Coastal saline soil dynamic monitoring and early warning system based on Internet of Things

Abstract

The invention relates to the technical field of soil ecological monitoring, in particular to a coastal saline soil dynamic monitoring and early warning system based on the Internet of Things, which comprises a collaborative monitoring unit, a fusion resolving analysis unit and an early warning trigger unit. The capillary water rising flux and the underground water seepage direction are synchronously monitored, the fusion resolving analysis unit dynamically corrects the capillary conductivity through a pore scale migration model in combination with three-dimensional pore structure parameters reconstructed through soil CT scanning, the salt exchange capacity is resolved, the salt increment is predicted through a long and short-term memory network, and a decision coefficient is calculated. The early warning trigger unit judges the secondary salinization risk when the conditions are met, activates drainage equipment and pushes early warning, the problems that traditional monitoring dimensions are single and the early warning precision is insufficient are solved, and the method is suitable for dynamic monitoring and risk prevention and control of coastal salinized soil.

Description

Technical Field

[0001] The present invention relates to the technical field of soil ecological monitoring, and in particular to a coastal saline soil dynamic monitoring and early warning system based on the Internet of Things. Background Art

[0002] Soil ecological monitoring is an important technology. In coastal agricultural production and ecological protection, this technology is a core means to ensure the quality of cultivated land and prevent soil degradation by sensing soil salinity changes in real time and predicting salinization risks.

[0003] However, traditional coastal saline soil monitoring technology suffers from core issues such as a single monitoring dimension and insufficient early warning accuracy. Existing solutions rely solely on surface soil salt concentration data and fail to establish a correlation model between vertical gradients and groundwater movement. As saltwater percolates upward through capillary action, these solutions fail to capture the migration patterns of salt across different soil layers, resulting in a one-sided assessment of the extent of salinization. When soil pore connectivity changes due to water changes, the system relies solely on a fixed threshold to trigger an early warning, failing to incorporate dynamic changes in soil pore structure. This leads to increased deviations in predicted salt migration rates and delayed early warnings. Furthermore, traditional systems lack closed-loop verification between monitoring data and actual salinization risk. When predicted results deviate significantly from measured values, model parameters cannot be dynamically adjusted, resulting in ineffective saltwater intrusion warnings, exacerbated soil salinization, reduced farmland productivity, and deteriorating ecological environments. Ultimately, this results in wasted land resources and increased agricultural economic losses in coastal areas, making effective management and sustainable utilization of saline soil difficult. To address this technical issue, we propose an IoT-based dynamic monitoring and early warning system for coastal saline soil. Summary of the Invention

[0004] The purpose of the present invention is to provide a coastal saline soil dynamic monitoring and early warning system based on the Internet of Things to solve the problems raised in the above background technology.

[0005] Since traditional monitoring has a single dimension and relies only on surface salt data, it is unable to capture the vertical migration law of salt. Therefore, the present invention uses collaborative monitoring units to collect vertical salt gradient data in layers, and simultaneously monitors capillary water flux and groundwater seepage direction, which can fully grasp the salt migration characteristics and improve monitoring integrity.

[0006] Due to the lack of accuracy of traditional early warnings and the fact that fixed thresholds do not incorporate dynamic changes in pore structure, resulting in delayed warnings, the present invention constructs a pore-scale migration model by integrating solution and analysis units, dynamically corrects capillary permeability, and combines long-short-term memory networks to predict salt increments. This can improve early warning accuracy and timely prevent and control salinization risks.

[0007] To achieve the above objectives, a dynamic monitoring and early warning system for coastal saline soil based on the Internet of Things is provided. This system includes a collaborative monitoring unit that uses a layered array of salt sensors deployed vertically at the monitoring points to collect real-time vertical gradient data on salt concentration. Soil tension sensors and groundwater flow meters are also synchronously configured to dynamically monitor capillary water flux and groundwater seepage direction. The fusion solution analysis unit inputs the concentration vertical gradient data, capillary water rising flux, and groundwater seepage direction into the pore-scale transport model through the data fusion module. The model dynamically corrects the initial capillary conductivity using the three-dimensional pore structure parameters reconstructed from soil CT scans. Based on the corrected capillary conductivity, the convection-diffusion equation in the dynamic solution module is used to solve the salt exchange capacity, and this is input into the correlation analysis module. The correlation analysis module uses the long-short-term memory network time series prediction algorithm to output the predicted value of soil salt increment with the salt exchange capacity as the input feature, and calculates the determination coefficient between the predicted value and the measured value of soil salt increment. When the determination coefficient is greater than the preset coefficient threshold and the soil salinity increment exceeds the salt increment threshold for a continuous period of time, the early warning trigger unit determines that there is a risk of secondary salinization caused by saltwater intrusion, activates the directional drainage equipment and pushes the early warning information.

[0008] As a further improvement of this technical solution, the method for obtaining capillary permeability adopts in-situ pore topology analysis technology: Undisturbed soil samples were drilled at the monitoring points, and pore structure tomographic images were obtained through soil CT scanning. A three-dimensional point cloud model of the soil pores was reconstructed based on the pore structure tomographic images. The pore throat network of the three-dimensional soil pore point cloud model was extracted using the octree spatial segmentation algorithm. The equivalent curvature radius distribution histogram of the throat network was calculated, and the curvature radius corresponding to the frequency peak was selected as the dominant pore radius. Finally, the dominant pore radius was combined with the fluid dynamics constitutive relationship to generate the initial capillary conductivity.

[0009] As a further improvement of this technical solution, the operation of the data fusion module adopts multimodal spatiotemporal coding fusion technology: The system receives the original signals of vertical gradient data of salt concentration, capillary water flux and groundwater seepage direction, performs three-dimensional orthogonal decomposition on the groundwater seepage direction, extracts the vertical seepage component and marks its direction sign, inputs the dominant pore radius into the capillary dynamics constraint model, and generates the critical threshold of capillary water flux. When the real-time capillary water flux exceeds the critical threshold and the vertical seepage component is directed upward, a saltwater intrusion intensity index classification signal is output.

[0010] As a further improvement of this technical solution, the three-dimensional pore structure parameter dynamic correction capillary permeability adopts a throat connectivity feedback mechanism: The three-dimensional point cloud model of soil pores reconstructed by soil CT scanning is topologically simplified, the number of throat nodes and connected paths is identified, and the cross-density of throat connected paths within a unit volume is counted to generate a quantitative index of pore throat connectivity. Based on the graded signal of the saltwater intrusion intensity index, the initial capillary permeability is dynamically adjusted through the pore structure deformation compensation function to form a corrected capillary permeability.

[0011] As a further improvement of this technical solution, the convection-diffusion equation is used to solve the salt exchange amount using physical field coupling driving technology: The corrected capillary conductivity is received as the migration rate control variable, and the soil texture database is called to match the clay content percentage of the current monitoring point, which is then mapped to the salt diffusion coefficient benchmark value. The vertical gradient data of the salt concentration is processed by central difference to generate the concentration gradient change rate. The capillary water rise flux and the concentration gradient change rate are superimposed, and the salt exchange amount per unit time is solved by the law of conservation of solute transport.

[0012] As a further improvement of this technical solution, the long short-term memory network time series prediction algorithm adopts a time lag compensation training architecture: A bidirectional long-short-term memory neural network model was constructed. The input layer received a continuous 24-hour time series of salt exchange, and the hidden layer implanted a lag effect compensation unit. The time lag characteristics of groundwater salt migration were learned through historical soil salt increment data. The output layer generated a predicted value of soil salt increment based on the time lag characteristics and simultaneously output a prediction confidence evaluation index.

[0013] As a further improvement of this technical solution, the calculation of the determination coefficient adopts a dynamic weight sliding window technology: A data sliding window is set up to capture the predicted and measured values ​​of soil salinity increment in real time. The data points in the window are dynamically weighted according to the corrected capillary conductivity, and the weighted Pearson correlation coefficient is calculated as the quantitative value of the determination coefficient, and the statistical significance mark is recorded simultaneously.

[0014] As a further improvement of this technical solution, when the quantized value of the determination coefficient exceeds a preset threshold, the pore structure self-calibration loop is activated: The absolute error between the predicted and measured soil salinity increments is converted into a pore throat connectivity correction coefficient. The throat connectivity quantification index is adjusted using the throat node deformation feedback function. The capillary conductivity dynamic correction process is re-executed based on the updated throat connectivity, forming a closed-loop capillary conductivity optimization.

[0015] As a further improvement to this technical solution, the association analysis module has a built-in contribution attribution analysis engine: When the early warning is triggered, the salt exchange amount, saltwater intrusion intensity index and vertical seepage component are extracted, and the contribution percentage of each parameter to the soil salt increment prediction result is quantified through the reverse gradient propagation technology. A parameter contribution ranking list is generated and the categories of the dominant influencing factors are marked.

[0016] As a further improvement of this technical solution, the early warning trigger unit implements a contribution-driven hierarchical response strategy: When the contribution of capillary conductivity exceeds the threshold set by the system, the shallow vacuum pumping device is started, and the pumping depth automatically matches the capillary climbing height corresponding to the dominant pore radius. When the contribution of the saltwater intrusion intensity index exceeds the threshold set by the system, the deep well pressure reduction drainage system is activated, and the drainage flow is adjusted according to the proportion of the vertical seepage component.

[0017] Compared with the prior art, the present invention has the following beneficial effects: In the dynamic monitoring and early warning system for coastal saline soil based on the Internet of Things, the fusion solution and analysis unit dynamically corrects the capillary permeability based on the three-dimensional pore structure parameters reconstructed from soil CT scans, combines the convection-diffusion equation to solve the salt exchange amount, and realizes accurate prediction of salt increment through the long-term and short-term memory network, thereby improving the ability to predict the evolution trend of salinization. The early warning trigger unit determines the risk based on the determination coefficient and the salt increment threshold, activates the directional drainage equipment and sends an early warning. Combined with the pore throat connectivity feedback mechanism and closed-loop verification, the model parameters are dynamically optimized to effectively avoid early warning lags. The system realizes the precise prevention and control of secondary salinization risks caused by saltwater intrusion, ensures the quality of cultivated land, improves the utilization rate of coastal land resources, and provides reliable support for agricultural production and ecological protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is an overall block diagram of the present invention.

[0019] The meaning of each number in the figure is: 1. Collaborative monitoring unit; 2. Fusion solution and analysis unit; 3. Early warning trigger unit. DETAILED DESCRIPTION

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0021] The present invention provides a coastal saline soil dynamic monitoring and early warning system based on the Internet of Things. Figure 1 As shown, it includes the following units: Collaborative monitoring unit 1 collects real-time vertical gradient data of salt concentration through a group of salt sensors layered and deployed vertically at the monitoring points. It also syncs with soil tension sensors and groundwater flow meters to dynamically monitor capillary water flux and groundwater seepage direction. The fusion solution analysis unit 2 inputs the concentration vertical gradient data, capillary water rising flux and groundwater seepage direction into the pore-scale transport model through the data fusion module. The model dynamically corrects the initial capillary conductivity through the three-dimensional pore structure parameters reconstructed by soil CT scanning. Based on the corrected capillary conductivity, the convection-diffusion equation in the dynamic solution module is used to solve the salt exchange capacity, and the calculated value is input into the correlation analysis module. The correlation analysis module uses the long-short-term memory network time series prediction algorithm to output the predicted value of soil salt increment with the salt exchange capacity as the input feature, and calculates the determination coefficient between the predicted value and the measured value of soil salt increment; In order to accurately obtain the capillary conductivity that reflects the water conductivity of soil pores and provide key parameters for subsequent salt migration models, the capillary conductivity is obtained using in-situ pore topology analysis technology: Undisturbed soil samples were drilled at the monitoring points, because only samples that maintain the original structure of the soil can truly reflect the natural state of the pores and avoid deformation or destruction of the pores caused by disturbance. Thin-walled soil samplers were used during drilling and slowly inserted into the soil to a preset depth. After being taken out, the samples were immediately wrapped with a sealing film to prevent water evaporation from causing changes in the pore structure and ensure that the physical properties of the samples were consistent with those of the in-situ soil. Pore structure tomographic images were obtained through soil CT scanning, because CT scanning can present the pore distribution characteristics inside the soil without loss. The undisturbed soil samples were fixed on the stage of the CT scanner, and the scanning parameters were set. The sample is subjected to continuous tomographic scanning to obtain a series of two-dimensional tomographic images along the vertical direction. Different grayscale values ​​in the image represent the difference between the soil solid phase and the pores. The soil pore three-dimensional point cloud model is reconstructed based on the pore structure tomographic image. Because the three-dimensional model can more comprehensively display the spatial distribution and connectivity of the pores, the continuous two-dimensional tomographic images are imported into the three-dimensional reconstruction software, and the images of each layer are aligned through image registration technology. The surface reconstruction algorithm is then used to extract the pore boundary of each layer, and the three-dimensional point cloud model is formed by stacking layer by layer. Each point in the model represents a coordinate position on the pore surface, which fully presents the three-dimensional shape of the pores. The size and interconnectedness of the pores are analyzed, and the octree space segmentation algorithm is used to extract the pore throat network of the soil pore three-dimensional point cloud model. Because the throat is the connecting channel between pores, its characteristics directly affect the migration capacity of water and salt. The octree space segmentation algorithm divides the three-dimensional space into 8 sub-cubes step by step, hierarchically refines the pore three-dimensional point cloud model, identifies the narrowest connection between the pores, that is, the throat, and constructs a throat network composed of throat nodes, that is, the connection points at both ends of the throat and the connection paths, so as to accurately capture the key channels for water flow in the pore system and calculate the equivalent of the throat network. In the curvature radius distribution histogram, the curvature radius corresponding to the frequency peak is selected as the dominant pore radius. This is because the dominant pore radius reflects the most representative pore size in the soil and is a key parameter determining capillary conduction capacity. For each throat in the throat network, the equivalent curvature radius is calculated based on its three-dimensional morphology. This is the radius of a circle with the same cross-sectional area as the throat. The equivalent curvature radius values ​​of all throats are counted and a distribution histogram is plotted grouped by radius size. The radius with the highest frequency in the histogram is the dominant pore radius. For example, if the histogram of a sample has a peak at 0.2 mm, the dominant pore radius is 0.2 mm indicates that the soil is dominated by pore throats of this radius. Finally, the initial capillary conductivity is generated by combining the dominant pore radius with the hydrodynamic constitutive relation. Because capillary conductivity is directly related to pore size, a theoretical relationship is needed to convert pore characteristics into conductivity parameters. The hydrodynamic constitutive relation is established based on experimental data and reflects the relationship between the dominant pore radius and capillary conductivity. For example, larger radius indicates higher capillary conductivity, because large pores offer less resistance to water. Substituting the dominant pore radius into this relationship, the calculated value is the initial capillary conductivity. This value quantifies the soil's ability to conduct water through capillary action, providing an initial benchmark for subsequent model modifications, ensuring that the initial capillary conductivity truly reflects the soil's inherent conductivity properties and laying the foundation for accurate simulation of salt transport.

[0022] In order to integrate monitoring data from different sources into a unified signal that can reflect the characteristics of saltwater intrusion, the data fusion module uses multimodal spatiotemporal coding fusion technology. The specific implementation method is as follows: Receive the original signals of vertical gradient data of salt concentration, capillary water rising flux and groundwater seepage direction, because these data reflect the soil state from three dimensions: salt distribution, water movement intensity and groundwater flow direction. Comprehensive analysis is required to judge the situation of saltwater intrusion. The vertical gradient data of salt concentration comes from the layered salt sensor group, which records the salt concentration of different soil layers by depth. The capillary water rising flux is measured by the soil tension sensor, which reflects the speed at which water rises through capillary action. The groundwater seepage direction is collected by the groundwater velocity meter, which records the three-dimensional direction information of groundwater flow. These original signals provide basic data for subsequent fusion analysis. The groundwater seepage direction is subjected to three-dimensional orthogonal decomposition, and the vertical seepage component is extracted and its direction sign is marked, because the vertical seepage direction (upward or downward) directly affects whether the salt water migrates to the surface soil. The three-dimensional orthogonal decomposition is to decompose the complex direction of groundwater seepage into components of two dimensions, horizontal and vertical. The vertical seepage component is extracted, that is, the vertical seepage component, and the direction sign is marked: upward flow is marked as positive, and downward flow is marked as negative. For example, when groundwater seeps upward, the vertical seepage component is positive, indicating that it may carry underground salt water to the surface. The dominant pore radius is input into the capillary dynamics constraint model to generate the critical threshold of capillary water flux, because the dominant pore radius is the capillary dynamics constraint model, and the capillary water flux critical threshold is generated. The guiding pore radius determines the maximum capillary water rise intensity that the soil can withstand. Exceeding this intensity may mean that saltwater intrusion is intensified. The capillary dynamics constraint model is established based on the relationship between soil pore characteristics and water movement. The previously obtained dominant pore radius is substituted into the model to calculate the maximum capillary water rise flux of the soil under normal conditions, that is, the critical threshold. For example, for soil with a dominant pore radius of 0.2 mm, the critical threshold may be set to 5 mm / hour, which means that capillary water rise exceeding this value may be related to saltwater intrusion. When the real-time capillary water rise flux exceeds the critical threshold and the vertical seepage component is directed upward, a saltwater intrusion intensity index classification signal is output. When these two conditions are met at the same time, it indicates that underground salt water is migrating to the surface soil through capillary action and upward seepage. The intrusion intensity needs to be quantified. The graded signal is divided into levels according to the proportion of the real-time capillary water rising flux exceeding the critical threshold. For example, if the excess ratio is within 20%, it is a mild intrusion (level 1), 20%-50% is a moderate intrusion (level 2), and more than 50% is a severe intrusion (level 3). Each level corresponds to a different warning intensity. For example, a moderate intrusion signal triggers moderate-intensity prevention and control measures, thereby achieving an accurate portrayal of the degree of salt water intrusion, making the judgment of salt water intrusion more comprehensive and accurate, and providing a key basis for the subsequent correction of capillary conductivity.

[0023] In order to enable capillary conductivity to dynamically adapt to changes in soil pore structure due to external influences and improve the accuracy of salt migration simulation, a throat connectivity feedback mechanism is used to dynamically modify capillary conductivity using three-dimensional pore structure parameters: The soil pore three-dimensional point cloud model reconstructed by soil CT scanning is topologically simplified. Because the three-dimensional point cloud model contains a large amount of redundant data, the simplification can more clearly highlight the core structure of the pore throat. The topological skeleton simplification is to extract the central line that best represents the pore connectivity in the pore three-dimensional point cloud model, remove the surface detail points, and form a skeleton model composed of lines. The skeleton can intuitively reflect the direction, branching and connection mode of the pore throat, laying the foundation for the subsequent identification of throat nodes and connection paths, so that the analysis focuses on the key structures that affect water migration, and identifies the number of throat nodes and connection paths. The throat node is the connection point between the throats, and the connection path is the channel between the throats. The number of throat nodes directly reflects the degree of pore connectivity. The simplified topological skeleton uses a node detection algorithm to identify the points where lines in the skeleton intersect as throat nodes, and at the same time count the number of independent lines between different nodes in the skeleton as the number of connected paths. For example, the skeleton of a certain area has 5 throat nodes and 8 connected paths, indicating that the pore connectivity in this area is good, providing basic data for quantifying the connectivity. The cross-density of the throat connection paths in the unit volume is counted to generate a quantitative index of the pore throat connectivity, because the cross-density can more accurately reflect the degree of interweaving of the throats in the unit space, and is the core indicator for measuring pore connectivity. When calculating, first determine the statistical soil volume (such as 1 cubic centimeter), and then count the number of intersections of all connected paths in the volume. The more intersections, the denser the throat network. The stronger the connectivity, the stronger the cross-points. The cross-point density is obtained by dividing the number of intersections by the volume, which is used as a quantitative indicator of pore throat connectivity. For example, if there are 20 intersections in 1 cubic centimeter, the connectivity index is 20. The larger the value of this index, the smoother the channel for water and salt migration in the soil. Based on the saltwater intrusion intensity index grading signal, the initial capillary conductivity is dynamically adjusted through the pore structure deformation compensation function to form a corrected capillary conductivity. Because saltwater intrusion will change the soil pore structure, such as salt crystallization may block the throat, the compensation function needs to be used to reflect the impact of this change on capillary conductivity. The pore structure deformation compensation function sets the adjustment rules according to the saltwater intrusion intensity level: when the invasion is mild, the connectivity index decreases slightly, and the capillary conductivity is adjusted by 10 %; in moderate invasion, the connectivity index drops significantly, and the capillary conductivity is reduced by 20%; in severe invasion, the connectivity index drops sharply, and the capillary conductivity is reduced by 30%. For example, the initial capillary conductivity is 10 mm / h, and it is adjusted to 8 mm / h in moderate invasion, so that the capillary conductivity can match the pore structure changes caused by saltwater intrusion in real time, ensuring that it reflects the actual water conductivity of the current soil. From the topological characteristics of the pore structure to the quantification of the connectivity, and then combining the intensity of saltwater intrusion to dynamically adjust the capillary conductivity, the corrected capillary conductivity is based on the inherent pore characteristics of the soil and can respond to the structural changes brought about by external invasion, providing a more practical parameter basis for the subsequent accurate calculation of salt exchange capacity.

[0024] In order to accurately calculate the migration and exchange amount of salt in the soil per unit time and reflect the migration intensity of salt in the soil, the convection-diffusion equation is used to solve the salt exchange amount using the physical field coupling driving technology: The corrected capillary conductivity is received as the migration rate control variable. Capillary conductivity directly determines the speed of salt migration with water and is the core parameter for solving the salt exchange capacity. The corrected capillary conductivity has been dynamically adjusted in combination with pore structure changes and saltwater intrusion intensity. It can truly reflect the current rate of salt migration carried by water in the soil. Including it as a control variable in the calculation can ensure that the subsequent calculation of salt migration rate fits the actual situation. The soil texture database is called to match the clay content percentage of the current monitoring point and map it to the salt diffusion coefficient benchmark value. The clay content will affect the diffusion ability of salt in the soil. The more clay there is, the more salt is adsorbed. The stronger the force, the slower the diffusion. The soil texture database stores the corresponding relationship between the clay content and salt diffusion coefficient of different soil types. The clay content percentage of the area is retrieved from the database through the geographical location information of the monitoring point. Then, according to the preset mapping rules, such as when the clay content increases by 10%, the diffusion coefficient baseline value decreases by 15%, the clay content is converted into the salt diffusion coefficient baseline value, and the vertical gradient data of the salt concentration is processed by central difference to generate the concentration gradient change rate. The concentration gradient change rate can reflect the change speed of salt concentration between different soil layers and is the driving force for salt diffusion. The central difference processing is the salt concentration of two adjacent layers in the vertical direction. The concentration data is calculated by subtracting the concentration of the lower layer from the concentration of the upper layer, and then dividing it by the distance between the two layers to obtain the concentration change value per centimeter of soil layer. For example, if the concentration of the upper layer is 5 g / kg and the concentration of the lower layer is 3 g / kg, and the distance between the two layers is 2 cm, then the concentration gradient change rate is (5-3) / 2=1 g / kg·cm. The larger the value, the stronger the salt diffusion tendency in the vertical direction. The capillary water upward flux and the concentration gradient change rate are superimposed, and the salt exchange amount per unit time is solved by the law of conservation of solute transport. The migration of salt in soil is affected by both convection (flow with water) and diffusion (due to concentration differences), and the effects of both need to be superimposed. Specifically, the capillary water rise flux (the intensity of convection) and the concentration gradient change rate (the intensity of diffusion) are superimposed according to their contribution ratio to salt migration. Then, combined with the corrected capillary conductivity and salt diffusion coefficient baseline values, according to the law of conservation of solute transport, that is, the salt change in a certain area per unit time is equal to the difference between the inflow and outflow of salt, the salt exchange amount of the area per unit time is calculated, that is, the difference between the inflow and outflow of salt. For example, the superimposed value is calculated to have a salt exchange amount of 2 grams / kilogram per hour, indicating that the net increase of salt in the area is 2 grams / kilogram per hour, which provides accurate input parameters for the subsequent prediction of salt increment.

[0025] In order to accurately predict soil salinity increments, and considering the time lag in groundwater salinity migration, the long short-term memory network time series prediction algorithm adopts a time lag compensation training architecture: A bidirectional long-short-term memory neural network model is constructed. Because the bidirectional structure can simultaneously utilize past and future salt exchange data, it can more comprehensively capture the dependencies in the time series and improve the completeness of the prediction. The model consists of an input layer, a hidden layer, and an output layer. The hidden layer contains multiple memory units. These units can selectively memorize or forget information through a gating mechanism. They are suitable for processing time series data with long-term dependencies and provide a structural basis for learning the time lag characteristics of salt migration. The input layer receives a continuous 24-hour salt exchange time series because the salt exchange volume directly reflects the migration of salt per unit time. It is the core basis for predicting salt increment. The time series is a data point per hour, with a total of 24 data points. Each data point records the salt exchange amount in the corresponding hour. These data are input into the model in chronological order to provide original materials for the model to learn the dynamic law of salt migration, ensuring that the input information can cover the complete cycle of salt changes within a day. The hidden layer is implanted with a lag effect compensation unit to learn the time lag characteristics of groundwater salt migration through historical soil salt increment data. Because groundwater flow and salt diffusion require time, changes in salt exchange amount will not be immediately reflected as salt increments. Time lag compensation can correct this time difference, and the lag effect The training process of the compensation unit is as follows: the time series of salt exchange volume in the past six months and the corresponding measured values ​​of salt increment, such as the salt increment 3 hours and 6 hours after the exchange occurs, are input into the unit. Through repeated iterative learning, the unit can recognize the time lag length under different circumstances. For example, when the salt exchange volume of a certain period is input, the unit can judge that the exchange volume will have the greatest impact on the salt increment after 4 hours, thereby automatically adjusting the time matching relationship in the model calculation. The output layer generates the predicted value of soil salt increment according to the time lag characteristics, and simultaneously outputs the prediction confidence evaluation index, because the time lag characteristics determine the actual impact of salt exchange on the increment. The actual impact time, and the confidence level can reflect the reliability of the prediction results. The output layer combines the time lag characteristics learned by the hidden layer, and performs weighted calculation on the 24-hour salt exchange amount according to its corresponding time lag duration to generate a predicted value of soil salt increment in the future period. For example, if the time lag of a certain exchange amount is 3 hours, then its contribution to the salt increment after 3 hours is the largest. At the same time, the prediction confidence evaluation index is generated according to the error rate of historical predictions. If the historical prediction error of a certain type of soil is small, the confidence level of the current prediction is higher (such as 90%), otherwise it is lower (such as 60%), providing a basis for subsequent judgment of the reliability of the prediction results.

[0026] In order to accurately measure the degree of agreement between the predicted soil salinity increment and the measured value and provide a reliable consistency indicator for subsequent early warning judgment, the determination coefficient is calculated using the dynamic weight sliding window technology: A data sliding window is set to capture the predicted and measured values ​​of soil salinity increments in real time. The sliding window can continuously track the changes in values ​​within a certain period of time, avoiding the interference of data fluctuations at a single time point on the results. The length of the data sliding window is set according to the time scale of salt migration. For example, if it is set to 24 hours, the window contains the predicted values ​​and measured values ​​for 24 consecutive hours, with one data point per hour. As time goes by, the window moves forward once an hour, discarding the earliest data point and incorporating the latest data point, always maintaining a scale of 24 data points to ensure that a recent continuous and complete data sequence is captured, providing a stable sample basis for subsequent calculations. The data points in the window are dynamically weighted according to the corrected capillary conductivity. The corrected capillary conductivity reflects the current water conductivity of the soil, and its size is related to the activity of salt migration. Data points with different activity levels should have different effects on the overall fit. The dynamic weight allocation rule is: the larger the corrected capillary conductivity, that is, the more active the water and salt migration, the more significant the impact of the corresponding time point data on the prediction accuracy, and therefore a higher weight is given. The smaller the capillary conductivity, the lower the weight of the data point. For example, the capillary conductivity corresponding to a data point is 10 mm / h, and the weight is set to 0.12; the capillary conductivity of another data point is 5 mm / h , the weight is set to 0.08, and the sum of the weights of all data points is 1. This distribution method makes the data of the active period occupy a larger proportion in the calculation, improving the response sensitivity of the results to the actual migration status. The weighted Pearson correlation coefficient is calculated as the quantitative value of the determination coefficient, and the statistical significance mark is recorded simultaneously. The weighted Pearson correlation coefficient can reflect the degree of linear correlation between the predicted value and the measured value under dynamic weight, and the statistical significance can determine whether this correlation is statistically significant. During the calculation, the deviation between each predicted value and the measured value in the window is weighted according to the corresponding weight, and then the weighted Pearson correlation coefficient is obtained by the correlation coefficient calculation formula. The closer the value is to 1, the higher the degree of agreement between the predicted value and the measured value, and the closer it is to 0, the lower the degree of agreement. At the same time, the statistical significance mark is set according to the size of the correlation coefficient. If the correlation coefficient is greater than 0.7, it is marked as "highly significant"; between 0.4-0.7, it is marked as "significant"; and less than 0.4, it is marked as "insignificant". This intuitively reflects the reliability of the determination coefficient and provides a key basis for the subsequent warning trigger unit 3 to judge the effectiveness of the prediction.

[0027] In order to promptly correct the model parameters when the predicted values ​​deviate significantly from the measured values ​​and ensure that the capillary conductivity always matches the actual soil conditions, the pore structure self-calibration loop is activated when the quantized value of the determination coefficient exceeds the preset threshold: The absolute error between the predicted value and the measured value of soil salinity increment is converted into the correction coefficient of pore throat connectivity. The absolute error reflects the degree of deviation between the predicted result and the actual situation. This deviation often comes from the estimation deviation of pore throat connectivity and needs to be adjusted by the correction coefficient. The absolute error value is calculated by subtracting the measured value from the predicted value and taking the absolute value. For example, the predicted value is 5 g / kg, the measured value is 3 g / kg, and the absolute error value is 2 g / kg. The conversion rule is: the larger the absolute error value, the larger the correction coefficient. For example, the error value of 2 g / kg corresponds to a correction coefficient of 0.2, and the error value is 0.1. The difference of 3 g / kg corresponds to a correction coefficient of 0.3. The value of the correction coefficient directly determines the adjustment range of the throat connectivity, thereby establishing a relationship between the error and the pore structure parameters. The throat connectivity quantitative index is adjusted through the throat node deformation feedback function. The deformation of the throat node will directly affect the patency of the connection path and is the core factor leading to changes in connectivity. The throat node deformation feedback function sets the adjustment rule according to the size of the correction coefficient: when the correction coefficient is 0.2, the original throat connectivity quantitative index, such as the unit volume cross density of 20, is reduced by 20% to 16; when the correction coefficient is 0.3 , down by 30% to 1, in order to simulate the impact of throat node deformation caused by salt crystallization, pore blockage, etc. on connectivity, so that the adjusted connectivity index is more in line with the channel state of salt migration in the actual soil. Based on the updated throat connectivity, the capillary conductivity dynamic correction process is re-executed to form a capillary conductivity closed-loop optimization. The change in connectivity will directly change the migration rate of water and salt. It is necessary to re-correct the capillary conductivity to ensure that it matches the current pore structure. The re-executed dynamic correction process is consistent with the previous one: according to the updated throat connectivity, combined with the saltwater intrusion intensity index Based on the graded signal, the initial capillary permeability is adjusted again through the pore structure deformation compensation function. For example, when the updated connectivity index is 16, corresponding to moderate invasion, the capillary permeability is further reduced by 20% based on the initial value, and the optimized capillary permeability is finally obtained. This process forms a closed loop of "error identification-connectivity adjustment-capillary permeability correction", so that the capillary permeability can be optimized in real time with the prediction deviation, continuously improving the accuracy of salt exchange volume solution and incremental prediction, effectively reducing the prediction deviation caused by the discrepancy between the initial assumptions of the model and the actual soil state, and providing self-calibration capability for the long-term stable operation of the system.

[0028] When the determination coefficient is greater than the preset coefficient threshold and the soil salinity increment exceeds the salt increment threshold for a continuous period of time, the early warning trigger unit 3 determines that there is a secondary salinization risk caused by saltwater intrusion, activates the directional drainage equipment and pushes an early warning message.

[0029] In order to take precise prevention and control measures for secondary salinization risks dominated by different factors, the early warning trigger unit 3 implements a contribution-driven hierarchical response strategy: First, determine the contribution threshold of each parameter. The system presets the capillary conductivity contribution threshold and the saltwater intrusion intensity index contribution threshold. When the contribution of a parameter exceeds the corresponding threshold, it indicates that the parameter is the dominant factor of the current salinization risk, and targeted prevention and control equipment needs to be started. When the capillary conductivity contribution exceeds the system-set threshold, it indicates that the connectivity change of the soil pore structure is the main reason for the intensification of salt migration. At this time, the shallow vacuum drainage device is started, because the salt migration dominated by capillary conductivity is significantly affected by capillary action and mostly occurs in shallow soil. The drainage depth automatically matches the capillary climbing height corresponding to the dominant pore radius. The capillary climbing height refers to the maximum height that water can rise through the dominant pore. The larger the dominant pore radius, the lower the climbing height. For example, when the dominant pore radius is 0.2 mm, the corresponding capillary climbing height is 30 cm, and the drainage depth is set to 30 cm to ensure drainage. The device can precisely act on shallow areas where salt accumulates. It uses vacuum negative pressure to reduce the moisture content of shallow soil, reducing the driving force of salt rising with capillary water, thereby curbing the spread of salinization. When the contribution of the saltwater intrusion intensity index exceeds the system's set threshold, it indicates that active underground saltwater intrusion is the main cause of salinization risk. At this time, the deep well pressure reduction and drainage system is activated. Because saltwater intrusion often results from excessive groundwater pressure, deep well drainage can lower the groundwater level and block the path for saltwater to seep upward. The drainage flow rate is adjusted in proportion to the size of the vertical seepage component. The larger the positive value of the vertical seepage component, the stronger the upward seepage and the larger the drainage flow rate. For example, if the vertical seepage component is 5 mm / h, the drainage flow rate is set to 10 cubic meters / h; when the component increases to 10 mm / h, the flow rate correspondingly increases to 20 cubic meters / h. By dynamically adjusting the flow rate, the groundwater pressure is adapted to the seepage direction, effectively reducing the intensity of saltwater intrusion.

[0030] Through this strategy of graded response based on contribution, shallow drainage is initiated for shallow salinization dominated by capillary conductivity, and deep well drainage is activated for deep infiltration dominated by saltwater intrusion. This can not only accurately match the causes of risks, but also avoid the waste of prevention and control resources, effectively curb salinization risks, and ensure the farming performance and ecological stability of coastal soils.

[0031] In the present invention, the collaborative monitoring unit 1 collects vertical gradient data of salt concentration in layers, and simultaneously monitors the capillary water rise flux and groundwater seepage direction. The fusion solution analysis unit 2 uses the pore-scale migration model and the three-dimensional pore structure parameters reconstructed by soil CT scanning to dynamically correct the capillary conductivity, calculate the salt exchange amount, predict the salt increment and calculate the determination coefficient through the long-short-term memory network, and the early warning trigger unit 3 determines the secondary salinization risk when the conditions are met, activates the drainage equipment and sends an early warning, which solves the problems of single-dimensional traditional monitoring and insufficient early warning accuracy. It is suitable for dynamic monitoring and risk prevention and control of coastal saline soil.

[0032] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. The dynamic monitoring and early warning system for coastal saline soil based on the Internet of Things is characterized by: The following units are included: The collaborative monitoring unit (1) collects the vertical gradient data of salt concentration in real time through the salt sensor group deployed in layers in the vertical direction of the monitoring point, and synchronously configures the soil tension sensor and the groundwater flow meter to dynamically monitor the capillary water rise flux and groundwater seepage direction; The fusion solution analysis unit (2) inputs the concentration vertical gradient data, capillary water rising flux and groundwater seepage direction into the pore-scale transport model through the data fusion module. The model dynamically corrects the initial capillary conductivity through the three-dimensional pore structure parameters reconstructed by soil CT scanning. Based on the corrected capillary conductivity, the convection-diffusion equation in the dynamic solution module is used to solve the salt exchange amount, and the salt exchange amount is input into the correlation analysis module. The correlation analysis module uses the long-short-term memory network time series prediction algorithm to output the soil salt increment prediction value with the salt exchange amount as the input feature, and calculates the determination coefficient between the soil salt increment prediction value and the measured value; When the determination coefficient is greater than the preset coefficient threshold and the soil salinity increment exceeds the salt increment threshold within a continuous time period, the early warning trigger unit (3) determines that there is a secondary salinization risk caused by saltwater intrusion, activates the directional drainage equipment and pushes the early warning information.

2. The coastal saline soil dynamic monitoring and early warning system based on the Internet of Things according to claim 1 is characterized by: The method for obtaining capillary permeability adopts in-situ pore topology analysis technology: Undisturbed soil samples were drilled at the monitoring points, and pore structure tomographic images were obtained through soil CT scanning. A three-dimensional point cloud model of the soil pores was reconstructed based on the pore structure tomographic images. The pore throat network of the three-dimensional soil pore point cloud model was extracted using the octree spatial segmentation algorithm. The equivalent curvature radius distribution histogram of the throat network was calculated, and the curvature radius corresponding to the frequency peak was selected as the dominant pore radius. Finally, the dominant pore radius was combined with the fluid dynamics constitutive relationship to generate the initial capillary conductivity.

3. The dynamic monitoring and early warning system for coastal saline soil based on the Internet of Things according to claim 2 is characterized by: The data fusion module operates using multimodal spatiotemporal coding fusion technology: The system receives the original signals of vertical gradient data of salt concentration, capillary water flux and groundwater seepage direction, performs three-dimensional orthogonal decomposition on the groundwater seepage direction, extracts the vertical seepage component and marks its direction sign, inputs the dominant pore radius into the capillary dynamics constraint model, and generates the critical threshold of capillary water flux. When the real-time capillary water flux exceeds the critical threshold and the vertical seepage component is directed upward, a saltwater intrusion intensity index classification signal is output.

4. The coastal saline soil dynamic monitoring and early warning system based on the Internet of Things according to claim 3 is characterized by: The three-dimensional pore structure parameter dynamic correction capillary permeability adopts the throat connectivity feedback mechanism: The three-dimensional point cloud model of soil pores reconstructed by soil CT scanning is topologically simplified, the number of throat nodes and connected paths is identified, and the cross-density of throat connected paths within a unit volume is counted to generate a quantitative index of pore throat connectivity. Based on the graded signal of the saltwater intrusion intensity index, the initial capillary permeability is dynamically adjusted through the pore structure deformation compensation function to form a corrected capillary permeability.

5. The coastal saline soil dynamic monitoring and early warning system based on the Internet of Things according to claim 4 is characterized by: The convection-diffusion equation is used to solve the salt exchange capacity using physical field coupling driving technology: The corrected capillary conductivity is received as the migration rate control variable, and the soil texture database is called to match the clay content percentage of the current monitoring point, which is then mapped to the salt diffusion coefficient benchmark value. The vertical gradient data of the salt concentration is processed by central difference to generate the concentration gradient change rate. The capillary water rise flux and the concentration gradient change rate are superimposed, and the salt exchange amount per unit time is solved by the law of conservation of solute transport.

6. The coastal saline soil dynamic monitoring and early warning system based on the Internet of Things according to claim 5 is characterized by: The long short-term memory network time series prediction algorithm uses a time-delay compensation training architecture: A bidirectional long-short-term memory neural network model was constructed. The input layer received a continuous 24-hour time series of salt exchange, and the hidden layer implanted a lag effect compensation unit. The time lag characteristics of groundwater salt migration were learned through historical soil salt increment data. The output layer generated a predicted value of soil salt increment based on the time lag characteristics and simultaneously output a prediction confidence evaluation index.

7. The coastal saline soil dynamic monitoring and early warning system based on the Internet of Things according to claim 6 is characterized by: The coefficient of determination is calculated using a dynamic weight sliding window technique: A data sliding window is set up to capture the predicted and measured values ​​of soil salinity increment in real time. The data points in the window are dynamically weighted according to the corrected capillary conductivity, and the weighted Pearson correlation coefficient is calculated as the quantitative value of the determination coefficient, and the statistical significance mark is recorded simultaneously.

8. The coastal saline soil dynamic monitoring and early warning system based on the Internet of Things according to claim 7 is characterized by: When the quantitative value of the coefficient of determination exceeds the preset threshold, the pore structure self-calibration loop is activated: The absolute error between the predicted and measured soil salinity increments is converted into a pore throat connectivity correction coefficient. The throat connectivity quantification index is adjusted using the throat node deformation feedback function. The capillary conductivity dynamic correction process is re-executed based on the updated throat connectivity, forming a closed-loop capillary conductivity optimization.

9. The coastal saline soil dynamic monitoring and early warning system based on the Internet of Things according to claim 8 is characterized by: The association analysis module has a built-in contribution attribution analysis engine: When the early warning is triggered, the salt exchange amount, saltwater intrusion intensity index and vertical seepage component are extracted, and the contribution percentage of each parameter to the soil salt increment prediction result is quantified through the reverse gradient propagation technology. A parameter contribution ranking list is generated and the categories of the dominant influencing factors are marked.

10. The coastal saline soil dynamic monitoring and early warning system based on the Internet of Things according to claim 9 is characterized by: The early warning trigger unit (3) implements a contribution-driven hierarchical response strategy: When the contribution of capillary conductivity exceeds the threshold set by the system, the shallow vacuum pumping device is started, and the pumping depth automatically matches the capillary climbing height corresponding to the dominant pore radius. When the contribution of the saltwater intrusion intensity index exceeds the threshold set by the system, the deep well pressure reduction drainage system is activated, and the drainage flow is adjusted according to the proportion of the vertical seepage component.

Images