An AI-based agricultural automation linkage control method and control system

Abstract

This invention proposes an artificial intelligence-based agricultural automation linkage control method and control system, belonging to the field of agricultural automation control technology. The method includes: acquiring agricultural environmental data and agricultural automation equipment data; preprocessing the acquired agricultural environmental data; extracting features from the preprocessed agricultural environmental data to construct an agricultural environmental feature vector; and analyzing the agricultural environmental state based on the agricultural environmental feature vector to obtain an agricultural environmental state assessment result. Through preliminary control simulation, reinforcement learning algorithm optimization of the control strategy, and linkage control simulation, the equipment control strategy is continuously iterated and optimized to make the control strategy more aligned with actual production needs, improve the accuracy and effectiveness of agricultural automation equipment control, reduce manual intervention, and lower production costs.

Description

Technical Field

[0001] This invention proposes an agricultural automation linkage control method and control system based on artificial intelligence, belonging to the field of agricultural automation control technology. Background Technology

[0002] Traditional agricultural production methods rely on manual experience for irrigation, fertilization, and environmental regulation. This is not only labor-intensive but also makes it difficult to accurately grasp the various conditions required for plant growth, resulting in low production efficiency and serious resource waste. While the development of agricultural automation equipment has improved production efficiency to some extent, the lack of effective coordination and linkage between different devices prevents comprehensive control based on the real-time needs of plant growth. Furthermore, current technologies have not adequately considered creating a simulated wild growing environment, which is detrimental to improving plant quality. Summary of the Invention

[0003] This invention provides an artificial intelligence-based agricultural automation linkage control method and control system to solve the problems mentioned in the background section above:

[0004] This invention proposes an artificial intelligence-based agricultural automation linkage control method, the method comprising:

[0005] S1. Acquire agricultural environmental data and agricultural automation equipment data; preprocess the acquired agricultural environmental data; extract features from the preprocessed agricultural environmental data to construct agricultural environmental feature vectors; perform agricultural environmental status analysis based on the agricultural environmental feature vectors to obtain agricultural environmental status assessment results; perform correlation analysis on agricultural automation equipment based on the agricultural environmental status assessment results and agricultural automation equipment data to obtain the correlation relationships between agricultural automation equipment; integrate the agricultural automation architecture based on the correlation relationships between agricultural automation equipment to obtain the agricultural automation linkage control architecture.

[0006] S2. Perform image recognition processing on agricultural environmental data, extract plant growth parameters based on the recognition results, combine different plant growth models and historical data, use machine learning algorithms to predict the growth trend of plant growth parameters, and obtain plant growth trend data; perform environmental demand analysis on agricultural environmental data based on plant growth trend data, and obtain plant growth environment demand data; classify the plant growth environment demand data to obtain classified environmental demand data.

[0007] S3. Conduct preliminary control simulations of agricultural automation equipment using an agricultural automation linkage control architecture to obtain preliminary control simulation data; based on the preliminary control simulation data and plant growth environment requirements data, optimize the control strategy using reinforcement learning algorithms to obtain an optimized equipment control strategy; conduct linkage control simulations of agricultural automation equipment based on the optimized equipment control strategy to obtain linkage control simulation data.

[0008] S4. Obtain agricultural production target data, conduct target feasibility assessment on agricultural production target data based on plant growth trend data, and obtain target feasibility assessment data; based on target feasibility assessment data and classified environmental demand data, use deep learning algorithms to predict resource demand and obtain resource demand prediction data; combine resource demand prediction data and equipment linkage control simulation data to optimize resource allocation and obtain an optimized resource allocation scheme.

[0009] S5. Based on the optimized resource allocation scheme and equipment linkage control simulation data, the agricultural production target data is further refined and decomposed to obtain equipment control target data; based on the equipment control target data and the agricultural automation linkage control architecture, the equipment control parameters are optimized using a genetic algorithm to obtain optimized equipment control parameters; based on the optimized equipment control parameters, an agricultural automation linkage control strategy is generated, and the control strategy is transmitted to the agricultural automation equipment control system to execute the linkage control task.

[0010] The artificial intelligence-based agricultural automation linkage control system proposed in this invention includes:

[0011] One or more processors;

[0012] Memory, used to store one or more programs.

[0013] Wherein, when the one or more programs are executed by the one or more processors, the one or more processors are made to implement the method described in any one of the above.

[0014] The beneficial effects of this invention are as follows: by acquiring agricultural environmental data and preprocessing it, extracting features to construct feature vectors, and then analyzing the state of the agricultural environment, we can comprehensively and accurately grasp the state of the agricultural environment, providing a reliable basis for subsequent precise control, avoiding control errors caused by incomplete or inaccurate environmental information, and helping to improve the stability and quality of agricultural production.

[0015] By using image recognition technology to obtain plant growth parameters, and combining them with growth models and historical data to predict growth trends, we can analyze and classify the environmental requirements of plant growth. This allows us to understand the plant growth status and required environmental conditions in advance, providing strong support for precise regulation of the agricultural environment to meet plant growth needs, and helping to improve crop yield and quality.

[0016] By using preliminary control simulation, reinforcement learning algorithms to optimize control strategies, and linkage control simulation, the equipment control strategies are continuously iterated and optimized to make the control strategies more in line with actual production needs, improve the accuracy and effectiveness of agricultural automation equipment control, reduce manual intervention, and lower production costs.

[0017] Assessing the feasibility of agricultural production goals based on plant growth trends, predicting resource demand using deep learning algorithms, and optimizing resource allocation by combining equipment linkage control simulation data can rationally allocate agricultural production resources, avoid over-investment or under-investment, improve resource utilization efficiency, and achieve sustainable development of agricultural production.

[0018] By breaking down agricultural production goals into detailed steps and combining them with genetic algorithms to optimize equipment control parameters, agricultural automation linkage control strategies can be generated and executed. This can transform macro-level production goals into specific and operable equipment control instructions, ensuring that agricultural automation equipment operates according to optimal parameters, accurately achieving agricultural production goals, and improving the level of refined management in agricultural production. Attached Figure Description

[0019] Figure 1 This is a diagram illustrating the steps of the method described in this invention. Detailed Implementation

[0020] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] One embodiment of the present invention, for example Figure 1 As shown, an artificial intelligence-based agricultural automation linkage control method includes:

[0022] S1: Acquire agricultural environmental data and agricultural automation equipment data. The agricultural environmental data includes temperature, humidity, light intensity, air pressure, and air quality. The agricultural automation equipment data includes the operating status and parameters of drones, automatic irrigation systems, intelligent greenhouse control equipment, etc. Preprocess the acquired agricultural environmental data. Use machine learning algorithms (e.g., random forest algorithm) to extract features from the preprocessed agricultural environmental data and construct agricultural environmental feature vectors. Analyze the agricultural environmental status based on the agricultural environmental feature vectors to obtain agricultural environmental status assessment results. Based on the agricultural environmental status assessment results and agricultural automation equipment data, perform correlation analysis on agricultural automation equipment using a deep learning model (e.g., graph neural network model) to obtain the correlation relationships between agricultural automation equipment. Integrate the agricultural automation architecture based on the correlation relationships between agricultural automation equipment to obtain an agricultural automation linkage control architecture.

[0023] S2: Perform image recognition processing on agricultural environmental data, using image recognition technology to identify the growth status, pest and disease conditions of different plants; extract plant growth parameters based on the recognition results, including plant height, leaf color, and number of flowers; combine different plant growth models and historical data, and use machine learning algorithms (such as convolutional neural networks) to predict the growth trends of plant growth parameters, obtaining plant growth trend data; perform environmental demand analysis on agricultural environmental data based on plant growth trend data, obtaining plant growth environment demand data; classify the plant growth environment demand data, distinguishing between conventional environmental demands and special environmental demands, such as the requirements of simulated wild environments like dawn / dusk, fog, day / night temperature difference, and air pressure difference, obtaining classified environmental demand data;

[0024] S3: Conduct preliminary control simulations of agricultural automation equipment using an agricultural automation linkage control architecture. Simulate the operation of different equipment under different control strategies to obtain preliminary control simulation data. Based on the preliminary control simulation data and plant growth environment requirements data, optimize the control strategy using reinforcement learning algorithms (such as policy gradient methods). Continuously adjust the control parameters of the equipment to maximize the satisfaction of plant growth requirements and obtain the optimized equipment control strategy. Based on the optimized equipment control strategy, conduct linkage control simulations of the agricultural automation equipment, considering the synergistic effects and mutual influences between equipment, to obtain linkage control simulation data.

[0025] S4: Acquire agricultural production target data, including yield and quality targets; conduct a feasibility assessment of the agricultural production target data based on plant growth trend data, analyze the possibility of achieving the production targets under current environmental conditions, and obtain target feasibility assessment data; based on the target feasibility assessment data and the classified environmental demand data, use a deep learning algorithm to predict resource demand, including predicting the amount of irrigation water, fertilizer, energy consumption, and other resources required to achieve the production targets, and obtain resource demand prediction data; combine the resource demand prediction data and equipment linkage control simulation data to optimize resource allocation and obtain an optimized resource allocation scheme.

[0026] S5: Based on the optimized resource allocation scheme and equipment linkage control simulation data, the agricultural production target data is further decomposed into specific control targets for each piece of equipment to obtain equipment control target data; based on the equipment control target data and the agricultural automation linkage control architecture, the equipment control parameters are optimized using a genetic algorithm (e.g., adaptive genetic algorithm) to further adjust the equipment operating parameters to achieve precise equipment control and obtain optimized equipment control parameters; based on the optimized equipment control parameters, an agricultural automation linkage control strategy is generated and transmitted to the agricultural automation equipment control system to execute the linkage control task.

[0027] The working principle and effects of the above technical solution are as follows: By accurately analyzing agricultural environmental data and applying optimized control strategies, precise linkage of agricultural automated equipment can be achieved, reducing resource waste and improving the efficiency of agricultural production; by adopting intelligent algorithms and automated control, the dependence on manual operation is reduced, labor costs are lowered, and a stable agricultural production environment can be maintained for a relatively long period of time.

[0028] Through optimized equipment linkage control strategies, it is possible to simulate and realize a wild-like plant growth environment, meet the special environmental needs of different plants, and improve plant growth quality and yield. Through resource demand prediction and resource allocation optimization, it is possible to ensure the precise allocation of resources such as irrigation and fertilization, avoid overuse of resources, reduce energy consumption, and improve resource utilization efficiency.

[0029] Based on consideration of plant growth trends and environmental requirements, a feasibility assessment of production targets and a forecast of resource needs were conducted, further increasing the likelihood of achieving the targets and ensuring the realization of agricultural production goals. Through the combined application of deep learning and reinforcement learning algorithms, the synergy between different agricultural automation equipment was enhanced, avoiding conflicts and redundant operation between equipment and ensuring efficient collaboration. By precisely controlling the plant growth environment, the growth status and health of plants were improved, thereby improving plant quality and market competitiveness while ensuring yield.

[0030] In one embodiment of the present invention, S1 includes:

[0031] S11. Real-time collection of agricultural environmental data is achieved through a distributed sensor network. The distributed sensors are deployed in different areas of the field and greenhouse. The agricultural environmental data includes temperature (accurate to ±0.5℃), humidity (accurate to ±2%RH), light intensity (unit: lux), spectrum (unit: nm), photosynthetic photon flux density (unit: umol / m2 / s), air pressure (unit: hPa), and air quality (including indicators such as PM2.5 and CO2 concentration). Data from agricultural automation equipment is collected through an IoT interface, which includes LoRa and 4G modules. The agricultural automation equipment data includes the operating status and parameters of drones (flight trajectory, battery life, spray flow rate), automatic irrigation systems (pump power, pipeline pressure, valve on / off status), and intelligent greenhouse control equipment (ventilation fan speed, shading net opening / closing degree), forming a raw data set.

[0032] S12. Perform multi-level cleaning on the collected agricultural environmental data to remove outliers caused by sensor malfunctions (identified by the 3σ principle); perform interpolation on missing values ​​(linear interpolation for time series and Kriging interpolation for spatial distribution); standardize data of different units and magnitudes (mapping to the [0,1] interval); and finally generate a well-organized agricultural environmental preprocessing dataset.

[0033] S13. Use machine learning algorithms to extract features from the preprocessed agricultural environmental data and screen out core features that are strongly correlated with plant growth. The core features include the daily temperature range, the duration of sustained high humidity, and the cumulative amount of light. Through principal component analysis, compress the high-dimensional environmental data into low-dimensional feature vectors with an information retention rate of ≥90%. Each vector contains 8-12 core environmental features, forming an agricultural environmental feature vector.

[0034] S14. Based on agricultural environmental feature vectors, construct a multi-dimensional assessment model, set environmental suitability indicators, including suitable temperature ranges for crop growth and light thresholds, calculate the environmental suitability score corresponding to each feature vector using fuzzy comprehensive evaluation method; combine the correlation between historical environmental data and crop growth to classify environmental status levels (excellent, medium, poor), and generate agricultural environmental status assessment results (including specific indicator scores and comprehensive level).

[0035] S15. Input the agricultural environmental status assessment results and agricultural automation equipment data into a deep learning model, construct an equipment association graph with equipment as nodes and the collaborative relationship between equipment (e.g., the linkage between irrigation system and greenhouse ventilation equipment to regulate humidity) as edges; learn the mapping relationship between environmental status changes and equipment operation through model training, identify strong association rules between equipment (e.g., the collaborative start probability of ventilation fan and shade net at high temperature), and obtain the association relationship of agricultural automation equipment.

[0036] S16. Based on the equipment relationships, design a hierarchical linkage control architecture, where the bottom layer is the equipment execution layer (directly controlling equipment actions), the middle layer is the collaborative control layer (handling the linkage logic between equipment), and the top layer is the decision layer (receiving environmental assessment results and issuing instructions); determine the data interaction protocol for each layer (e.g., the MQTT protocol for communication between the equipment layer and the control layer), and integrate them to form an agricultural automation linkage control architecture that can dynamically adapt to environmental changes.

[0037] The working principle and effects of the above technical solution are as follows: By accurately collecting agricultural environmental data, the system can more efficiently adjust the equipment status, realize automated management, reduce manual intervention, optimize the production process, and thus improve the overall production efficiency; through the intelligent control system, irrigation, fertilization and other resources can be allocated on demand, avoiding overuse or waste, reducing energy consumption and improving resource utilization efficiency.

[0038] The system optimizes the collaboration between devices through deep learning algorithms, reducing conflicts and redundant operations and ensuring efficient and coordinated operation. By adjusting environmental parameters in real time, it provides a more suitable growth environment for plants, thereby improving their health and quality.

[0039] Because the system can automatically monitor and adjust the agricultural environment, it reduces reliance on manual operation, thereby lowering labor costs and ensuring production stability over a long period of time. By rationally allocating resources and optimizing equipment linkage, agricultural production can be carried out in a more sustainable manner, reducing excessive dependence on natural resources and contributing to environmental protection.

[0040] In one embodiment of the present invention, S15 includes:

[0041] Extract specific indicator scores, comprehensive grades, and corresponding timestamps from the agricultural environmental status assessment results. Simultaneously extract operating status parameters, equipment identifiers, and collection timestamps from agricultural automation equipment data. Use timestamp matching to associate and bind the two types of data, remove data pairs with time misalignments exceeding 10 seconds, and generate a fused dataset.

[0042] Based on the agricultural environmental status assessment results (excellent, medium, poor) and specific environmental indicators (such as high temperature and high humidity), the equipment operation status in the fused dataset is classified into scenarios, such as "equipment collaboration in high temperature environment (temperature ≥35℃)" and "equipment collaboration in high humidity environment (humidity ≥85%)", and the list of equipment participating in collaboration in different scenarios is determined.

[0043] Using agricultural automation equipment involved in the scenario as nodes, the equipment type and operating parameter range are used as node attributes; based on the consistency of the actions of the equipment at the same timestamp (e.g., starting up at the same time, adjusting parameters at the same time), initial association edges are established between nodes, and the initial association frequency of the edges (the number of times the same action occurs synchronously) is marked.

[0044] The initial graph of device associations, which includes node attributes and initial associated edges, and the environmental state data of the corresponding scenario are input into the graph neural network model. The model learns the mapping relationship between environmental state changes and the strength of device association edges, and optimizes the weight values ​​of the associated edges through model iteration (the weight values ​​are dynamically adjusted according to the frequency of collaboration and the effect of environmental improvement).

[0045] Based on a preset weight threshold (≥0.8), the associated edges that meet the optimized weight values ​​are selected. Combined with the environmental conditions of the corresponding scenario, strong association rules between devices are extracted. For example, "when the environmental assessment level is poor and the temperature index exceeds the standard, the collaborative start weight of ventilation fan and shade net is ≥0.85", which are then integrated to form the association relationship of agricultural automation equipment.

[0046] The working principle and effect of the above technical solution are as follows: By classifying scenarios and refining device association rules, the system can optimize device collaboration strategies according to different environmental states, making device cooperation more accurate and efficient, and reducing conflicts and incoordination in device operation; Through the learning and iterative optimization of graph neural network models, the system can dynamically adjust the weight values ​​of device association edges to ensure the accuracy of device collaboration under different environmental states, thereby effectively reducing errors in the environmental control process.

[0047] By utilizing optimized equipment association rules, the system can automatically trigger coordinated equipment adjustment when environmental factors such as temperature and humidity exceed the standard, greatly improving the intelligent linkage of equipment and avoiding the independent operation of equipment under different environmental conditions, thereby improving the overall level of automation control. Through precise equipment collaboration and optimized scene classification, the system can respond to environmental changes in real time, making agricultural production environment more adaptable and helping to improve the stability and suitability of crop growth environment.

[0048] In equipment collaboration, by screening and optimizing collaboration strategies, the system can reduce unnecessary equipment operation and energy consumption, further improve energy utilization efficiency, and reduce ineffective equipment consumption. Through the optimized collaboration of automated equipment and intelligent analysis of environmental status assessment results, the degree of automation in agricultural production has been further improved, reducing the need for manual intervention and achieving efficient and intelligent agricultural environmental management.

[0049] In one embodiment of the present invention, S2 includes:

[0050] S21. Collect plant growth images regularly using high-definition cameras (deployed on poles in the field or carried by drones) (once at 9 am and once at 3 pm daily), covering panoramic views of the plant, close-ups of leaves, and details of flowers and fruits; preprocess the images to generate a standardized growth image dataset;

[0051] S22. Use a deep learning image recognition model to analyze the preprocessed image, and use a trained feature extraction network (e.g., ResNet50) to identify the plant growth status (leaf spread, stem uprightness) and pest and disease conditions (lesion shape, insect outline); output the identification result labels (e.g., "yellow leaves", "aphid infestation", "vigorous growth"), and mark the coordinates of abnormal areas;

[0052] S23. Based on the image recognition results, extract quantitative growth parameters through image measurement algorithms, calculate plant height based on binocular vision (error ≤ 2cm); analyze leaf color through HSV color space (extract the mean values ​​of R, G, and B channels); count the number of flowers using target detection boxes (accuracy ≥ 95%), and form a set of plant growth parameters containing multi-dimensional parameters.

[0053] S24. Input the plant growth parameters and the corresponding crop growth model (e.g., the Logistic growth model for rice and the accumulated temperature growth model for tomatoes) and historical growth data (parameter change curves for the last 3 growth cycles) into the machine learning prediction model, train the model to learn the change pattern of parameters over time (in days), predict the trend of growth parameter changes in the next 15 days (e.g., daily increase in plant height and change in leaf color), and obtain plant growth trend data.

[0054] S25. Based on growth trend data, construct a growth status-environmental factor correlation model, analyze the environmental data corresponding to the "good growth" stage in the trend data, determine the key environmental factor thresholds for each growth stage (e.g., suitable temperature 20-25℃ for seedling stage, suitable humidity 60-70% for flowering stage); combine with crop physiological characteristics (e.g., the light requirements for photosynthesis) to obtain the environmental requirements that meet the growth trend and generate plant growth environment requirement data.

[0055] S26. Divide the plant growth environment requirements data into dimensions. Classify the requirements that meet the needs of conventional agricultural production (such as constant temperature and stable humidity during the day) as conventional environmental requirements; classify the special conditions that simulate natural ecology (such as gradual changes in light intensity during dawn and dusk, a day-night temperature difference of 5-8℃, and humidity fluctuations in foggy environments) as wild-simulated environmental requirements; verify the classification boundaries through the K-means clustering algorithm to obtain the classified environmental requirements data.

[0056] The working principle and effect of the above technical solution are as follows: By combining high-definition cameras and deep learning image recognition models, the system can accurately acquire various data on plant growth, especially in the monitoring of pests and diseases and analysis of growth status, ensuring that key parameters in the plant growth process can be captured in real time and accurately; traditional plant growth monitoring requires a lot of manual intervention, while this technical solution significantly reduces labor costs through automated image acquisition and analysis, especially in large-scale agricultural production, where it can save a lot of time and human resources.

[0057] By accurately identifying plant growth status and pest and disease conditions, the system can promptly detect potential problems and provide scientific decision support for agricultural managers, helping to take measures in advance to reduce losses caused by pests and diseases. Combining plant growth models and historical growth data, the system can predict growth trends for the next 15 days, enabling farmers to prepare in advance for future environmental changes and plant growth needs, optimize resource allocation, and improve crop yield and quality.

[0058] By establishing a correlation model between growth status and environmental factors, the system can clearly define the environmental requirements at different growth stages and adjust the environment based on real-time monitoring data, thereby ensuring that crops grow in the optimal environment and reducing the impact of adverse environmental factors on crop growth. Based on the growth characteristics of different crops, the system can intelligently adjust environmental requirements and production methods, generate demand data for both simulated wild environments and conventional environments, and help farmers develop personalized planting plans according to the specific needs of crops, thereby improving the level of intelligence in agricultural production.

[0059] In one embodiment of the present invention, step S23 includes:

[0060] Based on the identification result labels and abnormal region coordinates obtained from S22, feature regions related to plant height, leaf color and flower quantity are extracted from the standardized growth image dataset, and the specific location and range of each feature region in the image are determined.

[0061] Using stereo images of the feature region obtained by a binocular vision system, the spatial distance between the bottom and top of the plant in the feature region is calculated by combining the principles of stereo geometry. Through an error correction mechanism (ensuring that the error is ≤2cm), quantitative data of the plant height is obtained.

[0062] The image of the leaf in the feature region is converted from RGB color space to HSV color space, and the mean values ​​of the R, G, and B channels in this space are extracted as the quantization parameters of the leaf color.

[0063] For flowers in the feature region, based on the detection box generated by the target detection algorithm, the number of flowers is counted by the counting algorithm to ensure a statistical accuracy of ≥95% and obtain quantitative data on the number of flowers.

[0064] The relevant data, including plant height, leaf color quantification parameters, and flower quantity, are summarized and organized according to a unified data format to form a set of plant growth parameters containing multi-dimensional parameters.

[0065] The working principle and effect of the above technical solution are as follows: By combining binocular vision system and solid geometry principle, plant height can be accurately measured, ensuring that the error is controlled within 2cm, which greatly improves the accuracy and reliability of plant height data; through automated flower counting algorithm and leaf color quantification technology, deviations that may occur in manual measurement are avoided, making flower count and leaf color analysis more objective and accurate, and reducing the impact of human error.

[0066] Organizing multi-dimensional growth parameters into a standardized data format facilitates comparisons of crop growth data at different times and locations, enhancing data comparability and analytical depth. By extracting information from key characteristic regions of plants and quantifying it into standardized data, we can not only understand the growth status of plants in real time but also accurately reflect their health, helping agricultural managers to make timely adjustments and responses.

[0067] With the addition of this quantitative data, the accuracy of prediction models based on historical growth data and environmental factors can be further improved, helping farmers to more accurately predict future plant growth trends and optimize planting plans. By precisely quantifying each characteristic region, the impact of environmental changes on measurement results is reduced, ensuring the stability and availability of plant growth data, thereby making agricultural management decisions more scientific and rational.

[0068] In one embodiment of the present invention, S3 includes:

[0069] S31. Based on the agricultural automation linkage control architecture, initial control parameters are set for each agricultural automation device. The initial control parameters include: the initial spraying flow rate of the drone is set according to the crop density (e.g., 0.5L per square meter); the initial watering cycle of the automatic irrigation system is set according to the soil moisture threshold (starts when the soil moisture is below 60%); the initial parameters of the intelligent greenhouse control equipment are set according to the normal environmental requirements (e.g., the ventilation fan starts when the temperature is >30℃), forming a set of initial control parameters for the equipment.

[0070] S32. In the digital twin agriculture scenario (a three-dimensional virtual model built based on actual farmland / greenhouse), input the initial control parameters and classified environmental demand data, simulate the operation of different equipment under a single control strategy, set up multiple simulation scenarios (e.g., high temperature weather, rainy weather), record the equipment operation status (e.g., whether the drone sprays according to the trajectory, whether the irrigation system starts on time) and environmental response (e.g., temperature change curve in the greenhouse), and obtain preliminary control simulation data of the equipment.

[0071] S33. Using preliminary control simulation data and plant growth environment requirement data as input, construct a reinforcement learning model (with "meeting environmental requirements" as the reward function and "equipment energy consumption" as the penalty function). Through interaction with the environment (adjusting equipment control parameters), continuously optimize the strategy (e.g., when the temperature is high, first activate the shade net and then reduce the ventilation fan speed to reduce energy consumption). After 500 rounds of iterative training, the strategy satisfaction rate (environmental requirement compliance rate) is ≥90%, and the optimized equipment control strategy is obtained.

[0072] S34. Based on the relationships between agricultural automation equipment, formulate equipment linkage control rules, clarify the master-slave equipment logic (e.g., after a drone inspection detects pests and diseases, it automatically triggers a linkage command to add pesticides to the irrigation system); set a priority mechanism (e.g., under extreme weather conditions, ventilation equipment is activated before irrigation equipment); define parameter coordination thresholds (e.g., when the light intensity is <5000 lux, greenhouse supplemental lighting and ventilation fans are activated synchronously), and form a linkage control rule base;

[0073] S35. Load the optimized control strategy and linkage rules in the digital twin scenario to simulate the collaborative operation of multiple devices, simulate the joint operation of drones and irrigation systems (drones locate pest and disease areas, and irrigation systems apply pesticides precisely), and simulate the linkage of multiple devices in the smart greenhouse (coordinated adjustment of temperature, humidity, and light); record data such as response delay (≤2s), environmental indicator compliance rate, and resource consumption between devices to obtain simulation data of device linkage control.

[0074] The working principle and effects of the above technical solution are as follows: By simulating digital twin agricultural scenarios and applying reinforcement learning models, equipment control strategies are optimized, enabling equipment to respond to environmental demands more accurately and efficiently, thereby significantly improving the operating efficiency of agricultural automation equipment and the management level of crop growth; through the optimization of reinforcement learning models, equipment energy consumption is effectively controlled; and by formulating equipment linkage control rules, collaboration between different devices becomes smoother and more precise. For example, when a drone detects pests and diseases, the irrigation system can automatically add pesticides, avoiding manual intervention and delays, and improving operational efficiency.

[0075] By testing the collaborative operation of equipment in simulated scenarios and recording the response latency between devices and the compliance rate of environmental indicators, the rapid response of equipment in complex environments is ensured, making agricultural management more flexible and efficient. Through automated equipment linkage and precise control, errors and reliance on manual operation are reduced, ensuring the accuracy and consistency of all aspects of agricultural production and reducing the risk of human error.

[0076] By constructing digital twin models and using reinforcement learning optimization strategies, agricultural managers can obtain more scientific data support, adjust equipment parameters and management strategies in real time, thereby improving the rationality and effectiveness of decision-making.

[0077] In one embodiment of the present invention, S34 includes:

[0078] S341. Based on the functional dependence and cooperation strength between devices in the relationship between agricultural automation equipment, divide the hierarchy into master devices and slave devices, clarify the logical conditions for the master device to trigger the slave device's actions (for example, when a drone is used as the master device for detecting pests and diseases, it automatically sends a pesticide irrigation instruction to the irrigation system (slave device) when it detects a pest and disease density of ≥5 pests / m²), and establish a master-slave device trigger mapping table.

[0079] S342. Analyze the urgency of different agricultural production scenarios (e.g., extreme high temperature (≥40℃) and rainstorm warning are Level 1 emergency scenarios, and routine growth regulation is Level 3 scenario), and assign operating priorities to equipment according to scenario level. In Level 1 scenarios, ventilation and cooling equipment has a higher priority than irrigation equipment, and in Level 2 scenarios, irrigation equipment has a higher priority than fertilization equipment, thus forming a scenario-priority correspondence rule.

[0080] S343. Based on the correlation between environmental factors (such as the synergistic effect of temperature and humidity on transpiration), and combined with the data on the environmental requirements of plant growth, define the synergistic threshold of equipment parameters; when the temperature in the greenhouse is ≥32℃ and the humidity is ≥80%, the ventilation fan and dehumidifier start simultaneously; when the light intensity is in the range of 1000-3000 lux, the supplemental light lamp and CO2 generator are turned on in conjunction, and a parameter synergistic threshold comparison table is generated.

[0081] S344. The master-slave device logical division results, scene priority mechanism, and parameter coordination threshold are structurally integrated and the IF-THEN rule format is adopted (e.g., "IF drone detects pest area THEN irrigation system locates the area and applies pesticide at 1.2 times the standard concentration"). Through rule conflict detection (e.g., device priority conflict under different scenarios) and correction, a linkage control rule library is constructed.

[0082] The working principle and effects of the above technical solution are as follows: by clarifying the relationship between the master and slave devices and setting trigger conditions, the collaborative work of each device is ensured; by assigning priorities to devices according to environmental scenarios, the system can respond quickly and effectively handle emergencies; by defining parameter thresholds for device collaboration, the system ensures that devices such as fans, dehumidifiers, and CO2 generators work in coordination, reducing energy waste and maintaining the optimal environment for plant growth, thus promoting more sustainable agricultural practices.

[0083] By integrating equipment operations through IF-THEN logic rules, conflicts between devices are avoided, ensuring that devices start up in priority order, thus improving system stability and reliability. Through automatic adjustment and refined equipment settings, errors in manual operation are reduced, ensuring more consistent and precise crop care, thereby increasing crop yield. By providing real-time data and simulations, the system helps agricultural managers make more accurate decisions, optimize equipment operation, resource allocation, and environmental adjustments, and improve the overall effectiveness of agricultural management strategies.

[0084] In one embodiment of the present invention, S344 includes:

[0085] The results of the logical division of master and slave devices, scene priority mechanism, and parameter collaboration threshold are standardized and parsed and elements are extracted to generate a device-level element table, scene priority parameter set, and collaboration threshold element library; the device-level element table, scene priority parameter set, and collaboration threshold element library are associated and mapped to construct a multi-dimensional rule element association matrix.

[0086] The triggering conditions and execution actions in the multi-dimensional rule element association matrix are transformed into rules to generate initial IF-THEN rule entries (e.g., "IF greenhouse temperature ≥ 32℃ and humidity ≥ 80% THEN ventilation fan and dehumidifier start synchronously"); the initial IF-THEN rule entries are classified and coded to form rule clusters divided by equipment type and scene level;

[0087] Conflict pattern recognition is performed on the rule entries in the rule cluster. Through scenario priority comparison and parameter threshold cross-validation, a rule conflict list is generated (e.g., "priority conflict when ventilation equipment and irrigation equipment are triggered simultaneously in a first-level emergency scenario"). Based on the conflict types in the rule conflict list, and combined with the core needs of agricultural production (e.g., crop survival takes precedence over growth regulation), the rules are corrected and priorities are rearranged to obtain a set of conflict-free rules.

[0088] The set of conflict-free rules is validated for integrity and redundant rules are removed. Rule entries for edge scenarios (such as scenarios where extreme weather and equipment failure overlap) are added to generate an optimized set of rules. The optimized set of rules is then stored in a structured manner and indexed to form a linkage control rule library that can be dynamically called and expanded.

[0089] The working principle and effects of the above technical solution are as follows: By standardizing and extracting the logical division of master and slave devices, scene priority mechanism, and collaborative threshold elements, a device hierarchy element table, scene priority parameter set, and collaborative threshold element library are generated to ensure the standardization of device operation and system coordination; the multi-dimensional rule element association matrix is ​​transformed into clear IF-THEN rule entries and classified and coded according to device type and scene level, making device operation more automated, reducing human intervention, and improving the system's response speed and accuracy.

[0090] By identifying and correcting conflict patterns in the rule set and rearranging priorities in conjunction with core agricultural needs, equipment conflicts and priority confusion are avoided, ensuring smooth operation of the system in complex situations and reducing the potential risk of equipment damage. By verifying the integrity of the rule set and removing redundancies, and supplementing rule entries for edge scenarios, the rule system is made more complete and can cope with complex situations such as extreme weather and equipment failures, improving the system's adaptability and operational flexibility.

[0091] The optimized rule set is structured and indexed to form a dynamically callable and expandable linkage control rule base, ensuring efficient rule invocation and providing a solid foundation for future system expansion and upgrades. Through the dynamic expansion and intelligent management of the rule base, agricultural equipment can make optimal decisions based on different production scenarios and real-time environmental data, improving the overall intelligence level and adaptability of the agricultural system, and further enhancing the efficiency and sustainability of agricultural production.

[0092] In one embodiment of the present invention, step S4 includes:

[0093] S41. Collect production target data through the agricultural management system, including yield targets (e.g., rice yield ≥ 600 kg / mu) and quality targets (e.g., soluble solids content of tomatoes ≥ 5%). Break down the target data according to growth stages (e.g., set sub-targets for seedling stage, flowering stage, and fruiting stage respectively) and quantify them into calculable indicators (e.g., convert yield targets into "number of fruits per plant × weight of single fruit") to form a standardized production target dataset.

[0094] S42. Based on plant growth trend data, construct a feasibility assessment model, input the gap between the current growth trend (e.g., plant height growth rate) and the production target, calculate the probability of achieving the target at each stage (e.g., the probability of achieving the flowering period target = current number of flower buds / target number of flower buds); analyze limiting factors (e.g., insufficient light may lead to a decrease in the yield rate), and generate target feasibility assessment data (including the probability of achieving each stage and key limiting factors).

[0095] S43. Based on the target feasibility assessment data and the classified environmental demand data, screen resource demand prediction factors, including basic factors (such as water requirements of crop varieties and fertilization standards during the growing season), dynamic factors (such as irrigation increases by 5% for every 1°C increase in ambient temperature), and special factors (such as the additional water resources required for fog simulation in the simulated wild environment), to form a prediction factor library.

[0096] S44. Input the prediction factor library and historical resource consumption data (irrigation water volume and fertilizer application volume in the same period of the past 3 years) into the deep learning prediction model, train the model to learn the nonlinear relationship between factors and resource consumption, and optimize the model parameters through cross-validation (validation set accuracy ≥ 85%). Based on the optimized model, predict the irrigation water volume (broken down by day / week), fertilizer application volume (by nitrogen, phosphorus and potassium ratio), energy consumption (equipment operating power) required to achieve production targets, and obtain resource demand prediction data.

[0097] S45. Combining resource demand forecast data and equipment linkage control simulation data, construct a resource allocation optimization model with the objectives of "maximizing resource utilization" (e.g., irrigation water utilization ≥ 80%) and "minimizing cost". The constraints are the maximum load of equipment (e.g., the maximum flow rate of water pumps) and the upper limit of environmental carrying capacity (e.g., the amount of fertilizer does not exceed the soil absorption threshold). Use the particle swarm optimization algorithm (PSO) to solve for the optimal solution and output the resource allocation quota of each device (e.g., the daily pesticide usage of drones and the watering time of each area of ​​the irrigation system) to obtain the optimized resource allocation scheme.

[0098] The working principle and effects of the above technical solution are as follows: by breaking down production targets into growth stages and quantifying them into calculable indicators, a standardized production target dataset is formed, ensuring the clarity and feasibility of production targets, making agricultural production management more scientific and operable; by combining plant growth trend data to construct a feasibility assessment model, the probability of achieving targets at each stage can be assessed in real time, helping farm managers to understand the possibility of achieving targets in a timely manner and make corresponding adjustments based on constraints, thereby improving the success rate of achieving production targets.

[0099] By constructing a predictive factor library and combining it with historical data and deep learning models, the resource consumption required for each production target can be accurately predicted, reducing resource waste and providing a scientific basis for farm resource planning. Furthermore, by constructing a resource allocation optimization model and using particle swarm optimization to solve it, resource utilization is maximized and costs are minimized. The optimized resource allocation scheme not only improves the efficiency of equipment and resource utilization but also reduces unnecessary waste and lowers production costs.

[0100] By leveraging deep learning and particle swarm optimization algorithms, the prediction and allocation of agricultural resources have become more intelligent, avoiding the limitations of traditional manual prediction and allocation, and making the decision-making process more scientific, accurate, and efficient. Through rational resource allocation and maximizing resource utilization, energy and water waste has been reduced, promoting environmental protection and sustainable development. Simultaneously, optimized cost control has made agricultural production more economical, enhancing the overall sustainability of the agricultural system.

[0101] Through real-time data analysis and dynamic model adjustments, agricultural management systems can adapt to constantly changing environmental conditions and production goals, providing farm managers with greater flexibility and ensuring the stability and efficiency of the production process.

[0102] In one embodiment of the present invention, S45 includes:

[0103] Heterogeneous data fusion processing is performed on resource demand forecast data (irrigation water volume, fertilizer application, etc.) and equipment linkage control simulation data (equipment collaborative operation parameters, load thresholds, etc.) to unify data dimensions and measurement standards (e.g., converting "pesticide application" to "liters / hectare"); core influencing parameters (e.g., resource consumption rate, equipment collaborative efficiency coefficient) are extracted to generate a standardized fusion parameter matrix;

[0104] Based on the objectives of maximizing resource utilization and minimizing costs, a dual-objective function model is constructed. The resource utilization function includes sub-indicators such as effective irrigation water absorption rate and fertilizer conversion rate, while the cost function covers elements such as water resource procurement cost and equipment energy consumption cost. The boundary values ​​of the constraints are clearly defined, including equipment physical limits (e.g., ventilation volume corresponding to the maximum speed of the fan) and environmental safety thresholds (e.g., the upper limit of soil nitrogen content), forming a constraint parameter library.

[0105] A layered modeling architecture is adopted to build a resource allocation optimization model. The bottom layer is the data input layer (connecting and integrating parameter matrices and constraint parameter libraries), the middle layer is the algorithm operation layer (integrating the core module of particle swarm optimization algorithm), and the top layer is the result output layer (defining the data format of resource allocation quotas). Key algorithm parameters (such as particle population size and maximum number of iterations) are set to complete the model framework.

[0106] The particle swarm optimization algorithm is started, and the particle swarm is initialized (each particle represents a set of resource allocation schemes). The quality of the particles is calculated by the fitness function (which comprehensively evaluates the value of the dual objective function and the degree of constraint satisfaction). Iterative optimization is performed, and the particles update their position and velocity by tracking the individual optimal solution and the global optimal solution. After each round of iteration, schemes that violate the constraints are eliminated until the convergence condition is met (the optimal solution does not change significantly after 5 consecutive rounds of iteration).

[0107] The optimal solution output by the algorithm is verified in multiple scenarios (such as extreme weather and partial equipment failure scenarios) to evaluate the robustness of the solution; the resource allocation amount is fine-tuned based on the verification results (such as increasing the irrigation redundancy by 10% to cope with the risk of drought) to generate the final optimized solution; the solution is broken down into equipment-level execution instructions (such as "fertilizer A runs from 9:00 to 11:00 every day, with a nitrogen application rate of 2 kg / mu") to form a resource allocation execution list that can be directly invoked.

[0108] The working principle and effects of the above technical solution are as follows: By fusing heterogeneous data such as resource demand forecasting data and equipment linkage control simulation data, not only are data dimensions and measurement standards unified, but core influencing parameters are also extracted to generate a standardized fusion parameter matrix. This greatly improves the comparability between data, ensuring the consistency and accuracy of the resource management system. The construction of the dual-objective function model considers both maximizing resource utilization (e.g., effective absorption rate of irrigation water and fertilizer conversion rate) and minimizing costs (e.g., water resource procurement costs and equipment energy consumption costs). This goal-oriented optimization approach makes the farm's resource allocation more rational and reduces overall operating costs.

[0109] By combining Particle Swarm Optimization (PSO) with multi-scenario validation, the adaptability of the optimization scheme is ensured in various environments. For example, the algorithm can adjust according to extreme weather or equipment failure, improving the robustness and adaptability of the scheme and ensuring efficient operation even in uncertain environments. With a hierarchical modeling architecture, the model can achieve a completely automated process from data input to result output, reducing manual intervention and improving the intelligence level of decision-making. Furthermore, the application of PSO in resource allocation avoids the inefficiency and errors associated with traditional manual adjustments.

[0110] When generating the final optimized solution, the model addresses natural risks such as drought by adding redundancy, ensuring sufficient resource supply and environmental sustainability. This helps farm managers prevent potential risks and maximize the stability and continuity of production. By breaking down the final optimized solution into equipment-level execution instructions, resource allocation plans can be directly translated into operational commands, improving execution accuracy and efficiency. This automation and refinement of the process ensures that resource management is not limited to theoretical models but can be efficiently implemented in practical operations.

[0111] In one embodiment of the present invention, step S5 includes:

[0112] S51. Develop rules for breaking down overall objectives and divide responsibility units according to equipment functions (e.g., drones are responsible for pest and disease control, and irrigation systems are responsible for water supply); based on the contribution weight of each piece of equipment to the production objectives (e.g., the irrigation system has a weight of 30% for output), break down the overall output and quality objectives into specific control objectives for each piece of equipment (e.g., drones need to control the incidence of pests and diseases to ≤5%, and irrigation systems need to maintain soil moisture at 60-70%).

[0113] S52. Based on the disassembly rules and combined with plant growth trend data, quantify the control targets of each device. The control targets of the drone include "inspection coverage rate ≥98%" and "pesticide spraying error ≤5%"; the control targets of the automatic irrigation system include "watering uniformity ≥90%" and "response delay ≤1min"; and form the device control target data (including specific values ​​and achievement time limits).

[0114] S53. Based on the equipment control target data and the agricultural automation linkage control architecture, define the optimization range of control parameters for each piece of equipment, including the parameter range for UAVs (flight altitude 5-10m, spraying flow rate 0.3-0.8L / m²); and the parameter range for smart greenhouses (ventilation fan speed 500-1500r / min, shade net opening / closing degree 0-100%). Ensure that the parameter range covers the adjustment space required to achieve the target, forming a parameter constraint interval.

[0115] S54. Input the parameter constraint interval and equipment control target data into the genetic algorithm. Using the target achievement rate as the fitness function, through selection (retaining the top 30% of parameter combinations with fitness), crossover (parameter fragment recombination), and mutation (randomly adjusting parameter values) operations, after multiple rounds of iteration, the optimal control parameter combination (e.g., drone flight altitude 7m + spray flow rate 0.5L / m², greenhouse ventilation fan speed 800r / min) is selected to obtain the optimized equipment control parameters.

[0116] S55. Based on the optimized equipment control parameters and linkage rule base, generate a structured control strategy, including trigger conditions (e.g., execute cooling linkage when the temperature is >35℃), execution steps (first open the shade net → then turn on the ventilation fan → finally adjust the humidifier), and feedback mechanism (collect environmental data every 10 minutes to verify the effect); verify the effectiveness of the strategy through a digital twin scenario (target achievement rate ≥95%), and form the final control strategy.

[0117] S56. The final control strategy is transmitted to the agricultural automation equipment control system (such as PLC controller, edge computing node) through the Internet of Things gateway to trigger the equipment to execute in linkage; at the same time, the equipment operating status (such as valve opening and closing, motor speed) and environmental changes (such as temperature and humidity curves) are tracked in real time through the monitoring platform. When a deviation occurs (such as the actual humidity being 5% lower than the target value), the backup parameter adjustment strategy is automatically called.

[0118] The working principle and effects of the above technical solution are as follows: By dividing responsibility units according to equipment functions, the overall yield and quality targets are precisely broken down into specific control targets for each piece of equipment, enabling each piece of equipment to perform its tasks in a targeted manner. This not only improves the execution efficiency of each piece of equipment but also makes the role and contribution of each piece of equipment clearer, thereby contributing to the effective allocation of resources. By combining plant growth trend data, quantitative control targets are set for each piece of equipment, making the operating standards of equipment such as drones and irrigation systems clearer. This ensures that the execution effect of the equipment can be evaluated through specific indicators, promoting the efficient operation of agricultural automation equipment.

[0119] By applying genetic algorithms, equipment control parameters are optimized, and the optimal control combination is selected through multiple rounds of iteration. This process ensures that each piece of equipment achieves the best results in actual operation, reduces resource waste, and improves production efficiency. The generated structured control strategy not only includes triggering conditions, execution steps, and feedback mechanisms, but also uses digital twin scenarios to verify the effectiveness of the strategy, ensuring that the control strategy can be executed smoothly under different environmental conditions. Through the linkage of intelligent devices, more precise and real-time resource allocation and environmental regulation can be achieved, improving the overall level of intelligence in agricultural production.

[0120] By leveraging IoT technology, the system tracks equipment status and environmental changes in real time, enabling dynamic adjustments to the equipment. When environmental deviations occur, the system can automatically invoke backup parameter adjustment strategies to ensure the stability and continuity of agricultural production, reduce human intervention and errors, and improve the ability to respond to emergencies. By combining equipment control strategies with IoT gateways, the system achieves automated execution and autonomous decision-making in the production process, which not only improves work efficiency but also reduces human intervention, ensuring a high degree of automation and precise control of the production process.

[0121] The final control strategies and optimization schemes can provide flexible and efficient solutions for agricultural production. In particular, when facing risks such as extreme weather or equipment failure, they can ensure the stability of agricultural production through automatic adjustment and flexible response, which helps to promote the development of sustainable agriculture.

[0122] One embodiment of the present invention provides an artificial intelligence-based agricultural automation linkage control system, comprising:

[0123] One or more processors;

[0124] Memory, used to store one or more programs.

[0125] Wherein, when the one or more programs are executed by the one or more processors, the one or more processors are made to implement the method described in any one of the above.

[0126] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An artificial intelligence-based agricultural automation linkage control method, characterized in that, The method includes: S1. Acquire agricultural environmental data and agricultural automation equipment data; preprocess the acquired agricultural environmental data; extract features from the preprocessed agricultural environmental data to construct agricultural environmental feature vectors; perform agricultural environmental status analysis based on the agricultural environmental feature vectors to obtain agricultural environmental status assessment results; perform correlation analysis on agricultural automation equipment based on the agricultural environmental status assessment results and agricultural automation equipment data to obtain the correlation relationships between agricultural automation equipment; integrate the agricultural automation architecture based on the correlation relationships between agricultural automation equipment to obtain the agricultural automation linkage control architecture. S2. Perform image recognition processing on agricultural environmental data, extract plant growth parameters based on the recognition results, combine different plant growth models and historical data, use machine learning algorithms to predict the growth trend of plant growth parameters, and obtain plant growth trend data; perform environmental demand analysis on agricultural environmental data based on plant growth trend data, and obtain plant growth environment demand data; classify the plant growth environment demand data to obtain classified environmental demand data. S3. Conduct preliminary control simulations of agricultural automation equipment using an agricultural automation linkage control architecture to obtain preliminary control simulation data; based on the preliminary control simulation data and plant growth environment requirements data, optimize the control strategy using reinforcement learning algorithms to obtain an optimized equipment control strategy; conduct linkage control simulations of agricultural automation equipment based on the optimized equipment control strategy to obtain linkage control simulation data. S4. Obtain agricultural production target data, conduct target feasibility assessment on agricultural production target data based on plant growth trend data, and obtain target feasibility assessment data; based on target feasibility assessment data and classified environmental demand data, use deep learning algorithms to predict resource demand and obtain resource demand prediction data; combine resource demand prediction data and equipment linkage control simulation data to optimize resource allocation and obtain an optimized resource allocation scheme. S5. Based on the optimized resource allocation scheme and equipment linkage control simulation data, the agricultural production target data is further refined and decomposed to obtain equipment control target data; based on the equipment control target data and the agricultural automation linkage control architecture, the equipment control parameters are optimized using a genetic algorithm to obtain optimized equipment control parameters; based on the optimized equipment control parameters, an agricultural automation linkage control strategy is generated, and the control strategy is transmitted to the agricultural automation equipment control system to execute the linkage control task.

2. The artificial intelligence-based agricultural automation linkage control method according to claim 1, characterized in that, S1 includes: S11. Collect agricultural environmental data in real time through a distributed sensor network; collect data from agricultural automation equipment through the device IoT interface to form a raw data set; S12. Perform multi-level cleaning on the collected agricultural environmental data to remove outliers caused by sensor malfunctions; interpolate missing values; standardize data of different units and magnitudes; and finally generate a well-organized agricultural environmental preprocessing dataset. S13. Use machine learning algorithms to extract features from the preprocessed agricultural environmental data, screen out the core features that are strongly related to plant growth, and compress the high-dimensional environmental data into low-dimensional feature vectors with an information retention rate of ≥90% through principal component analysis. Each vector contains 8-12 core environmental features, forming an agricultural environmental feature vector. S14. Based on agricultural environmental feature vectors, construct a multi-dimensional assessment model, set environmental suitability indicators, and calculate the environmental suitability score corresponding to each feature vector through fuzzy comprehensive evaluation method; combine the correlation between historical environmental data and crop growth, classify environmental status levels, and generate agricultural environmental status assessment results. S15. Input the agricultural environmental status assessment results and agricultural automation equipment data into a deep learning model, construct an equipment association graph with equipment as nodes and the cooperation relationship between equipment as edges; learn the mapping relationship between environmental status changes and equipment operation through model training, identify strong association rules between equipment, and obtain the association relationship of agricultural automation equipment. S16. Based on the equipment relationships, design a hierarchical linkage control architecture; determine the data interaction protocols at each level, and integrate them to form an agricultural automation linkage control architecture that can dynamically adapt to environmental changes.

3. The artificial intelligence-based agricultural automation linkage control method according to claim 1, characterized in that, The S2 includes: S21. Regularly collect plant growth images using a high-definition camera; preprocess the images to generate a standardized growth image dataset; S22. Use a deep learning image recognition model to analyze the preprocessed image, identify the plant growth status and pest and disease conditions through a trained feature extraction network; output the recognition result labels and mark the coordinates of abnormal areas. S23. Based on the image recognition results, extract quantitative growth parameters through image measurement algorithms, calculate plant height based on binocular vision, analyze leaf color through HSV color space, and count the number of flowers using target detection boxes to form a set of plant growth parameters containing multi-dimensional parameters. S24. Input the plant growth parameters, the corresponding crop growth model, and historical growth data into the machine learning prediction model, train the model to learn the change pattern of parameters over time, predict the change trend of growth parameters in the next 15 days, and obtain plant growth trend data. S25. Based on growth trend data, construct a growth status-environmental factor correlation model; combine crop physiological characteristics to obtain the environmental requirements that meet the growth trend, and generate plant growth environment requirement data; S26. Divide the plant growth environment requirement data into dimensions, classify the requirements that conform to conventional agricultural production into conventional environmental requirements, classify the special conditions that simulate natural ecology into simulated wild environment requirements, and verify the classification boundary through the K-means clustering algorithm to obtain the classified environmental requirement data.

4. The artificial intelligence-based agricultural automation linkage control method according to claim 3, characterized in that, S23 includes: Based on the identification result labels and abnormal region coordinates obtained from S22, feature regions related to plant height, leaf color and flower quantity are extracted from the standardized growth image dataset, and the specific location and range of each feature region in the image are determined. Using stereo images of feature regions acquired by a binocular vision system, the spatial distance between the bottom and top of the plant in the feature region is calculated by combining the principles of stereo geometry. Through an error correction mechanism, quantitative data of plant height is obtained. The image of the leaf in the feature region is converted from RGB color space to HSV color space, and the mean values ​​of the R, G, and B channels in this space are extracted as the quantization parameters of the leaf color. For flowers in the feature region, based on the detection boxes generated by the target detection algorithm, the number of flowers is counted by the counting algorithm to obtain quantitative data on the number of flowers; The relevant data are summarized and organized according to a unified data format to form a set of plant growth parameters containing multiple dimensions.

5. The artificial intelligence-based agricultural automation linkage control method according to claim 1, characterized in that, The S3 includes: S31. Based on the agricultural automation linkage control architecture, set initial control parameters for each agricultural automation device to form a set of initial control parameters for the device. S32. In the digital twin agriculture scenario, input the initial control parameters and classified environmental demand data, simulate the operation of different equipment under a single control strategy, set up multiple simulation scenarios, record the equipment operation status and environmental response, and obtain preliminary control simulation data of the equipment. S33. Using preliminary control simulation data and plant growth environment requirement data as input, construct a reinforcement learning model, and continuously optimize the strategy through interaction with the environment to obtain an optimized equipment control strategy. S34. Based on the relationships between agricultural automation equipment, formulate equipment linkage control rules, clarify the master-slave equipment logic, set a priority mechanism, define parameter coordination thresholds, and form a linkage control rule library; S35. Load the optimized control strategy and linkage rules in the digital twin scenario, simulate the collaborative operation of multiple devices, and obtain simulation data of device linkage control.

6. The artificial intelligence-based agricultural automation linkage control method according to claim 5, characterized in that, S34 includes: S341. Based on the functional dependencies and collaboration strength between devices in the agricultural automation equipment relationship, divide the hierarchy into master and slave devices, clarify the logical conditions for the master device to trigger the slave device's action, and establish a master-slave device trigger mapping table. S342. Analyze the urgency of different agricultural production scenarios, assign operating priorities to equipment according to scenario level, and form scenario-priority correspondence rules; S343. Based on the correlation between environmental factors and combined with plant growth environment requirements data, define the coordination threshold of equipment parameters; generate a parameter coordination threshold comparison table. S344. The master-slave device logical division results, scene priority mechanism, and parameter coordination threshold are structurally integrated, and the IF-THEN rule format is adopted. Through rule conflict detection and correction, a linkage control rule library is constructed.

7. The artificial intelligence-based agricultural automation linkage control method according to claim 6, characterized in that, S344 includes: The results of the logical division of master and slave devices, scene priority mechanism, and parameter collaboration threshold are standardized and parsed and elements are extracted to generate a device-level element table, scene priority parameter set, and collaboration threshold element library; the device-level element table, scene priority parameter set, and collaboration threshold element library are associated and mapped to construct a multi-dimensional rule element association matrix. The triggering conditions and execution actions in the multi-dimensional rule element association matrix are transformed into rules to generate initial IF-THEN rule entries; the initial IF-THEN rule entries are classified and coded to form rule clusters divided by device type and scene level; Conflict pattern recognition is performed on the rule entries in the rule cluster. A rule conflict list is generated by comparing scene priority and cross-validating parameter thresholds. Based on the conflict types in the rule conflict list and combined with the core needs of agricultural production, the rules are modified and their priorities are rearranged to obtain a set of conflict-free rules. The set of conflict-free rules is validated for integrity and redundant rules are removed. Rule entries for edge scenarios are added to generate an optimized set of rules. The optimized set of rules is then stored in a structured manner and indexed to form a linkage control rule library that can be dynamically called and expanded.

8. The artificial intelligence-based agricultural automation linkage control method according to claim 1, characterized in that, The S4 includes: S41. Collect production target data through the agricultural management system, break down the target data according to growth stages, and quantify it into calculable indicators to form a standardized production target dataset. S42. Based on plant growth trend data, construct a feasibility assessment model, input the gap between the current growth trend and the production target, calculate the probability of achieving the target at each stage, analyze the limiting factors, and generate target feasibility assessment data. S43. Based on the target feasibility assessment data and the classified environmental demand data, screen resource demand prediction factors to form a prediction factor library. S44. Input the predictor library and historical resource consumption data into the deep learning prediction model, train the model to learn the nonlinear relationship between factors and resource consumption, and optimize the model parameters through cross-validation; based on the optimized model, obtain resource demand prediction data. S45. Combining resource demand forecast data and equipment linkage control simulation data, construct a resource allocation optimization model, use particle swarm optimization algorithm to solve for the optimal solution, output the resource allocation quota for each device, and obtain the optimized resource allocation scheme.

9. The artificial intelligence-based agricultural automation linkage control method according to claim 1, characterized in that, The S5 includes: S51. Formulate overall target decomposition rules and divide responsibility units according to equipment functions; based on the contribution weight of each piece of equipment in the production target, decompose the overall output and quality targets into specific control targets for each piece of equipment; S52. Based on the disassembly rules and combined with plant growth trend data, quantify the control targets of each device to form device control target data; S53. Based on the equipment control target data and the agricultural automation linkage control architecture, define the optimization range of each equipment control parameter and form a parameter constraint interval; S54. Input the parameter constraint range and equipment control target data into the genetic algorithm. Through selection, crossover, and mutation operations, and after multiple rounds of iteration, the optimal combination of control parameters is selected to obtain the optimized equipment control parameters. S55. Based on the optimized equipment control parameters and linkage rule base, generate a structured control strategy; verify the effectiveness of the strategy through a digital twin scenario to form the final control strategy; S56. The final control strategy is transmitted to the agricultural automation equipment control system through the IoT gateway to trigger the equipment to execute in conjunction. At the same time, the equipment operating status is tracked in real time through the monitoring platform. When a deviation occurs, the backup parameters are automatically called to adjust the strategy.

10. An artificial intelligence-based agricultural automation linkage control system, including: One or more processors; Memory, used to store one or more programs. Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any one of claims 1 to 9.

Images