Soil erosion intelligent prediction and restoration decision-making method and system

Abstract

The invention discloses a soil erosion intelligent prediction and restoration decision-making method and system, and relates to the technical field of ecological environment monitoring and restoration, and the method comprises the steps: building a multi-source data space-time coupling database, and achieving the dynamic updating and refining of land utilization classification data; a land utilization classification-oriented soil erosion prediction model is constructed based on a neural network model, and sensor data time correction and trend fusion analysis are combined, so that the erosion intensity prediction precision is improved; and meanwhile, a land utilization type-erosion risk-restoration strategy knowledge base is established, differential restoration schemes are generated, and risk early warning is carried out. Through multi-source data fusion and intelligent model application, full-process management of soil erosion from dynamic prediction to precise restoration is realized, the restoration scheme mismatching rate is reduced, and scientific support is provided for ecological restoration decision making.

Description

Technical Field

[0001] This invention relates to the field of ecological environment monitoring and restoration technology, and more specifically to a method and system for intelligent prediction and restoration decision-making of soil erosion. Background Technology

[0002] Soil erosion is a significant factor leading to land degradation and ecological deterioration, and its intensity is closely related to land use type, topography, and climate conditions. Traditional soil erosion prediction methods (such as the USLE and RUSLE models) are mostly based on static land use data and do not fully consider the dynamic changes in land use classifications, resulting in discrepancies between predicted results and actual erosion levels, and thus low accuracy. Furthermore, existing remediation schemes often adopt a one-size-fits-all approach, failing to develop differentiated strategies for the erosion mechanisms and ecological needs of different land use scenarios (such as arable land, forest land, and grassland), leading to poor adaptability of remediation schemes to land types, resource waste, and unsatisfactory remediation effects.

[0003] Currently, common methods for monitoring soil erosion include manual sampling, remote sensing image interpretation, and single-sensor monitoring. However, these methods have the following shortcomings: manual sampling is time-consuming and labor-intensive, making it difficult to achieve large-scale, high-frequency dynamic monitoring; remote sensing image interpretation is greatly affected by weather and has limited data resolution, making it impossible to capture small-scale soil erosion details; and single-sensor monitoring (such as rain gauges and soil moisture sensors) can only obtain local environmental parameters and cannot comprehensively reflect the relationship between land use type and erosion factors, which can easily lead to misjudgments in predictions.

[0004] Therefore, how to propose an intelligent prediction and remediation decision-making method and system for soil erosion, combining multi-source dynamic data, intelligent prediction models and precise remediation strategies, to achieve full-process optimization of soil erosion from dynamic prediction to differentiated remediation, and improve prediction accuracy and remediation adaptability, is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method and system for intelligent prediction and remediation decision-making of soil erosion, solving the problems of low prediction accuracy and poor adaptability of remediation schemes in traditional methods. Through multi-source data fusion, sensor time correction, neural network prediction, and trend fusion analysis, it achieves dynamic and accurate prediction of soil erosion; by constructing a knowledge base based on land use types, it generates differentiated remediation schemes, providing scientific support for ecological restoration decision-making. To achieve the above objectives, the present invention adopts the following technical solution: A method for intelligent prediction and remediation decision-making of soil erosion includes: Real-time data collection of land use classification, environmental, and soil erosion correlation data in the monitored area is achieved through multiple types of sensors. The land use classification data, environmental data, and soil erosion correlation data corresponding to each sub-monitoring unit are grouped together to generate a linear map of the changes of each type of data over time. Obtain the data acquisition delay time of the sensor corresponding to each type of data, and perform time correction on the horizontal axis of the linear spectrum of each type of data based on the delay time; The time-corrected linear map is input into a pre-built neural network soil erosion prediction model to obtain the erosion intensity level and corresponding time interval for different land use types. Trend fusion analysis was performed on the erosion intensity level, land use type data, and environmental data of each sub-monitoring unit to calculate the comprehensive soil erosion risk value; Based on the knowledge base of land use type-erosion risk-remediation strategy, a differentiated remediation plan is generated by combining the comprehensive value of erosion risk.

[0006] Optionally, it also includes: risk monitoring of the power supply system and sensor deployment environment in the soil erosion monitoring area. If the risk monitoring is passed, the soil erosion data acquisition process is allowed to start; otherwise, the data acquisition process is prohibited from starting and a maintenance prompt is triggered. Real-time status data of the power supply lines in the monitoring area are collected at multiple nodes by power supply line sensors, and environmental characteristic data of the power supply line nodes and sensor deployment locations are collected by environmental sensors. Based on real-time status data of the power supply line and environmental characteristic data, calculate the operational hazard value of the power supply line and the reliability value of sensor data acquisition; If the operational hazard value of the power supply line is less than the preset safety threshold and the reliability value of the sensor data acquisition is greater than the preset reliability threshold, then the risk monitoring passes; otherwise, it fails.

[0007] Optionally, the operational hazards of the power supply line include: ; in, Indicates the hazardous value of the power supply line operation; This indicates the total number of monitoring nodes for the power supply lines; Indicates the first The total number of types of suspicious and abnormal data collected at each node; Indicates the first At the node, the first The impact weight of suspicious and abnormal data; Indicates the first At the node, the first Measured values ​​of suspicious and abnormal data; Indicates the first At the node, the first Standard values ​​for this type of data; Indicates the first At the node, the first Permissible deviation values ​​for this type of data; This indicates the calibration error value of the corresponding sensor; This indicates the impact of environmental characteristic data on sensor measurement errors.

[0008] Optionally, the pre-built neural network soil erosion prediction model includes: Acquire historical data on multiple sets of known land use types, environmental parameters, and corresponding soil erosion intensities. The historical data covers different climate zones, terrain types, and land use scenarios. Historical data is converted into a linear map that changes over time, and land use type labels and erosion intensity level labels are labeled to construct a training dataset; The training dataset is input into the basic neural network model for training. The model parameters are optimized through cross-validation to obtain a neural network soil erosion prediction model.

[0009] Optionally, the calculation of the comprehensive soil erosion risk value includes: ; in, This represents the comprehensive value of soil erosion risk; S represents the fusion influence factor of erosion intensity data. Indicates the first The number of abnormal data types of erosion intensity at time tc represents the end time of data sampling; P represents the number of time intervals with multiple abnormal data types; Gk represents the number of abnormal data types in the k-th abnormal time interval. This represents the duration of the k-th abnormal time interval; This represents the influence weight of the g-th type of outlier data in the k-th interval; This represents the measured average value of the g-th data in the k-th interval; This represents the standard value of the corresponding data; Indicates the trend influencing factor. =1; This represents the weight of the i-th abnormal segment; , These represent the maximum and minimum values ​​of the data in the i-th segment, respectively. It indicates the time it takes for the data to change from its minimum value to its maximum value.

[0010] Optionally, it also includes: periodically collecting soil erosion data after the implementation of the remediation plan, inputting it into the neural network soil erosion prediction model for model iteration and updating, and updating the land use type-erosion risk-remediation strategy knowledge base at the same time.

[0011] Optionally, the environmental data includes rainfall, slope, vegetation coverage, and soil texture, and the soil erosion-related data includes soil moisture content and measured erosion values.

[0012] Optionally, a soil erosion intelligent prediction and remediation decision-making system includes: Data acquisition module: used to collect land use classification data, environmental data and soil erosion correlation data of the monitoring area in real time through multiple types of sensors; Map generation module: used to group the land use classification data, environmental data and soil erosion correlation data corresponding to each sub-monitoring unit into a group, and generate a linear map of the changes of each type of data over time; Time correction module: used to obtain the data acquisition delay time of the sensor corresponding to each type of data, and to perform time correction on the horizontal axis of the linear spectrum of each type of data based on the delay time; Erosion prediction module: This module is used to input the time-corrected linear spectrum into a pre-built neural network soil erosion prediction model to obtain the erosion intensity level and corresponding time interval for different land use types. Risk Analysis Module: Used to perform trend fusion analysis on the erosion intensity level, land use type data and environmental data of each sub-monitoring unit, and calculate the comprehensive value of soil erosion risk; The restoration decision module is used to generate differentiated restoration plans based on the land use type-erosion risk-restoration strategy knowledge base and the comprehensive value of erosion risk.

[0013] Optionally, it also includes a risk monitoring module, which is equipped with an alarm unit. If the risk monitoring fails, the alarm unit triggers an audible and visual alarm and pushes the maintenance task to the management personnel terminal, while recording the location of the fault node and the fault type.

[0014] Optionally, the data acquisition module includes a data preprocessing unit, which performs outlier removal, missing value completion, and standardization on the acquired raw data to ensure data integrity and consistency.

[0015] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method and system for intelligent prediction and remediation of soil erosion, which has the following beneficial effects: This invention proposes an intelligent prediction and remediation decision-making method for soil erosion, comprising: real-time acquisition of land use classification data, environmental data, and soil erosion-related data of a monitoring area using multiple types of sensors; grouping the acquired land use classification data, environmental data, and soil erosion-related data corresponding to each sub-monitoring unit into a group, generating a linear spectrum of each data type changing over time; obtaining the acquisition data delay time of the sensors corresponding to each data type, and performing time correction on the horizontal axis of the linear spectrum of each data type based on the delay time; inputting the time-corrected linear spectrum into a pre-constructed neural network soil erosion prediction model to obtain the erosion intensity level and corresponding time interval for different land use types; performing trend fusion analysis on the erosion intensity level, land use type data, and environmental data of each sub-monitoring unit to calculate a comprehensive soil erosion risk value; and generating differentiated remediation schemes based on a land use type-erosion risk-remediation strategy knowledge base, combined with the comprehensive erosion risk value. This invention solves the prediction bias problem caused by static data in traditional methods by using multi-source data fusion, sensor time correction, and a CNN-LSTM neural network model, thereby improving the accuracy of erosion intensity prediction. Based on a knowledge base of land use type, erosion risk, and restoration strategies, differentiated solutions are generated for different scenarios to avoid a one-size-fits-all approach and reduce the rate of solution mismatch. A comprehensive risk value is calculated through trend fusion analysis, and solutions are ranked based on cost, benefit, and feasibility, providing managers with a quantitative basis for decision-making. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of a method for intelligent prediction and remediation decision-making of soil erosion provided by the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] This invention discloses an intelligent prediction and remediation decision-making method for soil erosion, such as... Figure 1 As shown, it includes: Real-time data collection of land use classification, environmental, and soil erosion correlation data in the monitored area is achieved through multiple types of sensors. The land use classification data, environmental data, and soil erosion correlation data corresponding to each sub-monitoring unit are grouped together to generate a linear map of the changes of each type of data over time. Obtain the data acquisition delay time of the sensor corresponding to each type of data, and perform time correction on the horizontal axis of the linear spectrum of each type of data based on the delay time; The time-corrected linear map is input into a pre-built neural network soil erosion prediction model to obtain the erosion intensity level and corresponding time interval for different land use types. Trend fusion analysis was performed on the erosion intensity level, land use type data, and environmental data of each sub-monitoring unit to calculate the comprehensive soil erosion risk value; Based on the knowledge base of land use type-erosion risk-remediation strategy, a differentiated remediation plan is generated by combining the comprehensive value of erosion risk.

[0020] In a specific implementation, a method for intelligent prediction and remediation decision-making of soil erosion includes the following steps: Step 1: Conduct risk monitoring on the power supply system and sensor deployment environment of the soil erosion monitoring area. If the risk monitoring passes, the soil erosion data acquisition process can be started; otherwise, the data acquisition process is prohibited and a maintenance prompt is triggered. Step 2: If the risk monitoring passes, land use classification data, environmental data and soil erosion correlation data of the monitoring area are collected in real time through multiple types of sensors. The environmental data includes rainfall, slope, vegetation coverage and soil texture. The soil erosion correlation data includes soil moisture content and measured values ​​of erosion. Step 3: Divide the land use classification data, environmental data and soil erosion correlation data corresponding to each sub-monitoring unit into a group, and generate a linear map of the change of each type of data over time; Step 4: Obtain the data acquisition delay time of the sensor corresponding to each type of data, and perform time correction on the horizontal axis of the linear spectrum of each type of data based on the delay time; Step 5: Input the time-corrected linear spectrum into the pre-built neural network soil erosion prediction model to obtain the erosion intensity level and corresponding time interval for different land use types; Step Six: Perform trend fusion analysis on the erosion intensity level, land use type data, and environmental data of each sub-monitoring unit to calculate the comprehensive soil erosion risk value; Step 7: Based on the land use type-erosion risk-restoration strategy knowledge base, generate differentiated restoration plans by combining the comprehensive value of erosion risk, and prioritize the restoration plans according to their implementation cost, ecological benefits and feasibility.

[0021] Furthermore, step one, which involves risk monitoring of the power supply system and sensor deployment environment, includes: (1) Real-time status data of power supply lines in the monitoring area are collected at multiple nodes by power supply line sensors, and environmental characteristic data of power supply line nodes and sensor deployment locations are collected by environmental sensors. The environmental characteristic data includes temperature, humidity and wind speed. (2) Calculate the operational hazard value of the power supply line and the reliability value of sensor data acquisition based on the real-time status data of the power supply line and the environmental characteristic data; (3) If the operating hazard value of the power supply line is less than the preset safety threshold and the reliability value of the sensor data acquisition is greater than the preset reliability threshold, then the risk monitoring is passed; otherwise, it is not passed.

[0022] Furthermore, the formula for calculating the operational hazard value of power supply lines is as follows: ; in, Indicates the hazardous value of the power supply line operation; This indicates the total number of monitoring nodes for the power supply lines; Indicates the first The total number of types of suspicious and abnormal data collected at each node; Indicates the first At the node, the first The impact weight of suspicious and abnormal data; Indicates the first At the node, the first Measured values ​​of suspicious and abnormal data; Indicates the first At the node, the first Standard values ​​for this type of data; Indicates the first At the node, the first Permissible deviation values ​​for this type of data; This indicates the calibration error value of the corresponding sensor; This indicates the impact of environmental characteristic data on sensor measurement errors.

[0023] Furthermore, the method for pre-constructing a neural network soil erosion prediction model is as follows: (1) Obtain historical data of multiple known land use types, environmental parameters and corresponding soil erosion intensity, the historical data covering different climate zones, terrain types and land use scenarios; (2) Convert historical data into a linear map that changes over time, label it with land use type and erosion intensity level labels, and construct a training dataset; (3) Input the training dataset into the basic neural network model (such as the CNN-LSTM fusion model) for training, optimize the model parameters through cross-validation, and obtain the neural network soil erosion prediction model.

[0024] Furthermore, the formula for calculating the comprehensive value of soil erosion risk in step six is ​​as follows: ; in, This represents the comprehensive value of soil erosion risk; S represents the fusion influence factor of erosion intensity data. Indicates the first The number of abnormal data types of erosion intensity at time tc represents the end time of data sampling; P represents the number of time intervals with multiple abnormal data types; Gk represents the number of abnormal data types in the k-th abnormal time interval. This represents the duration of the k-th abnormal time interval; This represents the influence weight of the g-th type of outlier data in the k-th interval; This represents the measured average value of the g-th data in the k-th interval; This represents the standard value of the corresponding data; Indicates the trend influencing factor. =1; This represents the weight of the i-th abnormal segment; , These represent the maximum and minimum values ​​of the data in the i-th segment, respectively. It indicates the time it takes for the data to change from its minimum value to its maximum value.

[0025] Furthermore, it also includes: regularly collecting soil erosion data after the implementation of the remediation plan, inputting it into the neural network soil erosion prediction model for model iteration and update, and updating the knowledge base of land use type-erosion risk-remediation strategy.

[0026] Furthermore, after generating differentiated remediation plans, it also includes: visually displaying the distribution of land use types, erosion risk levels, and remediation plan deployment locations in the monitored area based on the GIS system, and supporting user interaction to adjust remediation plan parameters.

[0027] In a specific implementation, a soil erosion intelligent prediction and remediation decision-making system includes: Risk monitoring module: Used to monitor the power supply system and sensor deployment environment of the monitoring area and determine whether to allow the data acquisition process to start; Data acquisition module: used to collect land use classification data, environmental data and soil erosion correlation data in real time through multiple types of sensors; Map generation module: used to group the collected data and generate linear maps of each data type over time; Time correction module: Used to correct the time axis of the linear spectrum based on sensor delay time; Erosion prediction module: Used to input the corrected map into the neural network model and output the erosion intensity level and the corresponding time interval; Risk Analysis Module: Used to calculate the comprehensive value of soil erosion risk through trend fusion analysis; Repair Decision Module: Used to generate differentiated repair solutions based on the knowledge base and prioritize them; Visualization and Interaction Module: Used for visualizing GIS data and solutions, and supports user parameter adjustments.

[0028] Furthermore, the risk monitoring module also includes an alarm unit: if the risk monitoring fails, the alarm unit triggers an audible and visual alarm and pushes the maintenance task to the management personnel terminal, while recording the location of the fault node and the fault type.

[0029] Furthermore, the data acquisition module includes a data preprocessing unit: performing outlier removal, missing value completion, and standardization on the acquired raw data to ensure data integrity and consistency.

[0030] In a specific embodiment, a method for intelligent prediction and remediation decision-making of soil erosion includes the following steps: Step 1: Risk Monitoring and Initiation Judgment Risk monitoring should be conducted on the power supply system and sensor deployment environment in the soil erosion monitoring area to ensure the stability and reliability of the data acquisition process.

[0031] Power supply system monitoring: Voltage and current sensors are deployed at key nodes of the power supply lines in the monitoring area (such as power output terminals, sensor access points, and line branch points) to collect real-time power supply status data; at the same time, temperature, humidity, and wind sensors are deployed around these nodes to collect environmental characteristic data and avoid environmental factors from affecting the stability of power supply.

[0032] Sensor environment monitoring: For soil moisture sensors, vegetation cover sensors, erosion monitoring sensors, etc., collect data such as soil compaction, terrain slope, and shading at their deployment locations, and evaluate the reliability of sensor data collection (e.g., vegetation shading may cause deviations in remote sensing sensor data).

[0033] Hazard value calculation and judgment: Calculate the power supply line operation hazard value BW and the sensor data acquisition reliability value according to the formula. If BW is less than the preset safety threshold and the reliability value is greater than the preset threshold, the data acquisition process is allowed to start; otherwise, the process is prohibited and a maintenance prompt (such as poor line contact, sensor obstruction, etc.) is pushed.

[0034] Step 2: Real-time acquisition of multi-source data

[0035] After risk monitoring is approved, multiple types of sensors are activated to collect data, covering three major categories of data: land use, environment, and erosion correlation. Land use classification data: Land use types (arable land, forest land, grassland, construction land, etc.) and dynamic change data (such as the time point when arable land is converted into construction land) in the monitored area are collected through high-resolution remote sensing sensors (such as multispectral cameras carried by UAVs) and ground survey sensors.

[0036] Environmental data: Rainfall amount and duration were collected using rain gauges; terrain slope was collected using slope sensors; vegetation cover was collected using vegetation cover sensors; and the proportions of sand, silt, and clay particles in the soil were collected using soil texture analyzers.

[0037] Soil erosion-related data: Soil moisture is collected through soil moisture sensors, measured values ​​of soil erosion are collected through erosion monitoring instruments (such as sensors for runoff plot method), and organic matter content (reflecting soil erosion resistance) is collected through soil nutrient sensors.

[0038] During the data collection process, a unique identifier is assigned to each sub-monitoring unit (such as a 1km×1km grid), and all data within that unit are grouped together to facilitate accurate subsequent analysis.

[0039] Step 3: Linear Spectrum Generation

[0040] The three types of data from each sub-monitoring unit are processed separately to generate a linear spectrum that changes over time: Using time as the horizontal axis (e.g., hours, days, months, adjusted according to monitoring frequency) and data values ​​as the vertical axis, we generate land use type change maps (e.g., time nodes marking the conversion of a sub-unit from cultivated land to forest land), rainfall-time maps, erosion-time maps, etc.

[0041] The data acquisition timestamps and corresponding sensor identifiers need to be marked in the graph to facilitate subsequent time correction and traceability.

[0042] Step 4: Time Correction

[0043] Because different sensors have different acquisition principles and response speeds, there are data delays (such as data transmission delays in remote sensing sensors and response delays in soil moisture sensors), so time correction is required for the linear spectrum. Delay time acquisition: The data acquisition delay time of each sensor is determined through laboratory calibration and field testing (e.g., 1 minute delay for rain gauge and 5 minutes delay for soil moisture sensor).

[0044] Map correction: Shift the linear map of a certain type of data in the negative direction along the time axis (horizontal axis) by the corresponding delay time (e.g., shift the soil moisture data map to the left by 5 minutes) to ensure that data from different sensors are aligned in the time dimension and avoid analysis errors caused by delay (e.g., time matching deviation between rainfall and erosion).

[0045] Step 5: Neural network predicts erosion intensity

[0046] The time-corrected linear spectrum is input into a pre-built neural network soil erosion prediction model to obtain the erosion intensity level and its corresponding time interval: (1) Model building: Dataset preparation: Collect historical data on different climate zones (such as humid and semi-arid regions), terrain types (mountains and plains), and land use scenarios, label land use types (such as cultivated land and irrigated land) and erosion intensity levels (slight, moderate, severe, and extremely severe, based on the "Classification and Grading Standards for Soil Erosion"), and construct a training dataset.

[0047] (2) Model training: The CNN-LSTM fusion neural network is used as the basic model. CNN is used to extract spatial features in the linear map (such as the impact of the spatial distribution of different land use types on erosion), and LSTM is used to extract temporal features (such as the cumulative effect of continuous increase in rainfall on erosion). The learning rate, number of iterations and other parameters are optimized through cross-validation so that the model can predict the erosion intensity of different scenarios with an accuracy of over 90%.

[0048] (3) Predictive output: The model outputs the erosion intensity level (e.g., forest land - light erosion) and the corresponding time interval (e.g., the rainy season from June to August 2024 is moderate erosion) for each sub-monitoring unit, and also marks the prediction confidence level (e.g., 95%).

[0049] Step 6: Trend Fusion Analysis and Risk Calculation

[0050] Trend fusion analysis was performed on the erosion intensity level, land use type data, and environmental data of each sub-monitoring unit to calculate the comprehensive soil erosion risk value FY: Trend fusion: Combining the formula, comprehensively consider the types of erosion intensity anomalies at different time points (e.g., excessive rainfall + insufficient vegetation cover occurring simultaneously), the duration of the anomaly time interval (e.g., moderate erosion for 15 consecutive days), and the data change trend (e.g., erosion from 5t / (hm²)). a) Increased to 15t / (hm²) (a) rate).

[0051] Risk classification: Soil erosion risk is divided into low risk (FY<0.3) and medium risk (FY<0.3) based on the FY value. FY<0.7), high risk (FY) 0.7), providing a basis for the formulation of subsequent repair plans.

[0052] Step 7: Generation and sorting of differentiated repair solutions

[0053] Based on a knowledge base of land use type, erosion risk, and remediation strategy, remediation plans are generated by combining comprehensive erosion risk values ​​and then prioritized. Knowledge base construction: This involves analyzing the erosion mechanisms and the suitability of restoration technologies for different land use types. For example, arable land is suitable for contour farming with straw mulch; forest land is suitable for reforestation and stand transformation; grassland is suitable for fencing and artificial grass planting; and construction land is suitable for slope greening and rain gardens. The knowledge base should also include the implementation costs, ecological benefits (such as the rate of vegetation cover increase), and applicable timeframes (such as short-term restoration <1 year, long-term restoration >5 years) of restoration technologies.

[0054] Solution generation: For a specific monitoring unit (e.g., arable land - high risk), a remediation strategy (e.g., straw mulching + contour tillage + increased application of organic fertilizer) is matched from the knowledge base, and parameters (e.g., increase drainage facilities if there is a lot of rainfall) and economic level (e.g., low-cost technology is preferred in poor areas) are adjusted (e.g., straw mulching thickness is 10cm).

[0055] Prioritization: The analytic hierarchy process (AHP) is used to score the solutions from three dimensions: implementation cost (weight 0.3), ecological benefits (weight 0.4), and feasibility (weight 0.3). After ranking, the optimal solution is recommended (e.g., straw mulching + contour tillage has low cost and high benefits, so it is recommended first).

[0056] Step 8: Model Iteration and Knowledge Base Update

[0057] Regularly (e.g., quarterly) collect soil erosion data after the implementation of the remediation plan (such as changes in erosion amount and improvement in vegetation coverage), input the data into the neural network soil erosion prediction model for iterative optimization, and update the knowledge base (e.g., add a remediation strategy for orchards with moderate erosion: grass cover + drip irrigation system) to improve prediction accuracy and plan adaptability.

[0058] In a specific embodiment, a soil erosion intelligent prediction and remediation decision-making system includes the following modules: 1. Risk monitoring module Function: Enables risk monitoring of the power supply system and sensor environment, including a power supply status acquisition unit, an environmental data acquisition unit, a hazard value calculation unit, and an alarm unit.

[0059] Alarm unit: If the risk monitoring fails, an audible and visual alarm is triggered and the maintenance task is pushed to the management personnel terminal (such as a mobile APP) via the Internet of Things. At the same time, the location of the fault node (such as latitude and longitude), the fault type (such as abnormal line voltage) and the suggested repair solution (such as replacing the line connector) are recorded.

[0060] 2. Data Acquisition Module

[0061] Function: Enables real-time acquisition and preprocessing of multi-source data, including sensor control unit, data transmission unit and data preprocessing unit.

[0062] Data preprocessing unit: Removes outliers from the collected raw data (e.g., zero values ​​from rainfall sensors for 10 consecutive days due to malfunction), fills in missing values ​​(e.g., uses linear interpolation to fill in missing soil moisture data), and performs standardization (e.g., converts slope data into a 0-1 range) to ensure data integrity and consistency.

[0063] 3. Atlas generation module

[0064] Function: Converts preprocessed grouped data into a linear graph, including data grouping units, graph plotting units, and label annotation units.

[0065] Labeling unit: Automatically labels land use type, data collection timestamp, and sensor identifier in the map to facilitate subsequent analysis and traceability.

[0066] 4. Time Correction Module

[0067] Function: Based on sensor delay time correction linear spectrum, including delay time storage unit (stores calibration delay time of various sensors), spectrum correction unit (performs time axis movement operation) and correction verification unit (verifies the time alignment accuracy of the corrected data).

[0068] 5. Erosion Prediction Module

[0069] Function: Runs a neural network soil erosion prediction model and outputs erosion intensity levels, including a model storage unit (stores the trained CNN-LSTM model), a data input unit (receives the corrected map), a prediction calculation unit, and a result output unit (outputs the level and time interval).

[0070] 6. Risk Analysis Module

[0071] Function: Calculates the comprehensive value of soil erosion risk, including a data fusion unit (integrating erosion level, land use, and environmental data), a formula calculation unit (performing FY value calculation), and a risk classification unit (outputting low / medium / high risk).

[0072] 7. Fixed the decision-making module

[0073] Function: Generate and sort differentiated remediation plans, including a knowledge base unit (storing the relationship between land use, risk, and strategy), a plan matching unit (matching strategies based on risk values), a parameter adjustment unit (optimizing plans based on local conditions), and a priority ranking unit (ranking based on the analytic hierarchy process).

[0074] 8. Visual Interaction Module

[0075] Function: Enables GIS visualization of data and solutions, including GIS map integration unit (loading monitoring area map), data layer unit (overlaying land use, erosion risk, and remediation solution layers), and user interaction unit (supporting users to adjust solution parameters, such as changing straw cover thickness from 10cm to 15cm, and displaying cost and benefit changes in real time).

[0076] In specific embodiments, the technical solution of the present invention will be described in detail: Example 1: Soil Erosion Monitoring and Remediation in a Hilly Area A hilly area of ​​500 km² has land use primarily consisting of arable land (40%), forest land (35%), grassland (20%), and construction land (5%). Soil erosion is a significant problem during the rainy season (June-August). The technical solution of this invention is used for monitoring and remediation. Risk monitoring: Voltage / current sensors were deployed at 5 key nodes of the power supply line. The measured voltage value at one node was Ssz=225V, the standard value was Bz=220V, and the allowable deviation was... =5V, calibration error Wc=0.5V, environmental impact value Hc=0.3V, influence weight =0.8; Calculate BW=0.8×(|225-220|-5-0.5-0.3)=-0.64 (less than the safety threshold 0), sensor confidence value 0.92 (greater than the threshold 0.8), risk monitoring passed.

[0077] Data Acquisition: Initiate sensor acquisition to obtain rainfall (cumulative 150mm in June), slope (15°), vegetation coverage (60%), and soil erosion (8t / (hm²)) for a specific monitoring unit (farmland). a)).

[0078] Map generation and correction: Generate the rainfall-time map and erosion-time map for this unit. Based on the rain gauge delay of 1 minute and the erosion sensor delay of 3 minutes, shift the maps to the left by 1 minute and 3 minutes respectively.

[0079] Erosion prediction: The corrected map was input into the neural network model, and the output showed that the erosion of this unit was moderate from June to August, with a confidence level of 93%.

[0080] Risk calculation: Combining farmland type and slope of 15° (high risk factor), FY=0.58 (medium risk) is calculated.

[0081] Solution generation: Match straw mulch (10cm thickness) + contour tillage from the knowledge base, cost 200 yuan / mu, ecological benefits (10% increase in vegetation coverage), priority is first, recommended for implementation.

[0082] Three months after implementation, the erosion rate in the monitored unit decreased to 5 t / (hm²) a) The repair effect is significant, verifying the effectiveness of the method and system of this application.

[0083] Example 2: System Deployment and Operation

[0084] Deploying the system described in this application at the ecological monitoring center includes: Hardware: Deploy 20 power line sensors, 50 environmental sensors, and 100 soil erosion-related sensors; equipped with 10 UAV remote sensing systems; and the server uses a high-performance GPU (supporting neural network model training).

[0085] Software: Develop a GIS visualization platform (supporting map overlay and scheme adjustment), integrate neural network prediction algorithms, and build a land use-risk-strategy knowledge base (containing 10 land use types, 5 risk levels, and 20 remediation technologies).

[0086] After one year of operation, the system has monitored a cumulative area of ​​10,000 km², generated more than 300 restoration plans, increased the plan adaptation rate from 65% to 92% using traditional methods, and improved the erosion prediction accuracy from 78% to 91%, significantly enhancing the scientific nature and efficiency of regional ecological restoration decision-making.

[0087] In summary, this invention addresses the core challenges of traditional soil erosion prediction and remediation through multi-source data fusion, intelligent model prediction, and differentiated remediation scheme generation. It combines dynamic land use changes with neural network models to improve prediction accuracy; constructs a knowledge base based on land use types to achieve precise adaptation of remediation schemes; and ensures monitoring reliability and scientific decision-making through risk monitoring and trend analysis. This method and system can be widely applied in agriculture, forestry, and ecological protection, providing technical support for soil erosion prevention and control and contributing to ecological civilization construction.

[0088] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0089] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for intelligent prediction and remediation decision-making of soil erosion, characterized in that, include: Real-time data collection of land use classification, environmental, and soil erosion correlation data in the monitored area is achieved through multiple types of sensors. The land use classification data, environmental data, and soil erosion correlation data corresponding to each sub-monitoring unit are grouped together to generate a linear map of the changes of each type of data over time. Obtain the data acquisition delay time of the sensor corresponding to each type of data, and perform time correction on the horizontal axis of the linear spectrum of each type of data based on the delay time; The time-corrected linear map is input into a pre-built neural network soil erosion prediction model to obtain the erosion intensity level and corresponding time interval for different land use types. Trend fusion analysis was performed on the erosion intensity level, land use type data, and environmental data of each sub-monitoring unit to calculate the comprehensive soil erosion risk value; Based on the knowledge base of land use type-erosion risk-remediation strategy, a differentiated remediation plan is generated by combining the comprehensive value of erosion risk.

2. The intelligent prediction and remediation decision-making method for soil erosion according to claim 1, characterized in that, Also includes: Risk monitoring is conducted on the power supply system and sensor deployment environment in the soil erosion monitoring area. If the risk monitoring is passed, the soil erosion data acquisition process is allowed to start; otherwise, the data acquisition process is prohibited and a maintenance prompt is triggered. Real-time status data of the power supply lines in the monitoring area are collected at multiple nodes by power supply line sensors, and environmental characteristic data of the power supply line nodes and sensor deployment locations are collected by environmental sensors. Based on real-time status data of the power supply line and environmental characteristic data, calculate the operational hazard value of the power supply line and the reliability value of sensor data acquisition; If the operational hazard value of the power supply line is less than the preset safety threshold and the reliability value of the sensor data acquisition is greater than the preset reliability threshold, then the risk monitoring passes; otherwise, it fails.

3. The intelligent prediction and remediation decision-making method for soil erosion according to claim 2, characterized in that, The hazardous values ​​for the operation of the power supply line include: ； in, Indicates the hazardous value of the power supply line operation; This indicates the total number of monitoring nodes for the power supply lines; Indicates the first The total number of types of suspicious and abnormal data collected at each node; Indicates the first At the node, the first The impact weight of suspicious and abnormal data; Indicates the first At the node, the first Measured values ​​of suspicious and abnormal data; Indicates the first At the node, the first Standard values ​​for this type of data; Indicates the first At the node, the first Permissible deviation values ​​for this type of data; This indicates the calibration error value of the corresponding sensor; This indicates the impact of environmental characteristic data on sensor measurement errors.

4. The intelligent prediction and remediation decision-making method for soil erosion according to claim 1, characterized in that, The pre-built neural network soil erosion prediction model includes: Acquire historical data on multiple sets of known land use types, environmental parameters, and corresponding soil erosion intensities. The historical data covers different climate zones, terrain types, and land use scenarios. Historical data is converted into a linear map that changes over time, and land use type labels and erosion intensity level labels are labeled to construct a training dataset; The training dataset is input into the basic neural network model for training. The model parameters are optimized through cross-validation to obtain a neural network soil erosion prediction model.

5. The intelligent prediction and remediation decision-making method for soil erosion according to claim 1, characterized in that, The calculation of the comprehensive value of soil erosion risk includes: ； in, This represents the comprehensive value of soil erosion risk; S represents the fusion influence factor of erosion intensity data. Indicates the first The number of abnormal data types of erosion intensity at time tc represents the end time of data sampling; P represents the number of time intervals with multiple abnormal data types; Gk represents the number of abnormal data types in the k-th abnormal time interval. This represents the duration of the k-th abnormal time interval; This represents the influence weight of the g-th type of outlier data in the k-th interval; This represents the measured average value of the g-th data in the k-th interval; This represents the standard value of the corresponding data; Indicates the trend influencing factor. =1; This represents the weight of the i-th abnormal segment; , These represent the maximum and minimum values ​​of the data in the i-th segment, respectively. It indicates the time it takes for the data to change from its minimum value to its maximum value.

6. The intelligent prediction and remediation decision-making method for soil erosion according to claim 4, characterized in that, Also includes: Soil erosion data are collected regularly after the implementation of remediation plans, and input into a neural network soil erosion prediction model for iterative updates. At the same time, the knowledge base of land use type-erosion risk-remediation strategy is updated.

7. The intelligent prediction and remediation decision-making method for soil erosion according to claim 1, characterized in that, The environmental data includes rainfall, slope, vegetation coverage, and soil texture, while the soil erosion-related data includes soil moisture content and measured erosion amounts.

8. A soil erosion intelligent prediction and remediation decision-making system, characterized in that, include: Data acquisition module: used to collect land use classification data, environmental data and soil erosion correlation data of the monitoring area in real time through multiple types of sensors; Map generation module: used to group the land use classification data, environmental data and soil erosion correlation data corresponding to each sub-monitoring unit into a group, and generate a linear map of the changes of each type of data over time; Time correction module: used to obtain the data acquisition delay time of the sensor corresponding to each type of data, and to perform time correction on the horizontal axis of the linear spectrum of each type of data based on the delay time; Erosion prediction module: This module is used to input the time-corrected linear spectrum into a pre-built neural network soil erosion prediction model to obtain the erosion intensity level and corresponding time interval for different land use types. Risk Analysis Module: Used to perform trend fusion analysis on the erosion intensity level, land use type data and environmental data of each sub-monitoring unit, and calculate the comprehensive value of soil erosion risk; The restoration decision module is used to generate differentiated restoration plans based on the land use type-erosion risk-restoration strategy knowledge base and the comprehensive value of erosion risk.

9. The intelligent prediction and remediation decision-making system for soil erosion according to claim 8, characterized in that, It also includes a risk monitoring module, which is equipped with an alarm unit. If the risk monitoring fails, the alarm unit triggers an audible and visual alarm and pushes the maintenance task to the management personnel terminal, while recording the location of the fault node and the fault type.

10. The intelligent prediction and remediation decision system for soil erosion according to claim 8, characterized in that, The data acquisition module includes a data preprocessing unit, which performs outlier removal, missing value completion, and standardization on the acquired raw data to ensure data integrity and consistency.

Images