Saline-alkali soil drainage method, device and equipment based on Internet of Things monitoring and medium

Abstract

The invention relates to a saline-alkali soil drainage method, device and equipment based on Internet of Things monitoring and a medium. The method comprises the following steps: acquiring multi-source monitoring data in saline-alkali soil, and performing space-time alignment on the multi-source monitoring data to obtain space-time aligned multi-source monitoring data; inputting the time-space alignment multi-source monitoring data into a preset water-salt coordinated regulation and control model to obtain salt flux pointing to a drainage layer from a root layer; according to the salinity flux and a preset crop salt tolerance threshold value, a physiological stress index used for representing the current salinity stress risk and a dynamic regulation threshold value used for representing the leaching drainage window period are obtained; and matching the physiological stress index, the dynamic regulation and control threshold value and a preset cooperative regulation and control rule to generate a cooperative regulation and control instruction set containing an irrigation trigger instruction, an irrigation and drip washing quota, a drainage valve opening condition and a drainage valve closing condition. By adopting the method, the problems of disjunction of irrigation and drainage operation time sequence and single decision basis can be effectively solved, and the salt elimination efficiency and the water resource utilization rate are improved.

Description

Technical Field

[0001] This invention belongs to the field of agricultural drainage technology, and in particular relates to a method, device, equipment and medium for drainage of saline-alkali land based on Internet of Things monitoring. Background Technology

[0002] With the development of precision agriculture and smart water management technologies, IoT monitoring has become an important means of modern agricultural resource management. In the field of saline-alkali land improvement, using sensor networks to acquire soil information and combining it with drainage measures for salinity control is currently a major research direction.

[0003] In traditional techniques, drainage of saline-alkali land mainly relies on engineering measures such as underground pipe drainage and open ditch drainage, combined with periodic irrigation for salt leaching. In practice, irrigation and drainage are often managed as two independent agronomic processes: irrigation is usually implemented first to leach salts from the root zone, and then the drainage system is activated according to a fixed schedule or experience. This model relies on preset programs or thresholds for a single soil indicator (such as moisture content) for control.

[0004] However, current methods have significant problems. Irrigation and drainage lack dynamic coordination based on real-time salt transport, potentially causing drainage to miss the optimal window for salt leaching to the drainage layer, resulting in low desalination efficiency and water waste. Spatially, it is difficult to implement differentiated and precise control based on soil salinity heterogeneity. Essentially, existing methods fail to achieve intelligent decision-making in the coordinated "water-salt conjugate" process, and the disconnect between irrigation and drainage restricts the efficient utilization of water and soil resources in saline-alkali land. Summary of the Invention

[0005] Therefore, it is necessary to provide methods, devices, equipment, and media for drainage of saline-alkali land based on Internet of Things (IoT) monitoring to address the aforementioned technical problems.

[0006] Firstly, this application provides a method for draining saline-alkali land based on Internet of Things (IoT) monitoring, comprising:

[0007] S1. Acquire multi-source monitoring data in saline-alkali land and perform spatiotemporal alignment on the multi-source monitoring data to obtain spatiotemporally aligned multi-source monitoring data;

[0008] S2. Input the spatiotemporally aligned multi-source monitoring data into the preset water-salt synergistic regulation model to obtain the salt flux from the root layer to the drainage layer.

[0009] S3. Based on the salt flux and the preset crop salt tolerance threshold, obtain the physiological stress index used to characterize the current salt stress risk and the dynamic regulation threshold used to characterize the leaching drainage window period.

[0010] S4. Match physiological stress indicators, dynamic control thresholds and preset collaborative control rules to generate a set of collaborative control instructions that includes irrigation trigger commands, irrigation leaching quotas, drainage valve opening conditions and drainage valve closing conditions.

[0011] The coordinated control rules include the triggering and interlocking logic for irrigation and rinsing operations and drainage operations; the coordinated control instruction set is used to start irrigation according to the irrigation and rinsing quota when the irrigation trigger instruction is met, to start drainage when the drainage valve opening condition is met, and to stop drainage when the drainage valve closing condition is met.

[0012] In one embodiment, spatiotemporal alignment of multi-source monitoring data is performed to obtain spatiotemporally aligned multi-source monitoring data, including:

[0013] S11. Perform outlier cleaning on the multi-source monitoring data to obtain the cleaned monitoring data;

[0014] S12. Perform spatial interpolation on the cleaned monitoring data to obtain regular gridded spatial distribution data;

[0015] S13. Perform time alignment on the spatially distributed data to obtain time series data with a uniform time step;

[0016] S14. Perform weighted optimization and fusion on the time series data to obtain spatiotemporally aligned multi-source monitoring data.

[0017] In one embodiment, spatiotemporally aligned multi-source monitoring data is input into a preset water-salt synergistic regulation model to obtain the salt flux from the root layer to the drainage layer, including:

[0018] S21. Input the spatiotemporally aligned multi-source monitoring data into the water-salt co-regulation model; the water-salt co-regulation model includes a feature extraction network and a deep learning network;

[0019] S22. Feature extraction is performed on the spatiotemporally aligned multi-source monitoring data through a feature extraction network, and key influencing factors are extracted through principal component analysis to obtain the dimensionality-reduced feature data.

[0020] S23. Predict the dynamic changes of soil salinity using the dimensionality-reduced feature data through a deep learning network to obtain salinity prediction results;

[0021] S24. Based on the salinity prediction results, the salinity flux from the root zone to the drainage layer is calculated.

[0022] The expression for salt flux is:

[0023]

[0024] In the formula, Salt flux, The salt diffusion coefficient is... This refers to the salt concentration in the soil solution. For soil depth, This refers to the water flow rate.

[0025] In one embodiment, based on salt flux and a preset crop salt tolerance threshold, physiological stress indicators characterizing the current salt stress risk and dynamic regulation thresholds characterizing the leaching drainage window period are obtained, including:

[0026] S31. Establish a salt stress response function based on crop growth stage and soil salinity content, and calculate the physiological stress index using the following formula:

[0027]

[0028] In the formula, As a physiological stress indicator, The stress response coefficient is... This represents the current soil salinity. This is a half-lethal salt concentration;

[0029] S32. Based on the root zone soil salinity and crop salt tolerance threshold, obtain the leaching demand coefficient;

[0030] S33. A dynamic threshold model is established based on the salinity change rate and leaching demand coefficient, and the dynamic control threshold is calculated through an adaptive algorithm.

[0031] In one embodiment, physiological stress indicators, dynamic control thresholds, and preset collaborative control rules are matched to generate a set of collaborative control instructions, including irrigation trigger commands, irrigation leaching quotas, drainage valve opening conditions, and drainage valve closing conditions, comprising:

[0032] S41. Based on physiological stress indicators and dynamic control thresholds, calculate irrigation initiation membership degree and drainage initiation membership degree to obtain the initial action decision vector;

[0033] S42. Input the initial action decision vector into the preset rule conflict resolution model. Based on the current soil moisture content, weather forecast data and historical action effectiveness evaluation results, arbitrate and correct the conflict decision to obtain the optimized action decision.

[0034] S43. Based on the certainty of irrigation initiation in the optimized action decision, combined with the leaching demand coefficient and real-time soil moisture data, the irrigation leaching quota is calculated through the water balance model.

[0035] S44. Based on the certainty of drainage initiation, the direction of salinity flux, and the saturation status of the drainage layer in the optimized action decision, the logical judgment expressions for the opening and closing conditions of the drainage valve are dynamically generated to obtain a set of coordinated control instructions.

[0036] In one embodiment, a deep learning network is used to predict dynamic changes in soil salinity from dimensionality-reduced feature data, resulting in a salinity prediction outcome, including:

[0037] S51. Input the dimensionality-reduced feature data into the deep learning network; the deep learning network is a spatiotemporal attention recurrent neural network.

[0038] S52. In the spatiotemporal attention recurrent neural network, the temporal attention mechanism assigns weights to the features of the dimensionality-reduced feature data at different time steps, obtaining the first weight assignment; the spatial attention mechanism assigns weights to the dimensionality-reduced feature data at different soil depths, obtaining the second weight assignment.

[0039] S53. Based on the first and second allocation weights, the dimensionality-reduced feature data is processed through the recurrent layer of the spatiotemporal attention recurrent neural network to capture the spatiotemporal dependence of soil water and salt transport and output a predicted sequence of soil solution salt concentration for multiple future time steps.

[0040] S54. Based on the predicted soil solution salinity concentration sequence and the synchronously predicted soil moisture content sequence, the predicted root zone soil salinity is calculated and used as the salinity prediction result.

[0041] Secondly, this application also provides a drainage device for saline-alkali land based on Internet of Things monitoring, comprising:

[0042] The multi-source data acquisition module is used to acquire multi-source monitoring data in saline-alkali land and perform spatiotemporal alignment on the multi-source monitoring data to obtain spatiotemporally aligned multi-source monitoring data.

[0043] The water and salt analysis module is used to input spatiotemporally aligned multi-source monitoring data into a preset water and salt synergistic regulation model to obtain the salt flux from the root layer to the drainage layer;

[0044] The salt tolerance analysis and drainage threshold analysis modules are used to obtain physiological stress indicators to characterize the current salt stress risk and dynamic regulation thresholds to characterize the leaching drainage window period based on salt flux and preset crop salt tolerance thresholds.

[0045] The drainage coordinated control instruction generation module is used to match physiological stress indicators, dynamic control thresholds and preset coordinated control rules to generate a set of coordinated control instructions, including irrigation trigger instructions, irrigation leaching quotas, drainage valve opening conditions and drainage valve closing conditions.

[0046] The coordinated control rules include the triggering and interlocking logic for irrigation and rinsing operations and drainage operations; the coordinated control instruction set is used to start irrigation according to the irrigation and rinsing quota when the irrigation trigger instruction is met, to start drainage when the drainage valve opening condition is met, and to stop drainage when the drainage valve closing condition is met.

[0047] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect.

[0048] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect.

[0049] The aforementioned IoT-based drainage method, device, equipment, and medium for saline-alkali land utilizes IoT to acquire multi-source monitoring data and achieve spatiotemporal alignment. Combined with a water-salt synergistic regulation model, it accurately obtains the salt flux from the root zone to the drainage layer. Based on the salt flux and crop salt tolerance threshold, it determines physiological stress indicators and dynamic regulation thresholds. Finally, it matches synergistic regulation rules to generate a set of instructions including irrigation and drainage triggers, quotas, and switching conditions. This effectively solves the problems of traditional methods lacking real-time dynamic coordination of salt movement during irrigation and drainage, and missing the optimal drainage window. It achieves intelligent decision-making for coordinated irrigation and drainage, while supporting spatially differentiated regulation through spatiotemporally aligned data. This avoids the drawbacks of extensive management in traditional models, improves salt drainage efficiency, reduces water waste, and promotes the efficient utilization of water and soil resources in saline-alkali land. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating a method for draining saline-alkali land based on IoT monitoring in one embodiment.

[0052] Figure 2 This is a schematic diagram of a saline-alkali land drainage device based on Internet of Things monitoring in one embodiment. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0054] refer to Figure 1 The document presents a flowchart illustrating the IoT-based drainage method for saline-alkali land provided in this application. The method includes the following steps:

[0055] S1. Obtain multi-source monitoring data in saline-alkali land and perform spatiotemporal alignment of the multi-source monitoring data to obtain spatiotemporally aligned multi-source monitoring data.

[0056] Specifically, the first step is to construct a monitoring network. This network adopts a hierarchical distributed deployment architecture, encompassing four core modules: soil parameter monitoring units, hydrological parameter monitoring units, meteorological parameter monitoring units, and location units. The soil parameter monitoring unit is the core sensing module, employing an array-style arrangement. The array density is dynamically adjusted based on the degree of soil heterogeneity in the target saline-alkali land. The spacing between monitoring points in highly heterogeneous and homogeneous areas is set according to the actual soil heterogeneity detection results. Sensors are vertically layered at each monitoring point, corresponding to the crop root layer, transition layer, and drainage layer, respectively. The drainage layer refers to the burial layer of the underground drainage system. Sensor types include soil conductivity sensors, soil moisture sensors, and soil temperature sensors. The soil conductivity sensor uses the four-electrode method to directly characterize soil salinity; the soil moisture sensor uses the time-domain reflectometry principle to synchronously acquire water transport data. The hydrological parameter monitoring unit is mainly deployed at the collection wells and main drainage pipe outlets of the drainage system. It is equipped with electromagnetic flowmeters and water level sensors. The electromagnetic flowmeters monitor drainage flow, while the water level sensors, using an immersion hydrostatic principle, monitor changes in the drainage layer water level. The meteorological parameter monitoring unit is deployed at a meteorological station located in the center of the target area, collecting parameters such as rainfall, evaporation, temperature, and wind speed. Rainfall is collected using a tipping bucket rain gauge, and evaporation is collected in real time using a combination of evaporation pans and sensors. The positioning unit uses a dual-mode GPS and BeiDou positioning module, adding spatiotemporal stamp information to the data from each monitoring point to ensure the spatiotemporal traceability of the data.

[0057] The acquisition of multi-source monitoring data adopts a combination of periodic acquisition and real-time triggered acquisition. The regular acquisition cycle for soil parameters and hydrological parameters, and the acquisition cycle for meteorological parameters are set according to monitoring needs. When a sudden change in soil conductivity value, rainfall, or drainage flow reaches a set abnormal threshold, a high-frequency acquisition mode is triggered to shorten the acquisition cycle and ensure data integrity under critical operating conditions. Data transmission adopts a two-level transmission architecture of "local edge computing node + remote cloud platform". Data collected by each sensor is aggregated to the local edge computing node via a bus. The edge computing node uses an industrial-grade embedded processor to complete the initial filtering and format conversion of the data. The filtering adopts the Kalman filter algorithm, and the filtering coefficient is dynamically adjusted according to the data type. After unified format conversion, the data is transmitted to the remote cloud platform through a wireless communication module. The core of spatiotemporal alignment processing is to solve the problems of asynchronous time dimension and spatial mismatch of multi-source data. Temporal alignment employs linear interpolation, using the highest acquisition frequency among all data points as the baseline time axis. Low-frequency data is interpolated and supplemented. For time periods with minimal missing data, linear interpolation of adjacent data is used; for time periods with significant missing data, these are marked as data outliers and supplemented using historical data from similar operating conditions. Spatial alignment is based on the rasterization processing of a geographic information system. The target saline-alkali land area is divided into raster cells. Using the center coordinates of each raster as a reference, inverse distance weighted interpolation is used to spatially interpolate the discrete monitoring point data, ensuring higher accuracy for raster data closer to the monitoring point. This ultimately generates rasterized spatiotemporally aligned multi-source monitoring data covering the entire area with one-to-one spatiotemporal coordinate correspondence. Data dimensions include time, spatial coordinates, soil electrical conductivity, soil moisture content, soil temperature, drainage flow rate, drainage layer water level, rainfall, and evaporation.

[0058] S2. Input the spatiotemporally aligned multi-source monitoring data into the preset water-salt coordinated control model to obtain the salt flux from the root layer to the drainage layer.

[0059] Optionally, the preset water-salt synergistic regulation model is constructed based on an improved version of the HYDRUS-1D water-salt transport mechanism model. Machine learning algorithms are integrated to dynamically optimize model parameters. The model input parameters are spatiotemporally aligned multi-source monitoring data, including soil moisture content, soil conductivity, soil temperature, drainage layer water level, rainfall, and evaporation at each layer. The output parameter is the salt flux from the root zone to the drainage layer. The core principle of the model is based on the Richards equation to describe the water transport process and the convection-dispersion equation to describe the salt transport process. Simultaneously, crop root water uptake and salt adsorption-desorption terms are introduced to achieve accurate simulation of water-salt synergistic transport.

[0060] The governing equation for water transport is: .in, Represents soil moisture content. Represents time, Represents vertical depth. This represents soil hydraulic conductivity, which is related to soil moisture content and is obtained by fitting a soil moisture characteristic curve. Represents soil water potential. This represents the water absorption rate of crop roots. The crop root water absorption rate is calculated using the FedDes model, and the calculation relationship is as follows: .in, This represents the maximum root water absorption rate, preset according to crop type. The value represents the water stress coefficient, which is dynamically adjusted based on the measured soil moisture content. When the soil moisture content is within a suitable range, the water stress coefficient is 1, and when the soil moisture content is lower than the wilting moisture content, the water stress coefficient is 0.

[0061] The governing equation for salt transport is: .in, This represents the soil solution salinity concentration, calculated from the soil electrical conductivity value. The conversion relationship is a linear correlation between soil solution salinity concentration and soil electrical conductivity value, with the correlation coefficient determined according to soil type. The salt dispersion coefficient represents the salt concentration, which is positively correlated with soil moisture content. It was determined through indoor experiments. Represents soil water flux. Represents soil bulk density. The adsorbed salt content in the soil is represented by the Langmuir adsorption model, which describes the adsorption-desorption process. This represents the salt source and sink term. The salt source and sink term is negative during irrigation and leaching, and positive during salt accumulation.

[0062] The model construction process includes two stages: offline calibration and online optimization. During offline calibration, soil samples from the target saline-alkali land are collected. Basic parameters such as soil bulk density, soil particle composition, and soil moisture characteristic curves are measured through indoor experiments and input into the HYDRUS-1D model for simulation. Combined with field-monitored salt transport data, a genetic algorithm is used to optimize sensitive parameters in the model, such as dispersion coefficient and hydraulic conductivity, ensuring that the error between the simulated and measured values ​​meets the set requirements. In the online optimization stage, the spatiotemporally aligned multi-source monitoring data obtained in step S1 is input into the model in real time. A long short-term memory (LSTM) network algorithm is used to dynamically correct the salt flux output by the model. The LTM network is trained using historical monitoring data and model output data. The initial salt flux value calculated by the model is compared with the real-time monitored drainage layer conductivity and drainage flow data. Correction coefficients are output through the LTM network to correct the initial salt flux, ultimately obtaining a high-precision salt flux result.

[0063] Specifically, firstly, the spatiotemporally aligned soil moisture content, electrical conductivity, and other data for each layer are input into the model, and the soil water flux at each depth is calculated using the Richards equation. Then, parameters such as soil water flux and soil solution salt concentration are substituted into the convection-diffusion equation to calculate the vertical migration rate of salt. Finally, the mass of salt that travels from the root zone through the transition layer to the drainage layer per unit time is integrated to obtain the salt flux. The expression for calculating the salt flux can be: .in, Represents salt flux. Represents the depth of the root layer bottom. Represents the depth of the top of the drainage layer. Represents the area of ​​the calculation unit. This represents the computation time step, which is consistent with the data acquisition cycle. The salt flux calculated by this model can reflect the migration status of salt from the root zone to the drainage layer in real time, providing core data support for subsequent determination of the optimal drainage window.

[0064] S3. Based on the salt flux and the preset crop salt tolerance threshold, obtain the physiological stress index used to characterize the current salt stress risk and the dynamic regulation threshold used to characterize the leaching drainage window period.

[0065] Optionally, the preset crop salt tolerance threshold refers to the upper limit of soil salinity that a specific crop can tolerate at different growth stages. This threshold needs to be pre-calibrated through field trials based on the type of crop to be planted in the target saline-alkali land. The salt tolerance conductivity threshold at each growth stage is converted into the corresponding soil solution salt concentration threshold and stored in the system parameter library, which can be updated in real time according to the dynamic changes in the crop growth stage.

[0066] For example, the construction of physiological stress indices is based on the coupled analysis of salt flux and crop salt tolerance threshold. The core principle is that when soil salinity approaches or exceeds the crop salt tolerance threshold, it inhibits root water and nutrient uptake, leading to physiological stress. Salt flux directly reflects the rate of salt accumulation or migration into the root zone, further influencing the trend of stress risk changes. Specifically, the physiological stress indices are calculated using a normalized method, comprehensively considering the influence of current root zone soil salinity concentration and salt flux. The calculation expression for the physiological stress indices can be expressed as: .in, Represents physiological stress indicators. , The representative weighting coefficients are determined using the analytic hierarchy process (AHP). The influence weight of the current salt concentration, The influence weight of salt migration rate, This represents the current salt concentration in the root zone soil. The soil solution salt concentration threshold representing crop salt tolerance. Represents salt flux. The maximum salinity flux threshold is preset and calibrated based on the drainage system design capacity of the target saline-alkali land. The physiological stress index ranges from 0 to 100, with different ranges corresponding to different levels of salinity stress risk. The lower range indicates no risk of salinity stress, while the higher the value, the higher the risk of mild, moderate, and severe stress.

[0067] The dynamic control threshold is used to accurately define the leaching drainage window period. When the salt flux reaches this threshold, it indicates that a large amount of salt has migrated from the root zone to the drainage layer. At this time, activating the drainage system can achieve efficient salt removal while reducing water waste. The determination of the dynamic control threshold needs to combine the trend of salt flux changes with the operating efficiency of the drainage system. It is calculated using a dynamic threshold model. The input parameters of this model include salt flux, root zone soil moisture content, drainage layer water level, and a preset drainage efficiency coefficient. The drainage efficiency coefficient is determined based on parameters such as the diameter and burial depth of the drainage pipe.

[0068] Specifically, the basic threshold is first determined, and its calculation expression can be expressed as: .in, Represents the basic threshold. This represents the optimal root zone salt concentration for crop growth, preset according to crop type. Represents the soil moisture content in the root zone. Represents the root layer thickness. This represents the rinsing cycle, preset according to local climate conditions. Then, the basic threshold is dynamically adjusted based on real-time monitoring of root zone soil moisture content and drainage layer water level. The adjustment formula is as follows: .in, This represents the corrected threshold. , Represents the correction factor. This represents the optimal root layer water content for crop growth, preset according to crop type. Represents the water level of the drainage layer. The appropriate water level for the drainage layer is determined based on the drainage system design. Finally, a drainage efficiency coefficient is introduced to calibrate the corrected threshold, yielding the final dynamic control threshold. The calibration formula can be expressed as: .in, This represents the final dynamic control threshold. This represents the drainage efficiency coefficient. When the real-time calculated salt flux is greater than or equal to the final dynamic control threshold, it indicates that the leaching drainage window has been entered. At this time, opening the drainage system can achieve efficient salt removal. When the salt flux is less than the final dynamic control threshold, it indicates that salt has not yet migrated to the drainage layer in large quantities. Opening the drainage system at this time will lead to water waste and low salt removal efficiency. By synergistically calculating physiological stress indicators and dynamic control thresholds, real-time assessment of crop salt stress risk and accurate determination of the drainage window can be achieved simultaneously, providing a scientific basis for the generation of subsequent synergistic control instructions.

[0069] S4. Match physiological stress indicators, dynamic control thresholds and preset collaborative control rules to generate a set of collaborative control instructions that includes irrigation trigger commands, irrigation leaching quotas, drainage valve opening conditions and drainage valve closing conditions.

[0070] Optionally, the coordinated control rules include the triggering and interlocking logic for irrigation and rinsing operations and drainage operations; the coordinated control instruction set is used to start irrigation according to the irrigation and rinsing quota when the irrigation trigger instruction is met, to start drainage when the drainage valve opening condition is met, and to stop drainage when the drainage valve closing condition is met.

[0071] Optionally, the coordinated regulation rules are a set of pre-defined logical rules based on crop growth needs, water and salt transport patterns, and drainage system operation characteristics. These rules clarify the triggering conditions, execution parameters, and interlocking logic for irrigation and drainage operations, ensuring the coordination and precision of irrigation and drainage operations.

[0072] Optionally, the preset coordinated regulation rules are represented by production rules, with the rule form being "IF condition set THEN operation set". The condition set includes the range of physiological stress indicators, the relationship between salinity flux and dynamic regulation threshold, the deviation between root soil moisture content and suitable moisture content, and meteorological forecast data. The operation set includes whether irrigation is triggered, the determination of irrigation leaching quotas, and the setting of conditions for opening / closing drainage valves. Simultaneously, the interlocking logic between irrigation and drainage is clearly defined: opening drainage valves is prohibited during irrigation to prevent direct loss of irrigation water; irrigation is prohibited during drainage to ensure efficient operation of the drainage system. Specifically, the core provisions of the coordinated regulation rules are set around the combination of different physiological stress risk levels with soil moisture and meteorological conditions. For severe stress risk and soil moisture content below the suitable range, emergency irrigation and leaching operations are triggered, and drainage triggering conditions are set after irrigation. For moderate stress risk and no effective rainfall forecast, regular irrigation and leaching operations are triggered, and the timing of drainage is dynamically determined based on salt flux after irrigation. For mild stress risk and soil moisture content near the suitable range, irrigation is not triggered, only salt flux is monitored, and drainage is initiated when the dynamic regulation threshold is reached. For no stress risk, irrigation is not triggered, only the regular monitoring status of the drainage system is maintained, and drainage is initiated when the drainage layer water level exceeds the suitable range to avoid water accumulation affecting crop root growth.

[0073] The irrigation leaching quota is determined using a combination of the water balance method and the salt balance method to ensure that the crop's water requirements are met while effectively leaching salts from the root zone. Specifically, the basic irrigation quota is first calculated using the water balance method. The formula for calculating the basic irrigation quota is as follows: .in, Represents the basic irrigation quota. Representative unit conversion factor, The evapotranspiration of crops over a specified future time period is calculated based on meteorological data and crop coefficients, which are determined using the FAO-recommended single-crop coefficient method. The effective rainfall amount, representing a future time period, is calculated from the predicted rainfall value and the effective rainfall coefficient. Then, the additional irrigation quota required for leaching is calculated using the salinity balance method. The formula for calculating the additional irrigation quota is: .in, Represents additional irrigation quota. This represents the salinity concentration of the irrigation water, and is a measured value. This represents the leaching efficiency, determined based on soil texture. The final irrigation leaching quota is the sum of the basic irrigation quota and the additional irrigation quota, while an upper limit is set for the final irrigation leaching quota, determined based on local water resource conditions to avoid excessive water consumption.

[0074] The opening and closing conditions of the drainage valves are set based on a coordinated assessment of salinity flux, drainage layer water level, and drainage flow rate. Specifically, the opening conditions of the drainage valves are divided into two operating conditions: Operating condition one is post-irrigation flushing drainage. After the irrigation operation is completed, the salinity flux is monitored in real time. When the salinity flux reaches the dynamic control threshold and the duration meets the set requirements, the drainage valve opening command is triggered, and the valve opening degree is set. Operating condition two is regular drainage. During non-irrigation periods, when the drainage layer water level exceeds the suitable range or the salinity flux reaches the dynamic control threshold, the drainage valve opening command is triggered. The valve opening degree is dynamically adjusted according to the drainage flow rate to avoid overloading the drainage system. The closing conditions of the drainage valves are: when the drainage layer water level is lower than the lower limit of the suitable range, the salinity flux is lower than the set proportion of the dynamic control threshold, and the drainage flow rate is lower than the set value for a continuous set time, the drainage valve closing command is triggered. The valve closing process adopts a gradual closing to avoid water hammer impact on the drainage pipes.

[0075] The generation process of the coordinated regulation command set is implemented using a rule engine. The rule engine uses a pattern matching algorithm to match real-time calculated physiological stress indicators, salinity flux, and other parameters with preset coordinated regulation rules. When a set of conditions for a certain rule is met, the corresponding set of operations is automatically triggered, generating standardized coordinated regulation commands. The command format uses a standardized data format, including command type, trigger time, execution parameters, monitoring requirements, and other information. The generated coordinated regulation command set is transmitted to the field controller via a remote communication module. The field controller uses a programmable logic controller (PLC) as the core control unit, driving the actions of actuators such as irrigation pumps and electromagnetic drainage valves according to the command content. Simultaneously, it provides real-time feedback of execution status data to the remote cloud platform, forming a closed-loop regulation process of "monitoring-decision-control-feedback." This ensures precise coordination of irrigation and drainage operations, achieving efficient utilization of water and soil resources in saline-alkali land.

[0076] In the aforementioned IoT-based drainage method for saline-alkali land, multi-source monitoring data is acquired and spatiotemporally aligned via the IoT, providing comprehensive and consistent data support for precise regulation. A water-salt synergistic regulation model accurately captures the salt flux from the root zone to the drainage layer, enabling real-time perception of soil salt transport status. Combining crop salt tolerance thresholds with physiological stress indicators and dynamic regulation thresholds allows for precise determination of salt stress risk and leaching drainage window periods. Based on synergistic regulation rules, intelligent matching and interlocking of irrigation and drainage operations are achieved, effectively solving the problems of lack of dynamic coordination in irrigation and drainage timing and missed optimal drainage windows in traditional methods. Simultaneously, it takes into account the spatial heterogeneity of soil salinity to achieve differentiated regulation, breaking the limitations of irrigation-drainage disconnect, improving salt drainage efficiency, reducing water waste, and promoting the efficient utilization of water and soil resources in saline-alkali land.

[0077] In an optional embodiment, spatiotemporal alignment of the multi-source monitoring data is performed to obtain spatiotemporally aligned multi-source monitoring data, including:

[0078] S11. Perform outlier cleaning on the multi-source monitoring data to obtain cleaned monitoring data.

[0079] Optionally, a dual judgment system is constructed using statistical criteria and physical attribute constraints. The statistical criteria employ the 3σ criterion or interquartile range method to screen for suspected outliers, while the physical attribute constraints are further verified by considering the reasonable range of monitoring parameters. For identified outliers, if the proportion is lower than the set standard, it is supplemented by linear interpolation of adjacent data; if the proportion is too high, it is directly deleted and the missing segment is marked, ultimately obtaining complete and accurate cleaned monitoring data. By eliminating outliers caused by sensor malfunctions, transmission interference, etc., data reliability is ensured.

[0080] S12. Spatial interpolation is performed on the cleaned monitoring data to obtain regular gridded spatial distribution data.

[0081] Optionally, to address the issue of incomplete coverage of discrete monitoring point data, the target area is first rasterized based on a geographic information system, with the raster size determined according to the control precision. An inverse distance-weighted interpolation method is used, based on the monitoring point data, to calculate raster parameter values ​​by assigning weights according to distance. For edge rasteres, if there are insufficient effective monitoring points, trend surface analysis is incorporated for correction, ensuring the formation of a continuous and consistent regular gridded spatial distribution data covering the entire area.

[0082] S13. Perform time alignment on the spatially distributed data to obtain time-series data with a uniform time step.

[0083] Optionally, to eliminate differences in data acquisition frequencies, the highest acquisition frequency among all data points is used as the baseline time step. Linear interpolation is used to supplement the parameter values ​​at the baseline time step for low-frequency data, while mean sampling is used to downsample the high-frequency data to the baseline time step. For missing data segments, short missing segments are filled using interpolation of data from adjacent time points, while long missing segments are filled using historical data from similar operating conditions during the same period, ultimately forming time-series data with a unified time step and one-to-one spatiotemporal coordinate correspondence.

[0084] S14. Perform weighted optimization and fusion on the time series data to obtain spatiotemporally aligned multi-source monitoring data.

[0085] Optionally, deep fusion of multi-source data can be achieved by optimizing weight allocation to improve data accuracy. Data reliability, information entropy, and correlation with target parameters are selected as evaluation indicators. The analytic hierarchy process (AHP) combined with entropy weighting is used to determine the comprehensive weight of each data point, taking into account both subjective professional understanding and objective data characteristics. Time-series data are weighted and calculated according to their weights, and fusion errors are verified in real time. If a threshold is exceeded, the weights are adjusted and the data is re-fused, ultimately obtaining standardized, highly reliable spatiotemporally aligned multi-source monitoring data.

[0086] In an optional embodiment, spatiotemporally aligned multi-source monitoring data is input into a preset water-salt synergistic regulation model to obtain the salt flux from the root layer to the drainage layer, including:

[0087] S21. Input the spatiotemporally aligned multi-source monitoring data into the water-salt coordinated regulation model.

[0088] Optionally, the water-salt co-regulation model adopts a dual-network architecture, including a feature extraction network and a deep learning network. The two networks work together to achieve accurate derivation of salinity flux from multi-source monitoring data. Spatiotemporally aligned multi-source monitoring data serves as the input data source for the model, providing a complete and standardized data foundation for subsequent feature extraction and salinity prediction.

[0089] S22. Feature extraction is performed on the spatiotemporally aligned multi-source monitoring data using a feature extraction network. Key influencing factors are extracted using principal component analysis to obtain dimensionality-reduced feature data.

[0090] Optionally, the feature extraction network is used to remove redundant information from multi-source data and focus on the core features related to soil salinity changes; principal component analysis transforms multi-dimensional monitoring data into a few linearly independent principal components through orthogonal transformation, retaining most of the effective information in the data, achieving data dimensionality reduction, reducing the computational load of subsequent deep learning networks, and improving the model prediction efficiency and accuracy, ultimately outputting feature data with simplified dimensions and concentrated information.

[0091] S23. Predict the dynamic changes of soil salinity using the dimensionality-reduced feature data through a deep learning network, and obtain the salinity prediction results.

[0092] Optionally, the deep learning network is used to perform a comprehensive analysis of the temporal and spatial dimensions of the dimensionality-reduced feature data based on the trained network parameters, accurately fit the variation law of soil salinity over time and space, and realize the prediction of soil salinity distribution and change trend in the future. The output salinity prediction results containing soil salinity information at different depths provide a direct basis for subsequent salinity flux calculation.

[0093] S24. Based on the salinity prediction results, the salinity flux from the root zone to the drainage layer is calculated.

[0094] The expression for salt flux is:

[0095]

[0096] In the formula, Salt flux, The salt diffusion coefficient is... This refers to the salt concentration in the soil solution. For soil depth, This refers to the water flow rate.

[0097] In the above formula, Salt flux is the mass of salt that migrates from the root zone to the drainage zone through a unit area per unit time. The salt diffusion coefficient characterizes the ability of salt to diffuse and migrate in the soil. The soil solution salinity concentration is derived from salinity prediction results. The soil depth is the vertical range from the root zone to the drainage layer. Water flux, reflecting the rate of water migration in the soil. This expression, combined with salinity prediction results... Value and corresponding depth Substitute water flow rate and salt diffusion coefficient The salt flux from the root layer to the drainage layer can then be calculated.

[0098] The model construction process includes two stages: offline calibration and online optimization. During offline calibration, soil samples were collected from the target saline-alkali land. Basic parameters such as soil bulk density, soil particle composition, and soil moisture characteristic curves were measured through indoor experiments and input into the model for training. Network parameters were optimized by combining data on salt transport from field monitoring, ensuring that the error between the model's predicted values ​​and the measured values ​​met the set requirements. In the online optimization stage, spatiotemporally aligned multi-source monitoring data was input into the model in real time. The prediction results were dynamically corrected by combining data from drainage layer monitoring to ensure the accuracy of salt flux calculation.

[0099] In an optional embodiment, based on salt flux and a preset crop salt tolerance threshold, physiological stress indicators characterizing the current salt stress risk and dynamic regulation thresholds characterizing the leaching drainage window period are obtained, including:

[0100] S31. Establish a salt stress response function based on crop growth stage and soil salinity content, and calculate the physiological stress index using the following formula:

[0101]

[0102] In the formula, As a physiological stress indicator, The stress response coefficient is... This represents the current soil salinity. This is the half-lethal salt concentration.

[0103] Optionally, since different crop growth stages exhibit varying tolerance to salt, a response function that accurately reflects the correlation between salt content and crop stress can be constructed by combining measured soil salinity data from each growth stage. In the calculation expression for the aforementioned physiological stress index, It is a physiological stress indicator used to quantify the current degree of salt stress on crops; The stress response coefficient is determined through field trials based on crop varieties and reflects the crop's sensitivity to salt stress. The current soil salinity is obtained from spatiotemporally aligned multi-source monitoring data; The half-lethal salt concentration, i.e., the soil salinity concentration at which crop growth is significantly inhibited and the survival rate drops to 50%, is determined in advance through variety trials. Substituting the parameters allows for the calculation of the physiological stress index corresponding to the growth stage.

[0104] S32. Based on the salt content of the root zone soil and the salt tolerance threshold of the crop, the leaching demand coefficient is obtained.

[0105] Optionally, the root zone soil salinity is calculated by multiplying the root zone soil salinity content by the root zone volume, directly reflecting the total amount of salt accumulated in the root zone. The crop salt tolerance threshold is the upper limit of salt tolerance for a specific crop at a corresponding growth stage. The difference between the two is used to determine the current degree of excessive root zone salinity, thereby determining the urgency of leaching. The core function of the leaching demand coefficient is to quantify the leaching intensity required to reduce root zone salinity to a suitable range. The larger the coefficient, the more urgent the leaching demand. It is calculated by the ratio of root zone salinity to the salinity corresponding to the crop salt tolerance threshold, providing a basis for determining the subsequent dynamic control threshold.

[0106] S33. A dynamic threshold model is established based on the salinity change rate and leaching demand coefficient, and the dynamic control threshold is calculated through an adaptive algorithm.

[0107] Optionally, the salinity change rate is calculated from the salinity flux at different times, reflecting the changing trend of the rate of salinity migration to the drainage layer. A dynamic threshold model integrates the salinity change rate and the leaching demand coefficient to establish a correlation between the two and the dynamic control threshold. An adaptive algorithm is used to optimize the model parameters in real time, enabling the model to adaptively adjust according to dynamic changes in soil salinity, ensuring that the output dynamic control threshold accurately matches the current leaching drainage demand. The dynamic control threshold calculated by this model can accurately determine the optimal time to initiate drainage, achieving efficient leaching removal of salinity.

[0108] In an optional embodiment, physiological stress indicators, dynamic control thresholds, and preset collaborative control rules are matched to generate a set of collaborative control instructions, including irrigation trigger commands, irrigation leaching quotas, drainage valve opening conditions, and drainage valve closing conditions, comprising:

[0109] S41. Based on physiological stress indicators and dynamic control thresholds, calculate the irrigation initiation membership degree and drainage initiation membership degree to obtain the initial action decision vector.

[0110] Optionally, the irrigation initiation membership degree is used to quantify the urgency of initiating irrigation operations, calculated through the deviation relationship between physiological stress indicators and crop salinity suitability zones; the drainage initiation membership degree is determined by comparing salinity flux with dynamic control thresholds, characterizing the necessity of initiating drainage operations. Using these two membership degree values ​​as core elements, an initial action decision vector is constructed. This vector intuitively reflects the initial judgment of the priority of irrigation and drainage actions under the current operating conditions, providing basic data support for subsequent decision optimization.

[0111] S42. Input the initial action decision vector into the preset rule conflict resolution model. Based on the current soil moisture content, weather forecast data and historical action effectiveness evaluation results, arbitrate and correct the conflict decision to obtain the optimized action decision.

[0112] Optionally, the core function of the rule conflict resolution model is to address potential conflicts between irrigation and drainage actions, and mismatches between actions and environmental conditions, in the initial decision-making process. It assesses water supply and demand based on current soil moisture content, predicts the impact of future precipitation on salt leaching based on meteorological forecast data, and arbitrates and corrects the initial decision based on historical action effectiveness evaluation results (such as salt removal efficiency and water resource utilization rate under similar past conditions). Finally, it outputs optimized action decisions adapted to the current actual conditions.

[0113] S43. Based on the certainty of irrigation initiation in the optimized action decision, combined with the leaching demand coefficient and real-time soil moisture data, the irrigation leaching quota is calculated through the water balance model.

[0114] Optionally, the irrigation initiation determination quantifies the clarity of irrigation initiation in the optimization decision-making process and serves as a weighted reference for calculating the leaching quota. The water balance model comprehensively considers the difference between real-time soil moisture and suitable crop moisture, the salt leaching water demand corresponding to the leaching demand coefficient, and also incorporates parameters such as crop evapotranspiration and effective rainfall. Through balance calculations, it determines the irrigation leaching quota that satisfies both water replenishment and salt leaching, avoiding water waste and insufficient leaching.

[0115] S44. Based on the certainty of drainage initiation, the direction of salinity flux, and the saturation status of the drainage layer in the optimized action decision, the logical judgment expressions for the opening and closing conditions of the drainage valve are dynamically generated to obtain a set of coordinated control instructions.

[0116] Optionally, the certainty of drainage initiation determines the sensitivity of valve opening, the direction of salt flux reflects the salt migration trend (opening preferentially when pointing towards the drainage layer), and the saturation status of the drainage layer directly affects drainage efficiency and crop root growth safety. By combining these three factors to construct a logical judgment expression, the triggering conditions for valve opening (e.g., certainty meets the standard + salt flux points towards the drainage layer + drainage layer is not saturated) and the closing conditions (e.g., salt flux is below the threshold + drainage layer reaches a suitable water level) are clearly defined. After integrating parameters such as irrigation leaching quotas, a complete set of coordinated control instructions is formed.

[0117] In an optional embodiment, a deep learning network is used to predict the dynamic changes in soil salinity from the dimensionality-reduced feature data, resulting in a salinity prediction outcome, including:

[0118] S51. Input the dimensionality-reduced feature data into the deep learning network.

[0119] Optionally, the deep learning network is a spatiotemporal attention recurrent neural network. This network integrates the spatiotemporal attention mechanism with the recurrent neural network structure, which can accurately capture the spatiotemporal dependence characteristics of soil salinity changes. The dimensionality-reduced feature data is used as network input and needs to be converted into tensor data adapted to the network input format in advance, including time, space (soil depth) and feature dimension information, to lay the data foundation for subsequent spatiotemporal weight allocation and dependency capture.

[0120] S52. In the spatiotemporal attention recurrent neural network, the temporal attention mechanism assigns weights to the features of the dimensionality-reduced feature data at different time steps, obtaining the first assigned weight; the spatial attention mechanism assigns weights to the dimensionality-reduced feature data at different soil depths, obtaining the second assigned weight.

[0121] Optionally, the temporal attention mechanism calculates the correlation between features at different time steps and the current prediction task, and assigns differentiated first allocation weights to the features at each time step of the dimensionality-reduced feature data, so that the network focuses on the time stage data that contributes more to salinity prediction; the spatial attention mechanism assigns second allocation weights to the dimensionality-reduced feature data at different soil depths, combined with the correlation of salinity transport in each soil layer, to strengthen the influence weight of key soil layer (such as root layer and transition layer) features and weaken the interference of irrelevant soil layers.

[0122] S53. Based on the first and second allocation weights, the dimensionality-reduced feature data is processed through the recurrent layer of the spatiotemporal attention recurrent neural network to capture the spatiotemporal dependence of soil water and salt transport and output a predicted sequence of soil solution salt concentration for multiple future time steps.

[0123] Optionally, the recurrent layer uses the weight allocation results to perform weighted aggregation of feature data, and captures the cross-time and cross-space dependencies in the process of soil water and salt transport through iterative calculations, namely the inheritance relationship of salt changes at different time steps and the mutual influence law of salt in different soil layers. Finally, it outputs the predicted sequence of soil solution salt concentration for multiple consecutive time steps in the future, realizing the time-series prediction of dynamic salt changes.

[0124] S54. Based on the predicted soil solution salinity concentration sequence and the synchronously predicted soil moisture content sequence, the predicted root zone soil salinity is calculated and used as the salinity prediction result.

[0125] Optionally, the synchronously predicted soil moisture content sequence is output synchronously by a spatiotemporal attention recurrent neural network, and the root zone soil salt storage is calculated by summing the products of soil solution salt concentration, soil moisture content and corresponding soil volume of each soil layer in the root zone. This result comprehensively reflects the accumulation state of salt in the root zone and provides a predictive basis for subsequent salt flux calculation and regulation decisions.

[0126] The aforementioned IoT-based drainage method for saline-alkali land acquires multi-source monitoring data and aligns it spatiotemporally through the IoT. It accurately obtains the salt flux from the root layer to the drainage layer based on a water-salt synergistic regulation model. Combined with crop salt tolerance thresholds, it generates physiological stress indicators and dynamic regulation thresholds. Then, through synergistic regulation rule matching, it generates a set of instructions including irrigation / drainage triggers, quotas, and valve start / stop conditions, achieving coordinated irrigation and drainage. This effectively solves the problems of traditional technologies, such as lack of real-time dynamic coordination of salt movement during irrigation and drainage, missed optimal drainage windows, low salt removal efficiency, and water waste. Simultaneously, it can adapt to the spatial heterogeneity of soil salinity to achieve precise regulation, overcoming the limitations of existing methods that disconnect irrigation and drainage. It realizes intelligent decision-making in the linkage of water-salt conjugate processes, improving the efficient utilization of water and soil resources in saline-alkali land.

[0127] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0128] Based on the same inventive concept, this application also provides an IoT-based saline-alkali land drainage device for implementing the above-mentioned IoT-based saline-alkali land drainage method. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more IoT-based saline-alkali land drainage device embodiments provided below can be found in the above-described limitations of the IoT-based saline-alkali land drainage method, and will not be repeated here.

[0129] In one exemplary embodiment, such as Figure 2 As shown, a schematic diagram of a saline-alkali land drainage device 10 based on Internet of Things monitoring is provided, including:

[0130] The multi-source data acquisition module 11 is used to acquire multi-source monitoring data in saline-alkali land and perform spatiotemporal alignment on the multi-source monitoring data to obtain spatiotemporally aligned multi-source monitoring data.

[0131] Water and salt analysis module 12 is used to input spatiotemporally aligned multi-source monitoring data into a preset water and salt synergistic regulation model to obtain the salt flux from the root layer to the drainage layer;

[0132] Salt tolerance analysis and drainage threshold analysis module 13 is used to obtain physiological stress indicators to characterize the current salt stress risk and dynamic regulation thresholds to characterize the leaching drainage window period based on salt flux and preset crop salt tolerance thresholds.

[0133] The drainage coordinated control instruction generation module 14 is used to match physiological stress indicators, dynamic control thresholds and preset coordinated control rules to generate a set of coordinated control instructions including irrigation trigger instructions, irrigation leaching quotas, drainage valve opening conditions and drainage valve closing conditions.

[0134] The coordinated control rules include the triggering and interlocking logic for irrigation and rinsing operations and drainage operations; the coordinated control instruction set is used to start irrigation according to the irrigation and rinsing quota when the irrigation trigger instruction is met, to start drainage when the drainage valve opening condition is met, and to stop drainage when the drainage valve closing condition is met.

[0135] Furthermore, the multi-source data acquisition module 11 is also used for:

[0136] S11. Perform outlier cleaning on the multi-source monitoring data to obtain the cleaned monitoring data;

[0137] S12. Perform spatial interpolation on the cleaned monitoring data to obtain regular gridded spatial distribution data;

[0138] S13. Perform time alignment on the spatially distributed data to obtain time series data with a uniform time step;

[0139] S14. Perform weighted optimization and fusion on the time series data to obtain spatiotemporally aligned multi-source monitoring data.

[0140] Furthermore, the water and salt analysis module 12 is also used for:

[0141] S21. Input the spatiotemporally aligned multi-source monitoring data into the water-salt co-regulation model; the water-salt co-regulation model includes a feature extraction network and a deep learning network;

[0142] S22. Feature extraction is performed on the spatiotemporally aligned multi-source monitoring data through a feature extraction network, and key influencing factors are extracted through principal component analysis to obtain the dimensionality-reduced feature data.

[0143] S23. Predict the dynamic changes of soil salinity using the dimensionality-reduced feature data through a deep learning network to obtain salinity prediction results;

[0144] S24. Based on the salinity prediction results, the salinity flux from the root zone to the drainage layer is calculated.

[0145] The expression for salt flux is:

[0146]

[0147] In the formula, Salt flux, The salt diffusion coefficient is... This refers to the salt concentration in the soil solution. For soil depth, This refers to the water flow rate.

[0148] Furthermore, the root salt tolerance analysis and drainage threshold analysis module 13 is also used for:

[0149] S31. Establish a salt stress response function based on crop growth stage and soil salinity content, and calculate the physiological stress index using the following formula:

[0150]

[0151] In the formula, As a physiological stress indicator, The stress response coefficient is... This represents the current soil salinity. This is a half-lethal salt concentration;

[0152] S32. Based on the root zone soil salinity and crop salt tolerance threshold, obtain the leaching demand coefficient;

[0153] S33. A dynamic threshold model is established based on the salinity change rate and leaching demand coefficient, and the dynamic control threshold is calculated through an adaptive algorithm.

[0154] Furthermore, the drainage coordinated control instruction generation module 14 is also used for:

[0155] S41. Based on physiological stress indicators and dynamic control thresholds, calculate irrigation initiation membership degree and drainage initiation membership degree to obtain the initial action decision vector;

[0156] S42. Input the initial action decision vector into the preset rule conflict resolution model. Based on the current soil moisture content, weather forecast data and historical action effectiveness evaluation results, arbitrate and correct the conflict decision to obtain the optimized action decision.

[0157] S43. Based on the certainty of irrigation initiation in the optimized action decision, combined with the leaching demand coefficient and real-time soil moisture data, the irrigation leaching quota is calculated through the water balance model.

[0158] S44. Based on the certainty of drainage initiation, the direction of salinity flux, and the saturation status of the drainage layer in the optimized action decision, the logical judgment expressions for the opening and closing conditions of the drainage valve are dynamically generated to obtain a set of coordinated control instructions.

[0159] Furthermore, the water and salt analysis module 12 is also used for:

[0160] S51. Input the dimensionality-reduced feature data into the deep learning network; the deep learning network is a spatiotemporal attention recurrent neural network.

[0161] S52. In the spatiotemporal attention recurrent neural network, the temporal attention mechanism assigns weights to the features of the dimensionality-reduced feature data at different time steps, obtaining the first weight assignment; the spatial attention mechanism assigns weights to the dimensionality-reduced feature data at different soil depths, obtaining the second weight assignment.

[0162] S53. Based on the first and second allocation weights, the dimensionality-reduced feature data is processed through the recurrent layer of the spatiotemporal attention recurrent neural network to capture the spatiotemporal dependence of soil water and salt transport and output a predicted sequence of soil solution salt concentration for multiple future time steps.

[0163] S54. Based on the predicted soil solution salinity concentration sequence and the synchronously predicted soil moisture content sequence, the predicted root zone soil salinity is calculated and used as the salinity prediction result.

[0164] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the IoT-based saline-alkali land drainage method as described above.

[0165] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0166] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0167] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A drainage method for saline-alkali land based on Internet of Things (IoT) monitoring, characterized in that, The method includes: S1. Acquire multi-source monitoring data in saline-alkali land, and perform spatiotemporal alignment on the multi-source monitoring data to obtain spatiotemporally aligned multi-source monitoring data; S2. Input the spatiotemporally aligned multi-source monitoring data into a preset water-salt synergistic regulation model to obtain the salt flux from the root layer to the drainage layer. S3. Based on the salt flux and the preset crop salt tolerance threshold, obtain the physiological stress index used to characterize the current salt stress risk and the dynamic regulation threshold used to characterize the leaching drainage window period. S4. Match the physiological stress index, the dynamic regulation threshold and the preset collaborative regulation rules to generate a set of collaborative regulation instructions that includes irrigation trigger instructions, irrigation leaching quotas, drainage valve opening conditions and drainage valve closing conditions. The coordinated control rules include triggering and interlocking logic for irrigation and rinsing operations and drainage operations; the coordinated control instruction set is used to start irrigation according to the irrigation and rinsing quota when the irrigation trigger instruction is met, to start drainage when the drainage valve opening condition is met, and to stop drainage when the drainage valve closing condition is met.

2. The method according to claim 1, characterized in that, The process of performing spatiotemporal alignment on the multi-source monitoring data to obtain spatiotemporally aligned multi-source monitoring data includes: S11. Perform outlier cleaning on the multi-source monitoring data to obtain cleaned monitoring data; S12. Perform spatial interpolation on the cleaned monitoring data to obtain regular gridded spatial distribution data; S13. Perform time alignment on the spatially distributed data to obtain time-series data with a uniform time step; S14. Perform weighted optimization and fusion on the time series data to obtain the spatiotemporally aligned multi-source monitoring data.

3. The method according to claim 2, characterized in that, The step of inputting the spatiotemporally aligned multi-source monitoring data into a preset water-salt coordinated regulation model to obtain the salt flux from the root layer to the drainage layer includes: S21. Input the spatiotemporally aligned multi-source monitoring data into the water-salt synergistic regulation model; the water-salt synergistic regulation model includes a feature extraction network and a deep learning network; S22. The feature extraction network is used to extract features from the spatiotemporally aligned multi-source monitoring data, and key influencing factors are extracted by principal component analysis to obtain dimensionality-reduced feature data. S23. Predict the dynamic changes of soil salinity using the dimensionality-reduced feature data through the deep learning network to obtain salinity prediction results; S24. Based on the salinity prediction results, the salinity flux from the root zone to the drainage layer is calculated. The expression for the salt flux is as follows: In the formula, Salt flux, The salt diffusion coefficient is... This refers to the salt concentration in the soil solution. For soil depth, This refers to the water flow rate.

4. The method according to claim 1, characterized in that, The process of obtaining physiological stress indicators to characterize the current salt stress risk and dynamic regulation thresholds to characterize the leaching drainage window period based on the salt flux and a preset crop salt tolerance threshold includes: S31. Establish a salt stress response function based on crop growth stage and soil salinity content, and calculate the physiological stress index using the following formula: In the formula, As a physiological stress indicator, The stress response coefficient is... This represents the current soil salinity. This is a half-lethal salt concentration; S32. Based on the root zone soil salinity and the crop's salt tolerance threshold, obtain the leaching requirement coefficient; S33. A dynamic threshold model is established based on the salinity change rate and the rinsing demand coefficient, and the dynamic control threshold is calculated by an adaptive algorithm.

5. The method according to claim 4, characterized in that, The process involves matching the physiological stress indicators, the dynamic control threshold, and preset collaborative control rules to generate a set of collaborative control instructions, including irrigation trigger commands, irrigation leaching quotas, drainage valve opening conditions, and drainage valve closing conditions. S41. Based on the physiological stress index and the dynamic regulation threshold, calculate the irrigation initiation membership degree and the drainage initiation membership degree to obtain the initial action decision vector; S42. Input the initial action decision vector into a preset rule conflict resolution model. Based on the current soil moisture content, weather forecast data and historical action performance evaluation results, arbitrate and correct the conflict decision to obtain an optimized action decision. S43. Based on the certainty of irrigation initiation in the optimized action decision, combined with the leaching demand coefficient and real-time soil moisture data, the irrigation leaching quota is calculated using a water balance model. S44. Based on the certainty of drainage initiation in the optimized action decision, the direction of salinity flux, and the saturation status of the drainage layer, dynamically generate logical judgment expressions for the opening conditions and closing conditions of the drainage valve to obtain the set of coordinated control instructions.

6. The method according to claim 3, characterized in that, The step of predicting soil salinity dynamics using the dimensionality-reduced feature data through the deep learning network to obtain salinity prediction results includes: S51. Input the dimensionality-reduced feature data into the deep learning network; the deep learning network is a spatiotemporal attention recurrent neural network. S52. In the spatiotemporal attention recurrent neural network, the temporal attention mechanism assigns weights to the features of the dimensionality-reduced feature data at different time steps to obtain a first weight; the spatial attention mechanism assigns weights to the dimensionality-reduced feature data at different soil depths to obtain a second weight. S53. Based on the first allocation weight and the second allocation weight, the dimensionality-reduced feature data is processed through the recurrent layer of the spatiotemporal attention recurrent neural network to capture the spatiotemporal dependence of soil water and salt transport and output a predicted sequence of soil solution salt concentration for multiple future time steps. S54. Based on the predicted soil solution salinity concentration sequence and the synchronously predicted soil moisture content sequence, the predicted root zone soil salinity is calculated and used as the salinity prediction result.

7. A drainage device for saline-alkali land based on Internet of Things (IoT) monitoring, characterized in that, The device includes: A multi-source data acquisition module is used to acquire multi-source monitoring data in saline-alkali land and perform spatiotemporal alignment on the multi-source monitoring data to obtain spatiotemporally aligned multi-source monitoring data. The water and salt analysis module is used to input the spatiotemporally aligned multi-source monitoring data into a preset water and salt collaborative regulation model to obtain the salt flux from the root layer to the drainage layer; The salt tolerance analysis and drainage threshold analysis module is used to obtain physiological stress indicators to characterize the current salt stress risk and dynamic regulation thresholds to characterize the leaching drainage window period based on the salt flux and the preset crop salt tolerance threshold. The drainage coordinated control instruction generation module is used to match the physiological stress index, the dynamic control threshold and the preset coordinated control rules to generate a set of coordinated control instructions including irrigation trigger instructions, irrigation leaching quotas, drainage valve opening conditions and drainage valve closing conditions. The coordinated control rules include triggering and interlocking logic for irrigation and rinsing operations and drainage operations; the coordinated control instruction set is used to start irrigation according to the irrigation and rinsing quota when the irrigation trigger instruction is met, to start drainage when the drainage valve opening condition is met, and to stop drainage when the drainage valve closing condition is met.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 6.

Images