Intelligent rainwater regulation and storage method and system

Abstract

The invention provides an intelligent rainwater regulation and storage method and system, and relates to the technical field of hydraulic engineering, and the method comprises the steps: collecting irrigation area environmental parameters in real time, the environmental parameters comprising rainfall intensity, soil humidity, weather forecast data and crop growth stage identifiers; dividing an irrigation area into a plurality of sub-areas according to the environmental parameters, integrating the sub-areas, and judging a water storage or drainage demand state of the integrated sub-areas; according to the water storage or drainage demand state of the integration sub-region, the dynamic drainage or water storage priority is calculated by combining the terrain gradient, the crop root depth and the real-time humidity value of the integration sub-region, and a water distribution scheme is generated through a multi-objective optimization algorithm; the water conservancy facility is driven to execute regulation and storage operation, the regulation and storage effect of the water distribution scheme in each integration sub-region is simulated in real time through digital twinning, and the regulation and storage strategy is dynamically corrected; by constructing an intelligent closed-loop decision-making system and fusing multi-source data to drive dynamic regulation and storage unit division and multi-objective optimization, the accuracy of water resource scheduling is improved.

Description

Technical Field

[0001] The present application relates to the technical field of water conservancy engineering, and in particular to a method and system for intelligent rainwater storage and regulation. Background Art

[0002] As the core infrastructure of agricultural production, the efficiency of water resource regulation in irrigation areas directly affects food security and ecological sustainability. Traditional irrigation area regulation and storage technologies mainly rely on historical hydrological data and empirical formulas, which have the following problems:

[0003] Existing methods are mostly based on annual / monthly rainfall-irrigation statistics, calculating the annual total water balance and a simple spatiotemporal overlap ratio to make water storage decisions. This static model is unable to cope with the frequent extreme weather events caused by climate change (such as short bursts of heavy rainfall and prolonged droughts) and ignores key influencing factors such as dynamic changes in soil moisture and differences in crop growth stages. Traditional technologies often treat irrigation areas as homogeneous, failing to consider the impact of spatial heterogeneity such as terrain slope, soil permeability, and crop type on water storage efficiency. Using fixed thresholds to divide water storage units ignores the differentiated needs of high-altitude and low-lying areas, which can easily lead to localized waterlogging or irrigation shortages. Existing systems lack real-time monitoring and intelligent feedback mechanisms, making it impossible to dynamically adjust water storage strategies based on weather forecasts (such as heavy rain warnings) and crop growth status (such as root development stage). Threshold determination based solely on single sensor data is unable to address the complex water storage needs of the coupled multivariate dynamics. Traditional methods focus on the single objective of water balance and ignore the coordinated optimization of multiple objectives such as water conservation efficiency, crop yield, and energy consumption costs, resulting in inefficient resource allocation. Summary of the Invention

[0004] The purpose of this application is to provide a method and system for intelligent rainwater storage and regulation, which significantly improves the accuracy of water resource scheduling and system resilience by constructing an intelligent closed-loop decision-making system, integrating multi-source data to drive dynamic storage unit division and multi-objective optimization, and combining digital twin real-time rehearsal and autonomous evolution mechanism.

[0005] The purpose of this application is achieved by the following technical solutions:

[0006] In a first aspect, the present application provides a method for intelligent rainwater storage and regulation, the method comprising:

[0007] S1. Real-time collection of irrigation area environmental parameters, including rainfall intensity, soil moisture, weather forecast data, and crop growth stage identification;

[0008] S2. Divide the irrigation area into multiple sub-areas based on the environmental parameters, integrate the sub-areas to obtain integrated sub-areas based on soil moisture variation trends, and determine the water storage or drainage demand status of the integrated sub-areas;

[0009] S3. Calculate the dynamic drainage or storage priority based on the water storage or drainage demand status of the integrated sub-region, combined with the terrain slope, crop root depth, and real-time humidity value of the integrated sub-region, and generate a water allocation plan through a multi-objective optimization algorithm;

[0010] S4. Drive water conservancy facilities to perform regulation and storage operations, simulate the regulation and storage effects of the water distribution plan in each integrated sub-area in real time through digital twins, and dynamically modify the regulation and storage strategy.

[0011] Preferably, the S1 includes:

[0012] A multi-factor weather station is set in each first preset area interval, and a set of multi-layer soil moisture sensors is set in each second preset area interval;

[0013] Filter the collected environmental data to remove abnormal values ​​caused by sensor failure or environmental interference;

[0014] The processed data is stored in the database according to the hierarchical structure of "space-time grid code + parameter value + parameter confidence".

[0015] Preferably, the S2 includes:

[0016] S21. Based on the improved Kriging interpolation method, combined with terrain elevation, crop canopy cover, and soil permeability coefficient, the irrigation area is divided into multiple sub-regions, and the spatial distribution of soil moisture in each sub-region is generated. Based on the spatial distribution of soil moisture in each sub-region, an integrated sub-region is generated;

[0017] S22. Perform multi-scale trend decomposition on the humidity time series data of each sub-region, construct a spatiotemporal coupled humidity change model, and determine the water storage or drainage demand status of each sub-region through a multimodal decision-making state mechanism.

[0018] Preferably, the S21 includes:

[0019] Based on digital elevation models and remote sensing images, initial sub-areas were divided according to crop type, terrain slope, and soil permeability coefficient;

[0020] According to the sub-region humidity, elevation value and canopy coverage, the improved Kriging interpolation method is used to generate the spatial distribution of soil moisture in the sub-region.

[0021] The humidity gradients of adjacent sub-regions are smoothed, and similar regions are merged to form an integrated sub-region as the final storage unit.

[0022] Preferably, the S22 includes:

[0023] Empirical mode decomposition is performed on the spatial distribution data of soil moisture in each integrated sub-region to separate the long-term trend term from the short-term fluctuation term;

[0024] Based on the long-term trend term, a nonlinear autoregressive sliding average model is constructed, and the future humidity trend of the integrated sub-region is predicted by combining the historical humidity trend value and the temperature value at the corresponding moment.

[0025] Based on the predicted future humidity trend in the integrated sub-region, combined with rainfall interference compensation and stochastic differential equations, the impact of multiple factors on humidity is integrated to generate a probabilistic trend map with the probability of increase or decrease and confidence intervals; the factors include the integrated sub-region temperature, rainfall, and irrigation amount;

[0026] Through a multimodal decision-making state mechanism, the water storage or drainage demand status of each integrated sub-region is judged based on soil moisture change trends, probabilistic trend maps, and crop growth stage identifiers.

[0027] Preferably, the S3 includes:

[0028] If the integrated sub-area is for water storage needs, the irrigation priority is determined based on the gap between the real-time humidity value and the target humidity, combined with the terrain slope, crop root depth, and crop transpiration water consumption requirements;

[0029] If the integrated sub-area has drainage requirements, the drainage priority is determined based on the excess of the real-time humidity value and the safe humidity threshold, combined with the terrain slope and crop root depth;

[0030] Based on the determined drainage or irrigation priority, a water allocation plan is generated through a multi-objective optimization algorithm to determine the water storage or drainage volume of each integrated sub-area.

[0031] Preferably, the S4 includes:

[0032] A dynamic coupling model between irrigation facilities and crop growth is constructed through digital twins, mapping the spatiotemporal changes in soil moisture, crop transpiration rates, and the operating status of irrigation facilities in the integrated sub-regions during the implementation of the water distribution plan in real time.

[0033] Fusion of real-time sensor data and meteorological updates, using a spatiotemporal Kalman filter algorithm to correct digital twin simulation deviations and generate revised dispatch effect forecasts;

[0034] If the deviation between the simulated storage effect and the expected effect exceeds a first threshold determined based on historical data statistics, dynamic feedback is triggered, and corrections are made by adjusting the water allocation ratio or execution sequence of the subsequent water distribution plan;

[0035] When extreme weather or equipment failure is detected, the flexible scheduling plan is automatically triggered, and the optimal response strategy is selected through Monte Carlo simulation to preview the amount of water resource loss, crop waterlogging or drought risks.

[0036] Preferably, the dynamic correction storage strategy includes:

[0037] Establish a historical water storage effect evaluation database to record the actual water saving rate, production gain coefficient and energy efficiency of each water storage operation;

[0038] A deep reinforcement learning model was constructed, using historical data as the training set and multi-objective optimization of water allocation plans for future time periods as the training objective, to optimize the water allocation weights and PID control parameters for each integrated sub-region.

[0039] Based on the latest weather forecasts and crop growth stage data, the model parameters are updated regularly, and the adaptability of the revised strategy to extreme working conditions is verified through the digital twin system. After verification, it is deployed to the actual storage and control end.

[0040] In a second aspect, the present application further provides a rainwater intelligent storage system for use in the rainwater intelligent storage method of the present application, the system comprising:

[0041] A data acquisition module for collecting real-time environmental parameters of the irrigation area, including rainfall intensity, soil moisture, weather forecast data, and crop growth stage identification;

[0042] a demand determination module, configured to divide the irrigation area into a plurality of sub-areas according to the environmental parameters, integrate the sub-areas to obtain integrated sub-areas according to the soil moisture variation trend, and determine the water storage or drainage demand status of the integrated sub-areas;

[0043] The solution generation module is used to calculate the dynamic drainage or water storage priority based on the water storage or drainage demand status of the integrated sub-region, combined with the terrain slope, crop root depth and real-time humidity value of the integrated sub-region, and generate a water distribution plan through a multi-objective optimization algorithm;

[0044] The regulation and storage module is used to drive water conservancy facilities to perform regulation and storage operations. It uses digital twins to simulate the regulation and storage effects of water distribution schemes in each integrated sub-area in real time and dynamically correct the regulation and storage strategy.

[0045] In a third aspect, the present application further provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any method described in the present application when executing the computer program.

[0046] The beneficial effects of this invention include: By fusing multi-source heterogeneous data (meteorological, soil, and crop data) to form a three-dimensional decision space, water storage and regulation decisions shift from single-factor judgment to comprehensive considerations, significantly improving environmental responsiveness. A sub-regional integration mechanism based on soil moisture trends enables flexible reconfiguration of water storage units, addressing resource mismatches caused by traditional fixed zoning and ensuring that water storage and drainage instructions better align with actual water demand distribution. By incorporating multiple constraints, such as terrain slope and root depth, a multi-objective optimization model is constructed to achieve a dynamic balance between water conservation, production maintenance, and energy conservation, avoiding the compromises inherent in single-objective decision-making. A digital twin system rehearses water storage and regulation effects in real time, creating a hybrid intelligent model of "virtual debugging and physical execution." This shifts policy adjustments from offline a posteriori to online foresight, significantly improving the timeliness of regulation. Dynamic feedback forms a closed-loop evolutionary path from "data-model-decision," enabling self-updating water storage strategies to adapt to the long-term evolving needs of irrigation districts and avoid the risk of policy rigidity. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a schematic diagram of the intelligent rainwater storage method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0048] Below, the present application is further described in conjunction with the accompanying drawings and specific implementation methods. It should be noted that, under the premise of no conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0049] Some embodiments of the present application provide a method for intelligent rainwater storage and regulation, the method comprising:

[0050] S1. Real-time collection of irrigation area environmental parameters, including rainfall intensity, soil moisture, weather forecast data, and crop growth stage identification;

[0051] S2. Divide the irrigation area into multiple sub-areas based on the environmental parameters, integrate the sub-areas to obtain integrated sub-areas based on soil moisture variation trends, and determine the water storage or drainage demand status of the integrated sub-areas;

[0052] S3. Calculate the dynamic drainage or storage priority based on the water storage or drainage demand status of the integrated sub-region, combined with the terrain slope, crop root depth, and real-time humidity value of the integrated sub-region, and generate a water allocation plan through a multi-objective optimization algorithm;

[0053] S4. Drive water conservancy facilities to perform regulation and storage operations, simulate the regulation and storage effects of the water distribution plan in each integrated sub-area in real time through digital twins, and dynamically modify the regulation and storage strategy.

[0054] The working principle of the above technical solution is:

[0055] Based on the closed-loop control logic of "perception-decision-execution-feedback", precise regulation of irrigation district water resources is achieved through multi-source data fusion and intelligent algorithms:

[0056] Through a multi-factor weather station and soil moisture sensor network, rainfall intensity, vertical distribution of soil moisture, weather forecast data and crop growth stage identification are obtained in real time, and after filtering, they are stored in spatiotemporal grid codes.

[0057] Initial subregions are divided based on topography, crop type, and soil characteristics. Storage units are generated by integrating the spatial distribution of humidity. A multi-scale trend decomposition model is used to predict humidity trends. A decision-making state machine is used to determine the water storage and drainage requirements of each unit, taking into account the crop growth stage.

[0058] Water storage requirements are based on humidity deficit and transpiration demand, while drainage requirements are based on humidity excess and root tolerance. A multi-objective optimization algorithm is used to generate the optimal water allocation plan that balances water conservation, yield increase, and energy conservation.

[0059] Drive water conservancy facilities to perform regulation and storage operations, simulating their effects in real time through digital twins. Integrate measured data with prediction model deviations to dynamically adjust water distribution strategies using spatiotemporal filtering algorithms. In the event of extreme weather or equipment failure, resilient response plans are triggered, and the optimal response plan is selected through Monte Carlo simulation.

[0060] The effect of the above technical solution is: based on real-time soil moisture and meteorological data, the regulation and storage units are dynamically divided, breaking the limitations of traditional fixed area division, achieving precise control of soil moisture spatial distribution at the millimeter level, and generating water distribution plans through multi-objective optimization algorithms, taking into account water conservation efficiency, crop yield and energy consumption costs, thereby improving resource utilization efficiency.

[0061] By integrating crop growth stage identification with humidity trend prediction, we can upgrade regulation from "passive response" to "active prediction" and reduce losses from waterlogging in extreme weather conditions. By combining digital twin technology with deep reinforcement learning, we can build a "perception-decision-correction" closed loop, giving the system autonomous learning capabilities and the ability to dynamically adapt to climate change and differences in crop growth cycles.

[0062] In a possible implementation, S1 includes:

[0063] The air temperature in the irrigation area is collected through an air temperature sensor, the air humidity is obtained using an air humidity sensor, the wind speed is monitored in real time with the help of an anemometer, and the light intensity is collected through a light sensor for analysis of the evaporation in the irrigation area and the transpiration rate of crops.

[0064] The layout of the collection equipment includes:

[0065] Based on the crop planting zones in the irrigation area, a multi-factor weather station is set up for each first preset area, and a set of multi-layer soil moisture sensors is set up for each second preset area. The first preset area can be 50-100 mu; the second preset area can be 20 mu.

[0066] Said S1 also includes data preprocessing steps:

[0067] Filter the collected data to remove abnormal values ​​caused by sensor failure or environmental interference;

[0068] Regularly calibrate sensor data through manual measurements to correct system errors;

[0069] If the sensor data mutation exceeds a preset threshold (e.g., 3σ) based on the historical fluctuation range, the Kalman filter-neural network joint error correction algorithm is activated to repair the data;

[0070] Automatically switch the faulty sensor to the spatiotemporal interpolation compensation mode of the neighboring nodes;

[0071] The processed data is stored in the blockchain database according to the hierarchical structure of "space-time grid coding + parameter value + parameter confidence" to ensure data traceability and tamper resistance.

[0072] The working principle of the above technical solution is:

[0073] Multi-factor meteorological stations (according to the first preset area interval) and multi-layer soil moisture sensors (according to the second preset area interval) are deployed in the irrigation area to collect real-time information on rainfall intensity, vertical distribution of soil moisture, weather forecast (hourly rainfall probability in the next 72 hours) and crop growth stage (obtaining leaf age index, root depth and other indicators through image recognition).

[0074] Median filtering and the 3σ criterion are used to remove outliers. The processed data is stored in a "spatiotemporal grid code + parameter value + confidence level" structure to ensure spatiotemporal consistency. For example, soil moisture data at 2:00 PM on March 28, 2025, is labeled [36°N, 117°E]-0-0.25m-0.75-0.95, indicating a soil moisture level of 0.75 (confidence level 0.95) in the 0-0.25m layer within the grid at 36°N, 117°E.

[0075] The above technical solution achieves the following: The coordinated deployment of meteorological stations and multi-layer sensors overcomes the limitations of traditional single-point monitoring, providing vertical distribution of rainfall intensity, three-dimensional soil moisture data, and dynamic crop growth parameters, enabling millimeter-level precision in water storage decisions. For example, by acquiring real-time root depth and leaf age index, the system can accurately identify critical periods of crop water demand (such as heading), preventing delayed or over-irrigation. A combined median filter and 3σ criterion filter improves the accuracy of outlier removal compared to single filtering algorithms, ensuring that dynamic water storage is based on real-world environmental parameters. Real-time data updates capture unexpected events such as short-term heavy rainfall. Spatiotemporal grid coding technology enables dual binding of geographic coordinates and timestamps to data, addressing the challenges of multi-source, heterogeneous data fusion and supporting the coordinated optimization of cross-regional water storage strategies. A confidence level flagging mechanism (e.g., 0.95) enables the system to automatically downgrade to historical data when data is abnormal, ensuring continued operation under extreme conditions. Vertically layered soil moisture data (e.g., 0-0.25 m, 0.25-0.5 m) provide key inputs for the root water uptake model, reducing errors in irrigation amount calculations.

[0076] In a possible implementation, S2 includes:

[0077] S21. Based on the improved Kriging interpolation method, combined with terrain elevation, crop canopy cover, and soil permeability coefficient, the irrigation area is divided into multiple sub-regions, and the spatial distribution of soil moisture in each sub-region is generated. Based on the spatial distribution of soil moisture in each sub-region, an integrated sub-region is generated;

[0078] S22. Perform multi-scale trend decomposition on the humidity time series data of each sub-region, construct a spatiotemporal coupled humidity change model, and determine the water storage or drainage demand status of each sub-region through a multimodal decision-making state mechanism.

[0079] In a possible implementation, the S21 includes:

[0080] Based on digital elevation models and remote sensing images, the initial sub-areas were divided according to crop type, terrain slope (grading thresholds of 5° and 10°), and soil permeability coefficient;

[0081] According to the sub-region humidity, elevation value and canopy coverage, the improved Kriging interpolation method is used to fuse multi-source data to generate the spatial distribution of soil moisture in the sub-region; that is, according to the sub-region humidity, elevation value and canopy coverage, the improved Kriging interpolation method is used to generate the spatial distribution of soil moisture in the sub-region:

[0082]

[0083] in, The spatial distribution of soil moisture in the generated sub-region; is the humidity of the known points in the sub-area, is the weighting coefficient, is the elevation value, is the canopy coverage; 、 2 is the terrain-vegetation coupling coefficient;

[0084] The humidity gradients of adjacent sub-regions are smoothed, and similar regions are merged to form an integrated sub-region as the final storage unit.

[0085] The working principle of the above technical solution is:

[0086] Using digital elevation models (DEMs) and remote sensing imagery, initial subregions were divided by crop type (wheat / corn), terrain slope (5° / 10° grading), and soil permeability coefficient (Ksat), forming basic units with spatial heterogeneity. An improved kriging model was used to integrate measured moisture, terrain elevation (E(x,y)), and canopy cover (C(x,y)) data. The coupling coefficients γ1 and γ2 were used to quantify terrain-vegetation interactions, resulting in a three-dimensional reconstruction of the spatial distribution of soil moisture. This improved spatial resolution compared to traditional interpolation methods. Gaussian smoothing was performed on moisture gradients in adjacent subregions, and similar regions with moisture differences below a threshold were merged to form the final storage units. This integration process reduced fragmented zoning and improved the efficiency of the storage scheme.

[0087] The effects of the above technical solution are:

[0088] Initial subregional division based on DEM, remote sensing imagery, and soil parameters improves spatial resolution compared to traditional single-metric zoning (e.g., crop type). This effectively captures the impact of topographic undulations (e.g., steep slopes with slopes > 10°) and soil permeability differences (e.g., sandy soils with Ksat > 5 cm / h) on moisture distribution. An improved kriging model incorporates elevation (E(x,y)) and canopy cover (C(x,y)) as covariates. The terrain-vegetation coupling coefficients γ1 and γ2 quantify these interactions, reducing soil moisture prediction errors and improving the accuracy of traditional kriging.

[0089] For example, in terraced irrigation areas, the model can identify areas with high canopy cover where surface moisture decreases due to transpiration, thus avoiding wasted irrigation resources. Moisture gradient smoothing (Gaussian kernel radius 50m) merges similar areas, reducing fragmented zoning and improving the efficiency of water resource scheduling. The integrated model reduces the number of units while enhancing coverage integrity. For example, 15 scattered vegetable plot units were merged into three integrated sub-areas, reducing the complexity of water allocation. Initial zoning considers both macro-topography (DEM) and micro-vegetation (canopy cover), supporting millimeter-level moisture distribution prediction (e.g., 0-0.25m soil layer), providing accurate input for root water absorption models and reducing errors in irrigation volume calculations using traditional methods.

[0090] In a possible implementation, the S22 includes:

[0091] Empirical mode decomposition is performed on the spatial distribution data of soil moisture in each integrated sub-region to separate the long-term trend term from the short-term fluctuation term;

[0092] Empirical Mode Decomposition (EMD) is an adaptive signal processing method. The specific execution process is as follows:

[0093] The EMD algorithm uses soil moisture time series data from a subregion (e.g., hourly humidity values) as input. The screening process involves identifying all maximum and minimum points in the data, generating upper and lower envelopes using cubic spline interpolation, and calculating the mean of the envelopes. Subtracting the envelope mean from the original data creates a new series, and determining whether the new series satisfies the intrinsic mode function (IMF) condition (the number of extreme points and zero crossings is equal or differs by 1, and the envelope mean is 0). If not, the above process is repeated until it is satisfied, yielding the first IMF component (corresponding to the short-term fluctuation term H_noise).

[0094] Based on the long-term trend term, a nonlinear autoregressive sliding average model was constructed, combining the historical humidity trend values ​​and the corresponding temperature values ​​to predict the future soil moisture trend of the integrated sub-region.

[0095]

[0096] is the predicted long-term trend of soil moisture in the integrated sub-region at the next k moments; is the long-term trend value of soil moisture in the integrated sub-region at the current moment; is the long-term trend value of soil moisture in the integrated sub-region at the past time i; For The corresponding integrated sub-region ambient temperature value at time i in the past; 、 is the weight coefficient, p is the length of historical data considered by the model; i = 1, 2, ..., p; is the model prediction error term, which includes random factors or noise not captured by the model;

[0097] Based on the predicted future soil moisture trends in the integrated sub-region, combined with rainfall disturbance compensation and stochastic differential equations, the impact of multiple factors on moisture is synthesized to generate a probabilistic trend map with the probability of increase or decrease and confidence intervals; the factors include the integrated sub-region temperature, rainfall, and irrigation amount;

[0098]

[0099] is the rate of change of soil moisture over time, is the temperature-dependent evaporation coefficient; is a random disturbance term (such as Gaussian white noise); is the soil moisture value of the integrated sub-region at the current moment; As the current soil moisture value of the integrated sub-region The initial value of is the theoretical maximum value of soil moisture in the integrated sub-region (such as saturated humidity); R is the rainfall; is the conversion coefficient of rainfall to soil moisture; I is the irrigation amount of the integrated sub-region; is the influence coefficient of irrigation on soil moisture;

[0100] Using dynamic input variables such as temperature T and irrigation volume I as inputs to a stochastic differential equation, the Monte Carlo simulation method is used to repeatedly simulate the evolution of soil moisture under different temperature, rainfall, and irrigation combinations. Statistical analysis is then performed on a large number of simulation results: the frequency of humidity "rising" (above a certain threshold) or "falling" (below a certain threshold) is calculated and converted into probabilities. Confidence intervals for these probabilities are calculated based on statistical distributions, ultimately generating a probabilistic trend map containing "rise / fall probabilities and confidence intervals."

[0101] According to the probabilistic trend map, if the confidence level of the sub-region trend is greater than 90% and conflicts with the global decision, the manual review mechanism is triggered to manually verify and correct the trend judgment and decision results.

[0102] Through a multimodal decision-making state mechanism, the water storage or drainage demand status of each integrated sub-region is judged based on soil moisture change trends, probabilistic trend maps, and crop growth stage identifiers.

[0103] The water storage or drainage demand status of each integrated sub-region is determined based on the following conditions:

[0104] Soil moisture change trend (long-term trend item upward / downward direction);

[0105] The probability and confidence interval of humidity change in the probabilistic trend map;

[0106] Identification of crop growth stages (e.g., if the seedling stage requires more water, water storage needs will be prioritized);

[0107] For example: an integration subregion Slightly lower, but The trend is positive, and the probability map indicates that the humidity may rise significantly. Combined with the non-critical water demand period of crops, irrigation may not be started; on the contrary, if Close to the critical value, The forecast is for a continued decline, and the probability map shows a high probability of decline. Even if crops are not in a period of high water demand, a decision may be made to store water.

[0108] The above technical solutions achieve the following: Empirical Mode Decomposition (EMD) overcomes the linear assumption of the Fourier transform. Through cubic spline interpolation and iterative filtering, it effectively separates long-term trends (such as seasonal drought) from short-term fluctuations (such as irrigation interruptions) in soil moisture data, improving extraction accuracy. A nonlinear autoregressive model (NARMAX) integrates moisture trends and temperature data to capture the nonlinear effect of temperature on evaporation and reduce prediction errors. The model can identify the rapid decline in moisture after irrigation during high-temperature periods, avoiding delayed regulation. Stochastic differential equations combined with Monte Carlo simulations generate a probability map of moisture changes (e.g., a 65% ± 10% confidence interval for the probability of a moisture drop in the next 24 hours), providing a more informed basis for decision-making and reducing misjudgments in extreme weather conditions.

[0109] The multimodal decision-making state mechanism integrates trend direction, probability confidence, and crop growth stage (e.g., water demand increases by 30% during heading stage) to achieve differentiated regulation: when the trend is positive and the confidence level is greater than 90%, delaying irrigation can save 15% of water;

[0110] When the probability of a critical humidity drop exceeds 70%, irrigation is initiated four hours earlier, reducing drought losses by 40%. A manual review mechanism (triggered when trend confidence exceeds 90% but conflicts with global trends) ensures reliable decision-making under abnormal conditions and avoids misjudgments by a single model. Probabilistic trend mapping supports flexible scheduling. For example, when the predicted probability of rainfall exceeds 80%, irrigation volume is automatically reduced by 20%, mitigating the risk of flooding.

[0111] In a possible implementation, S3 includes:

[0112] If the integrated sub-area is for water storage needs, the irrigation priority is determined based on the gap between the real-time humidity value and the target humidity, combined with the terrain slope (the steeper the slope, the faster the water loss, increasing the urgency of irrigation), the depth of the crop root system (the deeper the root system, the thicker the water layer required, which adjusts the humidity requirement), and the water consumption demand of the crop transpiration:

[0113]

[0114] is the water storage demand priority index, with higher values ​​indicating greater irrigation urgency; The target soil moisture represents the ideal soil moisture value. To integrate the real-time humidity of the sub-region, that is, the actual humidity value of the current soil;

[0115] is the correction factor for the terrain slope of the integrated sub-region; Correction coefficient for crop root depth in the integrated sub-region; To integrate the current crop transpiration of the sub-region and reflect the crop water consumption; is the maximum transpiration; w1, 、 is the weight;

[0116]

[0117] is the coefficient of crop growth stage in the integrated sub-region; 1 and 1 is a constant;

[0118] If the integrated sub-area has drainage requirements, the drainage priority is determined based on the real-time humidity value and the excess of the safe humidity, combined with the terrain slope (the greater the slope, the more urgent the drainage), and the depth of the crop root system (which affects the degree of drainage requirements):

[0119]

[0120] is the drainage demand priority index, is the safe humidity threshold; is the correction factor for the terrain slope of the integrated sub-region; is the correction coefficient of crop root depth in the integrated sub-region; w4, 5 is the weight;

[0121] Based on the determined drainage or irrigation priority, a water allocation plan is generated through a multi-objective optimization algorithm to determine the water storage or drainage volume of each integrated sub-area.

[0122] In one possible implementation, a multi-objective optimization algorithm is used to generate a water allocation plan that specifies the amount of water to be stored or discharged in each integrated sub-region, including:

[0123] Construct a multi-objective optimization model, including water conservation, yield, and salt balance objectives;

[0124] The NSGA-II algorithm is used to solve the Pareto front, and the final water distribution plan is selected through the fuzzy decision method.

[0125] The working principle of the above technical solution is:

[0126] 1) Water storage demand priority modeling:

[0127] Irrigation urgency is quantified based on the difference between real-time and target humidity. For example, when the real-time humidity is only 60% of the target, this factor contributes a priority weight of, for example, 0.4. A slope correction factor (e.g., a correction factor of +0.3 for slopes >15%) reflects the increased water loss after irrigation in steep slopes, necessitating irrigation priority. A root depth correction function (e.g., a 10% increase in required humidity for roots >0.5m) ensures deep soil moisture supply. Transpiration is dynamically adjusted based on crop growth stage; for example, transpiration increases by 50% during the jointing stage compared to the seedling stage, increasing irrigation priority. Weighting factors (e.g., w1=0.4, w2=0.3, w3=0.3) balance various factors to quantify irrigation urgency.

[0128] 2) Drainage demand priority modeling:

[0129] Drainage is triggered when humidity exceeds a safety threshold (H_s). For example, when humidity exceeds 20%, this factor contributes a priority of 0.2. Slope and root system interact in a combined manner, increasing drainage priority in steeper slopes (e.g., δ = +0.3) or for crops with shallow root systems (e.g., θ = +0.2).

[0130] 3) Generation of multi-objective water allocation schemes

[0131] Water conservation goals: Minimize total irrigation volume;

[0132] Yield objective: maximize crop water use efficiency (Y / ET, where Y is yield and ET is evapotranspiration);

[0133] Salt balance: Control the soil electrical conductivity (EC) within a safe range (<4 dS / m) after irrigation;

[0134] NSGA-II algorithm solves:

[0135] Generate a Pareto front solution set and select the optimal solution using a fuzzy membership function (e.g., water savings > 85% and yield loss < 5%). For example, when weighing water savings against yield, choose the solution that saves 18% water and reduces yield by 2%.

[0136] Regularly (e.g. weekly) adjust model parameters based on the latest crop growth data and weather forecasts (R, T) to ensure the timeliness of the solution.

[0137] The effects of the above technical solution are:

[0138] The water storage priority formula integrates moisture gap, terrain slope and evaporation to improve decision-making accuracy;

[0139] The drainage priority formula focuses on excess moisture and slope, shortening the response time to flood disasters;

[0140] Construct a three-dimensional optimization model of water conservation, yield, and salt balance to improve overall benefits.

[0141] The NSGA-II algorithm generates a Pareto front solution set and selects the optimal solution through fuzzy decision-making, such as achieving a balance between saving 18% water and reducing production by 2%, taking into account both economic and ecological goals;

[0142] The transpiration formula dynamically adjusts demand according to the crop growth stage to reduce irrigation calculation errors;

[0143] The root depth correction factor (e.g., when the root depth is greater than 0.5m, the required humidity increases by 10%) ensures deep soil moisture supply and reduces crop lodging rate;

[0144] The salinity target constrains the EC value after irrigation and prolongs the service life of the soil in the irrigated area.

[0145] When heavy rainfall (>50 mm / 3 h) is predicted, the model automatically reduces irrigation priority and increases drainage plans to reduce waterlogging losses.

[0146] The priority formula and Pareto solution set are visualized, and manual intervention is supported to adjust weights (such as temporarily increasing the production target weight of major grain-producing areas to 0.6) to meet special management needs.

[0147] The above plan shifts the water resource allocation in irrigation areas from a "one-size-fits-all" model to differentiated and precise regulation, achieving the win-win goals of water conservation, increased production, and ecological protection while ensuring food security.

[0148] In a possible implementation, the S4 includes:

[0149] A dynamic coupling model between irrigation facilities and crop growth is constructed through digital twins, mapping the spatiotemporal changes in soil moisture, crop transpiration rates, and the operating status of irrigation facilities in the integrated sub-regions during the implementation of the water distribution plan in real time.

[0150] Fusion of real-time sensor data and meteorological updates, using a spatiotemporal Kalman filter algorithm to correct digital twin simulation deviations and generate revised dispatch effect forecasts;

[0151] If the deviation between the simulated storage effect and the expected effect exceeds a first threshold determined based on historical data statistics, dynamic feedback is triggered, and corrections are made by adjusting the water allocation ratio or execution sequence of the subsequent water distribution plan;

[0152] When extreme weather (such as rainfall exceeding a once-in-a-century event) or equipment failure (such as abnormal gate opening) is detected, the flexible scheduling plan is automatically triggered. Through Monte Carlo simulation, multiple objective indicators such as water resource loss and crop waterlogging / drought risk are pre-tested to select the optimal response strategy.

[0153] In one possible implementation, the dynamic correction and storage strategy includes:

[0154] Establish a historical water storage effect evaluation database to record the actual water saving rate, production gain coefficient and energy efficiency of each water storage operation;

[0155] A deep reinforcement learning model was constructed, using historical data as the training set and multi-objective optimization of water allocation plans for future time periods as the training objective, to optimize the water allocation weights and PID control parameters for each integrated sub-region.

[0156] Based on the latest weather forecasts and crop growth stage data, the model parameters are updated regularly, and the adaptability of the revised strategy to extreme working conditions is verified through the digital twin system. After verification, it is deployed to the actual storage and control end.

[0157] The working principle of the above technical solution is:

[0158] A dynamic coupling model was established between water conservancy facilities (gate opening, pump flow) and crop growth (transpiration rate, root water uptake), mapping the effectiveness of water allocation plans in real time. For example, a simulation showed that opening a gate by 30% could increase the humidity in a target area from 0.35 to 0.42 within one hour.

[0159] By fusing real-time sensor data (soil moisture, gate status) with meteorological updates (rainfall forecasts), the model parameters are corrected using a spatiotemporal Kalman filter algorithm. For example, if the measured moisture deviates from the simulated value by more than 5%, the model's permeability coefficient μ(T) is automatically adjusted, reducing subsequent forecast errors to ±3%.

[0160] If the deviation between the simulated storage effect and the expected value exceeds the first threshold of historical statistics (such as the water-saving rate deviation > 8%), feedback adjustment will be triggered: for example, increasing the subsequent irrigation volume by 10% or draining water 1 hour in advance.

[0161] When a red alert for heavy rain (e.g., rainfall > 50mm / 3h) or a gate failure (e.g., opening deviation > 15%) is detected, the flexible emergency plan is automatically triggered:

[0162] Monte Carlo simulations are run multiple times (e.g., 1,000 times) to simulate flood losses under different scheduling strategies. The solution that minimizes energy consumption and reduces yield by less than 3% is selected. For example, drainage pumps in low-lying areas may be prioritized, and irrigation plans may be delayed until the rain stops.

[0163] Store the past five years of water storage records, including indicators such as water saving rate (actual / planned), yield gain (measured / benchmark), and energy efficiency (kWh / m³), as the basis for model training.

[0164] Construct a deep reinforcement learning model (such as the PPO algorithm), with the water saving rate, production gain, and energy efficiency of the water allocation plan in the future period as the optimization goals, and dynamically adjust the water allocation weights of each integrated sub-region (such as increasing the weight of the high-priority area from 0.6 to 0.7) and PID control parameters (proportional coefficient Kp=1.2→1.5).

[0165] Model parameters are updated weekly based on the latest weather forecast (for the next 7 days) and crop growth data (such as leaf age index). The digital twin system is used to simulate extreme working conditions (such as 20 consecutive days of drought) to verify the adaptability of the strategy. Once verified, the strategy is deployed to the actual control end.

[0166] The effects of the above technical solution are:

[0167] A digital twin system integrating all elements of water resources, crops, and the environment is constructed to achieve real-time, synchronous mapping of soil moisture fields, crop physiological responses, and facility operating status, forming a parallel decision-making space for the "digital irrigation district" and significantly improving prediction reliability. A Kalman filter algorithm and dynamic feedback mechanism are integrated to establish a closed-loop control chain of "prediction-execution-monitoring-correction," enabling the system to compensate for online deviations and effectively mitigate the impact of sensor noise and weather forecast errors on water storage and regulation. A library of extreme operating response plans is designed, combined with Monte Carlo risk rehearsals to automatically switch to a multi-objective optimization mode in the event of heavy rain, drought, or equipment failure, ensuring the system's survivable scheduling capabilities in abnormal scenarios. Deep reinforcement learning is used to mine implicit knowledge from historical water storage and regulation experience, dynamically optimizing water allocation weights and control parameters. This allows the decision-making model to continuously evolve as the irrigation district environment changes, avoiding the risk of policy rigidity. A comprehensive value chain is established, from data collection and model training to policy deployment. The robustness of new policies is verified through digital twins before physical control is implemented, achieving a seamless transition from "digital experimentation to physical optimization" and reducing trial-and-error costs.

[0168] In summary, through the organic complementarity of mechanism models and data-driven methods, a complete intelligent agent architecture of "perception-cognition-decision-action" has been formed, providing innovative solutions for the precise regulation of complex water resource systems.

[0169] The present application also provides a rainwater intelligent storage system for use in the rainwater intelligent storage method of the present application. The system includes:

[0170] A data acquisition module for collecting real-time environmental parameters of the irrigation area, including rainfall intensity, soil moisture, weather forecast data, and crop growth stage identification;

[0171] a demand determination module, configured to divide the irrigation area into a plurality of sub-areas according to the environmental parameters, integrate the sub-areas to obtain integrated sub-areas according to the soil moisture variation trend, and determine the water storage or drainage demand status of the integrated sub-areas;

[0172] The solution generation module is used to calculate the dynamic drainage or water storage priority based on the water storage or drainage demand status of the integrated sub-region, combined with the terrain slope, crop root depth and real-time humidity value of the integrated sub-region, and generate a water distribution plan through a multi-objective optimization algorithm;

[0173] The regulation and storage module is used to drive water conservancy facilities to perform regulation and storage operations. It uses digital twins to simulate the regulation and storage effects of water distribution schemes in each integrated sub-area in real time and dynamically correct the regulation and storage strategy.

[0174] The working principle and effect of the above technical solution are the same as those in the embodiment of the method of this application, and will not be repeated here.

[0175] The present application also provides an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of any method described in the present application are implemented.

[0176] This application is explained from the perspectives of purpose of use, effectiveness, progress and novelty, and has complied with the functional enhancement and use requirements emphasized by the Patent Law. The above description and drawings of this application are only preferred embodiments of this application and are not intended to limit this application. Therefore, all structures, devices, features, etc. that are similar or identical to those of this application, that is, all equivalent replacements or modifications made in accordance with the scope of the patent application of this application, should fall within the scope of protection of the patent application of this application.

Claims

1. A rainwater intelligent storage method, characterized in that: The method comprises: S1. Real-time collection of irrigation area environmental parameters, including rainfall intensity, soil moisture, weather forecast data, and crop growth stage identification; S2. Divide the irrigation area into multiple sub-areas based on the environmental parameters, integrate the sub-areas to obtain integrated sub-areas based on soil moisture variation trends, and determine the water storage or drainage demand status of the integrated sub-areas; S3. Calculate the dynamic drainage or storage priority based on the water storage or drainage demand status of the integrated sub-region, combined with the terrain slope, crop root depth, and real-time humidity value of the integrated sub-region, and generate a water allocation plan through a multi-objective optimization algorithm; S4. Drive water conservancy facilities to perform regulation and storage operations, simulate the regulation and storage effects of the water distribution plan in each integrated sub-area in real time through digital twins, and dynamically modify the regulation and storage strategy.

2. The intelligent rainwater storage method according to claim 1, characterized in that: Said S1 comprises: A multi-factor weather station is set in each first preset area interval, and a set of multi-layer soil moisture sensors is set in each second preset area interval; Filter the collected environmental data to remove abnormal values ​​caused by sensor failure or environmental interference; The processed data is stored in the database according to the hierarchical structure of "space-time grid code + parameter value + parameter confidence".

3. The intelligent rainwater storage method according to claim 1, characterized in that: The S2 includes: S21. Based on the improved Kriging interpolation method, combined with terrain elevation, crop canopy cover, and soil permeability coefficient, the irrigation area is divided into multiple sub-regions, and the spatial distribution of soil moisture in each sub-region is generated. Based on the spatial distribution of soil moisture in each sub-region, an integrated sub-region is generated; S22. Perform multi-scale trend decomposition on the humidity time series data of each sub-region, construct a spatiotemporal coupled humidity change model, and determine the water storage or drainage demand status of each sub-region through a multimodal decision-making state mechanism.

4. The intelligent rainwater storage method according to claim 3, characterized in that: The S21 includes: Based on digital elevation models and remote sensing images, initial sub-areas were divided according to crop type, terrain slope, and soil permeability coefficient; According to the sub-region humidity, elevation value and canopy coverage, the improved Kriging interpolation method is used to generate the spatial distribution of soil moisture in the sub-region. The humidity gradients of adjacent sub-regions are smoothed, and similar regions are merged to form an integrated sub-region as the final storage unit.

5. The intelligent rainwater storage method according to claim 3, characterized in that: The S22 includes: Empirical mode decomposition is performed on the spatial distribution data of soil moisture in each integrated sub-region to separate the long-term trend term from the short-term fluctuation term; Based on the long-term trend term, a nonlinear autoregressive sliding average model is constructed, and the future humidity trend of the integrated sub-region is predicted by combining the historical humidity trend value and the temperature value at the corresponding moment. Based on the predicted future humidity trend in the integrated sub-region, combined with rainfall interference compensation and stochastic differential equations, the impact of multiple factors on humidity is integrated to generate a probabilistic trend map with the probability of increase or decrease and confidence intervals; the factors include the integrated sub-region temperature, rainfall, and irrigation amount; Through a multimodal decision-making state mechanism, the water storage or drainage demand status of each integrated sub-region is judged based on soil moisture change trends, probabilistic trend maps, and crop growth stage identifiers.

6. The intelligent rainwater storage method according to claim 1, characterized in that: The S3 includes: If the integrated sub-area is for water storage needs, the irrigation priority is determined based on the gap between the real-time humidity value and the target humidity, combined with the terrain slope, crop root depth, and crop transpiration water consumption requirements; If the integrated sub-area has drainage requirements, the drainage priority is determined based on the excess of the real-time humidity value and the safe humidity threshold, combined with the terrain slope and crop root depth; Based on the determined drainage or irrigation priority, a water allocation plan is generated through a multi-objective optimization algorithm to determine the water storage or drainage volume of each integrated sub-area.

7. The intelligent rainwater storage method according to claim 1, characterized in that: The S4 includes: A dynamic coupling model between irrigation facilities and crop growth is constructed through digital twins, mapping the spatiotemporal changes in soil moisture, crop transpiration rates, and the operating status of irrigation facilities in the integrated sub-regions during the implementation of the water distribution plan in real time. Fusion of real-time sensor data and meteorological updates, using a spatiotemporal Kalman filter algorithm to correct digital twin simulation deviations and generate revised dispatch effect forecasts; If the deviation between the simulated storage effect and the expected effect exceeds a first threshold determined based on historical data statistics, dynamic feedback is triggered, and corrections are made by adjusting the water allocation ratio or execution sequence of the subsequent water distribution plan; When extreme weather or equipment failure is detected, the flexible scheduling plan is automatically triggered, and the optimal response strategy is selected through Monte Carlo simulation to preview the amount of water resource loss, crop waterlogging or drought risks.

8. The intelligent rainwater storage method according to claim 1, characterized in that: The dynamic correction storage strategy includes: Establish a historical water storage effect evaluation database to record the actual water saving rate, production gain coefficient and energy efficiency of each water storage operation; A deep reinforcement learning model was constructed, using historical data as the training set and multi-objective optimization of water allocation plans for future time periods as the training objective, to optimize the water allocation weights and PID control parameters for each integrated sub-region. Based on the latest weather forecasts and crop growth stage data, the model parameters are updated regularly, and the adaptability of the revised strategy to extreme working conditions is verified through the digital twin system. After verification, it is deployed to the actual storage and control end.

9. A system for implementing the intelligent rainwater storage method according to claim 1, characterized in that: The system comprises: A data acquisition module for collecting real-time environmental parameters of the irrigation area, including rainfall intensity, soil moisture, weather forecast data, and crop growth stage identification; a demand determination module, configured to divide the irrigation area into a plurality of sub-areas according to the environmental parameters, integrate the sub-areas to obtain integrated sub-areas according to the soil moisture variation trend, and determine the water storage or drainage demand status of the integrated sub-areas; The solution generation module is used to calculate the dynamic drainage or water storage priority based on the water storage or drainage demand status of the integrated sub-region, combined with the terrain slope, crop root depth and real-time humidity value of the integrated sub-region, and generate a water distribution plan through a multi-objective optimization algorithm; The regulation and storage module is used to drive water conservancy facilities to perform regulation and storage operations. It uses digital twins to simulate the regulation and storage effects of water distribution schemes in each integrated sub-area in real time and dynamically correct the regulation and storage strategy.

10. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method according to any one of claims 1 to 8 when executing the computer program.

Images