A Smart Irrigation Method and System for Dryland Maize Based on Dynamic Adaptation and Uncertainty Perception
By collecting real-time data and dynamically adjusting weights, and combining meteorological forecast errors to generate a collaborative decision-making index, the shortcomings of intelligent irrigation systems in dynamic adaptability and uncertainty perception are solved, thereby improving the environmental adaptability of irrigation decisions and the efficiency of water resource utilization.
Patent Information
- Application Number
- CN202611211112.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-11
- Publication Date
- 2026-11-10
AI Technical Summary
Existing intelligent irrigation systems are inadequate in terms of dynamic adaptability, uncertainty perception, and forward-looking decision-making, and cannot accurately reflect the maize growth period and environmental changes, leading to problems of over- or under-irrigation.
By collecting real-time meteorological, soil, and crop data, dynamically adjusting the weights of the water demand assessment model, combining meteorological forecast errors and long short-term memory neural network prediction models, generating a collaborative decision-making index, implementing a tiered supplementary irrigation control strategy, and recording the decision-making effects for model optimization.
This approach enhances the environmental adaptability of irrigation decisions, accurately quantifies the coupling relationship between maize water demand and rainfall, and improves water resource utilization and decision robustness.
Smart Images

Figure CN123071104A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent agricultural irrigation technology, and in particular to an intelligent supplementary irrigation method and system for dryland maize based on dynamic adaptation and uncertainty perception. Background Technology
[0002] Precision agriculture and intelligent irrigation technology are key technologies supporting the improvement of water use efficiency in dryland maize and ensuring food security. As an important food crop in my country, maize has a high water requirement during its growth process. However, dryland maize cultivation faces prominent problems such as uneven rainfall distribution, poor soil water retention capacity, and uncertainty in weather forecasts. Therefore, how to achieve scientific and rational irrigation management has become the core issue for improving dryland maize yield.
[0003] Currently, existing smart irrigation systems mainly suffer from the following technical shortcomings: The existing water demand calculation models suffer from poor dynamic adaptability. They use fixed weight coefficients, which cannot reflect the dynamic changes in the importance of each parameter under different growth stages and environmental conditions. For example, changes in the stem during the corn seedling stage are sensitive to water shortage, while evaporation contributes the most to water demand during the tasseling stage. Fixed weights cannot adapt to such dynamic changes.
[0004] Rainfall utilization forecasts are inaccurate because current rainfall prediction technologies do not take into account the dynamic changes in soil infiltration. Soil infiltration rate varies with water content; infiltration is slow in dry soil and fast in moist soil. Current technologies using fixed infiltration parameters cannot accurately reflect this dynamic process.
[0005] Lacking risk perception capabilities, current technologies employ deterministic decision-making methods, failing to consider the impact of weather forecast errors and prediction model errors on decision-making. In scenarios with high weather forecast uncertainty, deterministic decision-making may lead to problems such as over-irrigation or under-irrigation.
[0006] The decision-making process lacks foresight. Current irrigation decisions are mainly based on the current state and do not take into account the synergistic effects of future weather conditions and crop water demand trends, thus failing to achieve a shift from a "passive response" to a "proactive prevention" decision-making model.
[0007] Therefore, there is an urgent need for a technical solution to address the aforementioned problems in existing technologies. Summary of the Invention
[0008] This invention aims to provide a method and system for intelligent supplementary irrigation of dryland maize based on dynamic adaptation and uncertainty perception, which can effectively solve the problems of poor dynamic adaptability, lack of uncertainty perception, insufficient forward-looking decision-making, and inability of the model to self-optimize in the existing technology.
[0009] According to one aspect of the present invention, a smart irrigation method for dryland maize based on dynamic adaptation and uncertainty perception is provided, comprising: Real-time collection of meteorological data, soil moisture data, and crop physiological data to assess the accuracy of weather forecasts and the error of prediction models; Based on the current growth stage of maize, the temperature influence coefficient and growth stage regulation factor are determined. The real-time weight adjustment coefficient is iteratively updated by combining historical decision feedback, and the weight coefficient of the water demand assessment model is dynamically adjusted to calculate the maize water demand-rainfall coupling index. The dynamic infiltration time constant is updated based on the real-time soil moisture content, and the available rainfall is calculated and predicted based on the dynamic infiltration time constant. Based on future weather forecasts and crop coefficients, the crop water demand for future periods is predicted. Combined with the predicted rainfall, the water surplus / deficit forecast is calculated, and a meteorological-crop collaborative decision-making index is generated. Based on the accuracy of the weather forecast and the error of the long short-term memory neural network prediction model, a comprehensive uncertainty weight is calculated. The collaborative decision-making index is then corrected using the comprehensive uncertainty weight. A graded supplementary irrigation control strategy is then executed based on the corrected decision-making index. Record the results of decision execution and evaluate the effectiveness of the decision based on the prediction deviation. Update the model parameters when the deviation exceeds a preset threshold.
[0010] The weighting coefficients of the dynamically adjusted water demand assessment model include: The dynamic weighting coefficients include three sets of coefficients: stem diameter change rate, evaporation, and crop temperature. Each set of dynamic weighting coefficients is determined by the product of the basic weighting coefficient, the growth period regulation factor, and the real-time weighting adjustment coefficient. The growth period regulation factor is set according to the emergence period, jointing period, large trumpet stage, tasseling period, grain filling period, or maturity period of the corn. The real-time weight adjustment coefficient is iteratively updated based on feedback data from historical decisions.
[0011] According to some embodiments, the predicted rainfall can be calculated based on weather forecast rainfall, soil infiltration coefficient, vegetation interception coefficient, and the dynamic infiltration time constant. The dynamic infiltration time constant is dynamically updated according to the difference between the current soil moisture content and field capacity. When the soil is dry, the dynamic infiltration time constant increases and the infiltration slows down; when the soil is moist, the dynamic infiltration time constant decreases and the infiltration speeds up. The prediction can utilize rainfall P av The calculation formula is:
[0012] in, To predict the amount of rainfall at a given time, Ksoil K is the soil infiltration coefficient, reflecting the degree to which the current state of the soil hinders rainfall infiltration, and its value ranges from 0 to 1. veg τ is the vegetation interception coefficient, reflecting the crop canopy's ability to intercept rainfall, with a value ranging from 0 to 1. inf (ti) is the dynamic infiltration time constant.
[0013] The dynamic infiltration time constant τ inf (t i Updated in real time using the following formula:
[0014] in, The basic infiltration time constant is defined as 0.8–1.5 hours; k is the soil texture attenuation coefficient; and θ is the soil texture attenuation coefficient. v (t i θ represents the soil volumetric water content at the predicted time. FC It refers to field water holding capacity.
[0015] According to some embodiments, the generation process of the meteorological-crop collaborative decision-making index includes: Calculate a decision index that reflects the current water demand risk; The water surplus / deficit forecast function is used to transform the water surplus / deficit forecast values for future periods into decision signals. The collaborative decision-making index is obtained by weighted summation of the decision index and the decision signal. Specifically, when the water surplus / deficit prediction value is a water surplus, the output value of the water surplus / deficit prediction function is negative in order to reduce the collaborative decision-making index. The SynDCI (Synthesis Decision Index) is calculated using the following formula:
[0016] Wherein, cid is the decision index reflecting the current water demand risk, ΔWater is the predicted water surplus / deficit value for the future period, and Synergy(ΔWater) is the water surplus / deficit prediction function. This is the collaborative weighting coefficient.
[0017] According to some embodiments, the comprehensive uncertainty weight γ(t) is calculated using the following formula: γ(t) = 1 / (1 + σ forecast (t)² + σ LSTM (t)²) Where, σ forecast (t) represents the uncertainty of weather forecast, σ LSTM (t) represents the uncertainty of the prediction model error.
[0018] According to some embodiments, the step of implementing a tiered irrigation control strategy based on a modified decision index includes: The first threshold is set according to the current reproductive period. Second threshold ,in It is a positive value. It is a negative value; When the revised decision index is less than the second threshold At that time, it is determined that irrigation is not needed or only an early warning is issued; When the revised decision index is greater than or equal to the second threshold And less than the first threshold At that time, execute the micro-irrigation command; When the revised decision index is greater than or equal to the first threshold At that time, an active supplemental irrigation strategy is implemented.
[0019] According to another aspect of the present invention, a smart irrigation system for dryland maize is provided, the system comprising: The uncertainty perception data fusion module is used to collect multi-source data in real time through meteorological monitoring sensor arrays, soil moisture sensor arrays and crop physiological sensor arrays, and evaluate the accuracy of meteorological forecasts and the error of long short-term memory neural network prediction models based on rolling verification. The dynamic adaptive water demand assessment module is used to determine the temperature influence coefficient and growth period adjustment factor based on the current growth stage of maize, and iteratively update the real-time weight adjustment coefficient by combining historical decision feedback. The maize water demand-rainfall coupling index is calculated through the dynamic weight coefficient. The dynamic infiltration prediction module is used to update the dynamic infiltration time constant based on the real-time soil moisture content and calculate and predict the available rainfall. The meteorological-crop collaborative decision-making module is used to predict crop water demand trends and rainfall replenishment capacity in future periods, calculate water surplus and deficit forecasts, and generate collaborative decision-making indices. The uncertainty perception and decision execution module is used to calculate the comprehensive uncertainty weight based on the accuracy of weather forecasts and the error of prediction models, correct the collaborative decision index, and execute the graded irrigation strategy. The model self-optimization module is used to record the decision execution results and the corresponding measured soil moisture values, compare the prediction deviation, and update the dynamic weight adjustment coefficient and the synergistic weight coefficient when the deviation exceeds the limit.
[0020] According to some embodiments, the system further includes a regional feature database, which stores soil texture parameters, peak daily water demand, water demand modulus, and water sensitivity index for different ecological types of zones; The dynamic adaptive water demand assessment module and the dynamic infiltration prediction module call the corresponding parameters from the regional feature library according to the current ecological type zone.
[0021] According to some embodiments, the uncertainty-aware data fusion module includes: The meteorological monitoring sensor array includes an atmospheric temperature sensor, an atmospheric humidity sensor, a wind speed sensor, a solar radiation sensor, and an evaporation sensor, used to collect atmospheric temperature, humidity, wind speed, solar radiation, and evaporation. The soil moisture sensor array includes a soil volumetric water content sensor, a soil temperature sensor, and a soil electrical conductivity sensor, which are arranged in a grid pattern at different depths in the corn root zone to collect soil volumetric water content, temperature, and electrical conductivity. The crop physiological sensor array includes a stem diameter change sensor and a leaf surface temperature sensor, used to collect data on stem diameter changes and leaf surface temperature.
[0022] According to another aspect of the present invention, a computing device is provided, including a processor; and a memory storing a computer program that, when executed by the processor, implements the method as described in any of the preceding claims.
[0023] The technical solutions provided by the embodiments of the present invention have the following beneficial effects: This invention improves the environmental adaptability of irrigation decisions by collecting multidimensional data in real time and evaluating forecast and model errors. By dynamically adjusting weights using growth period factors and historical feedback, the coupling relationship between maize water demand and rainfall can be accurately quantified, solving the problem of fixed parameters in traditional models. Updating the dynamic infiltration time constant by real-time soil moisture content improves the accuracy of predicting available rainfall. A collaborative decision-making index is generated by combining future meteorological and crop coefficients, and a comprehensive uncertainty weight is introduced for correction, eliminating the uncertainty interference from a single prediction source and enhancing the robustness of the decision. Based on the corrected index, hierarchical control is implemented and model parameters are updated in a closed loop, achieving precise on-demand irrigation supply and system self-learning optimization, thus improving water resource utilization.
[0024] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description
[0025] Figure 1 This is a flowchart illustrating an example embodiment of the present invention of a method for intelligent supplementary irrigation of dryland maize based on dynamic adaptation and uncertainty perception.
[0026] Figure 2 This is a schematic diagram of the process for dynamically and adaptively calculating the water demand-rainfall coupling index for maize according to an example embodiment.
[0027] Figure 3 This is a schematic diagram of the collaborative decision-making computation process in an example embodiment of the present invention.
[0028] Figure 4 This is a schematic diagram illustrating the process of calculating the comprehensive uncertainty weight in an example embodiment of the present invention.
[0029] Figure 5 This is a schematic diagram of the system composition for intelligent supplemental irrigation of dryland maize according to an example embodiment.
[0030] Figure 6 This is a schematic diagram of the architecture of a smart supplementary irrigation system for dryland maize according to an example embodiment.
[0031] Figure 7 A block diagram of a computing device according to an exemplary embodiment of the present invention is shown. Detailed Implementation
[0032] To further understand the content of this invention, a detailed description of the invention is provided in conjunction with the accompanying drawings and embodiments. The invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the relevant invention and are not intended to limit the invention. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.
[0033] cwpc (Corn Water-Precipitation Coupling Index): The corn water demand-rainfall coupling index represents the combined coupling state of corn water demand urgency and rainfall replenishment capacity at the current moment.
[0034] SynDCI (Synergistic Decision Composite Index): A meteorological-crop collaborative decision index that integrates current water demand risk and future water surplus / deficit trends into a comprehensive decision-making indicator.
[0035] Pav (Predicted Available Precipitation): Predicted available rainfall refers to the effective portion of forecasted rainfall that can infiltrate into the root zone and be utilized by crops.
[0036] LAI (Leaf Area Index): A dimensionless parameter characterizing crop canopy coverage, determined by referring to a table based on the growth stage.
[0037] (Reference Evapotranspiration): A baseline evapotranspiration value calculated based on temperature, humidity, wind speed, and solar radiation, used as a reference crop evapotranspiration.
[0038] like Figure 1 As shown in the figure, this embodiment proposes an intelligent supplementary irrigation method for dryland maize based on dynamic adaptation and uncertainty perception.
[0039] In S101, meteorological data, soil moisture data, and crop physiological data are collected in real time to assess the accuracy of weather forecasts and the error of prediction models.
[0040] According to some embodiments, meteorological data, soil moisture data, and crop physiological data are collected in real time using meteorological monitoring sensor arrays, soil moisture sensor arrays, and crop physiological sensor arrays, and the accuracy of meteorological forecasts and the error of prediction models are evaluated. Specifically, the meteorological data includes atmospheric temperature, atmospheric humidity, wind speed, solar radiation, and evaporation; the soil moisture data includes soil volumetric water content, soil temperature, and soil electrical conductivity; and the crop physiological data includes changes in stem diameter and leaf surface temperature.
[0041] In S103, the temperature influence coefficient and growth period regulation factor are determined based on the current growth stage of maize. The real-time weight adjustment coefficient is iteratively updated in combination with historical decision feedback. The weight coefficient of the water demand assessment model is dynamically adjusted, and the maize water demand-rainfall coupling index is calculated. The real-time weight adjustment coefficient is iteratively updated based on the feedback data of historical decisions.
[0042] According to some embodiments, the formula for calculating the maize water demand-rainfall coupling index cwpc is as follows:
[0043] in, The dynamic weighting coefficient is Sd, where Sd is the stem diameter. Let be the rate of change of stem diameter, Ta(τ) be the atmospheric temperature, and E(τ) be the evaporation rate. This refers to the soil volumetric water content. Field holding capacity Rh(τ) represents leaf surface temperature, and Rh(τ) represents atmospheric relative humidity. The total duration of the entire growth period of corn, expressed in hours.
[0044] Specifically, the α(τ) dynamic weighting coefficient includes three sets of coefficients: stem diameter change rate, evaporation, and crop temperature. Each set of dynamic weighting coefficients is determined by the product of the basic weighting coefficient α, the growth stage adjustment factor f(stage), and the real-time weighting adjustment coefficient w(τ). The growth stage adjustment factor f(stage) is set according to the corn's emergence, jointing, tasseling, grain-filling, or maturity stages. Among these, the emergence stage emphasizes stem change signals, while the jointing stage emphasizes evaporation signals. The real-time weighting adjustment coefficient is iteratively updated based on feedback data from historical decisions. The calculation process for the water demand-rainfall coupling index cwpc is described in [link to calculation]. Figure 2 .
[0045] Specifically, , in, Based on the weighting coefficient, These are real-time weight adjustment coefficients updated based on historical decision feedback. These are regulatory factors determined based on the reproductive period.
[0046] According to some embodiments, the real-time weight adjustment coefficients are updated using a Q-learning algorithm, which learns the optimal weight adjustment strategy through interaction with the environment. Specifically, a state space is constructed, with the current soil moisture content, the maize growth stage (e.g., encoding different growth stages as discrete states), and the accuracy (or confidence level) of the weather forecast as state variables. An action space is constructed, defining the direction and magnitude of weight adjustment as action variables, such as increasing the stem weight, decreasing the evaporation weight, or keeping it unchanged as discrete actions. A reward function is designed, using the actual effect after the irrigation decision as the feedback signal. Specifically, the negative value of the prediction deviation or the improvement in water use efficiency after irrigation is used as the reward value. A positive reward is given when the prediction deviation decreases (i.e., the prediction is more accurate) or the crop growth status improves; conversely, a negative reward (penalty) is given.
[0047] Based on the current state, the selected action is executed with weight adjustment, and the state and reward value at the next moment are observed. The Q-value table is then updated. Through the above process, the system can automatically iteratively update the real-time weight adjustment coefficient based on the feedback data of historical decisions, so that the water demand assessment model can adapt to the current environmental changes.
[0048] The reproductive period regulation factor Based on different growth stages of maize, the focus during the emergence stage is on changes in stem stalks, as these changes are most sensitive to water shortage. Take the higher value; during the jointing stage, focus on the evaporation rate signal, as evaporation during this stage contributes the most to water demand. Take the higher value; temperature stress during the grain-filling period affects grain development. Take the higher value. The specific range of values is as follows: Emergence period: ; Jointing stage: ; Big trumpet mouth period: ; Erection period: ; Grouting period: ; Maturity stage: .
[0049] The temperature influence coefficient characterizes the nonlinear effect of atmospheric temperature on crop transpiration rate, with a value ranging from 0.01 to 0.05. The specific value is determined based on the temperature sensitivity of maize at different growth stages: 0.01 for emergence, 0.02 for jointing, 0.04 for tasseling, 0.03 for grain filling, and 0.02 for maturity. Considering that maize seedlings have weak transpiration during the emergence stage and are not sensitive to temperature changes, Take the lower value of 0.01; during the jointing stage, maize enters a rapid growth phase, and transpiration increases. The value was 0.02; the temperature during the large trumpet stage had the most significant impact on maize leaf transpiration and photosynthesis. The highest value was taken as 0.04; the temperature sensitivity during the male ejaculation period decreased slightly. Take 0.03; High temperature stress during the grain-filling period affects grain filling and dry matter accumulation, with temperature showing the highest sensitivity. Take the highest value of 0.05; During the maturity stage, the physiological activity of maize weakens, and its temperature sensitivity decreases. Take 0.02.
[0050] In S105, the dynamic infiltration time constant is updated based on the real-time soil moisture content, and the available rainfall is calculated and predicted based on the dynamic infiltration time constant.
[0051] According to some embodiments, the predicted rainfall can be calculated based on weather forecast rainfall, soil infiltration coefficient, vegetation interception coefficient, and the dynamic infiltration time constant.
[0052] The predicted available rainfall The calculation formula is:
[0053] Where i is the sequence number of the rainfall event, and n is the total number of rainfall events within the forecast period. Let Δt be the predicted time of the i-th rainfall event. i Let be the time elapsed since the i-th rainfall occurred. Forecast rainfall at the predicted time, unit: millimeters.
[0054] K soil K is the soil infiltration coefficient, reflecting the degree to which the current soil condition hinders rainfall infiltration. Its value ranges from 0 to 1, decreasing as soil volumetric water content increases. When the soil is completely dry (with no obstruction to infiltration), K... soil = 1, K when the water content is close to the field capacity. soil < 1.
[0055] K vegK is the vegetation interception coefficient, reflecting the crop canopy's ability to intercept rainfall. Its value ranges from 0 to 1, decreasing as the maize leaf area index increases. Specifically, before the jointing stage, maize has low vegetation cover. veg 1. High vegetation cover after the large trumpet stage, K veg Take a value of 0.8 to 0.95.
[0056] τ inf (ti) is the dynamic infiltration time constant, which is dynamically updated according to the difference between the current soil moisture content and field capacity. inf (t i Updated in real time using the following formula:
[0057] in, The basic infiltration time constant is denoted by θ, with a value ranging from 0.8 to 1.5 hours. k is the soil texture attenuation coefficient, determined by soil texture, which is determined by soil particle size distribution. When sandy soil particles are ≥0.05mm and account for >70%, k is 0.25–0.35; when loam particles are 0.002–0.05mm and account for >70%, k is 0.45–0.55; when clay particles are <0.002mm and account for >70%, k is 0.75–0.85. v (t i θ represents the soil volumetric water content at the predicted time. FC This refers to field holding capacity. When the soil is dry, the dynamic infiltration time constant increases, and infiltration slows down; when the soil is wet, the dynamic infiltration time constant decreases, and infiltration accelerates.
[0058] In S107, based on future weather forecasts and crop coefficients, the crop water demand for future periods is predicted. Combined with the predicted available rainfall, the water surplus / deficit forecast is calculated, and a meteorological-crop collaborative decision-making index is generated.
[0059] According to some embodiments, the generation process of the meteorological-crop collaborative decision-making index is described in [reference needed]. Figure 3 The decision index cid, reflecting the current water demand risk, is calculated. The water surplus / deficit prediction function Synergy is used to convert the predicted water surplus / deficit values for future periods into decision signals. The decision index cid and the decision signals are then weighted and summed to obtain the SynDCI (Synchronous Decision Index). Specifically, when the predicted water surplus / deficit value is a water surplus, the output value of the water surplus / deficit prediction function is negative to reduce the SynDCI.
[0060] The SynDCI (Synthesis Decision Index) is calculated using the following formula:
[0061] Among them, cid is a decision index reflecting the current water demand risk. , C is the normalized baseline parameter for cwpc, taken as 0.8~1.2 based on the reproductive period, P threshold The effective utilization threshold for rainfall is set at 25-40 mm, depending on the ecological type of the zone. ΔWater represents the predicted water surplus / deficit value for future periods, and Synergy(ΔWater) is the water surplus / deficit prediction function. The collaborative weighting coefficient, and =1; , ΔWater(t+Δt) is the predicted water surplus / deficit value for future periods, which is equal to the future crop water requirement ET. future (t+Δt) and predicted available rainfall Rain future The difference between (t+Δt); When ΔWater ≥ 0, Synergy(ΔWater) = exp(ΔWater / ET) threshold -1; When ΔWater < 0, Synergy(ΔWater) = 1 - exp(-ΔWater / ET) threshold ), of which ET threshold This represents the water requirement threshold during the reproductive period.
[0062] In S109, a comprehensive uncertainty weight is calculated based on the accuracy of the weather forecast and the error of the long short-term memory neural network prediction model. The comprehensive uncertainty weight is used to correct the collaborative decision index, and a graded irrigation control strategy is executed according to the corrected decision index.
[0063] According to some embodiments, the uncertainty-aware decision computation process is described below. Figure 4 Forecast accuracy assessment calculation σ forecast Calculate the error σ of the Long Short-Term Memory Neural Network Model LSTM The overall weight γ(t) reflects the credibility of the decision, and cid final To achieve uncertainty-aware decision correction, the comprehensive uncertainty weight γ(t) is calculated using the following formula:
[0064] in, Basic forecast error, forecast confidence For the confidence level of a weather forecast, Δt forecast For forecast timeliness; , Among them, RMSE LSTMThe root mean square error was obtained by statistical analysis of historical validation datasets to validate the LSTM model. Q is the score given for the quality of the current input data. train This is the average quality score for the training data.
[0065] The comprehensive uncertainty weight γ(t) is calculated based on the weather forecast error and the LSTM model prediction error. When the prediction error is large, γ(t) decreases, and the decision is more conservative. When the prediction error is small, γ(t) is close to 1, and the decision is executed normally.
[0066] The revised decision index cid final = γ(t) × SynDCI.
[0067] The first threshold is set according to the current reproductive period. Second threshold ,in It is a positive value, ranging from 0.3 to 0.5; The value is negative, ranging from -0.3 to 0. Each threshold is adjusted according to different growth stages of corn. The threshold can be adjusted and determined within the above range. For example, a stricter threshold is used during the critical water demand period (the large trumpet stage and the tasseling stage) to ensure sufficient water supply, while a more lenient threshold can be used during the seedling stage and the maturity stage to save water resources.
[0068] When the revised decision index is less than the second threshold When the decision index is greater than or equal to the second threshold, it is determined that irrigation is not needed or only an early warning is issued; And less than the first threshold When this occurs, a micro-supplementary irrigation command is executed, including adjusting the solenoid valve opening to 35% of the original setting, shortening the irrigation duration to 1 / 3 of the original setting, and reducing the drip irrigation tape working pressure to 0.08 MPa. When the corrected decision index is greater than or equal to the first threshold... At the same time, an active supplemental irrigation strategy is implemented, including increasing the opening of the solenoid valve to 170% of the original setting, implementing zoned rotational irrigation according to the spatial distribution differences of soil moisture, and adjusting the basic water requirement threshold according to the current growth stage of corn.
[0069] According to some embodiments, when performing active supplemental irrigation, the demonstration field is divided into several irrigation zones based on the gridded monitoring data of the soil moisture sensor array. The zoning principle is: when the difference in soil volumetric moisture content between adjacent monitoring nodes exceeds 5%, they are divided into different zones. Each zone is irrigated sequentially according to water demand priority, and drip irrigation tape is evenly distributed within the same zone. The working pressure is adjusted to 0.12 MPa to ensure that the irrigation uniformity reaches more than 85%.
[0070] In S111, the decision execution results are recorded, and the decision effect is evaluated based on the prediction deviation. When the deviation exceeds the preset threshold, the model parameters are updated.
[0071] According to some implementation methods, the time, data, decision parameters, and execution results of each re-irrigation decision are recorded, including the decision time, current growth stage, real-time monitored soil volumetric water content, weather forecast accuracy, LSTM model error, and cwpc value. Values, SynDCI values, uncertainty weights, decision execution results, etc. When the prediction deviation exceeds a preset threshold, a model parameter update process is automatically triggered, including updating the real-time weight adjustment coefficients. and the collaborative weighting coefficient .
[0072] 24 hours after the decision is implemented, soil moisture measurement data are obtained. The predicted and measured values of soil volumetric water content at the same monitoring node at the same time point are compared to calculate the prediction deviation. δ=(∣θv pred - θv meas ∣ / θv meas )×100%, Where θv pred θv represents the soil volumetric water content predicted by the model. meas This represents the volumetric water content measured by the soil moisture sensor.
[0073] The prediction threshold is matched to the measurement accuracy of the soil moisture sensor, and is set to 0.8 to 1.2 times the sensor's accuracy. When the prediction deviation exceeds the preset threshold, the model parameter update process is automatically triggered. If the deviation is manifested as a lower-than-expected increase in soil moisture content after irrigation, the corresponding parameter for stem change is increased. This enhances the contribution of stem signals; if it manifests as a lag in drought early warning, it increases the corresponding evaporation rate. If the water demand assessment under high-temperature stress deviates from the measured value, then adjust the corresponding temperature. If the current state assessment bias dominates the decision error, then the error will increase. If the deviation in future trend prediction dominates, then it will be improved. The updated parameters take effect immediately, allowing the model to gradually adapt to local planting conditions.
[0074] The present invention also provides a system for intelligent supplemental irrigation of dryland maize based on dynamic adaptation and uncertainty perception, for implementing the methods of the above embodiments.
[0075] See Figure 5The intelligent irrigation system for dryland maize includes an uncertainty-aware data fusion module 01, a dynamic adaptive water demand assessment module 02, a dynamic infiltration prediction module 03, a meteorological-crop collaborative decision-making module 04, an uncertainty-aware decision execution module 05, and a model self-optimization module 06. The system's workflow architecture is described below. Figure 6 The same methods described above will not be repeated here.
[0076] The uncertainty-aware data fusion module 01 is used to collect multi-source data in real time through a meteorological monitoring sensor array, a soil moisture sensor array, and a crop physiological sensor array, and to evaluate the accuracy of weather forecasts and the error of LSTM prediction models based on rolling verification. Specifically, the meteorological monitoring sensor array includes an atmospheric temperature sensor, an atmospheric humidity sensor, a wind speed sensor, a solar radiation sensor, and an evaporation sensor, used to collect atmospheric temperature, humidity, wind speed, solar radiation, and evaporation; the soil moisture sensor array includes a soil volumetric water content sensor, a soil temperature sensor, and a soil conductivity sensor, arranged in a grid at different depths in the maize root zone, used to collect soil volumetric water content, temperature, and conductivity; the crop physiological sensor array includes a stem diameter change sensor and a leaf surface temperature sensor, used to collect stem diameter changes and leaf surface temperature.
[0077] The dynamic adaptive water demand assessment module 02 determines the temperature influence coefficient and growth period adjustment factor based on the current growth stage of maize, iteratively updates the real-time weight adjustment coefficient based on historical decision feedback, and calculates the maize water demand-rainfall coupling index cwpc through the dynamic weight coefficient. The dynamic infiltration prediction module 03 updates the dynamic infiltration time constant based on real-time soil moisture content and calculates the predicted available rainfall P. av The meteorological-crop collaborative decision-making module 04 is used to predict crop water demand trends and rainfall replenishment capacity in future periods, calculate water surplus / deficit forecasts, and generate the SynDCI collaborative decision-making index. The uncertainty-aware decision execution module 05 is used to calculate comprehensive uncertainty weights based on meteorological forecast accuracy and prediction model errors, correct the collaborative decision-making index, and execute tiered irrigation strategies. The model self-optimization module 06 is used to record the decision execution results and corresponding measured soil moisture values, compare prediction deviations, and update the dynamic weight adjustment coefficient and collaborative weight coefficient when the deviation exceeds the limit.
[0078] According to some embodiments, the system further includes a regional feature library, which stores soil texture parameters, peak daily water demand, water demand modulus coefficient, and water sensitivity index for different ecological types. The dynamic adaptive water demand assessment module and the dynamic infiltration prediction module call the corresponding parameters in the regional feature library according to the current ecological type.
[0079] Furthermore, taking Liaoning Province as an example, located in Northeast China, it boasts diverse ecological types, spanning multiple climate zones from semi-arid to humid, from the northwestern sandy area to the eastern humid area. Significant differences exist in rainfall characteristics, soil conditions, and evaporation rates across these different regions. The water requirement for maize throughout its entire growth period is not a constant but exhibits a complex dynamic pattern as the growth stage progresses and meteorological conditions change. The aforementioned regional feature database stores data for three ecological types: the northwestern sandy area, the northern black soil area, and the eastern humid area. Specifically, the soil texture attenuation coefficient for the northwestern sandy area is 0.25–0.35, for the northern black soil area it is 0.45–0.55, and for the eastern humid area it is 0.75–0.85.
[0080] This invention adapts three sets of independent dynamic weighting coefficients to three types of parameters: stem variation, evaporation, and temperature stress. The weights are adjusted in real-time based on the growth stage and historical feedback, emphasizing stem signals during seedling emergence, evaporation signals during jointing, and temperature signals during grain filling, significantly improving the accuracy of water requirement assessment. The dynamic infiltration time constant is updated in real-time with soil moisture content, accurately depicting the physical mechanism of slower infiltration in dry soil and faster infiltration in moist soil. Combined with differentiated settings of soil texture attenuation coefficient and vegetation interception coefficient, the effective rainfall prediction accuracy is improved by more than 15% compared to fixed-parameter models. Uncertainty-aware decision-making: Meteorological forecast errors and LSTM model errors are integrated into a comprehensive uncertainty weight. When prediction reliability is low, the decision sensitivity is automatically reduced, achieving a fundamental shift from "deterministic decision-making" to "risk-aware decision-making," reducing decision-making errors under extreme weather conditions.
[0081] The present invention will be further described in detail below with reference to specific embodiments, but the scope of protection of the present invention is not limited to the following embodiments.
[0082] Example 1: Application scenario in the windy and sandy area of northwestern Liaoning.
[0083] This example was applied to a thousand-mu dryland corn demonstration field in the windy and sandy area of northwestern Liaoning Province. The average annual rainfall is 450-550mm, the soil is mainly sandy loam, which has good air permeability but weak water retention capacity. The corn variety planted is the drought-resistant type that is mainly promoted in the local area, and the entire growth period is about 130 days.
[0084] The construction of the data acquisition and uncertainty perception decision execution module is aimed at the meteorological monitoring sensor array: three standard meteorological stations are set up in the demonstration area, equipped with air temperature sensors (platinum resistance, -40~60℃, accuracy ±0.1℃), air humidity sensors (capacitive, 0~100%RH, accuracy ±2%), wind speed sensors, solar radiation sensors and evaporation sensors, with a sampling period of 10 minutes.
[0085] Soil moisture sensor array: 100 monitoring nodes are arranged with a grid spacing of 100m×100m. Each node is equipped with a time-domain reflectometry soil moisture sensor (accuracy ±2% volumetric water content), a soil temperature sensor and a soil conductivity sensor. The probes are buried at four depths of 10cm, 20cm, 40cm and 60cm.
[0086] Crop physiological sensor array: A stem diameter change sensor (accuracy 0.01 mm) and an infrared leaf surface temperature sensor (accuracy ±0.5 ℃) were installed at 10 representative observation points.
[0087] Uncertainty assessment: A rolling verification method is used to compare the weather forecast values with the observed values over the past 72 hours to calculate σ. forecast ; RMSE statistics based on historical validation data of the LSTM model LSTM Current input data quality score Q t The average quality score Q of the training data is calculated based on sensor integrity rate and data continuity. train These are historical statistics.
[0088] In the dynamic infiltration prediction module, the growth period adjustment factor and temperature influence coefficient are determined based on the current growth stage of maize.
[0089] The entire growth period of maize is divided into six stages: emergence (15 days), jointing (25 days), tasseling (20 days), grain filling (15 days), and maturity (25 days). The temperature influence coefficient β1 and the regulating factor values for each growth stage are shown in the table below:
[0090] Basic weighting coefficient Real-time weight adjustment coefficient The initial value is 1.0, and it is updated iteratively based on historical decision feedback.
[0091] Taking the large funneling stage as an example, assume the measured parameters at a certain moment are as follows: atmospheric temperature Ta = 28℃, evaporation rate E = 0.4 mm / h, and the volumetric water content of the 20cm soil layer. =30%, field holding capacity =40%, leaf surface temperature =30℃, relative humidity Rh=65%, stem diameter change rate =0.02mm / h, and the real-time weight adjustment coefficient is 1.0.
[0092] Calculate the dynamic weighting coefficients: =0.4×1.0×0.8=0.32; =0.4×1.0×1.3=0.52; =0.2 × 1.0 × 1.2 = 0.24 It can be seen that during the large funnel-shaped mouth stage, the relevant parameters of evaporation... The weight was the highest (0.52), followed by parameters related to stem changes. (0.32), and finally, the parameters related to leaf surface temperature. (0.24). This weight distribution is consistent with the technical characteristics of maize water demand during the large trumpet stage, which is mainly driven by transpiration.
[0093] Calculate the cwpc integral components: Stem weight =
[0094] Evaporation amount =
[0095] Temperature component =
[0096] In the dynamic infiltration prediction module, the soil texture in this embodiment is mainly sandy loam, with a sandy loam texture attenuation coefficient k=0.30 and a foundation infiltration time constant. =1.0h.
[0097] Formula for calculating dynamic infiltration time constant:
[0098] when =15% (drying), =1.19h, infiltration slows down; when =30% (moistened), =1.08h, infiltration speed is increased.
[0099] Soil infiltration coefficient K soil =1-0.2×(θ v / θ FC Vegetation interception coefficient K veg =1-0.1×LAI, where LAI is determined by referring to a table based on the reproductive period.
[0100] Combined with the predicted available rainfall P av The calculation formula assumes two rainfall events within the next 48 hours, with a forecast of 20mm of rainfall 12 hours later and 15mm 36 hours later. The soil is dry during the first rainfall event. τ inf =1.19h, effective infiltration ratio = 1 - exp(-12 / 1.19) ≈ 0.9999; the soil was already moist before the second rainfall, let θ v =28%, τ inf =1.09h, effective infiltration ratio ≈1.0. Therefore, P av =20×0.9999+15×1.0≈35.0mm.
[0101] In the meteorological-crop collaborative decision-making module, the crop water demand trend (ET) for the next 72 hours is predicted. future The crop coefficient for maize at its current growth stage is calculated as follows: 0.6 for emergence, 0.8 for jointing, 1.1 for tasseling, 1.2 for grain filling, and 0.7 for maturity. The average reference evapotranspiration for the next 72 hours is also provided. =5.0 mm / d, crop coefficient Kc=1.1 at the large trumpet stage, then the future water requirement ET future =1.1×5.0=5.5mm / d, the cumulative water requirement in 72 hours is 16.5mm.
[0102] The predicted water surplus / deficit value is ΔWater = 16.5 - 35.0 = -18.5 mm (water surplus).
[0103] The decision index for upstream water demand risk is cid = [cwpc / C] × (1-P). av / P threshold Normalized baseline parameter C=1.2 (large funnel stage), effective rainfall utilization threshold P threshold =28mm (Recommended value for windy and sandy areas in northwestern Liaoning).
[0104] Assuming cwpc=0.5, then cid=0.5 / 1.2×(1-35 / 28)=-0.125.
[0105] According to the rules of the moisture surplus / deficit prediction function: When ΔWater<0: Synergy(ΔWater)=1-exp(-ΔWater / ET threshold ) ΔWater = -18.5mm < 0, let ET threshold =5.0mm, then Synergy=1-exp(18.5 / 5.0)≈-39.4.
[0106] Collaborative weighting coefficient Then the SynDCI (Syn-Decision Index) is approximately 0.6 × (-0.125) + 0.4 × (-39.4) ≈ -15.8.
[0107] In the uncertainty-aware decision-making and execution module, the comprehensive uncertainty weight γ(t) is calculated, assuming σ forecast =0.15, σ LSTM =0.02, then the overall uncertainty weight γ(t) = 1 / (1 + 0.152 + 0.022) ≈ 0.978 Corrected decision index cid final =0.978×(-15.2)≈-14.9.
[0108] The thresholds are set according to different growth stages, for the conventional growth stage: = 0.3~0.5, = -0.3~0; for the critical water requirement stage (the large bell-mouth stage and the tasseling stage): = 0.4, = -0.2.
[0109] Set according to the graded supplementary irrigation threshold (large bell-mouth stage), since cid in this embodiment final <D2, the system determines that irrigation is not required, and records the decision log.
[0110] If the active supplementary irrigation execution strategy is implemented: when cid final ≥ , increase the solenoid valve opening to 170% of the original set value, implement zonal rotational irrigation according to the spatial distribution difference of soil moisture, and adjust the basic water requirement threshold according to the current growth stage of corn. The zonal rotational irrigation rule is: when the difference in soil volumetric water content between adjacent monitoring nodes exceeds 5%, they are divided into different zones; each zone is irrigated sequentially according to the water requirement priority, which is determined by the difference between the current soil water content and the field water holding capacity; the drip irrigation belts are evenly arranged in the same zone, and the working pressure is adjusted to 0.12MPa to ensure that the irrigation uniformity reaches more than 85%.
[0111] If the trace supplementary irrigation execution strategy is implemented: when , adjust the solenoid valve opening to 35% of the original set value, shorten the irrigation duration to 1 / 3 of the original set value, and reduce the working pressure of the drip irrigation belt to 0.08MPa.
[0112] The model self-optimization module records the parameters of this decision: decision time, growth stage (large bell-mouth stage), measured volumetric water content of 20cm soil layer 32%, meteorological forecast confidence, LSTM error, cwpc value, P av value, SynDCI value, uncertainty weight, decision result (no irrigation).
[0113] Obtain the measured rainfall data (34.8mm) and soil moisture change data 24 hours later, the prediction deviation = (measured change amount - predicted change amount) / predicted change amount ≈ 1.2%, which is less than the preset threshold of 2% (the sensor accuracy in the windy sand area of northwest Liaoning is ±2%, and the threshold is set to 2%), and the system keeps the existing parameters unchanged. If the deviation exceeds 3%, the parameter update process is triggered to adjust the real-time weight adjustment coefficient and the collaborative weight coefficient .
[0114] Embodiment 2: Application scenario in the black soil area of northern Liaoning.
[0115] The purpose of this embodiment is to verify the adaptability and effectiveness of the technical solution of the present invention in the Liaobei black soil region (loam, semi-humid climate), which is characterized by strong soil water retention capacity and large spatial variability. The focus is on examining the prediction accuracy of the dynamic infiltration model on soils with medium water retention capacity, the sensor grid layout strategy under soil spatial variability conditions, and the matching relationship between the model self-optimization threshold and sensor accuracy, to prove the feasibility of the regional feature library and differentiated configuration strategy.
[0116] Experimental area: A thousand-mu demonstration field in the black soil region of northern Liaoning Province, with an average annual rainfall of 550-650 mm and the soil is mainly black soil.
[0117] The main differences from Example 1 are: soil texture attenuation coefficient k = 0.50 (loam), loam infiltration rate is lower than sandy soil, and foundation infiltration time constant. =1.2h; the grid spacing of the soil moisture sensor was adjusted to 80m×80m (soil spatial variation is large in the black soil area).
[0118] According to parameters in the regional characteristic database, the water modulus coefficient for maize cultivation in the Liaobei black soil region is 6% at the seedling stage, 18% at the jointing stage, 28% at the tasseling stage, 22% at the tasseling stage, 17% at the grain-filling stage, and 9% at the maturity stage. The peak daily water requirements for each growth stage are as follows: 3.5 mm / d at seedling stage, 5.0 mm / d at jointing stage, 6.5 mm / d at tasseling stage, 6.0 mm / d at tasseling stage, 5.5 mm / d at grain-filling stage, and 4.0 mm / d at maturity stage.
[0119] The dynamic infiltration prediction module considers the water retention characteristics of black soil and adds an upper limit constraint on soil moisture content. The threshold for tiered irrigation is set during the tasseling stage. =0.45、 =-0.15, other reproductive periods refer to Example 1.
[0120] The prediction deviation threshold is set at 3% (sensor accuracy ±3% in black soil areas). When the soil moisture content exceeds 90% of the field capacity, the excess rainfall is not included in the effective rainfall to prevent over-irrigation.
[0121] Example 3: Application scenario in the humid region of Liaodong.
[0122] This embodiment aims to verify the adaptability and effectiveness of the technical solution of the present invention in the humid region of Liaodong (clay, rainy climate). It focuses on examining the prediction accuracy of the dynamic infiltration model on soils with high water retention and low infiltration rate, the drainage demand assessment function in rainy scenarios, and the effect of strict threshold setting in avoiding over-irrigation, thus proving the applicability and economy of the technical solution in humid agricultural areas.
[0123] Experimental area: A thousand-mu demonstration field in the humid eastern region of Liaoning Province, with an average annual rainfall of 800-1000 mm, and the soil is mainly brown soil with abundant rainfall.
[0124] The main differences from Example 1 are: soil texture attenuation coefficient k = 0.80 (clay), and foundation infiltration time constant. =1.5h; the grid spacing of the soil moisture sensor was increased to 120m×120m (soil conditions are uniform and spatial variation is small).
[0125] Regional characteristic parameters: water demand model coefficient during the large funnel-shaped mouth period is 22%, and the peak daily water demand is 5.5 mm / d.
[0126] Tiered irrigation thresholds: Due to abundant rainfall in this region, irrigation is mainly used to supplement water deficits during drought periods; therefore, stricter thresholds are adopted to avoid over-irrigation. (Large trumpet stage) =0.50、 =-0.10, other reproductive periods =0.45、 =-0.20 (using a stricter threshold to avoid over-irrigation).
[0127] Add a drainage demand assessment function. When the predicted rainfall exceeds the soil infiltration capacity, calculate the required drainage volume and activate the drainage valve. The drainage trigger condition is that the soil moisture content in the 20cm layer exceeds 85% of the field capacity for more than 12 hours.
[0128] The prediction deviation threshold is set to 3% (sensor accuracy ±3% in humid areas). The model self-optimization module is set to automatically trigger the weight parameter update process when the prediction deviation exceeds 3%. Newly collected measured data will be used as training samples to fine-tune the real-time weight adjustment coefficients.
[0129] The results of this embodiment demonstrate that the method and system of this invention can adapt to soil conditions in different regions. The dynamic adaptive and uncertainty perception technology framework applied to the intelligent irrigation system for dryland crops has broad agricultural application value and promotion prospects.
[0130] In summary, this invention achieves dynamic adaptation in water requirement assessment through dynamic weighting coefficients. Different parameter weights are used for different growth stages: stem changes are emphasized during seedling emergence, evaporation during jointing, and temperature stress is comprehensively considered during grain filling, thus improving assessment accuracy. A refined physical mechanism model of available rainfall is achieved through a dynamic infiltration time constant, which is updated in real time with soil moisture content, improving prediction accuracy. An uncertainty-aware decision-making mechanism incorporates meteorological forecast errors and model prediction errors into decision-making considerations, reducing decision risk. A meteorological-crop collaborative decision-making mechanism comprehensively considers current conditions and future trends, improving the foresight of irrigation decisions. A self-optimizing closed-loop model is established, continuously updating model parameters through feedback, solving the problem of fixed model parameters and inability to adapt to environmental changes in existing technologies.
[0131] Figure 7 A block diagram of a computing device according to an exemplary embodiment of the present invention is shown.
[0132] like Figure 7 As shown, the computing device 30 includes a processor 12 and a memory 14. The computing device 30 may also include a bus 22, a network interface card 16, and an I / O interface 18. The processor 12, memory 14, network interface card 16, and I / O interface 18 can communicate with each other via the bus 22. The computing device of the present invention can be applied to the above-described systems and methods.
[0133] Processor 12 may include one or more general-purpose CPUs (Central Processing Units), microprocessors, or application-specific integrated circuits, for executing relevant program instructions. According to some embodiments, computing device 30 may also include a high-performance display adapter (GPU) 20 for accelerating processor 12.
[0134] Memory 14 may include a machine system readable medium in the form of volatile memory, such as random access memory (RAM), read-only memory (ROM), and / or cache memory. Memory 14 is used to store one or more programs containing instructions, as well as data. Processor 12 may read the instructions stored in memory 14 to perform the methods described above according to embodiments of the present invention.
[0135] The computing device 30 can also communicate with one or more networks via the network interface card 16. The network interface card 16 can be a DPU smart network card.
[0136] Bus 22 can include address bus, data bus, control bus, etc. Bus 22 provides a path for exchanging information between components.
[0137] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.
[0138] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A smart irrigation method for dryland maize based on dynamic adaptation and uncertainty perception, characterized in that, include: Real-time collection of meteorological data, soil moisture data, and crop physiological data to assess the accuracy of weather forecasts and the error of prediction models; Based on the current growth stage of maize, the temperature influence coefficient and growth stage regulation factor are determined. The real-time weight adjustment coefficient is iteratively updated in combination with historical decision feedback. The weight coefficient of the water demand assessment model is dynamically adjusted, and the maize water demand-rainfall coupling index is calculated. The real-time weight adjustment coefficient is iteratively updated based on the feedback data of historical decisions. The dynamic infiltration time constant is updated based on the real-time soil moisture content, and the available rainfall is calculated and predicted based on the dynamic infiltration time constant. Based on future weather forecasts and crop coefficients, the crop water demand for future periods is predicted. Combined with the predicted rainfall, the water surplus / deficit forecast is calculated, and a meteorological-crop collaborative decision-making index is generated. Based on the accuracy of the weather forecast and the error of the long short-term memory neural network prediction model, a comprehensive uncertainty weight is calculated. The collaborative decision-making index is then corrected using the comprehensive uncertainty weight. A graded supplementary irrigation control strategy is then executed based on the corrected decision-making index. Record the results of decision execution and evaluate the effectiveness of the decision based on the prediction deviation. Update the model parameters when the deviation exceeds a preset threshold.
2. The method according to claim 1, characterized in that, The weighting coefficients of the dynamically adjusted water demand assessment model include: The dynamic weighting coefficients include three sets of coefficients: stem diameter change rate, evaporation, and crop temperature. Each set of dynamic weighting coefficients is determined by the product of the basic weighting coefficient, the growth period regulation factor, and the real-time weighting adjustment coefficient. The growth period regulation factor is set according to the corn's emergence stage, jointing stage, large trumpet stage, tasseling stage, grain filling stage, or maturity stage.
3. The method according to claim 1, characterized in that, The prediction can be calculated using the rainfall data based on the meteorological forecast rainfall, the soil infiltration coefficient, the vegetation interception coefficient, and the dynamic infiltration time constant. The dynamic infiltration time constant is dynamically updated according to the difference between the current soil moisture content and field capacity. When the soil is dry, the dynamic infiltration time constant increases and the infiltration slows down; when the soil is moist, the dynamic infiltration time constant decreases and the infiltration speeds up. The prediction can utilize rainfall P av The calculation formula is: in, To predict the amount of rainfall at a given time, K soil K is the soil infiltration coefficient, reflecting the degree to which the current state of the soil hinders rainfall infiltration, and its value ranges from 0 to 1. veg τ is the vegetation interception coefficient, reflecting the crop canopy's ability to intercept rainfall, with a value ranging from 0 to 1. inf (t i ) represents the dynamic infiltration time constant; The dynamic infiltration time constant τ inf (t i Updated in real time using the following formula: in, The basic infiltration time constant is defined as 0.8–1.5 hours; k is the soil texture attenuation coefficient; and θ is the soil texture attenuation coefficient. v (t i θ represents the soil volumetric water content at the predicted time. FC It refers to field water holding capacity.
4. The method according to claim 1, characterized in that, The generation process of the meteorological-crop collaborative decision-making index includes: Calculate a decision index that reflects the current water demand risk; The water surplus / deficit forecast function is used to transform the water surplus / deficit forecast values for future periods into decision signals. The collaborative decision-making index is obtained by weighted summation of the decision index and the decision signal. Specifically, when the water surplus / deficit prediction value is a water surplus, the output value of the water surplus / deficit prediction function is negative in order to reduce the collaborative decision-making index. The SynDCI (Synthesis Decision Index) is calculated using the following formula: Wherein, cid is the decision index reflecting the current water demand risk, ΔWater is the predicted water surplus / deficit value for the future period, and Synergy(ΔWater) is the water surplus / deficit prediction function. This refers to the collaborative weighting coefficient.
5. The method according to claim 1, characterized in that, The comprehensive uncertainty weight γ(t) is calculated using the following formula: γ(t) = 1 / (1 + σ forecast (t)² + σ LSTM (t)² Where, σ forecast (t) represents the uncertainty of weather forecast, σ LSTM (t) represents the uncertainty of the prediction model error.
6. The method according to claim 1, characterized in that, The step of implementing a tiered irrigation control strategy based on the revised decision index includes: A first threshold is set based on the current reproductive period. Second threshold ,in It is a positive value. It is a negative value; When the revised decision index is less than the second threshold At that time, it is determined that irrigation is not needed or only an early warning is issued; When the revised decision index is greater than or equal to the second threshold And less than the first threshold At that time, execute the micro-irrigation command; When the revised decision index is greater than or equal to the first threshold At that time, an active supplemental irrigation strategy is implemented.
7. A smart irrigation system for dryland maize implementing the method as described in any one of claims 1 to 6, characterized in that, include: The uncertainty perception data fusion module is used to collect multi-source data in real time through meteorological monitoring sensor arrays, soil moisture sensor arrays and crop physiological sensor arrays, and evaluate the accuracy of meteorological forecasts and the error of long short-term memory neural network prediction models based on rolling verification. The dynamic adaptive water demand assessment module is used to determine the temperature influence coefficient and growth period adjustment factor based on the current growth stage of maize, and iteratively update the real-time weight adjustment coefficient by combining historical decision feedback. The maize water demand-rainfall coupling index is calculated through the dynamic weight coefficient. The dynamic infiltration prediction module is used to update the dynamic infiltration time constant based on the real-time soil moisture content and calculate and predict the available rainfall. The meteorological-crop collaborative decision-making module is used to predict crop water demand trends and rainfall replenishment capacity in future periods, calculate water surplus and deficit forecasts, and generate collaborative decision-making indices. The uncertainty perception and decision execution module is used to calculate the comprehensive uncertainty weight based on the accuracy of weather forecasts and the error of prediction models, correct the collaborative decision index, and execute the graded irrigation strategy. The model self-optimization module is used to record the decision execution results and the corresponding measured soil moisture values, compare the prediction deviation, and update the dynamic weight adjustment coefficient and the synergistic weight coefficient when the deviation exceeds the limit.
8. The system according to claim 7, characterized in that, Also includes: A regional feature database, which stores soil texture parameters, peak daily water demand, water demand modulus, and water sensitivity index for different ecological types of regions; The dynamic adaptive water demand assessment module and the dynamic infiltration prediction module call the corresponding parameters from the regional feature library according to the current ecological type zone.
9. The system according to claim 7, characterized in that, The uncertainty-aware data fusion module includes: The meteorological monitoring sensor array includes an atmospheric temperature sensor, an atmospheric humidity sensor, a wind speed sensor, a solar radiation sensor, and an evaporation sensor, used to collect atmospheric temperature, humidity, wind speed, solar radiation, and evaporation. The soil moisture sensor array includes a soil volumetric water content sensor, a soil temperature sensor, and a soil electrical conductivity sensor, which are arranged in a grid pattern at different depths in the corn root zone to collect soil volumetric water content, temperature, and electrical conductivity. The crop physiological sensor array includes a stem diameter change sensor and a leaf surface temperature sensor, used to collect data on stem diameter changes and leaf surface temperature.
10. A computing device, characterized in that, include: processor; as well as A memory storing a computer program that, when executed by the processor, causes the processor to perform the method as described in any one of claims 1 to 6.
