Digital twinning-based crop full-life-cycle real-time monitoring and precise regulation and control system

Abstract

The invention discloses a crop full-life-cycle real-time monitoring and precise regulation and control system based on digital twinning, and particularly relates to the field of digital twinning. Comprising a multi-source data acquisition module, a data transmission module, a digital twin model building and updating module, a real-time monitoring and early warning module, a data analysis and prediction module, a precise regulation and control decision module and a control instruction execution module. According to the method, high-fidelity mapping and self-adaptive calibration of the growth process are realized by constructing the dynamically-updated crop digital twinborn body, the system deeply fuses multi-source heterogeneous data, and the accuracy of growth trend prediction and early pest and disease damage identification is remarkably improved by utilizing a mechanism model and an artificial intelligence algorithm; quantitative irrigation, fertilization and pesticide application decisions are generated based on model driving, intelligent closed-loop regulation and control from sensing to execution are achieved, the problems of resource waste and extensive regulation and control caused by traditional dependence on static rules are effectively solved, and accurate management of the whole life cycle is achieved.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, and more specifically, to a real-time monitoring and precise control system for the entire life cycle of crops based on digital twins. Background Technology

[0002] With the development of precision agriculture technology, existing crop monitoring systems have gradually incorporated sensor networks, wireless communication, and automated control equipment to achieve preliminary monitoring and regulation of the farmland environment and crop growth status. A typical system usually includes sensor nodes deployed in the field to collect environmental parameters such as air temperature and humidity, soil moisture, and light intensity, and combines these with image acquisition equipment to obtain images of the crop canopy. The collected data is transmitted wirelessly to a cloud platform or local server for storage and preliminary analysis. The system can issue alarms based on preset thresholds and generate control commands such as irrigation and fertilization based on simple rules, driving the execution equipment to complete the corresponding operations, forming a certain degree of closed-loop control.

[0003] However, the aforementioned systems still have significant shortcomings in practical applications: First, they largely rely on static models and fixed thresholds, lacking the ability to dynamically model and update the entire crop growth process in real time, making it difficult to adapt to the differentiated needs of different varieties, soils, and meteorological conditions. Second, data acquisition, transmission, and processing are relatively independent, failing to achieve deep fusion and collaborative analysis of multi-source data, resulting in limited prediction accuracy, especially in pest and disease identification and growth trend judgment. Furthermore, regulatory decisions are mostly based on empirical rules, lacking refined decision support driven by models and data, making it difficult to achieve true "on-demand regulation," leading to resource waste and environmental pollution. Therefore, there is an urgent need to construct a system solution that can span the entire crop life cycle and possess real-time modeling, intelligent prediction, and precise regulation capabilities. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, this invention provides a real-time monitoring and precise control system for the entire life cycle of crops based on digital twins, which solves the problems mentioned in the background art through the following solutions.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a real-time monitoring and precise control system for the entire life cycle of crops based on digital twins, comprising:

[0006] Multi-source data acquisition module: Collects raw data from the farmland to provide input for the digital twin model. It acquires environmental data, including air temperature, air humidity, soil moisture, and light intensity, through environmental parameter acquisition. It acquires crop plant height, leaf area index, and stem diameter through crop growth parameter acquisition. It also acquires crop canopy image data through image data acquisition. The acquired data is directly transmitted to the data transmission module.

[0007] Data transmission module: Receives raw data from the multi-source data acquisition module, compresses and encapsulates the data through a data encoding process, and then sends the data to the cloud platform or local server via wireless transmission. It saves the data using a data storage process and forms a two-way feedback with the multi-source data acquisition module and the digital twin model construction and update module.

[0008] Digital twin model construction and update module: Based on the data provided by the data transmission module, a digital twin of crop growth is constructed. An initial model is created according to crop variety and soil type through model initialization. Then, real-time data is integrated into the model through data assimilation, the model parameters are updated, and the digital twin model outputs crop state variables to the real-time monitoring and early warning module.

[0009] Real-time monitoring and early warning module: Based on the crop status output by the digital twin model construction and update module, it performs real-time visual monitoring, displays crop growth curves and environmental data changes through status monitoring, and sets key parameter thresholds through a threshold early warning process. When the data is abnormal, an early warning signal is triggered, and the early warning information is transmitted to the data analysis and prediction module.

[0010] Data Analysis and Prediction Module: Receives early warning data and historical data from the real-time monitoring and early warning module, analyzes them, predicts the future growth status of crops through growth trend analysis, identifies pest and disease risks through pest and disease prediction, and outputs the analysis results to the precision control decision module in the form of a report.

[0011] Precision control decision module: Based on the output of the data analysis and prediction module, it generates precise control instructions, calculates irrigation water demand through the irrigation decision process, determines the amount of nitrogen, phosphorus and potassium fertilizer through the fertilization decision, and formulates a pesticide application plan through the pest and disease control decision. The decision instructions are sent to the control instruction execution module.

[0012] Control command execution module: Receives commands from the precision regulation decision module, drives farmland execution equipment, operates the irrigation system through the irrigation control process, controls the fertilizer applicator through the fertilization control process, and operates the spraying device through the pest and disease control control process. The execution status is fed back to the multi-source data acquisition module in real time.

[0013] The technical effects and advantages of this invention are as follows:

[0014] 1. This invention overcomes the limitations of existing technologies that rely on static models by constructing and continuously updating a digital twin of crop growth. The system can dynamically correct model parameters based on real-time multi-source data using data assimilation technology, so that the digital twin always keeps in sync with the crop growth status of the physical farmland, significantly improving the system's adaptability to different varieties, soils and growth environments and the accuracy of state description.

[0015] 2. This invention integrates environmental data, crop physiological data, and image data, and uses advanced time-series analysis models and convolutional neural network models for collaborative analysis, achieving in-depth mining of data value. This system can not only predict the future growth trend of crops with high accuracy, but also identify the risk of pests and diseases at an early stage, overcoming the shortcomings of existing systems with limited prediction accuracy and single analysis dimensions, and providing a forward-looking decision-making basis for precise regulation.

[0016] 3. This invention abandons the traditional decision-making method that relies on experience rules. By integrating crop water requirement models, nutrient balance laws, and risk threshold decision-making mechanisms, it generates quantitative instructions for irrigation, fertilization, and pesticide application. This ensures that each regulatory operation has a clear objective and scientific basis, truly achieving "supply on demand," effectively avoiding resource waste and environmental pollution. Furthermore, through real-time feedback on the execution status, a complete "perception-decision-execution-feedback" closed loop is formed, continuously optimizing the regulatory strategy. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0018] Figure 2 This is a schematic diagram of the digital twin model construction and update module of the present invention.

[0019] Figure 3 This is a schematic diagram of the data analysis and prediction module structure of the present invention.

[0020] Figure 4 This is a schematic diagram of the precise control decision module structure of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] refer to Figures 1-4 The digital twin-based real-time monitoring and precision control system for the entire life cycle of crops, as shown, includes:

[0023] Multi-source data acquisition module: Collects raw data from the farmland to provide input for the digital twin model. It acquires environmental data, including air temperature, air humidity, soil moisture, and light intensity, through environmental parameter acquisition. It acquires crop plant height, leaf area index, and stem diameter through crop growth parameter acquisition. It also acquires crop canopy image data through image data acquisition. The acquired data is directly transmitted to the data transmission module.

[0024] The environmental parameters are collected by deploying a sensor network in a grid pattern in the farmland. Each grid node is equipped with a digital temperature sensor, a digital humidity sensor, a soil moisture sensor, and a light sensor. The sensor nodes automatically collect air temperature, air humidity, soil moisture, and light intensity data at a fixed frequency of once every five minutes. The collected analog signals are converted into digital signals by an ADC converter. All data is encapsulated in JSON format, which includes timestamp information, sensor identifiers, and numerical readings. The encapsulated data packets are sent to the data center for further processing via a wireless transmission module.

[0025] The crop growth parameters are collected by an automated mobile platform that periodically traverses the farmland. A non-contact laser rangefinder is used to measure crop height, a leaf area meter is used to measure leaf area index, and a stem diameter meter is used to obtain stem diameter data. The collected data is wirelessly transmitted to the local gateway via Bluetooth or ZigBee.

[0026] The image data acquisition utilizes an RGB camera or multispectral camera mounted on a drone or fixed pole to capture crop canopy images at fixed points. The image resolution is no less than 1920×1080 pixels. The acquisition frequency is adjusted according to the crop growth stage: once a day during the seedling stage and twice a week during the growing season. The image data is compressed using the JPEG compression algorithm and stored in a designated storage system along with GPS coordinates and shooting time.

[0027] Data transmission module: Receives raw data from the multi-source data acquisition module, compresses and encapsulates the data through a data encoding process, and then sends the data to the cloud platform or local server via wireless transmission. It saves the data using a data storage process and forms a two-way feedback with the multi-source data acquisition module and the digital twin model construction and update module.

[0028] The data encoding uses the LZ77 compression algorithm to compress the JSON format sensor data, while the image data is compressed using H.264 encoding; the encapsulation protocol uses MQTT, and the topic is designed as farm / data / env for environmental data and farm / data / growth for growth data.

[0029] The wireless transmission adopts a dual-mode network architecture adapted to LoRaWAN and 5G networks. Network selection is automatically completed by a network selection algorithm deployed on the farmland gateway. This algorithm prioritizes scanning the 5G network signal strength. When the received signal strength indication value is higher than -95dBm, the system automatically connects to the 5G network for data transmission. When the 5G signal strength is lower than -95dBm or cannot connect, the system switches to LoRaWAN network mode. In LoRaWAN mode, the gateway is configured to Class A working mode, with a fixed receiving window delay of 1 second. The data transmission interval of all sensors is synchronized at a five-minute cycle. Data transmission adopts an acknowledgment-based mechanism. After data is sent, a three-second timer is started to wait for application layer acknowledgment. If no acknowledgment is received within the timeout period, a retransmission process is triggered. The system performs a maximum of three retransmission attempts. The structure of each transmitted data frame is fixed, including a two-byte frame header, a four-byte sensor node unique identifier, a variable-length data payload field, and a four-byte cyclic redundancy check code. The wireless transmission process directly receives and processes the output data packets from the data encoding process and forwards the data packets successfully delivered to the cloud platform to the data storage process for subsequent operations via the MQTT protocol.

[0030] The data storage receives successfully transmitted data from the wireless transmission process. First, the data is written to the InfluxDB time-series database, and an index table is created within the database. The primary key of the index table consists of a combination of timestamp and sensor ID. The storage process includes data deduplication, eliminating duplicate entries by comparing the new data with the timestamp and sensor ID of the oldest record in the database. Simultaneously, data verification is performed, using the CRC32 algorithm to calculate the checksum of the data payload and comparing it with the checksum in the transmitted data frame. Once verification is successful, the data is stored. After storage, the data storage process feeds back the storage status to the wireless transmission process to confirm data integrity and provides a standardized data access interface for the digital twin model construction and update module, supporting queries of historical data by time range or sensor ID.

[0031] Digital twin model construction and update module: Based on the data provided by the data transmission module, a digital twin of crop growth is constructed. An initial model is created according to crop variety and soil type through model initialization. Then, real-time data is integrated into the model through data assimilation, and the model parameters are updated. The digital twin model outputs crop state variables to the real-time monitoring and early warning module.

[0032] The model initialization process begins by establishing a crop variety parameter database and a soil type database. The crop variety parameter database pre-sets crop growth parameters, while the soil type database contains the physicochemical properties of different soil textures. When the initialization process starts, the operator selects the crop variety and soil type corresponding to the model through the system interface, and the system automatically retrieves the corresponding parameter set from the database. Simultaneously, the model calls the farmland location information stored in the data transmission module and the initial environmental dataset collected by the multi-source data acquisition module on the first working day after sowing through the interface. This dataset includes initial soil moisture, basic nutrient content, and daily average temperature recorded by the weather station. Subsequently, the system loads the mechanistic model framework, describes the crop growth process in the form of differential equations, and uses the WOFOST model to substitute the initial parameters into the calculation, generating the initial state variables of the digital twin, including initial biomass, leaf area index, and root depth. Finally, the initialized digital twin model enters the ready-to-run state, providing an iteratively updatable basic model for the data assimilation process.

[0033] The data assimilation employs the Extended Kalman Filter (EKF) algorithm. The state variables are leaf area index (LAI) and soil moisture, while the observed variables are air temperature, air humidity, soil moisture, and light intensity. The assimilation process is performed hourly. First, a prediction step is performed, using the previous state estimate and the dynamic equations of the WOFOST model to calculate the prior state estimate for the current moment. Then, an update step is performed, comparing the real-time observation data acquired by the multi-source data acquisition module with the prior state estimate. The prior state estimate is corrected by calculating the Kalman gain to obtain the posterior state estimate. The Kalman gain is determined by the model prediction error covariance and the observation error covariance. The model prediction error covariance is also updated within each assimilation cycle. The assimilated state variables are then output to the real-time monitoring and early warning module.

[0034] set up For state estimation, z is the observed value, K is the Kalman gain, and h is the observation model, then , k|k-1 represents the prior estimate of the current time k based on the information of the previous time k-1, and k|k represents the posterior estimate obtained at the current time k after incorporating the latest observation data;

[0035] This algorithm is based on two information sources for a dynamic system: a mechanistic model describing the system's intrinsic evolution and sensor data from external observations. The mechanistic model is responsible for prediction, but it inherently contains uncertainties; sensor data reflects the true state but includes observation errors. The algorithm optimally fuses these two imperfect information sources by recursively executing two steps: prediction and update. The prediction step utilizes the dynamic equations of a crop growth model to advance from the optimal state estimate of the previous moment to the current moment, obtaining a priori state prediction value. This prediction value represents our best guess of the system state before obtaining new observation data. The update step, after obtaining new observation data, compares the predicted values ​​with the observed values. Its core is calculating the Kalman gain, a weighting coefficient used to determine the extent to which we should trust the model predictions and the new observations. This gain is jointly determined by the uncertainty of the model predictions (prediction error covariance) and the uncertainty of the observations (observation error covariance). If the model predictions are very reliable but the observation noise is high, the gain will be smaller, and the system will have more confidence in the model predictions; conversely, if the observation data quality is high but the model prediction uncertainty is high, the gain will be larger, and the system will be more inclined to use the observations to correct the model.

[0036] The state update formula is an educated and quantitative correction to the prior prediction. The difference between the observed value and the predicted value of the observation model is usually called the "innovation". It intuitively reflects the deviation between the model prediction and the actual measurement. The final output of the entire assimilation process, namely the posterior state estimate, is the optimal estimate that is statistically closest to the real physical state. It utilizes both the crop growth laws contained in the model and the real information provided by real-time data, so that the digital twin can more accurately depict the actual evolution of the physical entity.

[0037] Real-time monitoring and early warning module: Based on the crop status output by the digital twin model construction and update module, it performs real-time visual monitoring. The status monitoring displays crop growth curves and environmental data changes, and sets key parameter thresholds through a threshold early warning process. When the data is abnormal, an early warning signal is triggered, and the early warning information is transmitted to the data analysis and prediction module.

[0038] The status monitoring utilizes the integrated web visualization framework ECharts to construct a dynamic monitoring interface, which displays crop status data obtained from the digital twin model construction and update module in real time. First, the system extracts the latest environmental parameters and crop growth indicators, including air temperature, soil moisture, and leaf area index, from the digital twin model construction and update module. Then, through data binding technology, these data are mapped to visualization components, including dashboards and time-series curves. The environmental parameter curves display temperature change trends, and the crop growth indicator curves display dynamic leaf area index. The data is automatically refreshed every 5 seconds. Simultaneously, the status monitoring process marks abnormal data and passes it to the threshold warning process for further processing.

[0039] The threshold warning is based on crop growth models and agronomic knowledge to set threshold ranges for key parameters. The warning logic adopts a rule engine to monitor crop status data from the digital twin model construction and update module in real time. When the data exceeds the threshold, the rule engine automatically triggers the generation of a warning event. The warning event includes high, medium, and low levels, timestamps, and specific suggestions. The generated warning information is pushed to the data analysis and prediction module in real time through the RabbitMQ message queue for risk analysis and decision support.

[0040] The threshold warning system categorizes warning levels into high, medium, and low, based on the numerical range of key parameters. For temperature parameters, a low level is defined as 30°C to 35°C, a medium level as 35°C to 40°C, and a high level as exceeding 40°C. For soil moisture parameters, a low level is defined as 30% to 40%, a medium level as 20% to 30%, and a high level as below 20%. Specific recommendations are formulated according to the level and parameter type: For a low-level high temperature warning, it is recommended to monitor temperature changes and prepare shading measures; for a medium-level high temperature warning, it is recommended to implement shading and increase irrigation frequency; for a high-level high temperature warning, it is recommended to immediately implement emergency shading and sufficient irrigation to prevent heat damage; for a low-level low soil moisture warning, it is recommended to check soil moisture and consider light irrigation; for a medium level, it is recommended to increase irrigation volume; for a high level, it is recommended to immediately implement sufficient irrigation to avoid drought stress.

[0041] Data Analysis and Prediction Module: Receives early warning data and historical data from the real-time monitoring and early warning module, analyzes them, predicts the future growth status of crops through growth trend analysis, identifies pest and disease risks through pest and disease prediction, and outputs the analysis results to the precision control decision-making module in the form of a report.

[0042] The growth trend analysis employs a time series analysis method, specifically using the ARIMA model. It inputs historical leaf area index and environmental data to predict the growth trend for the next 7 days. The analysis steps first involve a data stationarity test, using the ADF test to confirm the time series data is stationary; then, parameter estimation is performed, determining the autoregressive parameters and lag order through maximum likelihood estimation; finally, prediction is conducted, using the estimated model to generate future growth trend values.

[0043] Let φ be the autoregressive parameter and p be the lag order, then the ARIMA model Where tp represents the p-th time point before the current time point, ε represents the predicted value at the current time point t, y represents the observed time series data, such as historical observations of leaf area index, and ε represents the predicted value. t This represents the random error term of the model at time point t; the formula characterizes the inherent inertia and continuity of crop growth; for example, the change in crop leaf area index is not an isolated event, and today's growth status is closely related to yesterday's, the day before yesterday's, and even earlier growth statuses. Each term in the formula... Each represents the contribution of the growth state at a specific point in the past (i) to the current state. Therefore, this model captures the temporal dependence of crop growth and infers its future development trend by analyzing the growth data sequence over a period of time.

[0044] The pest and disease prediction uses a convolutional neural network model. The input data includes crop canopy image data and environmental data obtained from a multi-source data acquisition module. The model output is the probability value of pest and disease occurrence. The model training process uses transfer learning and is optimized based on the ResNet architecture. By loading pre-trained weights and fine-tuning the fully connected layers, historical pest and disease records and corresponding images are used as the training set. The training process uses the backpropagation algorithm to minimize the cross-entropy loss function, requiring a prediction accuracy greater than 90%. The prediction results are accompanied by a confidence score, which is used to determine the priority in the precision control decision module. The image data is preprocessed before input, including resizing to 224x224 pixels and normalization, while the environmental data is standardized.

[0045] Precision control decision module: Based on the output of the data analysis and prediction module, it generates precise control instructions, calculates irrigation water demand through the irrigation decision process, determines the amount of nitrogen, phosphorus and potassium fertilizer through fertilization decision, and formulates a pesticide application plan through pest and disease control decision. The decision instructions are then sent to the control instruction execution module.

[0046] The irrigation decision is based on a crop water requirement model and real-time environmental data. It calculates the daily irrigation water requirement, first using the FAO Penman-Monteith equation to calculate crop evapotranspiration (ETc), and simultaneously obtaining effective rainfall (R) from historical meteorological data or real-time rainfall sensors. When real-time soil moisture is below a preset threshold, the irrigation amount is I = ETc - R. This threshold is dynamically adjusted based on the crop growth stage and variety characteristics, set at 60% of field capacity during the seedling stage and 80% during maturity. The calculation is performed every 24 hours, and the irrigation amount is confirmed to match crop needs by combining growth trend information provided by the data analysis and prediction module. Finally, the irrigation command, including specific irrigation time and water volume, is sent to the control command execution module through the precision control decision module, driving the irrigation system (such as solenoid valves or drip irrigation equipment) to operate, and feeding back the execution status to the multi-source data acquisition module.

[0047] The fertilization decision-making adopts the nutrient balance method. Based on the target yield data and soil nutrient measurement values ​​provided by the data analysis and prediction module, the fertilization amount is calculated. First, the nutrient uptake per unit yield, target yield, soil supply, and fertilizer utilization rate are obtained. Then, the fertilization amount is calculated, which is equal to the nutrient uptake per unit yield multiplied by the target yield minus the soil supply, and then divided by the fertilizer utilization rate. During the calculation process, the nutrient uptake per unit yield is retrieved from the database according to the crop variety and growth stage. The soil supply is updated in real time through soil sensor data. The fertilizer utilization rate is adjusted based on historical data and environmental conditions. The final fertilization instruction includes the specific dosage of nitrogen, phosphorus, and potassium, and is directly sent to the control instruction execution module to drive the variable fertilizer applicator to perform precise fertilization operations.

[0048] The pest and disease control decision is based on the pest and disease prediction probability results provided by the data analysis and prediction module. A threshold decision method is used for judgment, with a preset risk threshold of 0.7. When the pest and disease probability exceeds the threshold, it is judged as a high risk and a pesticide application command is triggered. The pesticide application amount is calculated based on the crop planting area and the severity of the pest and disease. The severity is directly mapped by the pest and disease probability value, which is determined based on the growth trend analysis. The specific pesticide application amount is determined by multiplying the basic application rate per unit area by the crop area and then by the severity coefficient. The basic application rate is preset according to the crop type and pesticide characteristics. The decision-making process also includes determining the application time and target area. After generating a complete command, it is transmitted to the control command execution module to drive the spraying device to perform precise pesticide application.

[0049] The severity coefficient is dynamically determined based on the real-time probability values ​​of pests and diseases through a linear mapping relationship. When the probability value is 0.7, the severity coefficient is 1.0; when the probability value is 1.0, the severity coefficient is 2.0. For any probability value between 0.7 and 1.0, the corresponding severity coefficient is calculated in real time using a linear interpolation formula.

[0050] Control command execution module: Receives commands from the precision regulation decision module, drives farmland execution equipment, operates the irrigation system through the irrigation control process, controls the fertilizer applicator through the fertilization control process, and operates the spraying device through the pest and disease control control process. The execution status is fed back to the multi-source data acquisition module in real time.

[0051] The irrigation control receives and parses instructions from the precision control decision module through a PLC controller, and then drives the solenoid valve or drip irrigation system to perform irrigation operations. The control logic includes time-based control and quantity-based control. During execution, the PLC controller monitors the equipment status to confirm that the opening and closing of the solenoid valve is consistent with the instructions, while the drip irrigation system maintains a stable water flow through a pressure sensor. After irrigation is completed, the system collects data in real time through a soil moisture sensor and transmits the feedback information to a multi-source data acquisition module to form a closed-loop control to confirm that the irrigation effect matches the crop's needs.

[0052] The fertilization control module receives fertilization instructions from the precision regulation decision module based on the control instruction execution module. It calculates the application amount and ratio of nitrogen, phosphorus, and potassium based on the target crop yield, soil nutrient measurement values, and real-time growth data. The fertilization control operates on a variable-rate fertilizer applicator, which adjusts the fertilizer application amount according to the instructions. The control signal uses pulse width modulation to control the motor or valve action, with a fertilization accuracy error of ±5%. During execution, the actual fertilizer application amount is monitored and recorded in real time by flow and weight sensors and fed back to the multi-source data acquisition module for calibrating the digital twin model and optimizing subsequent regulation strategies.

[0053] The pest and disease control process first involves a spraying robot receiving pest and disease control instructions from a decision-making module. These instructions include the GPS coordinates of the target area, the type of pesticide, and the recommended dosage. The autonomous spraying robot then automatically travels to the designated farmland area according to a pre-set GPS path, correcting positional deviations using an inertial navigation system during its journey. During the pesticide application phase, the robot adjusts the pesticide flow rate via an electric pump, and an integrated flow sensor monitors the spray volume in real time. Simultaneously, the robot is equipped with spray nozzles that adjust the spray angle and pressure based on the crop canopy height. Upon completion, the robot transmits the actual application data, including location, time, and dosage, to a multi-source data acquisition module via a wireless network.

[0054] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0055] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A real-time monitoring and precise control system for the entire life cycle of crops based on digital twins, characterized in that: include: Multi-source data acquisition module: Collects raw data from the farmland to provide input for the digital twin model. It acquires environmental data, including air temperature, air humidity, soil moisture, and light intensity, through environmental parameter acquisition. It acquires crop plant height, leaf area index, and stem diameter through crop growth parameter acquisition. It also acquires crop canopy image data through image data acquisition. The acquired data is directly transmitted to the data transmission module. Data transmission module: Receives raw data from the multi-source data acquisition module, compresses and encapsulates the data through a data encoding process, and then sends the data to the cloud platform or local server via wireless transmission. It saves the data using a data storage process and forms a two-way feedback with the multi-source data acquisition module and the digital twin model construction and update module. Digital twin model construction and update module: Based on the data provided by the data transmission module, a digital twin of crop growth is constructed. An initial model is created according to crop variety and soil type through model initialization. Then, real-time data is integrated into the model through data assimilation, the model parameters are updated, and the digital twin model outputs crop state variables to the real-time monitoring and early warning module. Real-time monitoring and early warning module: Based on the crop status output by the digital twin model construction and update module, it performs real-time visual monitoring, displays crop growth curves and environmental data changes through status monitoring, and sets key parameter thresholds through a threshold early warning process. When the data is abnormal, an early warning signal is triggered, and the early warning information is transmitted to the data analysis and prediction module. Data Analysis and Prediction Module: Receives early warning data and historical data from the real-time monitoring and early warning module, analyzes them, predicts the future growth status of crops through growth trend analysis, identifies pest and disease risks through pest and disease prediction, and outputs the analysis results to the precision control decision module in the form of a report. Precision control decision module: Based on the output of the data analysis and prediction module, it generates precise control instructions, calculates irrigation water demand through the irrigation decision process, determines the amount of nitrogen, phosphorus and potassium fertilizer through the fertilization decision, and formulates a pesticide application plan through the pest and disease control decision. The decision instructions are sent to the control instruction execution module. Control command execution module: Receives commands from the precision regulation decision module, drives farmland execution equipment, operates the irrigation system through the irrigation control process, controls the fertilizer applicator through the fertilization control process, and operates the spraying device through the pest and disease control control process. The execution status is fed back to the multi-source data acquisition module in real time.

2. The real-time monitoring and precise control system for the entire life cycle of crops based on digital twins as described in claim 1, characterized in that: The environmental parameters are collected by deploying a sensor network in a grid pattern in the farmland. Each grid node is equipped with a digital temperature sensor, a digital humidity sensor, a soil moisture sensor, and a light sensor. The sensor nodes automatically collect air temperature, air humidity, soil moisture, and light intensity data at a fixed frequency of once every five minutes. The collected analog signals are converted into digital signals by an ADC converter. All data is encapsulated in JSON format, which includes timestamp information, sensor identifiers, and numerical readings. The encapsulated data packets are sent to the data center for further processing via a wireless transmission module. The crop growth parameters are collected by an automated mobile platform that periodically traverses the farmland. A non-contact laser rangefinder is used to measure crop height, a leaf area meter is used to measure leaf area index, and a stem diameter meter is used to obtain stem diameter data. The collected data is wirelessly transmitted to the local gateway via Bluetooth or ZigBee. The image data acquisition utilizes an RGB camera or multispectral camera mounted on a drone or fixed pole to capture crop canopy images at fixed points. The image resolution is no less than 1920×1080 pixels. The acquisition frequency is adjusted according to the crop growth stage: once a day during the seedling stage and twice a week during the growing season. The image data is compressed using the JPEG compression algorithm and stored in a designated storage system along with GPS coordinates and shooting time.

3. The real-time monitoring and precise control system for the entire life cycle of crops based on digital twins as described in claim 2, characterized in that: The data encoding uses the LZ77 compression algorithm to compress the JSON format sensor data, while the image data uses H.264 encoding for compression; the encapsulation protocol uses MQTT, and the topic is designed as farm / data / env for environmental data and farm / data / growth for growth data; The wireless transmission uses LoRaWAN or 5G networks to transmit data, and the transmission interval is synchronized with the collection frequency. The data storage receives successfully transmitted data from the wireless transmission process, first writes the data into the InfluxDB time-series database, and creates an index table in the database. The storage process includes data deduplication and verification, and verifies data integrity through CRC32.

4. The real-time monitoring and precise control system for the entire life cycle of crops based on digital twins as described in claim 3, characterized in that: The model initialization uses the WOFOST model to substitute initial parameters into the calculation to generate the initial state variables of the digital twin, including initial biomass, leaf area index and root depth; The data assimilation uses the extended Kalman filter algorithm, with the state variables being leaf area index and soil moisture, and the observed variables being air temperature, air humidity, soil moisture, and light intensity. The assimilation process is performed once per hour.

5. The real-time monitoring and precise control system for the entire life cycle of crops based on digital twins as described in claim 4, characterized in that: The status monitoring uses the integrated web visualization framework ECharts to build a dynamic monitoring interface, including dashboards and time-series graphs, with data automatically refreshed every 5 seconds. The threshold warning is based on crop growth models and agronomic knowledge to set threshold ranges for key parameters. The warning logic uses a rule engine. When the data exceeds the threshold, the rule engine automatically triggers the generation of a warning event. The warning event includes high, medium, and low levels, a timestamp, and specific suggestions. The generated warning information is pushed to the data analysis and prediction module in real time through the RabbitMQ message queue.

6. The real-time monitoring and precise control system for the entire life cycle of crops based on digital twins as described in claim 5, characterized in that: The growth trend analysis employs a time series analysis method, specifically using the ARIMA model. It inputs historical leaf area index and environmental data to predict the growth trend for the next 7 days. The analysis steps first involve a data stationarity test, using the ADF test to confirm the time series data is stationary; then, parameter estimation is performed, determining the autoregressive parameters and lag order through maximum likelihood estimation; finally, prediction is conducted, using the estimated model to generate future growth trend values. The pest and disease prediction uses a convolutional neural network model. The input data includes crop canopy image data and environmental data obtained from a multi-source data acquisition module. The model output is the probability value of pest and disease occurrence. The model training process uses transfer learning method and is optimized based on the ResNet architecture. The prediction accuracy is required to be greater than 90%, and the prediction results are accompanied by a confidence score.

7. The real-time monitoring and precise control system for the entire life cycle of crops based on digital twins as described in claim 6, characterized in that: The irrigation decision is based on the crop water requirement model and real-time environmental data. The daily irrigation water requirement is calculated. First, the crop evapotranspiration ETc is calculated using the FAO Penman-Monteith equation. At the same time, the effective rainfall R is obtained based on historical meteorological data or real-time rainfall sensors. When the real-time soil moisture is lower than the preset threshold, the irrigation amount I = ETc - R. The fertilization decision-making adopts the nutrient balance method, which calculates the amount of fertilizer based on the target yield data and soil nutrient measurement values ​​provided by the data analysis and prediction module. The pest and disease control decision is based on the pest and disease prediction probability results provided by the data analysis and prediction module. The threshold decision method is used for judgment. The preset risk threshold is 0.

7. When the pest and disease probability exceeds the threshold, it is judged as a high risk and a pesticide application instruction is triggered. The amount of pesticide applied is calculated based on the crop planting area and the severity of pests and diseases.

8. The real-time monitoring and precise control system for the entire life cycle of crops based on digital twins as described in claim 7, characterized in that: The irrigation control receives and parses instructions from the precision control decision module through a PLC controller, and then drives the solenoid valve or drip irrigation system to perform irrigation operations. The control logic includes time-based control and quantity-based control. The fertilization control module receives fertilization instructions from the precision regulation decision module based on the control instruction execution module, and calculates the application amount and ratio of nitrogen, phosphorus and potassium based on the target crop yield, soil nutrient measurement values ​​and real-time growth data. The pest and disease control process first receives pest and disease control instructions from the decision-making module through a spraying robot. The instructions include the GPS coordinates of the target area, the type of pesticide to be applied, and the recommended dosage.

Images