Dynamic calculation method for water demand of crops and intelligent irrigation system

Abstract

The invention provides a crop water demand dynamic calculation method and an intelligent irrigation system, and relates to the technical field of agricultural irrigation. According to the crop water demand dynamic calculation method, a quantum dot fluorescent probe array is adopted, the soil moisture activity is accurately obtained with the spatial resolution of 0.1 mm, and compared with a traditional sensor, the precision is improved by 3-5 times; a terahertz spectrum technology is used for analyzing osmotic pressure of leaf cell sap, nondestructive testing of crop physiological water demand signals is achieved, destructive sampling is avoided, a multi-mode fusion model is used in combination with transfer learning and quantum correction factors, the water demand calculation error is smaller than or equal to 3%, the dynamic response speed is increased to the hour level, and a block chain technology ensures that data are real. The flow control precision of the magneto-rheological actuating mechanism reaches + / -2% FS, and the water saving rate exceeds 30%.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of agricultural irrigation technology, in particular to a crop water requirement dynamic calculation method and an intelligent irrigation system. BACKGROUND

[0002] In agricultural production, accurate calculation of crop growth water requirement is the core prerequisite for scientific irrigation, and the irrigation system based on accurate water requirement is the key to improve water resource utilization efficiency and ensure high-quality and high-yield crops. As water resources are indispensable basic resources for agricultural production, their reasonable use is directly related to the sustainable development of agriculture. With the increasingly prominent global water shortage problem and the promotion of agricultural scale and intensive planting.

[0003] Currently, there are significant limitations in crop water requirement calculation and irrigation control technology: traditional methods rely on soil moisture sensors (such as time domain reflectometry) and empirical formulas (such as Penman-Monteith), soil moisture measurement is mostly single-point volumetric water content with a spatial resolution of ≤5cm, which is difficult to reflect the water heterogeneity in the root zone microdomain; crop physiological water requirement signals are obtained by destructive sampling (such as leaf water potential measurement), which cannot be continuously monitored; existing irrigation systems have problems such as single data dimension and static calculation model: environmental parameter collection mostly uses traditional weather stations, and the spatial and temporal resolution of wind speed, solar radiation, etc. is insufficient; water requirement calculation does not consider the influence of crop variety specificity and soil porosity, and the dynamic response lag is ≥24 hours; at the same time, the magneto-rheological irrigation actuator mostly uses electromagnetic valves or frequency conversion water pumps, and the flow regulation precision is ≤±5%FS, and the data transmission process is easily disturbed, resulting in distorted calculation results, causing more than 30% of water resources to be wasted, which cannot meet the demand of dynamic and multi-dimensional water requirement sensing in precision agriculture. SUMMARY

[0004] Based on the technical problems in the background art, the present application proposes a crop water requirement dynamic calculation method and an intelligent irrigation system to solve the problems in the background art.

[0005] The present application proposes a crop water requirement dynamic calculation method, which comprises: S1, using a quantum dot fluorescence probe array to collect multi-dimensional water characteristic parameters of crop root zone soil, the quantum dot fluorescence probe array is composed of CdSe / ZnS core-shell structure quantum dots and mesoporous silica carriers, and soil water data is obtained through the nonlinear mapping relationship between fluorescence intensity decay rate and soil water activity; S2, scanning the crop canopy leaves using a terahertz time domain spectroscopy system to obtain the characteristic absorption peak intensity in the 0.3-3THz frequency band, and analyzing the crop physiological water requirement signal by establishing a quantitative relationship model between the terahertz absorption coefficient and the leaf cell sap osmotic pressure; S3, deploy a millimeter wave radar-based environment field monitoring module to collect environmental parameters for crop growth: real-time wind speed, atmospheric pressure, and solar radiation flux density, and the millimeter wave radar works in the 77GHz frequency band; S4, input the acquired soil moisture data, crop physiological water demand signal, and environmental parameters into a multi-modal fusion calculation model based on transfer learning, the multi-modal fusion calculation model corrects the traditional Penman-Monteith formula by introducing a quantum tunneling effect correction factor, and outputs the dynamic water demand value of the crop on an hourly scale; S5, use a blockchain node to perform real-time hash verification on the original data and intermediate results in the calculation process to ensure that the water demand calculation result is tamper-proof.

[0006] Preferably, in S1, the excitation wavelength of the quantum dot fluorescence probe array is 365-405nm, the emission wavelength is 520-680nm, the spatial resolution is not less than 0.1mm, and the probe surface is modified with soil colloid specific recognition groups.

[0007] Preferably, in S2, the time resolution of the terahertz time domain spectroscopy system is ≤5fs, the intercellular water transport rate is calculated by the propagation time delay difference of terahertz waves in leaf tissues, and the quantitative relationship model is trained by an improved support vector machine algorithm, and the input features include 128 terahertz absorption peak characteristic values.

[0008] Preferably, in S4, the source domain data of the transfer learning comes from a crop water demand dataset in a laboratory controllable environment, and the target domain is the actual growth environment data in the field. The distribution alignment is realized through an adversarial domain adaptation network, and the calculation formula of the quantum tunneling effect correction factor is: wherein k is the quantum tunneling effect correction factor, a is the soil porosity correction coefficient, β is the crop variety-specific parameter, E is the real-time monitored soil water potential, and E0 is the baseline water potential value.

[0009] Preferably, in S5, the blockchain node adopts a consortium chain architecture, includes 3 or more consensus nodes, is respectively deployed in a meteorological station, a soil monitoring terminal, and a crop growth monitoring terminal, and uses a practical Byzantine fault tolerance algorithm for the consensus mechanism. The original data and intermediate results in the calculation process are verified by the blockchain node in real time, the consensus verification is realized when the water demand calculation result is tamper-proof, and the block generation interval is ≤30s.

[0010] Preferably, it further includes S6, temperature compensation of the collected soil moisture data based on the temperature sensitivity of the quantum dot fluorescence probe, and the compensation formula is: wherein W t is the compensated water content, W0 is the original measured value, γ is the temperature coefficient, T is the real-time temperature, and T0 is the calibration temperature.

[0011] The application also provides a crop water requirement dynamic intelligent irrigation system, which comprises: A data acquisition layer comprising an environmental field monitoring module of a quantum dot fluorescent probe array, a terahertz time-domain spectroscopy system and a millimeter wave radar; An edge computing gateway for receiving raw data of the data acquisition layer and performing a multi-modal fusion calculation model; An irrigation control hub for receiving a dynamic water requirement value output by the edge computing gateway, generating a pulse width modulation signal based on a preset crop growth stage water requirement threshold value; A magneto-rheological irrigation actuator for adjusting irrigation flow and pressure by changing the yield stress of magneto-rheological fluid in response to the pulse width modulation signal, the magneto-rheological irrigation actuator being internally provided with a nanoscale flow sensor.

[0012] Preferably, the edge computing gateway adopts a heterogeneous computing architecture integrating FPGA and RISC-V processors, the FPGA is used for real-time Fourier transform of terahertz spectroscopy data, and the RISC-V processor runs a multi-modal fusion calculation model, and the data processing delay is ≤10ms.

[0013] Preferably, the magneto-rheological irrigation actuator comprises a ring-shaped electromagnetic coil and a deformable valve core, the coil current adjustment range is 0-2A, the corresponding irrigation flow adjustment range is 0-50L / h, and the flow control accuracy is ≤±2%FS.

[0014] Preferably, it further comprises a UAV inspection module, the UAV inspection module is provided with a terahertz imager, a large range of crop canopy is scanned, the scanning result is fused with the measurement result of the ground terahertz time-domain spectroscopy system, and the spatial distribution accuracy of the crop physiological water requirement signal is optimized.

[0015] The application has the following beneficial effects: the quantum dot fluorescent probe array is used to accurately obtain soil water activity with a spatial resolution of 0.1mm, the accuracy is improved by 3-5 times compared with traditional sensors; the terahertz spectroscopy technology analyzes the osmotic pressure of leaf cell sap, realizes non-destructive detection of crop physiological water requirement signal, and avoids destructive sampling.

[0016] The multi-modal fusion model combines transfer learning and quantum correction factors, so that the water requirement calculation error is ≤3%, the dynamic response speed is improved to the hour level, the blockchain technology ensures data authenticity, the flow control accuracy of the magneto-rheological actuator is ±2%FS, and the water saving rate is more than 30%.

[0017] The spatial accuracy is optimized by data fusion of UAV and ground, the processing delay of the heterogeneous computing gateway is ≤10ms, the real-time irrigation demand is met, and the time and space limitations and accuracy bottlenecks of traditional methods are fully broken through. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 A flowchart of a crop water requirement dynamic calculation method is provided for the present application. Figure 2 A structural block diagram of a crop water requirement dynamic intelligent irrigation system is provided for the present application. DETAILED DESCRIPTION

[0019] Reference Figure 1 With Figure 2 The present application provides a crop water requirement dynamic calculation method, and the calculation method is as follows: S1, a quantum dot fluorescence probe array is used to collect multi-dimensional water characteristic parameters of the crop root zone soil, and soil water data is obtained through a nonlinear mapping relationship between fluorescence intensity decay rate and soil water activity.

[0020] Among them, the multi-dimensional parameters can more comprehensively reflect the true situation of soil water, avoiding the limitations of single parameter measurement, thereby improving the accuracy and reliability of the soil water data; since the quantum dot fluorescence probe array has a spatial resolution of not less than 0.1mm, the multi-dimensional water characteristic parameters can capture the water change in the soil microdomain, which helps to understand the fine distribution of soil water in the root zone and can provide more detailed information for irrigation; different soils have different physical and chemical properties, and the multi-dimensional water characteristic parameters can consider the influence of these factors on soil water, and better adapt to various complex soil environments.

[0021] The quantum dot fluorescence probe array is composed of CdSe / ZnS core-shell structure quantum dots and mesoporous silica carriers, which is used to penetrate into the crop root zone and accurately capture the multi-dimensional water characteristic parameters in the soil. Its special structure can ensure stable operation in complex soil environment. The excitation wavelength of the quantum dot fluorescence probe array is 365-405nm, the emission wavelength is 520-680nm, the spatial resolution is not less than 0.1mm, and the probe surface is modified with soil colloid specific recognition groups. Through specific excitation wavelength and emission wavelength, the quantum dot fluorescence probe array can emit specific fluorescence. When contacting with soil water, the fluorescence intensity decays. By using the nonlinear mapping relationship between the fluorescence intensity and the soil water activity, combined with the high spatial resolution and the recognition groups on the surface, the soil water data can be more accurately obtained.

[0022] For example, in a certain soil environment, when the soil water activity changes, the fluorescence intensity of the quantum dot fluorescence probe array will change accordingly. By measuring this change and according to the pre-established nonlinear mapping model (experimental design: under different soil water activity conditions, use quantum dot fluorescence probe array to measure and obtain corresponding fluorescence intensity decay rate data; the soil water activity can be adjusted by controlling the water content, humidity and other factors of the soil; data collection: repeat the experiment under multiple different soil water activity levels and collect a large number of corresponding data pairs of fluorescence intensity decay rate and soil water activity; model selection and training: select a nonlinear mathematical model support vector regression, divide the collected data into training set and test set, use the training set to train the model, adjust the parameters of the model, so that the model can best fit the relationship between fluorescence intensity decay rate and soil water activity), the soil water content and other related data can be accurately calculated. The excitation wavelength of 365-405 nm and the emission wavelength of 520-680 nm can reduce the interference of other substances in the soil and improve the recognition of the fluorescence signal. The spatial resolution of not less than 0.1 mm can capture the water change in the soil microdomain. The specific recognition group of the surface modified soil colloid enhances the specific recognition ability of the soil water, further improving the accuracy and reliability of the soil water data.

[0023] S2, scanning the crop canopy leaves by using the terahertz time domain spectroscopy system to obtain the characteristic absorption peak intensity (0.3-3THz band, which refers to the intensity value of the characteristic absorption peak of the crop canopy leaves to the terahertz wave in the 0.3-3THz band, which is a single numerical value reflecting the strength of the absorption peak, directly related to the material composition and structure of the leaves) to realize the nondestructive detection of the crop physiological state, and to analyze the crop physiological water demand signal by establishing a quantitative relationship model between the terahertz absorption coefficient and the leaf cell sap osmotic pressure.

[0024] The time resolution of the terahertz time domain spectroscopy system is ≤5fs, the intercellular water transport rate is calculated by the propagation time delay difference of the terahertz wave in the leaf tissue, and the quantitative relationship model is trained by using the improved support vector machine algorithm, and the input features include 128 terahertz absorption peak feature values (the absorption peak feature value is a multidimensional vector that contains more information about the terahertz absorption peak and can more comprehensively describe the spectral characteristics of the leaves; when establishing the quantitative relationship model between the terahertz absorption coefficient and the leaf cell sap osmotic pressure, using 128 terahertz absorption peak feature values as input features can provide more data information for the model, thereby improving the accuracy and reliability of the model).

[0025] The improved support vector machine algorithm training specific operation is as follows: Calculate each terahertz absorption peak feature value (denoted as x iThe mutual information I(x i , y) between the 128 terahertz absorption peak features and the leaf cell sap osmotic pressure (denoted as y) is calculated, and sorted in descending order. ; P(x i , y) is the joint probability distribution, P(x i ), P(y) is the marginal probability distribution.

[0026] The top 30% of features with mutual information I(x i , y) ≥ threshold value are retained. First, the mutual information I(x i , y) corresponding to all 128 terahertz absorption peak feature values is calculated, and sorted in descending order of I(x i , y). The minimum mutual information value of the top 30% of features is calculated. The smaller of the 1.2 times the average mutual information of all features and the minimum mutual information value of the top 30% of features is taken as the final threshold value, ensuring that the I(x i , y) of the top 30% of features is ≥ the threshold value. The remaining 70% of features with mutual information < the threshold value are removed to achieve redundant feature removal.

[0027] From the 128 terahertz absorption peak feature values, the core features related to the leaf cell sap osmotic pressure are selected (usually 20-30 are retained), reducing the input dimension and improving the calculation efficiency.

[0028] The leaf cell sap osmotic pressure is divided into three intervals (low osmotic: < 0.5 MPa; medium osmotic: 0.5-1.5 MPa; high osmotic: > 1.5 MPa), each interval corresponds to different kernel function parameters. The hybrid kernel function "RBF kernel + polynomial kernel" is used, and the expression is: ; K(x, x j ) is the hybrid kernel function, which is designed to fit the segmented nonlinear relationship between the terahertz absorption peak feature values and the leaf cell sap osmotic pressure. α is the weight coefficient (low osmotic interval α = 0.3, medium osmotic interval α = 0.5, high osmotic interval α = 0.7, adjusted adaptively through training data); γ1 is the RBF kernel parameter (controls the local fitting accuracy), λ is the polynomial kernel degree (controls the global trend fitting, take 2-3), c is the offset (fixed as 1), x, x j are the 128 terahertz absorption peak feature vectors of two leaf samples (each dimension corresponds to the absorption peak intensity of 0.3-3 THz frequency band, i.e. "128 terahertz absorption peak feature values" in the invention).

[0029] The kernel function parameters are solved by the particle swarm optimization algorithm (PSO), with the goal of minimizing the mean average error (MAE) within the interval.

[0030] Aiming at the “piecewise nonlinear relationship between terahertz absorption coefficient and osmotic pressure”, a dynamic kernel function is designed to improve the fitting accuracy in different osmotic pressure ranges.

[0031] The input terahertz absorption peak characteristic value x i Assign weight w i (w i Mutual information with features I(x i ,y) is positively correlated, that is ); The improved loss function is: ; Where n is the total number of input terahertz absorption peak eigenvalues ​​(corresponding to 128 terahertz absorption peak eigenvalues ​​in S2, that is, n=128); x is the terahertz absorption peak eigenvector of a single crop canopy leaf sample (containing n terahertz absorption peak eigenvalues ​​x i , i=1,2,...,n), ε is set to 0.03MPa (i.e., the allowed osmotic pressure prediction error is ≤3%), and the weight w i High-information features (such as the 10.5 THz absorption peak directly related to the cell sap) account for a higher proportion in the loss calculation, and the influence of noise features is weakened. f(x) is the model's predicted value of x (this model is a quantitative relationship model trained based on the improved support vector machine (SVM) algorithm. Its core function is to predict the osmotic pressure of leaf cell sap through the input terahertz spectral features).

[0032] The training process of the improved support vector machine algorithm must meet clear constraints to ensure reproducibility: 1. Dataset: The source data are leaf samples grown in a controlled laboratory environment (500 groups, covering three crop varieties and five growth stages). Each group contains 128 terahertz absorption peak characteristic values ​​and measured cell osmotic pressure (measured using a freezing point osmometer with an accuracy of ±0.01 MPa).

[0033] 2. Parameter range: Penalty parameter C e [5, 20] (determined by 5-fold cross-validation, the optimal value is usually 10); the penalty parameter C is used to balance the control of the support vector machine model on the "training data fitting accuracy" and the "model generalization ability", for example: the larger the C value: the model punishes the error classification (or prediction deviation) in the training data more severely, and will try to fit all the training samples as much as possible, and may even over-learn the noise in the training data, leading to the model "over-fitting" (fitting very well for training data, but the prediction accuracy for new data decreases); the smaller the C value: the model is more tolerant to error classification, and tends to pursue a simple decision boundary, which may lead to "under-fitting" (lower fitting accuracy for both training data and new data).

[0034] RBF kernel parameter gamma 1 e [0.1, 1.0] (0.5 in the medium penetration interval, 0.8 in the high / low penetration interval); Polynomial kernel degree lambda = 2 (fixed, to avoid over-fitting caused by high-order terms).

[0035] 3. Training termination condition: the osmotic pressure prediction error MAE on the test set is less than or equal to 0.025 MPa (i.e. less than or equal to 2.5%, meeting the requirement that the overall water demand calculation error is less than or equal to 3%).

[0036] The terahertz time-domain spectroscopy system scans the leaf with a time resolution of ≤5 fs. When the terahertz wave propagates in the leaf tissue, different water content and physiological states will cause different propagation time delays. By measuring the time delay difference, the intercellular water transport rate is calculated. Combined with the quantitative relationship model trained by the improved support vector machine algorithm, the relationship between the terahertz absorption coefficient and the leaf cell sap osmotic pressure is established. The 128 terahertz absorption peak characteristic values of the leaf are obtained by the terahertz time-domain spectroscopy system, and input into the quantitative relationship model trained by the improved support vector machine algorithm. The process of converting the "leaf cell sap osmotic pressure value" output by the model into information that can directly reflect the crop water status (such as "mild water shortage", "moderate water shortage", "sufficient water", etc.) is to convert the abstract physiological parameter (osmotic pressure) into specific water demand signal that can guide irrigation decision-making.

[0037] The high time resolution of ≤5fs can accurately capture the subtle changes of terahertz waves in the leaves, improving the measurement accuracy; the 128 terahertz absorption peak characteristic values provide a rich data basis for the quantitative relationship model, and the improved support vector machine algorithm makes the model prediction more accurate, and can more accurately analyze the crop physiological water demand signal, for example, when the crop is in a water shortage state, the leaf cell sap osmotic pressure will change, and the propagation characteristics of terahertz waves in the leaf tissue will also change, and the absorption peak intensity and other characteristics detected by the terahertz time-domain spectroscopy system will also change, and through the quantitative relationship model analysis, the current physiological water demand signal of the crop can be obtained, realizing non-destructive and real-time monitoring, and ensuring the accuracy of the crop's own demand data.

[0038] S3, deploy a millimeter wave radar-based environment field monitoring module to collect environmental parameters for crop growth: real-time wind speed, atmospheric pressure and solar radiation flux density, and provide comprehensive environmental data. The millimeter wave radar works in the 77GHz frequency band and can accurately and real-timely obtain the above environmental parameters.

[0039] For example, through real-time monitoring of atmospheric pressure, the influence of the weather system on the crop growth environment can be understood, because the change of atmospheric pressure is often related to the change of weather, which in turn affects the water demand of crops; real-time monitoring of wind speed helps to evaluate the evaporation rate of water in the crop canopy and the soil surface, and when the wind speed is high, the water evaporation is accelerated, and the water demand of crops may increase; the solar radiation flux density directly affects the photosynthesis and transpiration of crops, and in turn affects the water demand of crops. By accurately measuring the solar radiation flux density, the water demand of crops under different light conditions can be more accurately calculated. By comprehensively collecting these environmental parameters, the richness and accuracy of the environmental data are ensured.

[0040] S4, input the obtained soil moisture data, crop physiological water demand signal and environmental data into a multi-modal fusion calculation model based on transfer learning, the model corrects the traditional Penman-Monteith formula by introducing a quantum tunneling effect correction factor, and outputs the dynamic water demand value of crops on an hourly scale.

[0041] The source domain data of transfer learning comes from the crop water demand data set in the laboratory controllable environment, and the target domain is the actual growth environment data in the field. The distribution alignment is realized through the adversarial domain adaptation network, so that the model can better adapt to the complex environment in the field. Transfer learning can use laboratory data to improve the adaptability of the model in the actual environment in the field, reduce the amount of data collection in the field and the training time of the model; the adversarial domain adaptation network realizes the distribution alignment of the source domain and the target domain data, and improves the generalization ability of the model; The calculation formula of the quantum tunneling effect correction factor is: Wherein, k is the quantum tunneling effect correction factor, a is the soil porosity correction coefficient, for example, the porosity of sandy soil is larger, and the water transport is faster, and the a value is different from the soil with smaller porosity such as clay; β is the crop variety specific parameter, different crop varieties have different physiological characteristics such as root structure and leaf stomatal density, and the demand and utilization efficiency of water are also different, and the parameter reflects the difference of water demand characteristics of different crop varieties; E is the real-time monitored soil water potential, which reflects the energy state of water in soil, and real-time monitoring of soil water potential can timely understand the dynamic change of soil water; E0 is the reference water potential value, which is used as a reference standard to calculate the correction factor. The quantum tunneling effect correction factor considers the soil porosity and crop variety specificity and other factors, so that the calculation of the modified traditional Penman-Monteith formula is more accurate, and the output of the hourly scale dynamic water demand value can timely reflect the change of crop water demand, and improve the timeliness and accuracy of irrigation.

[0042] For example, at a certain moment, the real-time soil water potential E is obtained by measurement, combined with the known soil porosity correction coefficient a, crop variety specificity parameter β and reference water potential value E0, the quantum tunneling effect correction factor k is calculated, which is substituted into the modified Penman-Monteith formula, so that the accurate hourly scale dynamic water demand value of the crop at that moment can be obtained.

[0043] S5, the original data and intermediate results in the calculation process are checked by the blockchain node in real time, so as to ensure that the water demand calculation result is tamper-proof.

[0044] The blockchain node adopts the alliance chain architecture, which contains more than 3 consensus nodes, which are respectively deployed in the meteorological station, the soil monitoring terminal and the crop growth monitoring terminal. The consensus nodes of the meteorological station, the soil monitoring terminal and the crop growth monitoring terminal respectively obtain their own original data and intermediate results, and the practical Byzantine fault tolerance algorithm is used for consensus verification, the data is checked in real time, and the checked result is stored on the blockchain, the block generation interval is ≤30s, so as to ensure that the data is tamper-proof (the consensus mechanism is used for the consensus verification of the original data and intermediate results in the calculation process by the blockchain node, so as to realize that the water demand calculation result is tamper-proof).

[0045] The alliance chain architecture guarantees the credibility and convenience of management of the blockchain nodes; the consensus nodes are distributed in different terminals, improving the reliability and security of data verification; the practical Byzantine fault tolerance algorithm can effectively deal with node failures and malicious attacks, ensuring the correctness of consensus; real-time hash verification and block generation interval ≤30s guarantee the real-time and tamper-proof nature of data, avoiding interference and distortion in the data transmission process, providing data security for the accuracy of water demand calculation results. For example, the real-time wind speed, atmospheric pressure and other environmental data collected by the meteorological station are uploaded to the blockchain network through the consensus nodes deployed by the station, and the soil moisture data collected by the soil monitoring terminal and the crop physiological water demand signal collected by the crop growth monitoring terminal are also uploaded through their respective consensus nodes.

[0046] S6, based on the temperature sensitivity of quantum dot fluorescent probes, temperature compensation is performed on the collected soil moisture data to eliminate the influence of temperature on the measurement results.

[0047] The compensation formula is: wherein W t is the compensated moisture content, W0 is the original measurement value, γ is the temperature coefficient, T is the real-time temperature, and T0 is the calibration temperature.

[0048] Considering the temperature sensitivity of quantum dot fluorescent probes, temperature compensation can eliminate the interference of temperature changes on soil moisture measurement results, making the obtained soil moisture data more accurate and reliable, providing more accurate basic data, and improving the calculation accuracy of the entire system. For example, at a certain moment, the real-time temperature T changes, if temperature compensation is not performed, the original measurement value W0 of the soil moisture measured by the quantum dot fluorescent probe may be deviated due to the influence of temperature. By using the compensation formula, the compensated moisture content W t is calculated by using the known temperature coefficient γ, calibration temperature T0 and real-time temperature T, which more truly reflects the actual soil moisture content and provides reliable soil moisture data.

[0049] In a specific application, a quantum dot fluorescent probe array is first used to obtain soil moisture data by leveraging the nonlinear mapping relationship between its fluorescence intensity decay rate and soil water activity. A terahertz time-domain spectroscopy system is then used to scan crop canopy leaves, and the crop's physiological water demand signals are analyzed based on an established quantitative relationship model between the terahertz absorption coefficient and the osmotic pressure of leaf cell fluid. The millimeter-wave radar's environmental field monitoring module then collects environmental parameters. This data is then input into a multimodal fusion calculation model, and a revised formula is used to calculate and output dynamic water demand values. Finally, data verification is performed through blockchain nodes to ensure the reliability of the results. The ability to collect data from multiple dimensions, covering the soil, the crop itself, and the environment, comprehensively reflects the crop's water demand. The use of quantum dot fluorescent probe arrays improves the accuracy of soil moisture data. The terahertz time-domain spectroscopy system enables non-destructive detection of crop physiological water demand signals, avoiding damage to crops caused by traditional sampling. The millimeter-wave radar's environmental field monitoring module can accurately and in real time acquire environmental parameters. The multimodal fusion calculation model improves the accuracy and dynamism of water demand calculations. Blockchain technology ensures the credibility of the data, ensuring scientific irrigation and significantly reducing water waste caused by inaccurate data.

[0050] Reference Figure 1 and Figure 2 The embodiment of the present invention further provides a dynamic intelligent irrigation system for crop water demand, including a data acquisition layer, an edge computing gateway, an irrigation control hub, and a magnetorheological irrigation actuator. Specifically: The data acquisition layer, which includes a quantum dot fluorescent probe array, a terahertz time-domain spectroscopy system, and an environmental field monitoring module of a millimeter-wave radar, is responsible for comprehensively and accurately collecting various types of data related to crop growth. It is the data source foundation of the entire system. The quantum dot fluorescent probe array, with its unique structure and performance, can penetrate deep into the crop root zone to accurately collect soil moisture data; the terahertz time-domain spectroscopy system performs non-destructive scanning of crop canopy leaves to obtain crop physiological water demand signals; the environmental field monitoring module of the millimeter-wave radar monitors parameters such as wind speed, atmospheric pressure, and solar radiation flux density of the crop growth environment in real time.

[0051] The edge computing gateway receives raw data from the data acquisition layer and executes multimodal fusion computing models to rapidly process data, reduce data transmission latency, and improve system response speed. The edge computing gateway utilizes a heterogeneous computing architecture, integrating FPGA and RISC-V processors. The FPGA performs real-time Fourier transforms on terahertz spectral data, while the RISC-V processor runs the multimodal fusion computing model. Data processing latency is ≤10ms. For example, when the data acquisition layer collects new data, the edge computing gateway can process and calculate the data in a very short time, promptly transmitting the results to the irrigation control center. This enables the irrigation system to quickly respond to changes in crop water demand and achieve precise irrigation.

[0052] An irrigation control hub receives dynamic water demand values output by an edge computing gateway, generates a pulse width modulation signal based on preset water demand thresholds for different crop growth stages, and controls the action of a magneto-rheological irrigation actuator. For example, when the dynamic water demand value exceeds the water demand threshold for the current crop growth stage, the irrigation control hub increases the duty cycle of the pulse width modulation signal to control the magneto-rheological irrigation actuator to increase irrigation flow and pressure to meet the water demand of the crop. Conversely, when the dynamic water demand value is below the threshold, the duty cycle of the pulse width modulation signal is reduced to reduce irrigation flow and pressure, avoiding waste of water resources.

[0053] The preset water demand thresholds for different crop growth stages in the above content are obtained by the following method: Collect laboratory controllable environment water demand data of target crops (such as wheat and corn) at different growth stages (seedling stage, jointing stage, filling stage, and mature stage) (determined by artificial weighing method and lysimeter method).

[0054] Collect historical data of actual field irrigation water demand of the same crop in the same region (combined with historical records of local weather stations and soil monitoring terminals).

[0055] Input the above data into the multi-modal fusion calculation model in S4, and determine the water demand threshold interval for each growth stage (such as the water demand threshold for the jointing stage of wheat is 10-15 L / ㎡・h) through model iterative optimization (with crop yield and water use efficiency as objective functions).

[0056] Adjust the threshold to match the actual water demand law of the crop through small-scale field verification (select 5% of the planting area to test the threshold adaptability).

[0057] A magneto-rheological irrigation actuator is used to respond to the pulse width modulation signal and adjust the irrigation flow and pressure by changing the yield stress of the magneto-rheological fluid. The magneto-rheological irrigation actuator has a nanoscale flow sensor built-in to achieve precise irrigation. The magneto-rheological irrigation actuator includes a ring-shaped electromagnetic coil and a deformable valve core. The coil current adjustment range is 0-2A, corresponding to an irrigation flow adjustment range of 0-50L / h, and the flow control accuracy is ≤±2%FS.

[0058] The irrigation control center generates a pulse width modulation signal input to the annular electromagnetic coil, the coil generates a magnetic field of corresponding intensity, the yield stress of the magneto-rheological fluid changes under the action of the magnetic field, the deformable valve core is pushed to change the opening degree, thereby adjusting the irrigation flow and pressure, the nanoscale flow sensor monitors the flow in real time and feeds back, the cooperation of the annular electromagnetic coil and the deformable valve core makes the flow regulation more flexible and accurate; the nanoscale flow sensor can monitor the flow in real time, realize closed-loop control, and improve the regulation accuracy, the flow control accuracy is ≤±2%FS, which is much higher than that of the traditional electromagnetic valve or variable frequency water pump; the coil current regulation range of 0-2A corresponds to the flow range of 0-50L / h, which can meet the irrigation needs of different crops and different growth stages, reduce water resource waste, and improve irrigation efficiency, for example, when the irrigation control center sends out a pulse width modulation signal to increase the irrigation flow, the annular electromagnetic coil current increases, a stronger magnetic field is generated, the yield stress of the magneto-rheological fluid changes, the opening degree of the deformable valve core increases, and the irrigation flow increases, at the same time, the nanoscale flow sensor monitors the flow change in real time and feeds back the flow data to the irrigation control center, if the flow does not reach the expected value, the irrigation control center will further adjust the pulse width modulation signal until the flow reaches the set value, realizing accurate irrigation, the precise flow control capability is much higher than that of the traditional electromagnetic valve or variable frequency water pump, the coil current regulation range of 0-2A corresponds to the flow range of 0-50L / h, which can meet the irrigation needs of different crops and different growth stages, effectively reduce water resource waste, and improve irrigation efficiency.

[0059] The crop water requirement dynamic intelligent irrigation system also comprises an unmanned aerial vehicle inspection module, the unmanned aerial vehicle is equipped with a terahertz imager, a large range of crop canopy scanning is realized, the scanning results are combined with the measurement results of the ground terahertz time-domain spectrum system, and the spatial distribution accuracy of the crop physiological water requirement signal is optimized.

[0060] The unmanned aerial vehicle inspection module can realize large range of crop canopy scanning and make up for the limitation of the measurement range of the ground terahertz time-domain spectrum system; the large-area data obtained by the terahertz imager is combined with the local detailed data on the ground, the spatial distribution of the crop physiological water requirement signal can be more comprehensively and accurately reflected, the accuracy of judging the water requirement conditions of crops in different regions is improved, more comprehensive basis for accurate irrigation is ensured, for example, in a large area of farmland, the ground terahertz time-domain spectrum system can only measure a limited number of points, while the unmanned aerial vehicle equipped with a terahertz imager can scan the crop canopy of the entire farmland and obtain large-area crop physiological water information, the large-range data obtained by the unmanned aerial vehicle scanning are combined with the local detailed data measured by the ground terahertz time-domain spectrum system, the water requirement conditions of crops in different regions can be more accurately judged, and more comprehensive basis is provided, through data fusion, it can be found that crops in some regions of the farmland may be in a water shortage situation, and these regions may be missed in ground measurement, so as to guide the irrigation system to carry out targeted irrigation on these regions, and improve the accuracy and effectiveness of irrigation.

[0061] The above description is only the preferred embodiment of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can make equivalent substitutions or changes according to the technical solution and the inventive concept of the present application within the technical range disclosed by the present application, which should be covered in the protection scope of the present application.

Claims

1. A method for dynamically calculating crop water requirements, characterized in that: The method includes: S1. Using a quantum dot fluorescent probe array to collect multi-dimensional moisture characteristic parameters of crop root zone soil, the quantum dot fluorescent probe array is composed of a composite of CdSe / ZnS core-shell quantum dots and a mesoporous silica carrier. Soil moisture data is obtained through a nonlinear mapping relationship between fluorescence intensity decay rate and soil water activity. S2. Use a terahertz time-domain spectroscopy system to scan crop canopy leaves, obtain the characteristic absorption peak intensity in the 0.3-3 THz frequency band, and analyze the crop physiological water demand signal by establishing a quantitative relationship model between the terahertz absorption coefficient and the osmotic pressure of the leaf cell fluid; S3. Deploy an environmental field monitoring module based on a millimeter-wave radar to collect environmental parameters for crop growth: real-time wind speed, atmospheric pressure, and solar radiation flux density. The millimeter-wave radar operates in the 77 GHz frequency band. S4. Inputting the acquired soil moisture data, crop physiological water requirement signals, and environmental parameters into a multimodal fusion computing model based on transfer learning. The multimodal fusion computing model modifies the traditional Penman-Monteith formula by introducing a quantum tunneling effect correction factor to output a dynamic water requirement value for the crop on an hourly scale. S5. Use blockchain nodes to perform real-time hash verification on the original data and intermediate results in the calculation process to ensure that the water demand calculation results cannot be tampered with.

2. A method for dynamically calculating crop water requirements according to claim 1, characterized in that: The quantum dot fluorescent probe array described in S1 has an excitation wavelength of 365-405 nm, an emission wavelength of 520-680 nm, a spatial resolution of not less than 0.1 mm, and the probe surface is modified with a soil colloid-specific recognition group.

3. A method for dynamically calculating crop water requirements according to claim 1, characterized in that: In S2, the time resolution of the terahertz time-domain spectroscopy system is ≤5fs. The intercellular water transfer rate is calculated by the propagation delay difference of the terahertz wave in the leaf tissue. The quantitative relationship model is trained using an improved support vector machine algorithm, and the input features include 128 terahertz absorption peak characteristic values.

4. A method for dynamically calculating crop water requirements according to claim 1, characterized in that: In S4, the source domain data for transfer learning comes from a crop water requirement dataset in a controlled laboratory environment, and the target domain is the actual field growth environment data. Distribution alignment is achieved through an adversarial domain adaptation network. The calculation formula for the quantum tunneling effect correction factor is: , where k is the quantum tunneling effect correction factor, α is the soil porosity correction coefficient, β is the crop variety-specific parameter, E is the real-time monitored soil water potential, and E0 is the benchmark water potential value.

5. A method for dynamically calculating crop water requirements according to claim 1, characterized in that: In S5, the blockchain nodes adopt a consortium chain architecture, which includes more than 3 consensus nodes, deployed in meteorological stations, soil monitoring terminals and crop growth monitoring terminals respectively. The consensus mechanism adopts a practical Byzantine fault-tolerant algorithm, which is used by blockchain nodes to perform real-time hash verification on the original data and intermediate results in the calculation process, and to achieve consensus verification when the water demand calculation results cannot be tampered with. The block generation interval is ≤30s.

6. A method for dynamically calculating crop water requirements according to claim 1, characterized in that: Also includes: S6. Based on the temperature sensitivity of the quantum dot fluorescent probe, the collected soil moisture data is temperature compensated. The compensation formula is: , where W t is the moisture content after compensation, W0 is the original measurement value, γ is the temperature coefficient, T is the real-time temperature, and T0 is the calibration temperature.

7. A crop water requirement dynamic intelligent irrigation system, according to a crop water requirement dynamic calculation method according to any one of claims 1 to 6, characterized in that: The system includes: The data acquisition layer includes a quantum dot fluorescent probe array, a terahertz time-domain spectroscopy system, and an environmental field monitoring module using millimeter-wave radar; Edge computing gateway, used to receive raw data from the data acquisition layer and execute multimodal fusion computing models; The irrigation control hub receives the dynamic water demand value output by the edge computing gateway and generates a pulse width modulation signal based on the preset water demand threshold of the crop growth stage; The magnetorheological irrigation actuator is used to respond to the pulse width modulation signal and adjust the irrigation flow and pressure by changing the yield stress of the magnetorheological fluid. The magnetorheological irrigation actuator has a built-in nanoscale flow sensor.

8. The crop water demand dynamic intelligent irrigation system according to claim 7, characterized in that: The edge computing gateway adopts a heterogeneous computing architecture and integrates FPGA and RISC-V processors. The FPGA is used for real-time Fourier transform of terahertz spectral data, and the RISC-V processor runs a multimodal fusion computing model. The data processing delay is ≤10ms.

9. The crop water demand dynamic intelligent irrigation system according to claim 7, characterized in that: The magnetorheological irrigation actuator includes an annular electromagnetic coil and a deformable valve core. The coil current adjustment range is 0-2A, the corresponding irrigation flow adjustment range is 0-50L / h, and the flow control accuracy is ≤±2%FS.

10. The crop water demand dynamic intelligent irrigation system according to claim 7, characterized in that: It also includes a drone inspection module, which is equipped with a terahertz imager to perform large-scale scanning of the crop canopy, fuse the scanning results with the measurement results of the ground terahertz time-domain spectroscopy system, and optimize the spatial distribution accuracy of the crop physiological water demand signal.

Images