Intelligent agricultural plant protection operation control system based on big data Internet of Things

Abstract

The invention discloses an intelligent agricultural plant protection operation control system based on a big data Internet of Things, and the system comprises a sensing layer which is an agricultural environment parameter collection module which collects the environment parameters of a farmland in combination with a fine-grained expert splitting technology; the crop growth state monitoring module is used for monitoring the growth state of crops in real time; a network layer: an Internet of Things communication module which is responsible for data transmission between the sensing layer and the platform layer; the platform layer comprises a big data processing and analysis module which is used for cleaning, integrating, analyzing and mining data acquired by the sensing layer in combination with a hybrid expert model in a MoE architecture; the intelligent decision support module is used for constructing a dynamic decision optimization model and dynamically adjusting a decision scheme; the plant protection operation management module is used for designing a hybrid solver of a space-time heuristic rule and integer programming, generating a preliminary operation plan, and then adjusting and optimizing the preliminary operation plan in real time according to real-time data in combination with a dynamic decision optimization model and an expert model in the MoE architecture.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent agriculture, and particularly to an intelligent agricultural plant protection operation control system based on big data and the Internet of Things. Background Art

[0002] With the rapid development of information technology, big data and Internet of Things technologies have gradually penetrated into all walks of life, and the agricultural field is no exception. An intelligent agricultural plant protection operation control system based on big data and the Internet of Things has emerged as the times require. The intelligent agricultural plant protection operation control system can monitor the farmland environment and crop growth status in real time, timely warn of abnormal situations such as pests and diseases, and thus quickly respond and take effective measures to improve agricultural production efficiency.

[0003] In the prior art, basic machine learning models may seem inadequate when dealing with complex and changing data. Especially in the field of intelligent agriculture, the farmland environment and crop growth status are affected by multiple factors, including climate, soil conditions, pests and diseases, etc. The interaction between these factors makes the data extremely complex. Basic machine learning models may not be able to accurately capture these complex relationships, resulting in insufficient model expression ability, and further leading to mistakes in plant protection operation decisions, which reduces the practicality of the system. Therefore, an intelligent agricultural plant protection operation control system based on big data and the Internet of Things is proposed. Summary of the Invention

[0004] The purpose of the present invention is to solve the deficiencies existing in the prior art, and to propose an intelligent agricultural plant protection operation control system based on big data and the Internet of Things.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions:

[0006] An intelligent agricultural plant protection operation control system based on big data and the Internet of Things, comprising:

[0007] Perception layer: Agricultural environment parameter acquisition module: Using a sensor network, combined with fine-grained expert splitting technology, to acquire environmental parameters such as temperature, humidity, light intensity, and soil pH of farmland, and reducing the data transmission volume through communication overhead compression technology; Crop growth status monitoring module: Through cameras and image recognition technology, to monitor the growth status of crops in real time;

[0008] Network layer: Internet of Things communication module: Responsible for data transmission between the perception layer and the platform layer, ensuring the real-time and integrity of data;

[0009] Platform layer: Big data processing and analysis module: Combining the mixture-of-experts model in the MoE architecture, it cleans, integrates, analyzes, and mines the data collected by the perception layer; Intelligent decision support module: Constructs a dynamic decision optimization model, and dynamically adjusts the decision-making plan according to the analysis results of the big data processing and analysis module and real-time data;

[0010] Application layer: Plant protection operation management module: Designs a hybrid solver of spatio-temporal heuristic rules and integer programming. According to historical data, decision-making plans, and preset rules, it generates a preliminary operation plan. Then, in combination with the dynamic decision optimization model and the expert model in the MoE architecture, it adjusts and optimizes the preliminary operation plan in real time according to real-time data (such as farmland environmental parameters, crop growth status, pest and disease conditions, etc.); User interface: Provides a user interface.

[0011] The above technical solution further includes:

[0012] Further, the specific steps for the agricultural environment parameter acquisition module to collect the environmental parameters of the farmland by using a sensor network and combining fine-grained expert splitting technology are as follows:

[0013] Sensor deployment: Deploy temperature sensors, humidity sensors, light intensity sensors, and soil pH sensors at different positions in the farmland;

[0014] Data acquisition: The raw data collected by the sensors is transmitted to the edge computing node, and the data acquisition frequency is set according to actual needs, such as once per hour or once per minute;

[0015] Data preprocessing: Clean the collected raw data, remove outliers and noise, and perform normalization processing on the data;

[0016] Fine-grained expert splitting: On the edge computing node, input the collected environmental parameter data into the fine-grained expert splitting model. The fine-grained expert splitting model splits the data into different expert subsets according to the characteristics of the data, and each expert subset focuses on processing a certain type of data (such as temperature data, humidity data, etc.);

[0017] Data fusion: The data segments processed by the fine-grained expert splitting technology are fused on the edge computing node to form the monitoring results of the farmland environmental parameters. Before fusion, each data segment is compressed, and the monitoring results are transmitted to the big data processing and analysis module of the platform layer through the Internet of Things communication module.

[0018] Further, the specific steps for the agricultural environment parameter acquisition module to reduce the data transmission volume through communication overhead compression technology are as follows:

[0019] Data preprocessing: Before data transmission, the original data is preprocessed. The preprocessed data will be divided into multiple data segments, and each data segment contains specific farmland environmental parameter information, such as temperature, humidity, light intensity, etc.;

[0020] Application of communication overhead compression technology: For each data segment, communication overhead compression technology is applied, including data quantization, model filtering, and model low-rank processing;

[0021] Data quantization: The amount of data transmitted is reduced by reducing the precision of the data. For example, converting floating-point numbers to integers or converting high-precision numbers to low-precision numbers. Suppose the original temperature data is 32.12345 °C. After quantization, it can be approximated as 32 °C or 32.1 °C, thus reducing the amount of data transmitted;

[0022] Model filtering: If some data segments do not change much over a period of time or have little impact on plant protection operations, these data segments are selectively filtered to reduce the amount of data transmitted. For example, during the night or on cloudy days, the change in light intensity data may be small, so this part of the data can be selected not to be transmitted;

[0023] Model low-rank processing: By using low-rank decomposition technology, the original data matrix is decomposed into the product of multiple small matrices, reducing the amount of data transmitted. This method is particularly effective when transmitting large datasets.

[0024] Furthermore, the big data processing and analysis module constructs sub-models in combination with the mixture-of-experts model in the MoE architecture, including a temperature expert model, a humidity expert model, a light intensity expert model, a soil pH expert model, and a crop growth status expert model;

[0025] The temperature expert model focuses on analyzing farmland temperature data and extracting the laws and trends of temperature changes;

[0026] The humidity expert model focuses on analyzing farmland humidity data and studying the impact of humidity on crop growth;

[0027] The light intensity expert model analyzes light intensity data to understand the impact of light on crop photosynthesis;

[0028] The soil pH expert model studies the impact of soil pH on crop growth and nutrient absorption;

[0029] The crop growth status expert model analyzes the growth status of crops;

[0030] The big data processing and analysis module uses an ensemble learning method to fuse and optimize the results of these sub-models.

[0031] Further, the IoT communication module sends data to nearby gateways or base stations through LoRa or NB-IoT communication technologies, and then the gateways or base stations upload the data to the platform layer.

[0032] Further, the specific steps for the intelligent decision support module to build a dynamic decision optimization model to dynamically adjust the decision-making plan are as follows:

[0033] Data acquisition and preprocessing: Obtain the analysis results and real-time data from the big data processing and analysis module and perform preprocessing;

[0034] Build a dynamic decision optimization model: Based on the expert models in the support vector machine and the MoE architecture, build a dynamic decision optimization model. The dynamic decision optimization model outputs a decision-making plan according to the input data;

[0035] Input data analysis: Deeply analyze the preprocessed data and extract the key information and features related to the plant protection operation;

[0036] Dynamically adjust the decision-making plan: Input the extracted key information and features into the dynamic decision optimization model. The dynamic decision optimization model outputs a decision-making plan according to this information, and dynamically adjusts the decision-making plan according to the real-time data and new analysis results;

[0037] Output the decision result: Output the adjusted decision-making plan to the plant protection operation management module.

[0038] Further, the plant protection operation management module designs a hybrid solver of spatio-temporal heuristic rules and integer programming, and generates a preliminary operation plan according to the historical data, decision-making plan and preset rules, including the following steps:

[0039] Data collection and integration: Extract historical operation data (such as operation time, location, crop type, pest and disease situation, etc.) from the database, combine the decision-making plan (such as spraying plan, operation path, etc.) provided by the intelligent decision support module, and the system preset rules (such as operation priority, resource allocation, etc.), and perform data integration and preprocessing;

[0040] Build spatio-temporal heuristic rules: Analyze the correlation between operation time, location and crop growth status, pest and disease situation in the historical data, and combine the preset rules to build a set of spatio-temporal heuristic rules. These rules include the priority order of operation time, the selection strategy of operation location, the impact of crop growth status on the operation, etc.;

[0041] Integer programming model construction: Define decision variables (such as the allocation of job tasks, the usage of resources, etc.), and set the objective function (such as minimizing job costs, maximizing job efficiency, etc.). At the same time, according to spatio-temporal heuristic rules and actual situations, set constraint conditions (such as job time limits, resource quantity limits, etc.) to form an integer programming model;

[0042] Hybrid solver design: Combine spatio-temporal heuristic rules and integer programming algorithms to design a hybrid solver. The hybrid solver first uses spatio-temporal heuristic rules to preliminarily screen and simplify the problem, and then uses integer programming algorithms to solve the simplified problem. During the solving process, continuously adjust and optimize the solution according to spatio-temporal heuristic rules;

[0043] Generate a preliminary job plan: Convert the solution obtained by the hybrid solver into a specific job plan, including the allocation of job tasks, the arrangement of job times, the usage of resources, etc.

[0044] Furthermore, the plant protection operation management module combines a dynamic decision optimization model and an expert model in the MoE architecture to perform real-time adjustment and optimization of the preliminary job plan according to real-time data, including the following steps:

[0045] Real-time data collection and processing: Real-time collect environmental parameters in the farmland (such as temperature, humidity, light intensity, etc.), crop growth status (such as plant height, leaf area index, etc.), and pest and disease situations (such as pest and disease types, quantities, distributions, etc.), and perform preprocessing operations;

[0046] Dynamic decision optimization model construction: Based on neural networks, construct a dynamic decision optimization model for processing real-time data. The dynamic decision optimization model automatically adjusts the parameters in the job plan (such as job time, job location, resource allocation, etc.) according to the changes in real-time data to optimize the job effect;

[0047] Expert model design in the MoE architecture: Design one or more expert models for each specific problem (such as crop pest and disease control, crop nutrition management, etc.). The expert models provide professional suggestions and solutions for the problem according to real-time data and professional knowledge;

[0048] Real-time adjustment and optimization of the preliminary job plan: Input the real-time data into the dynamic decision optimization model to obtain preliminary adjustment suggestions. Combine these suggestions with the professional suggestions provided by the expert models in the MoE architecture to form a final adjustment plan. According to the adjustment plan, perform real-time adjustment and optimization of the parameters in the preliminary job plan;

[0049] Execution and feedback of the job plan: Send the adjusted job plan to the plant protection operation execution system (such as drones, spraying vehicles, etc.) for actual plant protection operations.

[0050] The present invention has the following beneficial effects:

[0051] In the present invention, by integrating the MoE architecture and the fine-grained expert splitting technology, tasks such as identifying the types of pests and diseases, evaluating the severity of pests and diseases, and formulating plant protection plans are split into multiple fine-grained experts, and each expert is responsible for processing one of the links, thereby improving the response speed and accuracy of the entire system. This means that the system can more accurately understand the farmland environment and the growth status of crops, and thus make more intelligent decisions. Combining the hybrid solver technology of spatio-temporal heuristic rules and integer programming improves the efficiency and effect of plant protection operations. This enables the system to reduce the operation cost and time cost while ensuring the operation quality. Brief Description of the Drawings

[0052] Figure 1 It is a system block diagram of a smart agriculture plant protection operation control system based on big data Internet of Things proposed by the present invention. Detailed Embodiments

[0053] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0054] Please refer to Figure 1 As shown, the present invention is a smart agriculture plant protection operation control system based on big data Internet of Things, including:

[0055] Perception layer: Agricultural environment parameter acquisition module: Using a sensor network, combined with fine-grained expert splitting technology, to collect environmental parameters such as temperature, humidity, light intensity, and soil pH of farmland, and reducing the data transmission volume through communication overhead compression technology; Crop growth status monitoring module: Real-time monitoring of the growth status of crops through cameras and image recognition technology;

[0056] Network layer: Internet of Things communication module: Responsible for data transmission between the perception layer and the platform layer to ensure the real-time and integrity of data;

[0057] Platform layer: Big data processing and analysis module: Combining the hybrid expert model in the MoE architecture, cleaning, integrating, analyzing, and mining the data collected by the perception layer; Intelligent decision support module: Building a dynamic decision optimization model, and dynamically adjusting the decision-making plan according to the analysis results of the big data processing and analysis module and real-time data;

[0058] Application layer: Plant protection operation management module: Design a hybrid solver of spatio-temporal heuristic rules and integer programming. According to historical data, decision-making schemes, and preset rules, generate a preliminary operation plan. Then, in combination with the dynamic decision optimization model and the expert model in the MoE architecture, adjust and optimize the preliminary operation plan in real time based on real-time data (such as farmland environmental parameters, crop growth status, pest and disease conditions, etc.); User interaction interface: Provide a user interaction interface.

[0059] In one embodiment, the specific steps for the agricultural environment parameter acquisition module to acquire the environmental parameters of the farmland by using a sensor network in combination with the fine-grained expert splitting technology are as follows:

[0060] Sensor deployment: Deploy temperature sensors, humidity sensors, light intensity sensors, and soil pH sensors at different positions in the farmland;

[0061] Data acquisition: The raw data collected by the sensors is transmitted to the edge computing node, and the data acquisition frequency is set according to actual needs, such as once per hour or once per minute;

[0062] Data preprocessing: Clean the collected raw data, remove outliers and noise, and perform normalization processing on the data;

[0063] Fine-grained expert splitting: On the edge computing node, input the collected environmental parameter data into the fine-grained expert splitting model. The fine-grained expert splitting model splits the data into different expert subsets according to the characteristics of the data, and each expert subset focuses on processing a certain type of data (such as temperature data, humidity data, etc.);

[0064] Data fusion: The data segments processed by the fine-grained expert splitting technology are fused on the edge computing node to form the monitoring results of the farmland environmental parameters. Before fusion, each data segment is compressed, and the monitoring results are transmitted to the big data processing and analysis module in the platform layer through the Internet of Things communication module.

[0065] In one embodiment, the specific steps for the agricultural environment parameter acquisition module to reduce the data transmission volume through communication overhead compression technology are as follows:

[0066] Data preprocessing: Before data transmission, preprocess the raw data. The preprocessed data will be divided into multiple data segments, and each data segment contains specific farmland environmental parameter information, such as temperature, humidity, light intensity, etc.;

[0067] Application of communication overhead compression technology: For each data segment, apply communication overhead compression technology, including data quantization, model filtering, and model low-rank processing;

[0068] Data quantization: Reduce the transmission volume by reducing the precision of data. For example, convert floating-point numbers to integers or convert high-precision numbers to low-precision numbers. Suppose the original temperature data is 32.12345 °C. After quantization, it can be approximated as 32 °C or 32.1 °C, thus reducing the amount of data transmitted;

[0069] Model filtering: If some data segments do not change much over a period of time or have little impact on the plant protection operation, selectively filter out these data segments to reduce the transmission volume. For example, during the night or on cloudy days, the change in light intensity data may be small, so this part of the data can be selected not to be transmitted;

[0070] Model low-rank processing: Decompose the original data matrix into the product of multiple small matrices through low-rank decomposition technology to reduce the amount of data transmitted. This method is particularly effective when transmitting large datasets.

[0071] In one embodiment, the big data processing and analysis module constructs sub-models in combination with the mixture-of-experts model in the MoE architecture, including a temperature expert model, a humidity expert model, a light intensity expert model, a soil pH expert model, and a crop growth status expert model;

[0072] The temperature expert model focuses on analyzing farmland temperature data and extracting the laws and trends of temperature changes;

[0073] The humidity expert model focuses on analyzing farmland humidity data and studying the impact of humidity on crop growth;

[0074] The light intensity expert model analyzes light intensity data to understand the impact of light on crop photosynthesis;

[0075] The soil pH expert model studies the impact of soil pH on crop growth and nutrient absorption;

[0076] The crop growth status expert model analyzes the growth status of crops;

[0077] The big data processing and analysis module uses an ensemble learning method to fuse and optimize the results of these sub-models.

[0078] In one embodiment, the IoT communication module sends data to a nearby gateway or base station through LoRa or NB-IoT communication technology, and then the gateway or base station uploads the data to the platform layer.

[0079] In one embodiment, the specific steps for the intelligent decision support module to construct a dynamic decision optimization model to dynamically adjust the decision-making scheme are as follows:

[0080] Data acquisition and preprocessing: Obtain the analysis results and real-time data from the big data processing and analysis module and perform preprocessing;

[0081] Construct a dynamic decision optimization model: Based on the support vector machine and the expert models in the MoE architecture, construct a dynamic decision optimization model. The dynamic decision optimization model outputs a decision plan according to the input data.

[0082] Analysis of input data: Conduct in-depth analysis on the preprocessed data to extract key information and features related to plant protection operations.

[0083] Dynamically adjust the decision plan: Input the extracted key information and features into the dynamic decision optimization model. The dynamic decision optimization model outputs a decision plan according to this information, and dynamically adjusts the decision plan according to real-time data and new analysis results.

[0084] Output the decision result: Output the adjusted decision plan to the plant protection operation management module.

[0085] In one embodiment, the plant protection operation management module designs a hybrid solver for spatio-temporal heuristic rules and integer programming, and generates a preliminary operation plan according to historical data, decision plans, and preset rules, including the following steps:

[0086] Data collection and integration: Extract historical operation data (such as operation time, location, crop type, pest and disease situation, etc.) from the database, combine the decision plans (such as spraying plans, operation paths, etc.) provided by the intelligent decision support module, and the preset rules of the system (such as operation priorities, resource allocation, etc.) to conduct data integration and preprocessing.

[0087] Construct spatio-temporal heuristic rules: Analyze the correlation between operation time, location and crop growth status, pest and disease situation in historical data, and combine preset rules to construct a set of spatio-temporal heuristic rules, which include the priority order of operation time, the selection strategy of operation location, the impact of crop growth status on operations, etc.

[0088] Integer programming model construction: Define decision variables (such as the allocation of operation tasks, the usage amount of resources, etc.), and set the objective function (such as minimizing operation costs, maximizing operation efficiency, etc.). At the same time, according to spatio-temporal heuristic rules and actual situations, set constraint conditions (such as operation time limits, resource quantity limits, etc.) to form an integer programming model.

[0089] Hybrid solver design: Combine spatio-temporal heuristic rules and integer programming algorithms to design a hybrid solver. The hybrid solver first conducts preliminary screening and simplification of the problem according to spatio-temporal heuristic rules, and then uses integer programming algorithms to solve the simplified problem. During the solution process, continuously adjust and optimize the solution according to spatio-temporal heuristic rules.

[0090] Generate a preliminary operation plan: Convert the solution obtained by the hybrid solver into a specific operation plan, including the assignment of operation tasks, the arrangement of operation time, the use of resources, etc.

[0091] In one embodiment, the plant protection operation management module combines a dynamic decision optimization model and an expert model in the MoE architecture to adjust and optimize the preliminary operation plan in real time according to real-time data, including the following steps:

[0092] Real-time data collection and processing: Real-time collect environmental parameters in the farmland (such as temperature, humidity, light intensity, etc.), crop growth status (such as plant height, leaf area index, etc.), and pest and disease conditions (such as types, quantities, distributions of pests and diseases, etc.), and perform preprocessing operations;

[0093] Construction of a dynamic decision optimization model: Based on a neural network, construct a dynamic decision optimization model for processing real-time data. The dynamic decision optimization model automatically adjusts the parameters in the operation plan (such as operation time, operation location, resource allocation, etc.) according to the changes in real-time data to optimize the operation effect;

[0094] Design of an expert model in the MoE architecture: Design one or more expert models for each specific problem (such as crop pest and disease control, crop nutrition management, etc.). The expert model provides professional suggestions and solutions for this problem according to real-time data and professional knowledge;

[0095] Real-time adjustment and optimization of the preliminary operation plan: Input the real-time data into the dynamic decision optimization model to obtain preliminary adjustment suggestions, combine these suggestions with the professional suggestions provided by the expert model in the MoE architecture to form a final adjustment plan, and adjust and optimize the parameters in the preliminary operation plan in real time according to the adjustment plan;

[0096] Execution and feedback of the operation plan: Send the adjusted operation plan to the plant protection operation execution system (such as drones, spraying vehicles, etc.) for actual plant protection operations.

[0097] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A smart agricultural plant protection operation control system based on big data and the Internet of Things, characterized in that, Including: Perception layer: Agricultural environment parameter acquisition module: Using a sensor network and combining fine-grained expert splitting technology, it acquires the environmental parameters of farmland and reduces the data transmission volume through communication overhead compression technology; Crop growth status monitoring module: Through cameras and image recognition technology, it monitors the growth status of crops in real time; Network layer: Internet of Things communication module: Responsible for data transmission between the perception layer and the platform layer; Platform layer: Big data processing and analysis module: Combining the mixture-of-experts model in the MoE architecture, it cleans, integrates, analyzes, and mines the data collected by the perception layer; Intelligent decision support module: Constructs a dynamic decision optimization model and dynamically adjusts the decision-making plan according to the analysis results of the big data processing and analysis module and real-time data; Application layer: Plant protection operation management module: Designs a hybrid solver of spatio-temporal heuristic rules and integer programming. According to historical data, decision-making plans, and preset rules, it generates a preliminary operation plan. Then, in combination with the dynamic decision optimization model and the expert model in the MoE architecture, it adjusts and optimizes the preliminary operation plan in real time according to real-time data; User interaction interface: Provides a user interaction interface.

2. The intelligent agricultural plant protection operation control system based on big data Internet of Things according to claim 1, wherein The specific steps for the agricultural environment parameter acquisition module to acquire the environmental parameters of farmland by using a sensor network and combining fine-grained expert splitting technology are as follows: Sensor deployment: Deploy temperature sensors, humidity sensors, light intensity sensors, and soil pH sensors at different positions in the farmland; Data acquisition: The raw data collected by the sensors is transmitted to the edge computing node; Data preprocessing: The collected raw data is cleaned to remove outliers and noise, and the data is normalized; Fine-grained expert splitting: On the edge computing node, the collected environmental parameter data is input into the fine-grained expert splitting model. The fine-grained expert splitting model splits the data into different expert subsets according to the characteristics of the data, and each expert subset focuses on processing a certain type of data; Data fusion: The data fragments processed by the fine-grained expert splitting technology are fused on the edge computing node to form the monitoring results of the farmland environmental parameters. Before fusion, each data fragment is compressed. The monitoring results are transmitted to the big data processing and analysis module of the platform layer through the Internet of Things communication module.

3. A smart agricultural plant protection operation control system based on big data Internet of Things according to claim 1, characterized in that, The specific steps for the agricultural environment parameter acquisition module to reduce the data transmission volume through communication overhead compression technology are as follows: Data preprocessing: Before data transmission, the raw data is preprocessed. The preprocessed data will be divided into multiple data fragments, and each data fragment contains specific farmland environmental parameter information; Application of communication overhead compression technology: For each data fragment, communication overhead compression technology is applied, including data quantization, model filtering, and model low-rank processing; Data quantization: Reduce the transmission volume by reducing the precision of the data; Model filtering: If some data fragments do not change much within a certain period of time or have little impact on plant protection operations, selectively filter out these data fragments; Model low-rank processing: Through low-rank decomposition technology, the original data matrix is decomposed into the product of multiple small matrices to reduce the amount of data transmitted.

4. The intelligent agricultural plant protection operation control system based on big data Internet of Things according to claim 1, characterized in that, The big data processing and analysis module constructs sub-models in combination with the Mixture of Experts (MoE) architecture, including a temperature expert model, a humidity expert model, a light intensity expert model, a soil pH expert model, and a crop growth status expert model; The temperature expert model focuses on analyzing farmland temperature data and extracting the laws and trends of temperature changes; The humidity expert model focuses on analyzing farmland humidity data and studying the impact of humidity on crop growth; The light intensity expert model analyzes light intensity data to understand the impact of light on crop photosynthesis; The soil pH expert model studies the impact of soil pH on crop growth and nutrient absorption; The crop growth status expert model analyzes the growth status of crops; The big data processing and analysis module uses an ensemble learning method to fuse and optimize the results of these sub-models.

5. The intelligent agricultural plant protection operation control system based on big data Internet of Things according to claim 1, characterized in that, The Internet of Things (IoT) communication module sends data to a nearby gateway or base station through LoRa or NB-IoT communication technology, and then the gateway or base station uploads the data to the platform layer.

6. The intelligent agricultural plant protection operation control system based on big data Internet of Things according to claim 1, characterized in that, The specific steps for the intelligent decision support module to build a dynamic decision optimization model to dynamically adjust the decision-making scheme are as follows: Data acquisition and preprocessing: Obtain the analysis results and real-time data from the big data processing and analysis module and perform preprocessing; Build a dynamic decision optimization model: Based on the support vector machine and the expert model in the MoE architecture, build a dynamic decision optimization model. The dynamic decision optimization model outputs a decision-making scheme according to the input data; Input data analysis: Conduct in-depth analysis of the preprocessed data to extract key information and features related to plant protection operations; Dynamically adjust the decision-making scheme: Input the extracted key information and features into the dynamic decision optimization model. The dynamic decision optimization model outputs a decision-making scheme according to this information and dynamically adjusts the decision-making scheme according to real-time data and new analysis results; Output decision results: Output the adjusted decision-making scheme to the plant protection operation management module.

7. The intelligent agricultural plant protection operation control system based on big data Internet of Things according to claim 1, characterized in that, The plant protection operation management module designs a hybrid solver for spatio-temporal heuristic rules and integer programming. According to historical data, decision-making schemes, and preset rules, it generates a preliminary operation plan, including the following steps: Data collection and integration: Extract historical operation data from the database, combine the decision-making schemes provided by the intelligent decision support module, and the preset rules of the system for data integration and preprocessing; Build spatio-temporal heuristic rules: Analyze the correlation between operation time, location, crop growth status, and pest and disease conditions in historical data, and combine preset rules to build a set of spatio-temporal heuristic rules; Integer programming model construction: Define decision variables and set the objective function. At the same time, according to spatio-temporal heuristic rules and actual situations, set constraint conditions to form an integer programming model; Hybrid solver design: Combine spatio-temporal heuristic rules and integer programming algorithms to design a hybrid solver. The hybrid solver first preliminarily screens and simplifies the problem using spatio-temporal heuristic rules, and then uses the integer programming algorithm to solve the simplified problem. During the solution process, continuously adjust and optimize the solution according to spatio-temporal heuristic rules; Generate a preliminary operation plan: Convert the solution obtained by the hybrid solver into a specific operation plan.

8. A smart agricultural plant protection operation control system based on big data Internet of Things according to claim 1, characterized in that, The plant protection operation management module combines the dynamic decision optimization model and the expert models in the MoE architecture to adjust and optimize the preliminary operation plan in real time according to real-time data, including the following steps: Real-time data collection and processing: Real-time collect the environmental parameters, crop growth status, and pest and disease conditions in the farmland, and perform preprocessing operations; Construction of the dynamic decision optimization model: Based on neural networks, construct a dynamic decision optimization model for processing real-time data. The dynamic decision optimization model automatically adjusts the parameters in the operation plan according to the changes in real-time data to optimize the operation effect; Design of expert models in the MoE architecture: Design one or more expert models for each specific problem. The expert models provide professional suggestions and solutions for the problem according to real-time data and professional knowledge; Real-time adjustment and optimization of the preliminary operation plan: Input the real-time data into the dynamic decision optimization model to obtain preliminary adjustment suggestions. Combine these suggestions with the professional suggestions provided by the expert models in the MoE architecture to form a final adjustment plan. According to the adjustment plan, adjust and optimize the parameters in the preliminary operation plan in real time; Execution and feedback of the operation plan: Send the adjusted operation plan to the plant protection operation execution system for actual plant protection operations.

Images