Intelligent control method and system for high-pressure atomization irrigation and storage medium thereof

Abstract

The invention relates to an intelligent control method and system for high-pressure atomization irrigation and a storage medium of the intelligent control method and system, and aims to solve the problem that the quality is inconsistent due to the fact that traditional water and fertilizer management of high-value crops is difficult to sense the internal physiological state of the crops in real time. Randomly selecting a plurality of offering samples in each management partition for destructive sampling, and obtaining crop truth value data through physiological and biochemical detection; then, an irrigation rule base is established and dynamically updated based on the truth value data, and then variable control instructions are generated for the management partitions in combination with the real-time environment data; the invention further provides a system for implementing the method and a storage medium, accurate regulation and control starting from real demands of crop bodies are achieved, the consistency of crop quality and the resource utilization efficiency are effectively improved, and self-learning and predictive prevention and control capabilities are achieved.

Description

TECHNICAL FIELD

[0001] The application relates to the field of irrigation technology, in particular to an intelligent control method and system for high-pressure atomization irrigation and a storage medium thereof. BACKGROUND

[0002] The planting of high-value crops needs to consider traditional improvement and modern technological innovation, relies on modern agricultural planting technology to ensure the stability of yield and crop quality, has multiple and efficient utilization values, and is beneficial to popularization and application.

[0003] Based on the difference between high-value crops and ordinary crops, the high-value crops have higher requirements for soil conditions, temperature and humidity, and light and other growth conditions, because they need to be managed more technically, for example, in the process of planting Chinese herbal medicines, the proportion of nitrogen, phosphorus and potassium needs to be accurately controlled to ensure the content of effective ingredients, and attention also needs to be paid to avoiding problems such as root burning and reduction of effective ingredient content caused by excessive fertilization and watering.

[0004] In the process of planting high-value crops, the quality of high-value crops is improved by scientific regulation and control of water and fertilizer, for example, in the control of watering amount, soil moisture content and crop water demand are monitored in real time by means of soil moisture sensors and crop canopy temperature sensors, and intelligent irrigation systems such as drip irrigation and sprinkler irrigation are used, and for the control of fertilization amount, soil nutrient content is detected by soil testing and formula fertilization technology, and personalized fertilization schemes are customized according to the differences in the demand for nitrogen, phosphorus and potassium and trace elements in different growth periods of crops (such as the root and stem swelling period of Chinese herbal medicines and the flower bud differentiation period of fruit trees), and efficient fertilization methods such as slow-release fertilizer, water-soluble fertilizer or foliar fertilizer are used to supply nutrients to crops.

[0005] However, in actual application, the inventors have found that even if the above scientific crop planting method is used to realize intelligent control of processes such as crop fertilization and watering, due to the influence of factors such as genetic characteristics of crop maternal plants, growth environment of crops in later stages and human interference, the content of nutrient components in mature crops obtained by using the preferentially customized watering and fertilization life cycle management method has a large deviation in different planting batches, which affects the product quality (for example, the content of nutrient components in high-value crops obtained by scientific planting has no large deviation from the content of nutrient components in conventional crops), and is not conducive to improving the quality of crops. SUMMARY

[0006] The purpose of the present application is to improve the quality of the planted crops, and the application provides an intelligent control method and system for high-pressure atomization irrigation and a storage medium thereof.

[0007] In order to achieve the above-mentioned purpose, the intelligent control method and system for high-pressure atomization irrigation and the storage medium thereof provided by the present application adopt the following technical solutions: In a first aspect, the present application discloses an intelligent control method for high-pressure atomization irrigation, dividing the irrigation area into a plurality of management partitions, and randomly selecting a number of crop plants in each management partition as sacrifice samples; destructive sampling of the sacrifice samples according to a preset sampling time sequence to obtain crop tissue samples, and performing broken physiological and biochemical index detection on the crop tissue samples to obtain at least one true value data representing the water state, nutrient state or health state of the crops; establishing an irrigation rule library, defining a mapping relationship between the true value data range and the irrigation control instruction through the established irrigation rule library, and obtaining the true value data based on the current sampling time sequence, and generating or updating personalized irrigation rules for each management partition; Based on the latest personalized irrigation rules, combined with real-time environmental data obtained by the environmental sensor, variable control instructions for variable watering, variable fertilization, variable dosing or variable insecticide for each management partition are generated; Control the high-pressure atomization sprinkler irrigation system to execute the generated variable control instructions.

[0008] Preferably, the irrigation rule library establishment method comprises: fuse and standardize the true value data obtained by the current sampling with the historical sampling data, historical irrigation operation records corresponding environmental data, construct a spatiotemporal correlation data set; Use the association rule learning algorithm to mine the spatiotemporal correlation data set to generate candidate rules with true value data and environmental data as condition items and irrigation control instructions as result items; Calculate the success probability of each candidate rule reaching the expected effect after being executed based on the historical sampling data and the historical irrigation operation records, and use the success probability as the confidence of the association rule learning algorithm; Add candidate rules with confidence higher than the preset threshold to the irrigation rule library, and mark the applicable crop growth period and environmental condition range for each candidate rule.

[0009] Preferably, the variable fertilization control instruction is generated, comprising: Based on the true value data, obtain the physiological and biochemical detection results of the sacrifice samples, extract the measured content values of several target nutrient components of the sacrifice samples; Compare the measured content values of the several target nutrient components with the current optimal nutrient target interval as the species and growth period one by one to obtain the surplus and deficiency amount of each nutrient component, which includes the missing item and the excess item; Input the surplus and deficiency amount into the nutrient trade-off optimization model to obtain the optimal fertilizer formula and application amount; Generate variable fertilization control instructions according to the optimal fertilizer formula and application amount.

[0010] Preferably, the variable insect control instruction is generated, including: The crown multispectral image of the crop is automatically acquired by the image acquisition device deployed in the management partition, the crown multispectral image is processed based on the target detection model, the type of insect pest is identified, and the coordinate information of the insect pest occurrence concentration point is marked in the image; The coordinate information of the concentration point is fused with the GIS geographic information of the management partition, and the insect pest distribution digital map of the management partition is generated, and the types and severity levels of the insect pests of different coordinate information are marked on the insect pest distribution digital map; According to the insect pest distribution digital map, the operation path of the high-pressure atomization irrigation system is planned; According to the operation path, the high-pressure atomization irrigation system is controlled to open the spray head when located at the current insect pest concentration point, and to close the spray head when located at the non-insect pest concentration point.

[0011] Preferably, it also includes a preventive control method for insect pests, specifically including: The microenvironment sensor network is distributedly deployed in the management partition, and the microenvironment data of temperature, humidity, light intensity and volatile organic compound concentration inside the crop canopy are continuously monitored and acquired; The microenvironment data and the historical insect pest occurrence records are associated and analyzed, and a time series prediction model is used to predict the probability and hot area of insect pest occurrence in a specific future time period; When the predicted insect pest occurrence probability exceeds the preset risk threshold, a preventive variable pesticide application instruction is automatically generated; The high-pressure atomization irrigation system sprays the preventive pesticide liquid in the predicted insect pest hot area.

[0012] Preferably, after the high-pressure atomization irrigation system executes the variable control instruction, it further includes: After the execution of the variable control instruction, the same management partition is sampled again at the next sampling time sequence, and the sacrifice sample is sampled and the sacrifice sample true value data is acquired, to obtain the verified data; The relative improvement rate η of the verified data and the true value data before the execution of the variable control instruction on the key indicators is calculated, and the calculation formula is η=( - ) / ×100% Wherein, is the measurement value of a certain indicator in the verified data, is the measurement value of the corresponding indicator in the true value data; The calculated improvement rate η is compared with the preset expected improvement rate threshold [ , ] in the individualized irrigation rule; If η< If the decision fails to achieve the expected results, the system will automatically impose a confidence penalty on the personalized irrigation rule that triggered the decision and mark it as a rule to be optimized. like ≤η≤ If so, the confidence level of the personalized irrigation rule is maintained; If η> If the decision-making effect is significant, the confidence level of the personalized irrigation rule will be improved. For rules that need optimization, the system initiates a reinforcement learning process, storing the corresponding entire decision-making process data as a failure case in a specific dataset. This data drives the irrigation rule base to prioritize learning such cases when mining candidate rules in the next iteration.

[0013] Preferably, it also includes cross-cycle dynamic optimization strategies aimed at the ultimate quality goal, including: Ultimate quality target setting: At the beginning of the crop growth cycle, set a clear ultimate quality target vector for the current planting batch. =[ , ,..., ],in , Quantitative indicators such as the concentration of specific functional components; Reverse path planning: Utilizing a digital twin model of crop growth to vectorize the ultimate quality target By decomposing the data in reverse to each critical reproductive stage, the target range of intermediate physiological states that each reproductive stage i needs to reach is calculated. =[ , ]; Real-time tracking and correction: After sacrificial sampling is performed at each reproductive period i, the obtained ground truth data is compared with the target interval of the intermediate physiological state for that period. Perform comparisons and generate variable control instructions; The objective function F of the variable control command is defined as minimizing the weighted sum of squares deviation between the current state and the target state, i.e. F=

[0014] Where j represents different physiological and biochemical indicators. The importance weight of the corresponding indicator j; The implementation of cross-cycle dynamic optimization strategies enables the crop growth trajectory to eventually converge to the ultimate quality target vector. .

[0015] Preferably, soil sampling and analysis are also required: collecting soil samples of the rhizosphere region of the current sacrifice sample while destructively sampling the sacrifice sample; performing metagenomic sequencing analysis on the soil samples of the rhizosphere region to detect the functional gene abundance of pathogenic bacteria in the soil and analyze the soil microbial community structure; establishing a soil-borne disease occurrence risk index R, and the calculation formula is

[0016] wherein, is the abundance of specific pathogenic bacteria genes, is the abundance of beneficial microorganisms genes, is a stress factor calculated from soil humidity and temperature; when the risk index R exceeds a safety threshold, it is determined that the management partition has a high risk of soil-borne disease, and the system generates a variable dosing instruction; controlling the high-pressure atomizing sprinkler system to apply functional water solution with soil disinfection function to the rhizosphere region of the crop to regulate the soil micro-ecology.

[0017] In a second aspect, the application discloses a high-pressure atomizing irrigation system, which is applied to the intelligent control method of high-pressure atomizing irrigation. The intelligent control method of high-pressure atomizing irrigation comprises the following steps: A first module is configured to randomly select a plurality of crops in each management partition as sacrifice samples; A second module is configured to destructively sample the sacrifice samples according to a preset sampling time sequence to obtain crop tissue samples, and perform broken physiological and biochemical index detection on the crop tissue samples to obtain at least one true value data representing the water state, nutrient state or health state of the crop; A third module is configured to establish an irrigation rule library, define a mapping relationship between the true value data range and the irrigation control instruction through the established irrigation rule library, and obtain the true value data based on the current sampling time sequence to generate or update the individualized irrigation rule for each management partition; A fourth module is configured to generate variable control instructions for variable watering, variable fertilization, variable dosing or variable insect killing for each management partition based on the latest individualized irrigation rule and real-time environmental data obtained by the environmental sensor; A fifth module is configured to control the high-pressure atomizing sprinkler system to execute the generated variable control instructions.

[0018] In a third aspect, the application discloses a storage medium storing a computer program capable of being loaded and executed by a processor to implement the intelligent control method of high-pressure atomizing irrigation according to the first aspect. ​​

[0019] Compared with the prior art, the application provides an intelligent control method and system of high-pressure atomization irrigation and a storage medium thereof, and has the following beneficial effects: 1. By introducing "sacrificial sample" destructive sampling and true value data acquisition, the transformation from "indirect inference based on external environment or crop phenotype characteristics" to "direct perception based on real physiological and biochemical state of crops" is realized, which can directly respond to the most essential water, fertilizer and pesticide needs of crops, thereby significantly improving the quality consistency and resource utilization efficiency of high-value crops; 2. Through association rule mining and confidence evaluation, the system can continuously learn from the success and failure of historical decisions, constantly optimize and adjust the control strategy, effectively overcome the poor adaptability and easy obsolescence of traditional preset models, and ensure the accuracy and robustness of long-term use; 3. Through the spatial precise matching mechanism, the coincidence degree of pesticide application range and pest distribution area is improved to more than 90%, and the pesticide coverage area of non-target area is reduced by 60%-75%, realizing the synchronous optimization of pesticide utilization rate and pest killing rate. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is an execution step flowchart of an intelligent control method of high-pressure atomization irrigation according to an embodiment of the application.

[0021] Figure 2 is a method flowchart for establishing an irrigation rule base of an intelligent control method of high-pressure atomization irrigation according to an embodiment of the application.

[0022] Figure 3 is a variable fertilization control instruction method flowchart of an intelligent control method of high-pressure atomization irrigation according to an embodiment of the application.

[0023] Figure 4 is a method flowchart for generating variable pesticide control instructions and execution of an intelligent control method of high-pressure atomization irrigation according to an embodiment of the application.

[0024] Figure 5 is a flowchart of a pest predictive prevention and control method of an intelligent control method of high-pressure atomization irrigation according to an embodiment of the application. DETAILED DESCRIPTION

[0025] The following will be described in conjunction with the accompanying Figures 1-5The technical solutions in the present application are described clearly and completely. Obviously, the described embodiments are only part of the embodiments of the present application, not all. The components of the present application described and shown in the drawings can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application. It should be noted that: similar numbers and letters represent similar items in the following drawings, therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. Meanwhile, in the description of the present application, the terms "first", "second" and the like are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0026] In the traditional high-value crop planting system, the static water and fertilizer management strategy based on the preset threshold value is difficult to adapt to the nonlinear state change in the growth process of crops, resulting in the deviation of plant physiological and biochemical indicators from the optimal interval. The fixed control model relies on historical experience data for construction and cannot perceive the real demand of crops in real time, causing the absorption efficiency of nutrient elements and the synthesis path of effective components to be blocked. The environmental parameters collected by the sensor network are decoupled from the actual physiological state of the crops, and the traditional non-destructive detection means are limited by the complexity of the canopy structure, making it difficult to obtain real metabolic data at the tissue level, resulting in water and fertilizer decision lagging behind the actual demand changes of crops.

[0027] For example, in a traditional Chinese medicinal material planting base, a monitoring system is constructed using soil moisture sensors and canopy temperature probes, combined with a preset drip irrigation control logic to implement water and fertilizer management. Irrigation is triggered when the soil water content reaches the set threshold, and standard formula water-soluble fertilizer is applied according to the growth stage division. It is found in actual operation that during the rootstock expansion period of Angelica sinensis, the content of effective component ferulic acid fluctuates within ±30%, and the detection shows that the difference in nitrate nitrogen metabolism rate of different plants reaches 2.8 times. The traditional system fails to timely capture the differences in nutrient absorption efficiency among plants, and the continuous application of homogenized fertilizers leads to ion antagonism effect in some plants. The correlation between canopy temperature monitoring data and leaf stomatal conductance only maintains at 0.65 level, which cannot accurately reflect the water stress degree, resulting in irrigation decision error. Destructive detection is usually carried out after the problem appears, resulting in a lag of 14-21 days for correction measures.

[0028] Among the above problems, the harm to high-value crops is that the effective component cannot have a qualitative leap during the maturation period, and when facing improper management, the crops will have a nutritional imbalance phenomenon, thus causing a metabolic pathway deviation, resulting in a decrease in the synthesis efficiency of target compounds, and when water management is improper, abnormal accumulation of secondary metabolites may be induced, and the superposition of such problems causes the cultivated crops to not be able to highlight their “high value” characteristics.

[0029] Therefore, in a first aspect, the application discloses an intelligent control method for high-pressure atomization irrigation.

[0030] With reference to Figure 1 An intelligent control method for high-pressure atomization irrigation comprises the following steps: S1, dividing an irrigation area into a plurality of management partitions, and randomly selecting a plurality of crops in each management partition as sacrifice samples; S2, destructively sampling the sacrifice samples according to a preset sampling time sequence to obtain crop tissue samples, and performing broken physiological and biochemical index detection on the crop tissue samples to obtain at least one true value data representing the water state, nutrient state or health state of the crops; S3, establishing an irrigation rule library, defining a mapping relationship between the true value data range and the irrigation control instruction through the established irrigation rule library, and generating or updating individualized irrigation rules for each management partition based on the current sampling time sequence and the true value data; S4, based on the latest individualized irrigation rules, combining real-time environmental data obtained by an environmental sensor, generating variable control instructions for variable watering, variable fertilization, variable dosing or variable insect killing for each management partition; S5, controlling a high-pressure atomization sprinkler irrigation system to execute the generated variable control instructions.

[0031] Among them, the “sacrifice sample” refers to randomly selecting a plurality of plants in the management partition, and the selected plants are in the same growth environment as other plants in the management partition and have randomness, and the physiological and biochemical true value data of the crop tissue are directly obtained by destructively sampling the randomly selected sacrifice plants.

[0032] Meanwhile, the true value data includes but is not limited to water state data, nutrient state data, health and stress data.

[0033] As a preferred embodiment, the implementation manner of the scheme of the application in the specific implementation process is: First, a 100-mu irrigation area is divided into 10 10-mu management partitions; Secondly, according to the preset sampling timing, the plants in the current management partition are randomly selected as the sacrifice samples according to the type and growth condition, for example, three days, four days, or several hours of time interval, and the selected samples are placed in liquid nitrogen for quick freezing preservation.

[0034] Further, the frozen leaf samples are subjected to crushing treatment. The leaves are ground into fine powder using a ball mill, and indicators such as chlorophyll, soluble sugar, and nitrate reductase are extracted. The content of these indicators is determined by spectrophotometer, high performance liquid chromatograph, and the like, to obtain true value data representing the water state and nutrient state of the crops.

[0035] Then, an irrigation rule library is established. A machine learning algorithm such as decision tree or random forest is used to train a mapping model between the true value data range and the irrigation control instructions. For example, when the relative water content of the leaves is lower than 75%, an instruction to increase the irrigation amount is triggered; when the nitrate reductase activity is lower than 0.5 μmol NO 2- / g·h, an instruction to increase the nitrogen fertilizer application amount is triggered.

[0036] Thus, the true value data obtained in the current sampling period generates or updates the personalized irrigation rules for each management partition. If it is detected that the growth state of the crops in a management partition changes significantly, the irrigation rule parameters of the management partition are adjusted accordingly.

[0037] Specifically, in combination with the real-time environmental data obtained by the temperature and humidity sensors, light sensors, and the like installed in each management partition, variable control instructions are generated according to the latest personalized irrigation rules. For example, when the soil moisture in a management partition is lower than the threshold value and there is no rainfall, a variable watering instruction for the management partition is generated.

[0038] Finally, the variable control instructions generated are executed by a high-pressure atomizing sprinkler irrigation system. The system adjusts the opening time, water amount, and atomizing particle size of the sprinkler heads of different management partitions according to the instructions, to achieve precise irrigation. At the same time, variable fertilization is completed by adding water-soluble fertilizers of different formulations into the irrigation water.

[0039] In some of the above schemes of the present application, an irrigation method based on the generation of variable control instructions from true value data and an irrigation rule library is proposed. However, in the process of establishing the irrigation rule library, there are problems such as insufficient fusion of historical sampling data and historical irrigation operation records with real-time environmental data, lack of dynamic adaptability of rule generation, and inability to quantitatively verify the rule execution effect.

[0040] To this end, the present application further proposes a method for establishing an irrigation rule library, with reference to Figure 2 , comprising the following steps: S31, true value data obtained in the current sampling is fused with historical sampling data and historical irrigation operation records to obtain a new rule training data set. The corresponding environmental data is fused and standardized to construct a spatio-temporal correlation dataset; S32, using a correlation rule learning algorithm to mine the spatio-temporal correlation dataset to generate candidate rules taking true value data and environmental data as condition items and taking irrigation control instructions as result items; S33, calculating the success probability of reaching the expected effect after the execution of each candidate rule based on historical sampling data and historical irrigation operation records, and taking the success probability as the confidence of the correlation rule learning algorithm; S34, adding candidate rules with a confidence higher than a preset threshold to the irrigation rule base, and marking the applicable crop growth period and environmental condition range for each candidate rule.

[0041] Specifically, in the construction of the spatio-temporal correlation dataset, the dimensional differences of different sensor data are eliminated through data cleaning and format conversion, for example, the soil moisture percentage and temperature in Celsius are uniformly converted into standardized Z values. In the candidate rule mining process, the generation of multi-dimensional condition combinations uses feature intersection technology, for example, combining light intensity intervals and air humidity thresholds to form composite condition items. When calculating the confidence, a time decay factor is introduced, for example, giving a weight of 0.8 to historical cases three months ago and a weight of 1.2 to cases within a week, to enhance the timeliness of the irrigation rule base. In the candidate rule marking stage, the decision tree model automatically identifies the boundary conditions for the rule to take effect, for example, when the day and night temperature difference exceeds 8℃, a certain irrigation rule is invalid. Through the dynamically updated dataset, the rule base is continuously optimized, for example, the incremental learning of the irrigation rule base is triggered by the newly added sampling data every week, so that the irrigation control instructions can adapt to the gradual change of the crop growth stage and the fluctuation of the environmental parameters, thereby improving the stability of the nutrient content of high-value crops.

[0042] In the process of generating variable fertilization control instructions based on true value data to realize precision fertilization, only relying on the mapping relationship between the true value data range and the irrigation instructions cannot dynamically analyze the difference between the actual nutrient demand of the crop and the target state, resulting in the lack of quantitative compensation mechanism for different nutrient content surplus and deficiency amounts in the generation of fertilization formula, and it is difficult to realize the coordinated optimization of multiple nutrient elements. Therefore, with reference to Figure 3 , the application further proposes to generate variable fertilization control instructions, including the following steps: S41, obtaining the physiological and biochemical detection results of the sacrifice sample based on the true value data, and extracting the measured content values of several target nutrient components of the sacrifice sample; S42, comparing the measured content values of the several target nutrient components with the optimal nutrient target interval of the current crop type and growth period one by one to obtain the surplus and deficiency amount of each nutrient component, which includes the missing item and the excess item; S43, inputting the surplus and deficiency amount into a nutrient trade-off optimization model to obtain the optimal fertilizer formula and application amount; S44, generating variable fertilization control instructions according to the optimal fertilizer formula and the application amount.

[0043] The measured content value of the target nutrient component is obtained through destructive physiological and biochemical detection, and the detection object includes but is not limited to nitrogen, phosphorus, potassium, iron, zinc and boron. The optimal nutrient target interval is determined by combining the crop variety database and the growth period model, for example, the target interval of potassium element is set to 2.3 mg / g-2.8 mg / g dry weight during the root and stem enlargement period of Chinese herbal medicine. The nutrient optimization model adopts a multi-objective programming algorithm, in which the missing item corresponds to the compensation constraint condition, and the excess item corresponds to the inhibition constraint condition. The objective function is set to minimize the weighted combination of the minimum deviation degree of the comprehensive nutrition and the minimum fertilizer cost. During the generation of the fertilizer formula, when the missing amount of a certain element exceeds the threshold value, the compensation mechanism of the element is triggered preferentially, for example, when the missing amount of potassium exceeds 15% of the lower limit of the target interval, a water-soluble fertilizer containing potassium≥45% is automatically matched as the base material.

[0044] Specifically, after obtaining the crop tissue sample through destructive sampling, the inductively coupled plasma mass spectrometry method is used to determine the content of each element. When comparing the measured data with the optimal target interval, an element deviation matrix is established, in which the relative deviation value of each element is calculated independently, for example, when the measured value of magnesium element is 0.8 mg / g and the target interval is 1.0 mg / g-1.2 mg / g, the missing amount is quantified as-20%. After receiving the profit and loss data, the nutrient optimization model first models the interaction between elements, for example, the calcium-magnesium antagonism coefficient is set to 0.7, to ensure that the compensation scheme will not cause element antagonism. When generating the fertilizer formula, if phosphorus is detected to be in excess and nitrogen is missing, a compound fertilizer with a nitrogen-phosphorus ratio≥3:1 is automatically selected, and the application amount is calculated to make the nitrogen supplement amount reach 110%-120% of the missing amount. When the finally generated variable fertilization instructions are executed through the high-pressure atomization system, the fertilizer solution concentration is dynamically adjusted according to the application amount, for example, the atomization liquid nitrogen concentration is 0.5% corresponding to the nitrogen application amount of 5 kg per hectare. This scheme accurately matches the actual demand of crops by establishing a multi-element independent compensation mechanism.

[0045] As a preferred embodiment, the scheme of the present application is implemented as follows: Based on the true value data, the physiological and biochemical detection results of the sacrifice sample are obtained, and the measured content values of the target nutrient components such as nitrogen, phosphorus, potassium, calcium and magnesium of the sacrifice sample are extracted. For example, the potassium, calcium and magnesium contents are determined by atomic absorption spectrophotometry, the nitrogen content is determined by Kjeldahl nitrogen determination method, and the phosphorus content is determined by molybdenum blue colorimetry.

[0046] The measured content values of target nutrients such as nitrogen, phosphorus, potassium, calcium, magnesium, etc. are compared one by one with the optimal nutrient target interval of the current crop type and growth period. For example, for tomato plants in the flowering and fruiting period, the measured nitrogen content of the leaves is 3.5%, which is compared with the optimal target interval of 4.0%-5.0%, and it is found that the nitrogen element is missing by 0.5%-1.5%; the measured phosphorus content of the leaves is 0.6%, which is compared with the optimal target interval of 0.3%-0.5%, and it is found that the phosphorus element is over-provided by 0.1%-0.3%, so that the gain and loss of each nutrient is obtained in this way.

[0047] The gain and loss are input into the nutrient trade-off optimization model to obtain the optimal fertilizer formula and application amount. The model considers the antagonism and synergy between different nutrient elements and the absorption efficiency of plants to different elements, and calculates the optimal formula balancing the demand for each element through a linear programming algorithm.

[0048] According to the optimal fertilizer formula and application amount, a variable fertilization control instruction is generated. The instruction contains information such as fertilizer type, ratio, application amount, and fertilization time, and can be directly used to control the operation of intelligent fertilization equipment.

[0049] As a preferred embodiment, the specific implementation of the scheme of the present application is, The irrigation rule base establishment method comprises the following steps: First, the true value data obtained by the current sampling is fused and standardized with the environmental data corresponding to the historical sampling data and the historical irrigation operation records to construct a spatio-temporal correlation data set. For example, the environmental data such as soil moisture, air temperature and light intensity at different time points can be time-aligned and spatially matched with the true value data such as crop water content and chlorophyll content at the corresponding time points to form a unified data format.

[0050] Secondly, the spatio-temporal correlation data set is mined using an association rule learning algorithm to generate candidate rules with true value data and environmental data as condition items and irrigation control instructions as result items. For example, the Apriori algorithm or FP-Growth algorithm can be used to find frequent item sets in the data, and association rules are generated according to the support and confidence thresholds.

[0051] Then, the success probability of achieving the expected effect after the execution of each candidate rule is calculated based on the historical sampling data and the historical irrigation operation records, and the success probability is taken as the confidence of the association rule learning algorithm. For example, the number of times each candidate rule is executed in the historical sampling data and the historical irrigation operation records can be counted, and the number of times the crop physiological indicators improve after execution can be counted to calculate the success probability.

[0052] Finally, candidate rules with a confidence higher than a preset threshold are added to the irrigation rule base, and each candidate rule is marked with applicable crop growth stages and environmental condition ranges. For example, rules with a confidence greater than 0.8 can be included in the irrigation rule base, and the candidate rules are marked with the data features used when the candidate rules are generated, indicating that the candidate rules are applicable to a specific growth stage of a certain crop and a specific temperature and humidity range.

[0053] Through the above technical solutions, the present application realizes dynamic optimization and adaptive updating of the irrigation rule base. Due to the fusion of multi-source heterogeneous data, the irrigation rule base can comprehensively reflect the complex relationship between crop growth and environmental factors, and through historical sampling data and historical irrigation operation records verification and confidence mechanism, the reliability and effectiveness of the candidate rules are ensured.

[0054] Moreover, the marking of the applicable conditions ensures that the candidate rules are triggered in appropriate scenarios, improving the accuracy of irrigation decisions.

[0055] Further, with reference to Figure 4 a variable insect control instruction is generated, including: S401, automatically acquiring a crop canopy multispectral image through an image acquisition device deployed in a management partition, processing the canopy multispectral image based on a target detection model, identifying a pest type and marking pest occurrence concentration point coordinate information in the image; S402, fusing the concentration point coordinate information with GIS geographic information of the management partition, generating a pest distribution digital map of the management partition, and marking the pest type and severity level of different coordinate information on the pest distribution digital map; S403, planning a running path for the high-pressure atomization irrigation system according to the pest distribution digital map; S404, according to the running path, controlling the high-pressure atomization irrigation system to open the spray head when located at the current pest concentration point and to close the spray head when located at the non-pest concentration point.

[0056] The multispectral image acquisition device can be set as a camera with near-infrared band capture capability, for example, using a sensor array with a wavelength range of 450nm-900nm to enhance the recognizability of pest features.

[0057] The target detection model can use a deep learning architecture based on a convolutional neural network, and during training, the input includes a labeled data set containing different pest types and severity levels, and the output layer can include pest class confidence and bounding box coordinate parameters.

[0058] During the generation of the pest distribution digital map, the conversion accuracy of the geographic coordinate system can be controlled within ±0.5 meters, ensuring that the positioning error of the spray head does not exceed the allowable threshold.

[0059] The spray head control logic adopts a coordinate matching mechanism. When the irrigation system moves to the preset triggering range of the target point, for example, a circular area with a radius of 1.5 meters centered at the point coordinate, the corresponding spray head group starts the pesticide application program.

[0060] Specifically, the crop canopy multispectral image enhances the reflection difference between pests and healthy tissues through specific wavelength combinations, such as identifying abnormal chlorophyll regions at 680 nm and detecting egg aggregation characteristics at 850 nm.

[0061] Further, the target detection model performs pixel-by-pixel analysis on the image and outputs a pest heat map containing coordinate information, such as detecting pest spots larger than 2 mm in diameter per square centimeter resolution.

[0062] Correspondingly, in the GIS geographic information fusion stage, the pest coordinates are mapped to the management partition geographic coordinate system to form a spatial distribution map with elevation data, such as using the UTM projection coordinate system to ensure the accuracy of the plane distance calculation.

[0063] At the same time, the path planning module automatically generates a zigzag or spiral traversal trajectory based on the pest distribution density, such as enabling continuous spraying mode when the pest point spacing is less than 5 meters and switching to jump mode when the spacing is greater than 10 meters. The spray head opening and closing control is matched with the preset coordinates through the real-time positioning system. When the device moves to the target point within ±0.3 meters, the electromagnetic valve opening instruction is triggered.

[0064] This scheme improves the coincidence degree of the pesticide application range and the pest distribution area to more than 90% through spatial precise matching mechanism, while reducing the pesticide coverage area of non-target area by 60%-75%, achieving synchronous optimization of pesticide utilization rate and pest killing rate.

[0065] As a preferred embodiment, the scheme of the present application is implemented as follows: Multiple high-resolution multispectral cameras are deployed within the management partition. The cameras automatically take canopy images every 2 hours, with an image resolution of 4000x3000 pixels. The collected images are transmitted in real time to the central processing server through a wireless network.

[0066] A pre-trained deep learning target detection model runs on the server. The model is based on the YOLOv5 architecture and has been specially trained for common crop pests. The model processes the received images, identifies common pest types, and labels the specific location coordinates of the pests in the images.

[0067] The system fuses the identified pest coordinate information with the GIS geographic information system of the management partition. The GIS system pre-stores spatial data such as the terrain and soil type of the management partition. After fusion, a pest distribution digital map is generated. The map adopts a grid data structure, and the spatial resolution is 0.5 meters. On the digital map, different colors represent different pest types, and the color depth represents the severity of the pest.

[0068] Based on the generated digital map, the system plans an optimal operation path for the high-pressure atomization irrigation equipment. The path planning adopts an improved ant colony algorithm, with the optimization objective of covering all pest points and minimizing the total travel distance. The planning result is output as a series of coordinate point sequences as navigation instructions for the equipment.

[0069] The high-pressure atomization irrigation system operates according to the planned path and is equipped with a GPS positioning module to obtain real-time current position coordinates. When the GPS coordinates match the pest point coordinates, the control system sends an opening instruction to the corresponding spray head, and the spray head starts spraying pesticide. When leaving the pest point, the corresponding spray head automatically closes. The spray heads in non-pest areas are always closed.

[0070] In some of the above schemes of the present application, a technical means of generating variable pest control instructions based on a pest distribution digital map is proposed. However, this scheme can only passively kill pests at specific points after the pests have occurred, and cannot predict risk areas before the pests outbreak and preventively intervene, resulting in a lag in actual prevention and control and possibly causing irreversible damage to crops.

[0071] To this end, the present application further proposes a pest predictive prevention and control method, with reference to Figure 5 , specifically including: S4011, a micro-environment sensor network is distributedly deployed in the management partition to continuously monitor and obtain micro-environment data of temperature, humidity, light intensity and volatile organic compound concentration inside the crop canopy; S4012, the micro-environment data is associated and analyzed with historical pest occurrence records, and a time series prediction model is used to predict the probability and hot spot area of pest occurrence in a specific future time period; S4013, when the predicted probability of pest occurrence exceeds a preset risk threshold, a preventive variable pesticide application instruction is automatically generated; S4014, a high-pressure atomization sprinkler system is used to spray preventive pesticide liquid in the predicted pest hot spot area.

[0072] Among them, the micro-environment sensor network is arranged to be deployed in the crop canopy at a height of 0.5-1.2 meters with a density of 3-5 nodes per square meter, and the monitoring parameters include temperature range 10-40℃, relative humidity 30%RH-95%RH, light intensity 0-2000μmol·m -2 ·s-1 -1500 pmol.m-2.s-1 -2 ·s -1 and volatile organic compound concentration 0 ppb-500 ppb.

[0073] The time series prediction model adopts a long short-term memory network architecture. The input layer receives a sequence of environmental data from the past 72 hours, and the output layer generates a prediction value of the probability of pest occurrence in the next 24 hours, with a prediction accuracy error controlled within ±5%.

[0074] The risk threshold is set to trigger preventive instructions when the probability value exceeds 65%, and the spraying liquid concentration is dynamically adjusted to 0.1%-0.5% emulsifier solution according to the predicted pest type.

[0075] Specifically, crop canopy microenvironment data is uploaded to the central processor every 5 minutes through a wireless transmission protocol, and is spatio-temporally matched with the stored pest history database. When the temperature is between 25℃-32℃ and the humidity is above 80% for 6 hours, the system automatically marks it as a high-risk environmental combination.

[0076] The prediction model analyzes the rate of change of environmental parameters, such as when the volatile organic compound concentration rises by more than 200 ppb within 3 hours, it is determined that the incubation period of the pest has begun. High-pressure atomizing nozzles spray liquid at a pressure of 0.2 MPa in the predicted hotspot area, with a spraying coverage radius set to 1.5 meters, ensuring complete infiltration of the potential egg distribution area.

[0077] After the preventive pesticide application instructions are generated, the system automatically calls the sprinkler equipment corresponding to the coordinates in the GIS map, and completes the pesticide coverage 24-48 hours before the outbreak of the pest. This prediction mechanism forms a double protection with the existing pest identification system. When the image recognition module does not detect pests, the prediction model can still start prevention and control in advance based on environmental abnormalities, reducing the crop damage rate to below 3%.

[0078] As a preferred embodiment, the scheme of the present application is implemented as follows: Distributed deployment of microenvironment sensor network within the management partition. A sensor node can be placed every 10 meters inside the crop canopy, each containing a temperature sensor, a humidity sensor, a light intensity sensor, and a volatile organic compound concentration sensor. The sensor collects data every 5 minutes and transmits it to the central control system through a wireless network.

[0079] Continuous monitoring and acquisition of microenvironment data inside the crop canopy. The central control system receives and stores data from all sensor nodes, forming a time series database. The data includes temperature (℃), relative humidity (%), light intensity (lux), and volatile organic compound concentration (ppb).

[0080] Correlate microenvironment data with historical pest occurrence records. The system calls past 3 years of pest occurrence records, including pest type, occurrence time, occurrence location, etc. information. Through machine learning algorithms such as random forest or support vector machine, a correlation model between microenvironment parameters and pest occurrence probability is established.

[0081] Using time series prediction model, predict the probability of pest occurrence and hot spot area in future specific period. Using deep learning algorithms such as long short-term memory network (LSTM), based on historical data and current microenvironment parameters, the pest occurrence probability of each area in the next 7 days is predicted. At the same time, through spatial interpolation algorithm, pest risk heat map is generated.

[0082] When the predicted probability of pest occurrence exceeds the preset risk threshold, automatically generate preventive variable application instructions. Set the risk threshold to 60%, when the pest occurrence probability of a certain area in the next 7 days exceeds the threshold, the system automatically triggers the preventive application process. According to different pest types and prediction probability, the system selects appropriate pesticide type and concentration.

[0083] Through high pressure atomization sprinkler system, spray preventive pesticide liquid in the predicted pest hot spot area. The system controls the high pressure atomization nozzle to carry out directional spraying in the predicted high risk area, and adjusts the nozzle angle and pressure in real time during the spraying process to ensure that the pesticide liquid uniformly covers the target area.

[0084] In some of the above schemes of the application, a method of generating variable control instructions based on individualized irrigation rules is proposed to realize precise irrigation. However, due to the dynamic nature of crop growth, environmental disturbance, and insufficient initial experience of rule base, etc., the actual execution effect of part of the irrigation rules may deviate from the expectation. To this end, the application further includes a decision effectiveness quantitative verification and adaptive learning stage after the high pressure atomization irrigation system executes the variable control instructions.

[0085] After the execution of the variable control instructions, the same management partition is sampled again at the next sampling time sequence, and the sacrifice sample is sampled and the sacrifice sample true value data is obtained, to obtain the verified data; Calculate the relative improvement rate η of the verified data and the true value data before executing the variable control instructions on the key indicators, and the calculation formula is η=( - ) / ×100% Wherein, is the measurement value of a certain indicator in the verified data, is the measurement value of the corresponding indicator in the true value data; Compare the calculated improvement rate η with the expected improvement rate threshold preset in the individualized irrigation rule , ] is compared; if η < 0 , it is determined that the decision effect is not expected, and the system automatically applies a confidence penalty to the personalized irrigation rule triggering this decision and marks it as a rule to be optimized; if 0 < η < 0.5 , and η > 0.5 , the confidence of the personalized irrigation rule is maintained; if η > 0.5 , it is determined that the decision effect is significant, and the confidence of the personalized irrigation rule is increased; For the rule to be optimized, the system starts the reinforcement learning process, and stores the decision-making process data corresponding to this decision as a failure case in a specific data set, which is used to drive the irrigation rule library to preferentially learn such situations when mining candidate rules next time.

[0086] In the verification data acquisition stage, the time continuity sampling design ensures the causal relationship between the verification data and the original decision, for example, sampling within a fixed time interval after the execution of the variable control instruction. The destructive sampling can be whole-plant sampling or specific organ removal.

[0087] Specifically, in the verification stage after the execution of the variable control instruction, continuous sampling of the same management partition avoids the interference of spatial heterogeneity on the evaluation results. The true value data obtained by destructive sampling can reflect the real changes of the internal physiological state of the crop, for example, the water transport efficiency is verified by analyzing the composition of the xylem sap of the root system. The relative improvement rate calculation converts multi-dimensional indicators into a single quantitative parameter, for example, the comprehensive improvement value is calculated by weighting the nitrogen absorption rate and the transpiration efficiency. The confidence penalty mechanism reduces the priority of inefficient rules in the decision tree, for example, rules with a confidence level below 60% are removed from the current growth period application list. The reinforcement learning process optimizes the rule generation algorithm using failure cases, for example, increasing the negative sample weight coefficient in association rule mining, so that new rules can avoid historical error decision patterns. Through the closed-loop feedback mechanism, the system can dynamically eliminate invalid rules and strengthen efficient rules, for example, rules with a confidence level consistently higher than 80% for three consecutive growth cycles will be marked as preferred strategies.

[0088] As a preferred embodiment, the scheme of the present application is implemented as follows: After the execution of the variable control instruction of the high-pressure atomization irrigation system, the decision effectiveness quantitative verification and adaptive learning are performed. First, at the next sampling time sequence after the execution of the variable control instruction, the same management partition is sampled again and the true value data is obtained, obtaining the post-verification data. Then, the relative improvement rate η of the post-verification data and the true value data before the execution of the variable control instruction on the key indicators is calculated.

[0089] Furthermore, the calculated improvement rate η is compared with the preset expected improvement rate threshold in the personalized irrigation rules. , Comparisons can be made. Specifically, settings can be configured. 5%, It is 20%. When η At 5%, the decision is deemed to have failed to meet expectations. The system automatically applies a confidence penalty to the rule that triggered this decision, for example, reducing its confidence by 0.1 and marking it as a rule to be optimized. When 5% ≤ η ≤ 20%, the rule's confidence remains unchanged. When η At 20%, the decision-making effect is considered significant, and the confidence level of the rule is increased, for example, by increasing its confidence level by 0.1.

[0090] In some of the solutions mentioned above in this application, although the method of dynamically adjusting irrigation rules based on real-time sampling data achieves precise water and fertilizer management, the lack of a cross-growth period collaborative control mechanism based on the ultimate quality goal results in uncontrollable deviations in core indicators such as the concentration of key functional components between different planting batches.

[0091] In response, this application further proposes a cross-cycle dynamic optimization strategy oriented towards the ultimate quality goal, which includes three stages: setting the ultimate quality goal, reverse path planning, and real-time tracking and correction.

[0092] Among them, the ultimate quality target setting stage sets a clear quantitative indicator vector for the current planting batch at the beginning of the crop growth cycle. For example, in the planting of Chinese medicinal herbs, saponin content and polysaccharide concentration can be used as target vector elements.

[0093] In the reverse path planning stage, the ultimate goal is decomposed into each growth stage using a digital twin model. Specifically, the chlorophyll content range during the flowering stage can be set as [2.3 mg / g, 2.8 mg / g], and the soluble solids content range during the fruit enlargement stage can be set as [12%, 15%].

[0094] In the real-time tracking and correction stage, true data is obtained through destructive sampling. For example, the measured value of anthocyanin concentration during the fruit color-changing period is 1.2 mg / L. After comparing it with the target range [1.5 mg / L, 1.8 mg / L], a control instruction is generated.

[0095] Specifically, an ultimate quality target vector is set at the early stage of crop transplanting. For example, the target for saponin content in ginseng cultivation is set at 4.2% and polysaccharide content at 28%. Through the reverse transmission of the digital twin model, the target for maturity is decomposed into the root activity target [35μg / g·h-40μg / g·h] for the seedling stage and the leaf nitrogen content target [3.5%-4.0%] for the growth stage.

[0096] At each growth stage, the actual data obtained from destructive sampling is compared with the decomposed target. When the root activity of the seedling stage is detected to be only 30 μg / g·h, the system automatically calculates the weighted squared deviation, where the root activity weight coefficient is set to 0.6 and the leaf area index weight is set to 0.4.

[0097] Based on the deviation calculation results, variable control instructions for increasing phosphorus and potassium fertilizer application are generated, and the nutrient supply in the rhizosphere microzone is precisely adjusted through a high-pressure atomization system. This cross-cycle optimization mechanism ensures that local regulation at each growth stage always targets the ultimate quality goal. In ginseng cultivation practice, the fluctuation range of saponin content between different batches can be reduced from ±15% in traditional methods to within ±5%.

[0098] As a preferred embodiment, the solution of this application is specifically implemented as follows: At the start of the crop growth cycle, set the ultimate quality target vector for the current planting batch. =[ , ,..., For example, for a certain Chinese medicinal herb, a setting can be made. =[Active ingredient A content 80mg / g, active ingredient B content 50mg / g, dry matter content 30%].

[0099] Using a digital twin model of crop growth, The process is decomposed in reverse to the critical growth stages. Assuming the crop has three critical growth stages, the target range of intermediate physiological states for each growth stage i is calculated using the model. .For example: =[Chlorophyll content 40SPAD-50SPAD, plant height 20cm-25cm]; =[Root activity 500μg / (g·h)-600μg / (g·h), stem diameter 1.5cm-2cm]; =[Photosynthetic rate 15μmol / ( ·s)-20μmol / ( •s), leaf area index 4-5.

[0100] After sampling the sacrificial offerings during each reproductive period, obtain ground truth data and... Comparison was performed. Assume the sampling results during growth period 2 are: root activity 480 μg / (g·h), stem diameter 1.8 cm.

[0101] According to F= Construct the objective function F: F = in , Importance weights of root activity and stem diameter, respectively.

[0102] Based on the calculation results of F, the system generates variable control instructions, such as increasing the application amount of root system promoters, adjusting the water-fertilizer ratio, etc. Through continuous real-time tracking and correction, the crop growth trajectory eventually converges to the target trajectory. .

[0103] In some of the above schemes of the present application, a method for generating variable control instructions based on physiological and biochemical detection of sacrifice samples and environmental data is proposed to realize precise irrigation and pest control. However, in this process, the risk of soil-borne diseases caused by abnormal soil microbial community structure is not considered. Traditional schemes only rely on the state of above-ground crops and environmental data, and cannot identify potential diseases caused by soil pathogens, resulting in irreversible damage to crops due to root diseases and affecting the achievement of the ultimate quality target.

[0104] To this end, the present application further proposes to collect soil samples in the rhizosphere region of the current sacrifice sample while destructively sampling the sacrifice sample; perform metagenomic sequencing analysis on the soil samples in the rhizosphere region to detect the functional gene abundance of pathogenic bacteria in the soil and analyze the soil microbial community structure; establish a soil-borne disease occurrence risk index R, and the calculation formula is ; wherein, ( ) is the abundance of a specific pathogenic gene, ( ) is the abundance of beneficial microorganisms, ( ) is a stress factor calculated from soil humidity and temperature.

[0105] When the risk index R exceeds the safety threshold, it is determined that there is a high risk of soil-borne disease in the management partition, and the system generates a variable administration instruction; a high-pressure atomizing sprinkler system is used to apply a functional water solution with soil disinfection function to the rhizosphere region of the crop to regulate the soil micro-ecology.

[0106] wherein, the collection of rhizosphere soil samples is synchronized with the destructive sampling of sacrifice samples, so that the physiological state data of above-ground crops and the microbial community data of underground soil have spatiotemporal consistency. Metagenomic sequencing analysis can use the Illumina NovaSeq platform for double-end sequencing, with a sequencing depth of 10M reads / sample. The Kraken2 algorithm is used for species annotation to calculate the gene abundance ratio of pathogenic bacteria and beneficial microorganisms.

[0107] The safety threshold of risk index R can be set to the interval of 2.5-3.0, when the soil humidity is higher than 85% of the field water capacity and the temperature is between 25-30℃, the stress factor The functional water solution can contain 0.1% amino oligosaccharide and 0.05% Bacillus subtilis complex preparation, and the fog droplets with a particle size of 5-10 μm are formed by high-pressure atomization to penetrate into the rhizosphere soil.

[0108] Specifically, during the destructive sampling process, the rhizosphere soil with a radius of 15 cm and a depth of 20 cm is collected by a ring soil sampler centered on the sacrificial sample main root to ensure the spatial representativeness of the microbial community data.

[0109] After bioinformatics analysis of the metagenomic sequencing results, the sum of the gene abundance of Fusarium, Phytophthora and other pathogenic fungi is taken as The sum of the gene abundance of Trichoderma, Pseudomonas and other beneficial bacteria is taken as .

[0110] Stress factor Through real-time data calculation by soil humidity sensor and temperature sensor, when the average humidity is ≥80% and the temperature is ≥28℃ for 3 consecutive days, the stress factor is automatically adjusted to 1.3. When the R value exceeds the threshold value, the system triggers the precise pesticide application instruction, controls the high-pressure atomization nozzle to atomize the pesticide liquid to the target area at a pressure of 0.8 MPa, and adjusts the pesticide application amount dynamically according to the R value exceeding range, for example The pesticide application amount is 80 ml / m 2 When R=4.0, it increases to 80 ml / m 2 . This scheme detects microbial activity and uses a dynamic risk model to implement targeted intervention during the incubation period of soil-borne diseases, which reduces the amount of pesticide used by 40%-60% compared with the traditional regular pesticide application method, while maintaining the abundance of soil beneficial bacteria at not less than 85% of the initial value.

[0111] As a preferred embodiment, the scheme of the present application is implemented as follows: while the crop selected as a sacrificial sample is being destructively sampled, rhizosphere soil samples are collected within a range of 0-5 mm from the surface of the main root system, with a sample weight controlled between 50 g-100 g, and Illumina NovaSeq sequencing platform is used for metagenomic sequencing.

[0112] By KEGG database comparison, the gene sequences of fusarium oxysporum and rhizoctonia solani in the soil sample are identified, the abundance percentage of the total microbial community is calculated, and the beneficial bacteria gene abundance of trichoderma harzianum and pseudomonas fluorescens is detected. In the stress factor calculation module, when the soil humidity is higher than 85% of the field water holding capacity for three consecutive days and the temperature is between 25-30℃, the stress factor is 1.5. The threshold value of risk index R is set to 2.0, and when the calculated value exceeds the threshold value, the system automatically generates a soluble powder mixing instruction containing 0.1% niclosamide and 0.05% bacillus subtilis, and implements fixed-point spraying in the target plant rhizosphere radius of 15 cm through high-pressure atomizing nozzle, and the single application amount is controlled in 200ml .

[0113] In the second aspect, the application discloses a high-pressure atomizing irrigation system, which is applied to the intelligent control method of high-pressure atomizing irrigation in the first aspect, and comprises: A first module is used for randomly selecting a plurality of crops in each management partition as sacrifice samples; A second module is used for destructively sampling the sacrifice samples according to a preset sampling time sequence to obtain crop tissue samples, and performing broken physiological and biochemical index detection on the crop tissue samples to obtain at least one true value data representing the water state, nutrient state or health state of the crops; A third module is used for establishing an irrigation rule library, defining a mapping relationship between the true value data range and the irrigation control instruction through the established irrigation rule library, and obtaining the true value data based on the current sampling time sequence to generate or update the individualized irrigation rule for each management partition; A fourth module is used for generating variable control instructions for variable watering, variable fertilization, variable drug delivery or variable insect killing for each management partition based on the latest individualized irrigation rule and the real-time environmental data obtained by the environmental sensor; A fifth module is used for controlling the high-pressure atomizing sprinkling irrigation system to execute the generated variable control instructions.

[0114] The first module randomly selects a plurality of crops in each management partition as sacrifice samples, wherein the random selection algorithm adopts a combination of spatial grid division and Monte Carlo sampling to ensure that the samples cover different light intensity gradients and soil moisture regions.

[0115] The second module performs a destructive sampling operation through a mechanical arm, and the sampling time point is set before the morning transpiration of the crops is started. After sampling, the crop tissue samples are transferred to a detection cabin after being rapidly frozen in liquid nitrogen, and microwave-assisted breaking technology is used to realize efficient cell wall lysis. The spectrum analysis unit built in the detection cabin can simultaneously determine the physiological and biochemical indexes of the sacrifice samples.

[0116] The irrigation rule base of the third module adopts a dynamic updating mechanism based on confidence, and when the mapping relationship between the newly acquired true value data and the existing rule base deviates by more than 15%, a rule reconstruction process is triggered.

[0117] The variable control instruction generation unit of the fourth module integrates a fuzzy logic controller, quantizes the temperature, humidity and light intensity parameters in the real-time environmental data into membership functions of 0-1, and performs weighted fusion with the basic instructions output by the irrigation rule base.

[0118] The fifth module communicates with the high-pressure atomizing nozzle array through the CAN bus protocol, each nozzle is equipped with an independent electromagnetic valve and a pressure sensor, and realizes partition control with a response accuracy of 0.1 seconds.

[0119] Through the above technical solutions, the present application realizes precise closed-loop management of high-value crops, and solves the problem of control instruction lag caused by low modularization of traditional irrigation systems. The synergistic effect of the true value data obtained by destructive sampling and the dynamic rule base shortens the water and fertilizer regulation response time to within 2 hours, effectively maintaining the batch stability of the effective ingredient content of the crops. The data linkage mechanism between modules overcomes the interference of environmental variables, improves the accuracy of irrigation decisions to more than 93%, reduces the amount of invalid pesticide application by 40%, and ensures the uniformity of crop quality indicators.

[0120] In a third aspect, the present application discloses a storage medium storing a computer program capable of being loaded by a processor and executing an intelligent control method of high-pressure atomizing irrigation.

[0121] Among them, the storage medium solidifies the intelligent control method into an executable program, forming a stable algorithm running carrier. This medium can be compatible with control method versions including dynamic selection of destructive sampling data, true value data fusion analysis, and self-learning mechanism of irrigation rule base, such as versions supporting integrated pest predictive prevention and control modules or versions of cross-cycle dynamic optimization strategies. During program execution, each sampling data, rule update record and execution effect feedback is recorded completely, forming a traceable data processing chain to provide basic data support for subsequent algorithm iteration. By executing the irrigation decision-making process through programming, the irrigation rule base is continuously optimized, for example, after adjusting the rule confidence each time, the reinforcement learning process is automatically triggered to update the candidate rule set.

[0122] As a preferred embodiment, the storage medium adopts an embedded flash chip of UFS3.1 specification, which is connected with a processor of ARM Cortex-A72 architecture through a PCIe interface. A computer program is stored in the medium in the form of a binary executable file, which contains an instruction sequence for the high-pressure atomization irrigation intelligent control method of the first aspect, and the instruction sequence is divided into a data acquisition thread, a rule generation thread and an execution control thread according to a preset priority. When the program is loaded into the L3 cache of the processor, the destructive sampling time sequence table, the truth data fusion algorithm parameters and the irrigation rule confidence threshold are loaded into the working memory through memory mapping. In the process of program execution, the storage medium continuously records the difference data packet before and after each rule library update, and the difference data packet contains the logical expression of the newly added rule, the hash value of the eliminated rule and the statistical information of the rule trigger number. When the program is running, if it is detected that the currently loaded control method version contains a pest prediction prevention module, the microenvironment sensor data stream analysis interface is automatically called; if it is detected that the cross-cycle dynamic optimization strategy is activated, the digital twin model parameter synchronization process is started.

[0123] The above are preferred embodiments of the present application, and are not intended to limit the protection scope of the present application, therefore: any equivalent changes made on the structure, shape, principle of the present application shall be covered within the protection scope of the present application.

Claims

1. An intelligent control method of high-pressure atomized irrigation, characterized in that: an irrigation area is divided into a plurality of management partitions, and a plurality of crops in each management partition are randomly selected and defined as sacrifice samples; destructive sampling is performed on the sacrifice samples according to a preset sampling timing to obtain crop tissue samples, and a broken physiological and biochemical index detection is performed on the crop tissue samples to obtain at least one true value data representing a water state, a nutrient state or a health state of the crops; an irrigation rule library is established, a mapping relationship between a true value data range and an irrigation control instruction is defined through the established irrigation rule library, and true value data is obtained based on a current sampling timing, and individualized irrigation rules are generated or updated for each management partition; based on the latest individualized irrigation rules, in combination with real-time environmental data obtained by an environmental sensor, variable control instructions for variable watering, variable fertilization, variable dosing or variable insecticide for each management partition are generated; a high-pressure atomized sprinkler irrigation system is controlled to execute the generated variable control instructions.

2. The intelligent control method of high-pressure atomizing irrigation according to claim 1, characterized in that, The irrigation rule library establishment method comprises: fusing and standardizing the true value data obtained by the current sampling, historical sampling data, historical irrigation operation records and corresponding environmental data to construct a spatiotemporal correlation data set; using an association rule learning algorithm to mine the spatiotemporal correlation data set to generate candidate rules with the true value data and environmental data as condition items and the irrigation control instruction as a result item; calculating the success probability of achieving the expected effect after the execution of each candidate rule based on the historical sampling data and the historical irrigation operation records, and taking the success probability as the confidence of the association rule learning algorithm; adding the candidate rules with a confidence higher than a preset threshold to the irrigation rule library, and marking the applicable crop growth period and environmental condition range for each candidate rule. The generation of the variable fertilization control instruction comprises:

3. The intelligent control method of high-pressure atomizing irrigation of claim 2, wherein, obtaining physiological and biochemical detection results of the sacrifice samples based on the true value data, and extracting measured content values of a plurality of target nutrient components of the sacrifice samples; comparing the measured content values of the plurality of target nutrient components with the optimal nutrient target interval of the current crop species and growth period one by one to obtain the profit and loss amount of each nutrient component, the profit and loss amount including a missing item and an excess item; inputting the profit and loss amount into a nutrient trade-off optimization model to obtain an optimal fertilizer formula and application amount; generating a variable fertilization control instruction according to the optimal fertilizer formula and application amount. The generation of the variable control instructions for variable watering, variable fertilization, variable dosing or variable insecticide for each management partition comprises:

4. The intelligent control method of high-pressure atomizing irrigation of claim 1, wherein, automatically obtaining a crop canopy multispectral image through an image acquisition device deployed in the management partition, processing the canopy multispectral image based on a target detection model, identifying the type of insect pests and marking the coordinates of the insect pest occurrence concentration points in the image; fusing the coordinates of the concentration points with the GIS geographic information of the management partition to generate a digital map of insect pest distribution of the management partition, and marking the type and severity level of the insect pests with different coordinate information on the digital map of insect pest distribution; planning a running path for the high-pressure atomized irrigation system according to the digital map of insect pest distribution. ​ According to the operation path, the high-pressure atomization irrigation system is controlled to open the spray head when located at the current pest concentration point and to close the spray head when located at the non-pest concentration point.

5. The intelligent control method of high-pressure atomizing irrigation of claim 4, wherein, The method also includes a pest predictive prevention and control method, which specifically includes: Distributing a micro-environment sensor network in the management partition to continuously monitor and obtain micro-environment data of temperature, humidity, light intensity and volatile organic compound concentration inside the crop canopy; Correlatively analyzing the micro-environment data and historical pest occurrence records, using a time series prediction model to predict the probability and hotspot area of pest occurrence in a specific future time period; When the predicted probability of pest occurrence exceeds a preset risk threshold, automatically generating a preventive variable pesticide application instruction; Spraying a preventive pesticide liquid in the predicted pest hotspot area through the high-pressure atomization sprinkler irrigation system.

6. The intelligent control method of high-pressure atomizing irrigation of claim 1, wherein, After the high-pressure atomization irrigation system executes the variable control instruction, the method further includes: After the execution of the variable control instruction, the sacrificial sample is sampled again in the same management partition at the next sampling time sequence to obtain verified data; Calculating the relative improvement rate η of the verified data and the true value data before the execution of the variable control instruction on the key indicators, and the calculation formula is η=( - ) / ×100% wherein, is the measured value of the indicator in the post-data, is the measured value of the indicator in the true-value data; comparing the calculated improvement rate with a preset expected improvement rate threshold value in the personalized irrigation rule , ] If If the decision effect is not as expected, the system automatically applies a confidence penalty to the personalized irrigation rule that triggered the decision and marks it as a rule to be optimized.​ If ≤ η ≤ then the personalized irrigation rule confidence is maintained; If < / , it is determined that the decision effect is significant, and the confidence of the personalized irrigation rule is improved. For the to-be-optimized rule, the system starts a reinforcement learning process, stores the corresponding decision-making process data of this time as a failure case in a specific data set, and uses the failure case to drive the irrigation rule base to preferentially learn such situations when mining candidate rules next time.

7. The intelligent control method of high-pressure atomizing irrigation of claim 1, wherein, The method also includes a cross-cycle dynamic optimization strategy for the ultimate quality target, which includes: Ultimate quality target setting: setting an explicit ultimate quality target vector for the current planting batch at the beginning of the crop growing cycle [ , ,..., ], wherein , represent the concentration equivalent quantification indicators of specific functional ingredients; Reverse path planning: using the crop growth digital twin model, the ultimate quality target vector is decomposed into intermediate physiological state target intervals for each key growth stage i Reverse decomposition to each key growth stage, calculate the intermediate physiological state target interval that needs to be reached for each growth stage i [ , ] Real-time tracking and correction: after the sacrifice sampling at each gestation period i, the true value data obtained is compared with the intermediate physiological state target interval of the current period Comparison is made to generate variable control instructions; The objective function F of the variable control instruction is defined as the minimization of the weighted square sum deviation of the current state and the target state, i.e. = ， wherein, represent different physiological and biochemical indicators, are the importance weights of the corresponding indicators . The implementation of the cross-cycle dynamic optimization strategy causes the crop growth trajectory to eventually converge on the ultimate quality target vector .

8. The intelligent control method of high-pressure atomizing irrigation of claim 5, wherein, The soil also needs to be sampled and analyzed: While destructively sampling the sacrificial sample, a soil sample in the rhizosphere region of the current sacrificial sample is collected; The soil sample in the rhizosphere region is subjected to metagenomic sequencing analysis to detect the functional gene abundance of pathogenic bacteria in the soil and analyze the soil microbial community structure; A soil-borne disease occurrence risk index R is established, and the calculation formula is ， wherein, is the abundance of the specific pathogen gene, For the abundance of beneficial microorganisms genes, a stress factor calculated from soil humidity, temperature; When the risk index R exceeds a safety threshold, it is determined that the management partition has a high risk of soil-borne disease, and the system generates a variable pesticide application instruction; The high-pressure atomization sprinkler irrigation system is controlled to apply a functional water solution with soil disinfection function to the rhizosphere region of the crop to regulate the soil micro-ecology.

9. A high-pressure atomizing irrigation system applied to the intelligent control method of high-pressure atomizing irrigation according to any one of claims 1-8, characterized in that, The method includes: A first module for randomly selecting a number of crops in each management partition as sacrificial samples; A second module for destructively sampling the sacrificial samples according to a preset sampling time sequence to obtain crop tissue samples, and performing broken physiological and biochemical index detection on the crop tissue samples to obtain at least one true value data representing the water state, nutrient state or health state of the crop; A third module for establishing an irrigation rule base to define the mapping relationship between the true value data range and the irrigation control instruction through the established irrigation rule base, and to generate or update individualized irrigation rules for each management partition based on the current sampling time sequence. The fourth module is configured to generate variable control instructions for variable watering, variable fertilization, variable dosing or variable insect killing of each management partition based on the latest personalized irrigation rules and real-time environmental data obtained by the environmental sensors. The fifth module is configured to control the high-pressure atomizing sprinkling irrigation system to execute the generated variable control instructions.

10. A storage medium, characterized by A computer program is stored, which can be loaded by a processor and execute the intelligent control method of the high-pressure atomizing irrigation according to any one of claims 1-8.

Images