modulus value set, until an Nth comparison sequence is constructed according to the Nth attribute of the water-salt migration influencing factor record deviation modulus value set; According to the benchmark data sequence, the first comparison sequence to the Nth comparison sequence, a grey correlation matrix is ​​constructed, grey correlation analysis is performed, and the key factors of water-salt migration with correlation greater than or equal to a correlation threshold are extracted.

[0063] Furthermore, the key factor acquisition module also includes the following execution steps: According to the water-salt distribution topology record value deviation set and the water-salt migration influencing factor record deviation modulus value set, the water-salt migration influencing factors are sorted by water-salt migration correlation to obtain the key factors of water-salt migration in the first fixed area; The first fixed area cyclic analysis is updated until the key factor of water-salt migration in the Mth fixed area is obtained, where M is an integer and M≥100; The key factor of water and salt migration in the first fixed area is counted until the trigger frequency of the key factor of water and salt migration in the Mth fixed area is greater than or equal to the key factor of water and salt migration with a trigger frequency threshold, wherein the trigger frequency threshold is greater than or equal to The floor value of .

[0064] Furthermore, the region segmentation module 11 further includes the following execution steps: Extract the water-salt migration key factor from the first fixed area water-salt migration key factor to the first key factor association degree set of the Mth fixed area water-salt migration key factor to the Qth key factor association degree set; Adding the first key factor association degree set to the Qth key factor association degree set to obtain an association degree sum value; Traversing the first key factor association degree set until the Qth key factor association degree set, respectively summing them up, obtaining the sum of the first key factor association degrees until the Qth key factor association degree sum, respectively comparing them with the association degree sum values, and obtaining the weight distribution of water-salt migration key factors; According to the weight distribution of the key factors of water and salt migration, a weighted Euclidean distance evaluation function is constructed for the key factors of water and salt migration. Combined with the Euclidean distance threshold, the area to be predicted is aggregated and segmented to obtain the segmentation result of the area to be predicted.

[0065] Furthermore, the first-order training module 12 includes the following execution steps: According to the plurality of coordinate regions, extracting a first coordinate region and constructing a spatial coordinate domain; Extracting first water-salt initial distribution information from the plurality of water-salt initial distribution information, distributing it in the spatial coordinate domain, and obtaining a water-salt initial distribution vector; Taking the spatial coordinate domain and the water-salt initial distribution vector as constraints, collecting water-salt migration monitoring data based on the water-salt migration key factors, wherein the water-salt migration monitoring data includes the time series information of the key factor characteristic value records and the time series information of the water-salt migration position records; Based on the key factor deviation threshold, the key factor characteristic value record time series information is segmented in the time domain to obtain the key factor characteristic value record values ​​of the first time zone to the key factor characteristic value record values ​​of the Kth time zone; According to the key factor characteristic value record values ​​of the first time zone to the key factor characteristic value record values ​​of the Kth time zone, the water-salt distribution vector at the end time of the first time zone to the water-salt distribution vector at the end time of the Kth time zone is extracted from the water-salt migration position time series information, and the water-salt distribution vector at the end time of the first time zone is extracted to construct the sub-region water-salt distribution mark; Taking the recorded values ​​of the key factor characteristic values ​​of the first time zone to the recorded values ​​of the key factor characteristic values ​​of the Kth time zone as input and the water-salt distribution identifier of the sub-region as supervision, a first-order predictor of water-salt migration in the first coordinate region is trained based on a long short-term memory neural network, and the several first-order predictors of water-salt migration are added.

[0066] Furthermore, the first-order training module 12 also includes the following execution steps: Using K value as the number of neurons, a first-level neural network architecture is built based on the long short-term memory neural network; Taking the recorded values ​​of the key factor characteristic values ​​of the first time zone to the recorded values ​​of the key factor characteristic values ​​of the Kth time zone as input, taking the water-salt distribution identifier of the sub-region as supervision, training a first-level neural network architecture, and obtaining a first-order predictor of water-salt migration in the first-level first coordinate region; When the first-order predictor of water-salt migration in the first-order coordinate area does not meet the convergence condition, increase The number of neurons is determined by building a two-level neural network architecture based on the long short-term memory neural network to perform cyclic training until the first-order predictor of water and salt migration in the first coordinate area of ​​level Y meets the convergence condition, and the first-order predictor of water and salt migration in the first coordinate area of ​​level Y is set as the first-order predictor of water and salt migration in the first coordinate area.

[0067] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0068] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0069] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1A device that provides the functions specified in a block or multiple blocks.

[0070] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0071] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0072] Although preferred embodiments of the present invention have been described, additional changes and modifications may occur to these embodiments once those skilled in the art understand the basic inventive concepts.

[0073] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention belong to the scope of the present invention and its equivalent technologies, the present invention is also intended to include these changes and variations.

Claims

1. A method for predicting water and salt migration in double-ridge ditch with full-film film on saline-alkali land, characterized in that: include: Based on the key factors of water and salt migration, the region to be predicted is aggregated and segmented to obtain a segmentation result of the region to be predicted, wherein the segmentation result of the region to be predicted includes a plurality of coordinate regions, and the key factors of water and salt migration belonging to the same coordinate region are consistent; Traversing the plurality of coordinate regions, placing the key factors of water-salt migration in an input layer based on the initial distribution information of water and salt, placing the water-salt distribution identifiers of the plurality of sub-regions in an output layer, and training a plurality of first-order predictors of water-salt migration; Based on the graph neural network, the several coordinate areas are used as neuron nodes, topological simulation is performed on the segmentation results of the area to be predicted, a second-order water and salt migration predictor architecture is built, the outputs of the several first-order water and salt migration predictors are set as the outputs of the neuron nodes, the overall regional water and salt distribution identifier is placed in the output layer, and the second-order water and salt migration predictor is trained; The plurality of first-order water-salt migration predictors are used to replace the neuron nodes of the second-order water-salt migration predictor, and a water-salt migration predictor is obtained to perform water-salt migration prediction for double-ridges in full-film saline-alkali land in the predicted area.

2. The method according to claim 1, characterized in that Based on the key factors of water and salt migration, the predicted area is aggregated and segmented to obtain the segmentation results of the predicted area, which previously included: Based on the water-salt migration influencing factor preset by the user end, collecting the water-salt migration monitoring log of the first fixed area, wherein any one of the water-salt migration monitoring logs includes a water-salt migration influencing factor record value and a water-salt distribution topology record value; Performing pairwise same-attribute deviation calculations on the water-salt migration monitoring logs to obtain a water-salt migration influencing factor record deviation modulus value set and a water-salt distribution topology record value deviation set; The water-salt migration influencing factors are sorted by water-salt migration correlation according to the water-salt distribution topology record value deviation set and the water-salt migration influencing factor record deviation modulus value set to obtain the water-salt migration key factors.

3. The method according to claim 2, characterized in that The water-salt migration influencing factors are sorted by water-salt migration correlation according to the water-salt distribution topology record value deviation set and the water-salt migration influencing factor record deviation modulus value set to obtain the water-salt migration key factors, including: Building a benchmark data sequence according to the water-salt distribution topology record value deviation set; A first comparison sequence is constructed according to the first attribute of the water-salt migration influencing factor record deviation modulus value set, until an Nth comparison sequence is constructed according to the Nth attribute of the water-salt migration influencing factor record deviation modulus value set; According to the benchmark data sequence, the first comparison sequence to the Nth comparison sequence, a grey correlation matrix is ​​constructed, grey correlation analysis is performed, and the key factors of water-salt migration with correlation greater than or equal to a correlation threshold are extracted.

4. The method according to claim 2, characterized in that The water-salt migration influencing factors are sorted by water-salt migration correlation according to the water-salt distribution topology record value deviation set and the water-salt migration influencing factor record deviation modulus value set to obtain the water-salt migration key factors, including: According to the water-salt distribution topology record value deviation set and the water-salt migration influencing factor record deviation modulus value set, the water-salt migration influencing factors are sorted by water-salt migration correlation to obtain the key factors of water-salt migration in the first fixed area; The first fixed area cyclic analysis is updated until the key factor of water-salt migration in the Mth fixed area is obtained, where M is an integer and M≥100; The key factor of water and salt migration in the first fixed area is counted until the trigger frequency of the key factor of water and salt migration in the Mth fixed area is greater than or equal to the key factor of water and salt migration with a trigger frequency threshold, wherein the trigger frequency threshold is greater than or equal to The floor value of .

5. The method according to claim 4, characterized in that Based on the key factors of water and salt migration, the area to be predicted is aggregated and segmented to obtain the segmentation results of the area to be predicted, including: Extract the water-salt migration key factor from the first fixed area water-salt migration key factor to the first key factor association degree set of the Mth fixed area water-salt migration key factor to the Qth key factor association degree set; Adding the first key factor association degree set to the Qth key factor association degree set to obtain an association degree sum value; Traversing the first key factor association degree set until the Qth key factor association degree set, respectively summing them up, obtaining the sum of the first key factor association degrees until the Qth key factor association degree sum, respectively comparing them with the association degree sum values, and obtaining the weight distribution of water-salt migration key factors; According to the weight distribution of the key factors of water and salt migration, a weighted Euclidean distance evaluation function is constructed for the key factors of water and salt migration. Combined with the Euclidean distance threshold, the area to be predicted is aggregated and segmented to obtain the segmentation result of the area to be predicted.

6. The method according to claim 1, characterized in that Traversing the plurality of coordinate regions, based on the plurality of water-salt initial distribution information, placing the water-salt migration key factors in the input layer, placing the plurality of sub-region water-salt distribution identifiers in the output layer, and training a plurality of first-order water-salt migration predictors, including: According to the plurality of coordinate regions, extracting a first coordinate region and constructing a spatial coordinate domain; Extracting first water-salt initial distribution information from the plurality of water-salt initial distribution information, distributing it in the spatial coordinate domain, and obtaining a water-salt initial distribution vector; Taking the spatial coordinate domain and the water-salt initial distribution vector as constraints, collecting water-salt migration monitoring data based on the water-salt migration key factors, wherein the water-salt migration monitoring data includes the time series information of the key factor characteristic value records and the time series information of the water-salt migration position records; Based on the key factor deviation threshold, the key factor characteristic value record time series information is segmented in the time domain to obtain the key factor characteristic value record values ​​of the first time zone to the key factor characteristic value record values ​​of the Kth time zone; According to the key factor characteristic value record values ​​of the first time zone to the key factor characteristic value record values ​​of the Kth time zone, the water-salt distribution vector at the end time of the first time zone to the water-salt distribution vector at the end time of the Kth time zone is extracted from the water-salt migration position time series information, and the water-salt distribution vector at the end time of the first time zone is extracted to construct the sub-region water-salt distribution mark; Taking the recorded values ​​of the key factor characteristic values ​​of the first time zone to the recorded values ​​of the key factor characteristic values ​​of the Kth time zone as input and the water-salt distribution identifier of the sub-region as supervision, a first-order predictor of water-salt migration in the first coordinate region is trained based on a long short-term memory neural network, and the several first-order predictors of water-salt migration are added.

7. The method according to claim 6, characterized in that Taking the first time zone key factor characteristic value record values ​​to the Kth time zone key factor characteristic value record values ​​as input, taking the sub-region water-salt distribution identifier as supervision, training the first coordinate region water-salt migration first-order predictor, adding the several water-salt migration first-order predictors, including: Using K value as the number of neurons, a first-level neural network architecture is built based on the long short-term memory neural network; Taking the recorded values ​​of the key factor characteristic values ​​of the first time zone to the recorded values ​​of the key factor characteristic values ​​of the Kth time zone as input, taking the water-salt distribution identifier of the sub-region as supervision, training a first-level neural network architecture, and obtaining a first-order predictor of water-salt migration in the first-level first coordinate region; When the first-order predictor of water-salt migration in the first-order coordinate area does not meet the convergence condition, increase The number of neurons is determined by building a two-level neural network architecture based on the long short-term memory neural network to perform cyclic training until the first-order predictor of water and salt migration in the first coordinate area of ​​level Y meets the convergence condition, and the first-order predictor of water and salt migration in the first coordinate area of ​​level Y is set as the first-order predictor of water and salt migration in the first coordinate area.

8. A prediction system for water and salt migration in double furrows with full-film film on saline-alkali land, characterized in that: The system is used to implement a method for predicting water and salt migration in double furrows with full-film film on saline-alkali land as claimed in any one of claims 1 to 7, and comprises: A regional segmentation module is used to aggregate and segment the area to be predicted based on the key factors of water and salt migration to obtain a segmentation result of the area to be predicted, wherein the segmentation result of the area to be predicted includes a plurality of coordinate areas, and the key factors of water and salt migration belonging to the same coordinate area are consistent; A first-order training module is used to traverse the plurality of coordinate regions, place the key factors of water-salt migration in an input layer, place the water-salt distribution identifiers of a plurality of sub-regions in an output layer based on a plurality of water-salt initial distribution information, and train a plurality of first-order predictors of water-salt migration; A second-order training module is used to perform topological simulation on the segmentation results of the area to be predicted based on a graph neural network, taking the several coordinate areas as neuron nodes, building a second-order water and salt migration predictor architecture, setting the outputs of the several first-order water and salt migration predictors as the outputs of the neuron nodes, placing the overall regional water and salt distribution identifiers in the output layer, and training the second-order water and salt migration predictor; The prediction execution module is used to use the several first-order water-salt migration predictors to replace the neuron nodes of the second-order water-salt migration predictor, and obtain the water-salt migration predictor to execute the water-salt migration prediction of the full-film double-ridge ditch in the saline-alkali land in the predicted area.

Images