Agricultural product distribution box remote temperature control strategy scheduling method and device

Abstract

The application relates to the technical field of agricultural product cold chain logistics, and particularly relates to a remote temperature control strategy scheduling method and device for an agricultural product distribution box, which comprises the following steps: acquiring micro-environment data, equipment state data and geographical position data in real time; constructing a real-time synchronous digital twin for each distribution box; the digital twin receives the micro-environment data, the equipment state data and the geographical position data, and combines a preset agricultural product temperature control demand knowledge graph and a distribution box performance energy consumption model to simulate and simulate temperature and humidity and energy consumption trends under different scheduling strategies; a multi-agent collaborative decision-making framework is constructed, the temperature control set points and the refrigeration power of each box body are dynamically adjusted based on the simulation results, and a dynamic variable temperature strategy or a multi-category partition collaborative temperature control strategy is selectively executed according to the distribution agricultural product categories and the mixed loading conditions; and the digital twin and the multi-agent collaborative decision-making framework are iteratively trained and updated. The application helps to improve the precision of agricultural product temperature control.

Description

Technical Field

[0001] This application relates to the field of cold chain logistics technology for agricultural products, and in particular to a method and device for remote temperature control strategy scheduling of agricultural product delivery boxes. Background Technology

[0002] Cold chain logistics for agricultural products is a crucial link in ensuring the quality of fresh agricultural products, reducing post-harvest losses, and connecting production and sales, playing a vital supporting role in the modernization of agricultural product circulation. With the rapid development of new business models such as fresh food e-commerce and community group buying, the demand for cross-regional, long-distance, and multi-batch delivery of agricultural products continues to grow, placing higher demands on the temperature control accuracy and responsiveness of cold chain delivery boxes.

[0003] Existing temperature control technology for agricultural product delivery boxes still has significant shortcomings: First, the temperature control strategy is rigid, mostly adopting a fixed constant temperature mode, relying on manual settings, and cannot be dynamically adjusted according to the characteristics of agricultural products, making it difficult to meet the needs of staged temperature variation preservation and ripening; second, the sensing dimensions are limited, mostly monitoring only the temperature inside the box, lacking the collection of multiple parameters such as humidity, gas concentration, equipment power consumption, and geographical location, resulting in incomplete data and the possibility of "temperature meeting the standard but quality being damaged"; third, the response is passive and lagging, generally using threshold trigger control, unable to predict temperature fluctuations in advance, and untimely temperature control adjustments, which can easily lead to the spoilage of agricultural products; fourth, there is insufficient coordination among multiple boxes, mostly controlling each box independently, making it difficult to achieve coordinated scheduling of multiple boxes, and easily causing thermal interference.

[0004] In conclusion, there is an urgent need for a temperature control method that can move from "passive insulation" to "active intelligent control" and from "isolated single container" to "network collaboration" in order to solve the above problems and improve the accuracy of temperature control for agricultural products. Summary of the Invention

[0005] Therefore, it is necessary to provide a method and device for remote temperature control strategy scheduling of agricultural product delivery boxes that can improve the accuracy of temperature control for agricultural products, in order to address the above-mentioned technical problems.

[0006] In a first aspect, this application provides a method for remote temperature control strategy scheduling of agricultural product delivery boxes, the method comprising: The sensor array deployed in the delivery box acquires real-time data on the microenvironment inside the box, equipment status data, and geographical location data; A real-time synchronized digital twin is constructed for each delivery box. The digital twin receives the microenvironment data, the equipment status data, and the geographical location data. Combined with a preset knowledge graph of agricultural product temperature control requirements and a delivery box performance and energy consumption model, the temperature, humidity, and energy consumption trends under different scheduling strategies are simulated. A multi-agent collaborative decision-making framework is constructed. The multi-agent collaborative decision-making framework dynamically adjusts the temperature control setpoint and cooling power of each box based on simulation results. At the same time, according to the type of agricultural products being delivered and the mixed packaging situation, it selectively executes dynamic temperature change strategy or multi-category zone collaborative temperature control strategy, and generates decision instructions to be sent to the delivery box for execution. After each delivery task is completed, the digital twin and the multi-agent collaborative decision-making framework are iteratively trained and updated based on the actual execution data.

[0007] In one embodiment, the construction of a multi-agent collaborative decision-making framework, which dynamically adjusts the temperature control setpoint and cooling power of each box based on simulation results, and selectively executes a dynamic temperature-changing strategy or a multi-category zone collaborative temperature control strategy according to the type and mixed packaging of the delivered agricultural products, and generates decision instructions to be sent to the delivery boxes for execution, includes: Deploy a box-level intelligent agent for each delivery box. The box-level intelligent agent acquires the temperature, humidity and energy consumption trends output by the digital twin of the box in real time, and monitors the instantaneous microenvironmental changes and equipment status of the box. A coordinating agent is deployed in the regional dispatch center. The coordinating agent receives the status data of the box itself uploaded by all box-level agents, the simulation results output by the digital twins of each box, external environmental information, and delivery route planning information. Based on the box-level intelligent agent and the coordinating intelligent agent, a multi-agent collaborative decision-making framework is constructed. Based on the simulation results, the temperature control setpoint and cooling power of each box are dynamically adjusted through the multi-agent collaborative decision-making framework. At the same time, according to the type of agricultural products being delivered and the mixed packaging situation, dynamic temperature change strategy or multi-category zone collaborative temperature control strategy is selectively executed, and decision instructions are generated and sent to the delivery box for execution.

[0008] In one embodiment, the step of selectively implementing a dynamic temperature control strategy or a multi-category zoned collaborative temperature control strategy based on the type and mixed packaging of the delivered agricultural products, and generating a decision instruction to be sent to the delivery box for execution, includes: The coordinating agent first queries a preset knowledge graph of agricultural product temperature control requirements based on the type of agricultural product being delivered, in order to determine whether the current delivery task requires the execution of a dynamic temperature control strategy. If a dynamic temperature change strategy needs to be implemented, the coordinating agent will decompose the temperature change curve of the corresponding agricultural product category into multiple stages of temperature setpoints, and generate temperature switching instructions for each stage in combination with the expected delivery time. If the same delivery box contains multiple types of agricultural products, the coordinating agent will further execute a multi-category zoned collaborative temperature control strategy and generate corresponding temperature control instructions based on the mixed loading situation. The decision command formed by the temperature switching command and / or the temperature control command is sent to the delivery box for execution.

[0009] In one embodiment, if the same delivery box contains multiple types of agricultural products, the coordinating agent further executes a multi-category zoned collaborative temperature control strategy, generating corresponding temperature control instructions based on the mixed loading situation, including: If the same delivery box contains multiple types of agricultural products, the coordinating agent obtains the suitable temperature range and ethylene sensitivity parameters of each type of agricultural product in the same delivery box, and calls the preset clustering algorithm to allocate different types of agricultural products to different independent physical partitions in the box according to temperature similarity and ethylene tolerance, and generates corresponding temperature and humidity target values ​​for each partition. The coordinating agent invokes the fluid dynamics model of the air duct inside the box, with the target temperature and humidity values ​​as constraints, and with the dual optimization objectives of minimizing thermal interference between zones and minimizing total cooling energy consumption, iteratively calculates the optimal fan speed combination and damper opening sequence required to maintain independent temperature control of each zone. The coordinating agent will combine the calculated optimal fan speed combination and damper opening sequence with the temperature and humidity target values ​​of each zone to generate a temperature control command for multi-category zone collaborative temperature control.

[0010] In one embodiment, the remote temperature control strategy scheduling method for agricultural product delivery boxes further includes: Based on the external environmental information, the predicted environmental temperature along the route is obtained, and based on the delivery route planning information and the predicted environmental temperature along the route, the trend of environmental temperature change that each box will enter in real time is determined. If it is predicted that the chamber will enter a high-temperature zone, a pre-cooling command will be issued to the corresponding chamber in advance, so that the temperature inside the chamber will be reduced to a preset pre-cooling value lower than the current set point before entering the high-temperature zone. If traffic congestion is predicted to cause delivery delays, an energy-saving command will be issued to the corresponding container within the acceptable short-term temperature fluctuation range for agricultural product quality. This will allow the container to relax its temperature control precision and enter energy-saving mode to reduce energy consumption.

[0011] In one embodiment, the construction of a real-time synchronized digital twin for each delivery box, wherein the digital twin receives the microenvironment data, the equipment status data, and the geographical location data, and combines a preset knowledge graph of agricultural product temperature control requirements and a delivery box performance and energy consumption model, simulates the temperature, humidity, and energy consumption trends under different scheduling strategies, including: A corresponding digital twin instance is created in the cloud for each delivery box, and the microenvironment data, the device status data and the geographical location data are synchronized to the digital twin instance in real time, so that the digital twin and the physical delivery box maintain the same status. The digital twin queries a pre-defined knowledge graph of agricultural product temperature control requirements based on the agricultural product category information of this delivery task, and obtains the appropriate temperature and humidity range, chilling injury threshold, ethylene sensitivity and temperature change curve of the corresponding category as the temperature control constraints for simulation. The digital twin invokes the performance and energy consumption model of the delivery box, which is used to calculate the energy consumption and temperature response under different operating conditions. After loading the temperature control constraints and the performance and energy consumption model of the delivery box, the temperature, humidity and energy consumption trends under different scheduling strategies are simulated using the digital twin.

[0012] In one embodiment, after loading the temperature control constraints and the performance and energy consumption model of the delivery box, the simulation of temperature, humidity, and energy consumption trends under different scheduling strategies using the digital twin includes: By integrating real-time synchronized microenvironmental data, equipment status data, geographical location data, and predicted ambient temperature along the route, a dynamic simulation input field is constructed. Based on the dynamic simulation input field, the temperature control constraints, and the performance and energy consumption model of the delivery box, the digital twin calls the built-in hybrid driving model to generate a variety of candidate scheduling strategies. The hybrid driving model is composed of a physical law model and a machine learning prediction model in parallel. The physical law model is responsible for the basic thermodynamic calculations, and the machine learning model is responsible for compensating for nonlinear deviations and accelerating the simulation speed. For each candidate scheduling strategy, the digital twin adopts a parallel simulation mechanism to simultaneously input the multiple candidate scheduling strategies into multiple hybrid driving model instances, deduce the dynamic change curve of temperature and humidity inside the box and the cumulative energy consumption value within a preset time under each strategy, and output temperature, humidity and energy consumption trend data.

[0013] In one embodiment, the microenvironment data, the device status data, and the geographic location data are uploaded to the cloud using an adaptive communication strategy that combines LPWAN and cellular networks. The adaptive communication strategy selects a LoRa, NB-IoT, 4G, or 5G transmission channel based on the current signal strength, data urgency, and pricing strategy.

[0014] Secondly, this application also provides a remote temperature control strategy scheduling device for agricultural product delivery boxes. The device includes: The data acquisition module is used to acquire real-time data on the microenvironment inside the delivery box, equipment status data, and geographical location data through a sensor array deployed on the delivery box; The simulation module is used to build a real-time synchronized digital twin for each delivery box. The digital twin receives the microenvironment data, the equipment status data, and the geographical location data, and combines them with a preset knowledge graph of agricultural product temperature control requirements and a delivery box performance and energy consumption model to simulate the temperature, humidity, and energy consumption trends under different scheduling strategies. The decision instruction generation module is used to construct a multi-agent collaborative decision-making framework. The multi-agent collaborative decision-making framework dynamically adjusts the temperature control setpoint and cooling power of each box based on simulation results. At the same time, according to the type of agricultural products being delivered and the mixed packaging situation, it selectively executes a dynamic temperature change strategy or a multi-category zone collaborative temperature control strategy, and generates decision instructions to be sent to the delivery box for execution. The iterative update module is used to iteratively train and update the digital twin and the multi-agent collaborative decision-making framework based on the actual execution data after each delivery task is completed.

[0015] In summary, this application includes the following beneficial technical effects: By synchronously collecting multi-dimensional real-time data on the internal environment, equipment operation, and geographical location of the delivery box using a sensor array, the single data collection mode is abandoned, laying a solid and reliable data foundation for precise temperature control and reducing temperature control deviations caused by missing information. A real-time synchronized digital twin is constructed for each delivery box, and simulations are performed using a pre-set knowledge graph of agricultural product temperature control requirements and a delivery box performance and energy consumption model. This allows for the prediction of temperature and humidity changes in advance, matching and adapting temperature control boundaries according to the characteristics of the agricultural products, avoiding temperature control errors caused by fixed parameter adjustments and differences in equipment operating conditions, and achieving precise pre-planning of temperature control strategies. Using a multi-agent collaborative decision-making framework that includes box-level agents and coordinating agents, dynamic temperature-changing strategies or multi-category zoned collaborative temperature control strategies can be selectively executed based on the type and mixed packaging of agricultural products being delivered. This can meet the phased temperature-changing needs of single-product storage and transportation, while also taking into account the differentiated temperature control requirements of mixed packaging of multiple products. This solves the problems of poor adaptability and rigid control of traditional unified temperature control, and significantly improves the precision of temperature control in complex scenarios. Based on actual execution data, iterative training and updating of the digital twin and the multi-agent collaborative decision-making framework can steadily optimize the temperature control effect and ensure the long-term stability and reliability of temperature control accuracy in agricultural product storage and transportation. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a remote temperature control strategy scheduling method for agricultural product delivery boxes in one embodiment. Figure 2 This is a flowchart illustrating the remote temperature control strategy scheduling method for agricultural product delivery boxes in another embodiment; Figure 3 This is a structural block diagram of a remote temperature control strategy scheduling device for agricultural product delivery boxes in one embodiment. Detailed Implementation

[0017] This invention provides a method and apparatus for remote temperature control strategy scheduling of agricultural product delivery boxes.

[0018] The embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the accompanying drawings and embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0019] In the description of the embodiments disclosed in this invention, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0020] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the remote temperature control strategy scheduling method for agricultural product delivery boxes in this invention includes: The S100 acquires real-time data on the microenvironment inside the delivery box, equipment status data, and geographical location data through a sensor array deployed in the delivery box.

[0021] Specifically, a multi-parameter sensor array is installed inside each delivery box and on key components. This sensor array includes at least: a high-precision digital temperature and humidity sensor for monitoring temperature and humidity in different zones within the box; an equipment status sensor for monitoring the power consumption of the refrigeration or heating module and the compressor status; and a GPS or BeiDou positioning module for acquiring the real-time geographical location information of the delivery box. The sensors continuously collect data at a preset sampling frequency (e.g., once every 10 seconds) and aggregate the data to a local data acquisition unit via a communication interface. The collected microenvironmental data includes, but is not limited to, box temperature, humidity, ethylene concentration, and CO2 concentration; equipment status data includes the instantaneous power of the refrigeration module, compressor start / stop status, and remaining battery power; and geographical location data includes longitude, latitude, altitude, and driving speed. This data serves as the input basis for subsequent digital twin construction and simulation.

[0022] S200 constructs a real-time synchronized digital twin for each delivery box. The digital twin receives micro-environment data, equipment status data, and geographical location data, and combines them with a pre-set knowledge graph of agricultural product temperature control requirements and a delivery box performance and energy consumption model to simulate the temperature, humidity, and energy consumption trends under different scheduling strategies.

[0023] Specifically, a real-time synchronized digital twin is built in the cloud for each physical delivery box. Micro-environmental data, equipment status data, and geographic location data collected by sensor arrays are continuously and in real-time synchronized to the digital twin, ensuring a high degree of consistency between the digital model and the physical box in terms of status, parameters, and dynamic changes. The digital twin further integrates a pre-built knowledge graph of agricultural product temperature control requirements to accurately match suitable temperature and humidity ranges, tolerance boundaries, and preservation characteristics for different types of agricultural products. Simultaneously, it incorporates a delivery box performance and energy consumption model to characterize the box's thermodynamic response, cooling capacity decay, and the quantitative relationship between energy consumption and operating conditions. Based on this, the digital twin simulates and extrapolates various alternative scheduling strategies, outputting the future trends of temperature and humidity changes and energy consumption changes within the box under each strategy. This enables pre-assessment of temperature control effectiveness and strategy optimization, allowing control to move beyond post-event feedback and instead achieve advance planning, precise matching, and risk prediction based on model prediction, significantly improving the predictability, stability, and accuracy of temperature control.

[0024] S300 constructs a multi-agent collaborative decision-making framework. Based on simulation results, the multi-agent collaborative decision-making framework dynamically adjusts the temperature control setpoint and cooling power of each box. At the same time, according to the type of agricultural products being delivered and the mixed packaging situation, it selectively executes dynamic temperature change strategy or multi-category zone collaborative temperature control strategy, and generates decision instructions to be sent to the delivery box for execution.

[0025] Specifically, a multi-agent collaborative decision-making framework is constructed, comprising box-level agents and coordinating agents. Each delivery box and its digital twin are defined as a box-level agent, responsible for local state awareness and rapid response. The regional dispatch center is defined as the coordinating agent, responsible for receiving state information from all box-level agents, simulation results from the digital twins, external environmental information, and delivery route planning information. Based on the simulation results of the digital twins, this multi-agent collaborative decision-making framework dynamically adjusts the temperature control setpoints and cooling power of each delivery box. Simultaneously, based on the types of agricultural products in the current delivery task and whether multiple types of agricultural products are mixed in the box, the framework selectively executes dynamic temperature-changing strategies (e.g., for single products requiring phased temperature changes) or multi-category zoned collaborative temperature control strategies (e.g., for multiple categories with different temperature zone requirements mixed in the same box). Finally, the multi-agent collaborative framework issues the generated decision commands to the corresponding delivery boxes for execution, achieving precise temperature scheduling within the boxes.

[0026] After each delivery task is completed, the S400 iteratively trains and updates the digital twin and multi-agent collaborative decision-making framework based on the actual execution data.

[0027] Specifically, a closed-loop optimization mechanism for delivery tasks is established. After each delivery task is completed, the actual execution data of the entire task process is automatically collected, including actual temperature and humidity curves, equipment energy consumption data, control command execution records, environmental disturbances, and feedback data related to agricultural product quality. Based on this real-world operational data, the digital twin is iteratively trained and its parameters are updated to continuously correct model biases, optimize thermodynamic response accuracy, and adapt to equipment aging and operational drift. Simultaneously, the multi-agent collaborative decision-making framework is updated in terms of strategy and logic optimization to improve the adaptability, stability, and optimization capabilities of the decision-making rules. Through continuous iteration driven by closed-loop tasks and data, the simulation accuracy of the digital twin and the control performance of the decision-making framework are continuously improved, gradually converging errors and adapting to long-term operational changes. This ensures long-term and stable temperature control accuracy for agricultural products throughout the entire process, avoiding the problems of accuracy decay and control failure after long-term use of traditional fixed models. This achieves continuous evolution of cold chain temperature control capabilities and a steady improvement in quality assurance levels.

[0028] In one embodiment, such as Figure 2 As shown, S300 includes: S310 deploys a box-level intelligent agent for each delivery box. The box-level intelligent agent obtains the temperature, humidity and energy consumption trends output by the digital twin of the box in real time, and monitors the instantaneous microenvironmental changes and equipment status of the box. S320 deploys a coordinating agent in the regional dispatch center. The coordinating agent receives the status data of the box itself uploaded by all box-level agents, the simulation results output by the digital twins of each box, external environmental information, and delivery route planning information. S330, based on box-level intelligent agents and coordinating intelligent agents, constructs a multi-agent collaborative decision-making framework; The S340 uses a multi-agent collaborative decision-making framework to dynamically adjust the temperature control setpoint and cooling power of each box based on simulation results. At the same time, it selectively executes dynamic temperature change strategy or multi-category zone collaborative temperature control strategy according to the type of agricultural products being delivered and the mixed packaging situation, and generates decision instructions to be sent to the delivery box for execution.

[0029] Specifically, a box-level intelligent agent is deployed for each delivery box. This agent can acquire real-time temperature, humidity, and energy consumption trend data output by the box's digital twin, while continuously monitoring instantaneous microenvironmental changes within the box (such as sudden temperature rises or humidity changes due to door opening) and equipment status (such as cooling module power, compressor start / stop, and battery charge). The box-level intelligent agent possesses local rapid response capabilities, enabling it to independently execute preliminary control actions based on preset emergency rules in the event of communication interruptions or delays in coordinating agent commands. For example, it can temporarily increase fan speed or activate backup cold storage release. Secondly, a coordinating agent is deployed at the regional dispatch center (e.g., a cloud server or local dispatch platform). This coordinating agent is responsible for receiving box status data (including real-time temperature and humidity, equipment status, and geographical location) uploaded by all box-level intelligent agents, simulation results output by each box's digital twin (such as future temperature and humidity change curves and predicted energy consumption), external environmental information (such as real-time weather, traffic conditions, and ambient temperature forecasts along the route), and delivery route planning information (such as estimated arrival times for each segment). The coordinating agent, acting as the core of global decision-making, aggregates information from multiple sources and maintains a global state view. Then, based on the deployed box-level agents and the coordinating agent, a complete multi-agent collaborative decision-making framework is constructed. This framework defines the communication protocol, data interaction format, and decision-making authority boundaries between the box-level agents and the coordinating agent. The box-level agents are responsible for local perception and rapid response, while the coordinating agent is responsible for global optimization and strategy coordination. The two exchange data bidirectionally via wireless networks (such as 4G / 5G, LoRa, etc.), forming a distributed intelligent decision-making network. Finally, through this multi-agent collaborative decision-making framework, the coordinating agent dynamically calculates and adjusts the temperature control setpoint and cooling power of each delivery box based on the simulation results of the digital twins of each box. Specifically, the coordinating agent aims to minimize total global energy consumption, minimize total cargo loss rate, and maximize on-time delivery rate. Combining the current state of each box and external environmental constraints, it solves for a set of optimal temperature control parameters and distributes them to the corresponding box-level agents via the network. Meanwhile, the framework selectively executes corresponding temperature control strategies based on the types of agricultural products loaded in the current delivery task (e.g., whether it is a single type or multiple types) and whether the contents of the box are mixed: if it is a single type with periodic temperature changes, a dynamic temperature change strategy is executed; if it is a mixed type of products, a multi-type zoned collaborative temperature control strategy is executed. The framework translates the above strategies into specific decision instructions (e.g., setting temperature, cooling power, temperature change time points, zoned control parameters, etc.) and sends them to the actuators of the delivery box. The box-level intelligent agent is responsible for local instruction parsing and execution, thereby achieving precise dynamic scheduling of the temperature inside the box.

[0030] In one embodiment, depending on the type and mixed nature of the agricultural products being delivered, a dynamic temperature control strategy or a multi-category zoned collaborative temperature control strategy is selectively implemented, and a decision instruction is generated and sent to the delivery box for execution, including: The coordinating agent first queries a pre-defined knowledge graph of agricultural product temperature control requirements based on the type of agricultural product being delivered to determine whether a dynamic temperature control strategy is required for the current delivery task. If a dynamic temperature control strategy is required, the coordinating agent decomposes the temperature curve of the corresponding agricultural product type into multiple temperature setpoints and generates temperature switching instructions for each stage based on the estimated delivery time. If the same delivery box contains multiple types of agricultural products, the coordinating agent further executes a multi-type zoned collaborative temperature control strategy, generating corresponding temperature control instructions based on the mixed loading situation. The decision instructions formed by the temperature switching instructions and / or temperature control instructions are then sent to the delivery box for execution.

[0031] Specifically, firstly, the coordinating agent queries a pre-defined knowledge graph of agricultural product temperature control requirements based on the product category information loaded in the current delivery task. It then retrieves the corresponding temperature control requirement data from the knowledge graph and determines whether a dynamic temperature control strategy is needed for this delivery. Dynamic temperature control strategies are typically applicable to agricultural products with post-ripening physiological characteristics (such as mangoes, tomatoes, and bananas). These products require lower temperatures in the early stages of transportation to inhibit ripening, while later stages require appropriate increases in temperature to promote ripening or maintain flavor. Maintaining a constant temperature throughout the process may actually lead to a decline in quality. If the determination indicates that a dynamic temperature control strategy is needed, the coordinating agent extracts the corresponding temperature curve for the agricultural product from the knowledge graph (e.g., 12-15℃ in the early stages, 18-20℃ in the later stages). Then, the coordinating agent decomposes this temperature curve into multiple stages, each containing a temperature setpoint and the duration or start and end times of that stage. Simultaneously, based on the estimated delivery time (based on route planning information and real-time vehicle speed estimation), the coordinating agent generates specific temperature switching instructions for each stage, such as "10 hours after departure, switch the set temperature from 14℃ to 19℃". These instructions ensure that the temperature control strategy is precisely matched with the delivery progress. On the other hand, if the coordinating agent detects that the same delivery box contains multiple types of agricultural products (e.g., strawberries and apples at the same time), it determines that a multi-category zone temperature control strategy needs to be implemented. At this time, the coordinating agent generates corresponding temperature control instructions based on the differences in temperature and humidity requirements of each category. These instructions typically involve setting different target temperatures and humidity levels for different physical zones within the box and controlling actuators such as cooling ducts and fans to reduce thermal interference between zones, thereby achieving independent control of multiple temperature zones within the same box. Finally, the coordinating agent merges the temperature switching instructions (from the dynamic temperature change strategy) and / or temperature control instructions (from the multi-category zone collaborative temperature control strategy) generated in the above steps into a final decision instruction set and sends it to the corresponding delivery box for execution via the communication network. If only a dynamic temperature-changing strategy is required (for a single product category with temperature-changing needs), the decision instruction only includes a temperature switching instruction. If only a multi-product zone-based collaborative temperature control strategy is required (for mixed multi-product packs without phased temperature changes), the decision instruction only includes a temperature control instruction. If both strategies are required simultaneously (e.g., a single product in a mixed multi-product pack also requires temperature changes), the decision instruction includes both types of instructions. After receiving the instruction, the box-level intelligent agent is responsible for parsing and controlling local actuators such as the cooling module, dampers, and fans to complete precise temperature control scheduling.

[0032] In one embodiment, if the same delivery box contains multiple types of agricultural products, the coordinating agent further executes a multi-category zoned collaborative temperature control strategy, generating corresponding temperature control instructions based on the mixed loading situation, including: If a delivery box contains multiple types of agricultural products, the coordinating agent obtains the suitable temperature range and ethylene sensitivity parameters for each type of agricultural product within the same delivery box. It then uses a pre-defined clustering algorithm to assign different types of agricultural products to different independent physical zones within the box based on temperature zone similarity and ethylene tolerance, generating corresponding temperature and humidity target values ​​for each zone. The coordinating agent then uses the fluid dynamics model of the air duct within the box, with the temperature and humidity target values ​​as constraints and minimizing thermal interference between zones and total cooling energy consumption as dual optimization objectives. It iteratively calculates the optimal fan speed combination and damper opening sequence required to maintain independent temperature control for each zone. Finally, the coordinating agent combines the calculated optimal fan speed combination and damper opening sequence with the temperature and humidity target values ​​for each zone to generate temperature control commands for coordinated temperature control of multiple product zones.

[0033] Specifically, firstly, the coordinating agent obtains the suitable temperature range and ethylene sensitivity parameters of various agricultural products within the same delivery box. This information is obtained by querying a pre-defined knowledge graph of agricultural product temperature control requirements. For example, strawberries have a suitable temperature range of 0-4℃ and are highly sensitive to ethylene, so they should not be stored with apples, which produce high levels of ethylene. The coordinating agent then calls a pre-defined clustering algorithm to group different agricultural products according to temperature similarity and ethylene tolerance. The goal of clustering is to assign products with similar temperature ranges and minimal ethylene interaction to the same independent physical partition, while assigning products with large temperature differences or conflicting ethylene sensitivities to different partitions. After clustering, the coordinating agent generates corresponding temperature and humidity target values ​​for each partition, serving as the basis for setting independent temperature control for that partition. Secondly, the coordinating agent calls a fluid dynamics model of the internal air ducts. This model is a digital description of the internal air duct structure, fan arrangement, damper positions, and airflow paths of the delivery box, capable of simulating the distribution of cooling capacity and temperature distribution in each partition under different combinations of fan speeds and damper openings. With the generated target temperature and humidity values ​​for each zone as constraints, the coordinating agent uses the dual optimization objectives of "minimizing thermal interference between zones" and "minimizing total cooling energy consumption" to iteratively calculate the optimal fan speed combination and damper opening sequence required to maintain independent temperature control for each zone. Specifically, the coordinating agent employs a multi-objective optimization algorithm (such as NSGA-II or particle swarm optimization) to search within the feasible region, minimizing the deviation between the actual temperature and the target value for each zone, while minimizing heat exchange (thermal interference) between zones and minimizing compressor energy consumption. Finally, the coordinating agent combines the calculated optimal fan speed combination and damper opening sequence with the target temperature and humidity values ​​for each zone to generate temperature control commands for multi-category zone collaborative temperature control. These commands include the target opening of each damper, the target speed of each fan, and the target temperature and humidity for each zone. The commands are sent to the actuators in the delivery box via the communication network. The box-level agent is responsible for local parsing and driving the damper motors and fan speed control modules, thereby achieving precise independent control of multiple temperature zones within the same box and effectively solving the temperature zone conflict problem when multiple categories are mixed.

[0034] In one embodiment, the remote temperature control strategy scheduling method for agricultural product delivery boxes further includes: Based on external environmental information, the system obtains predicted ambient temperature data along the route. Using delivery route planning information and the predicted ambient temperature data, it determines in real-time the ambient temperature trend each container will soon enter. If it predicts entering a high-temperature zone, it issues a pre-cooling command to the corresponding container in advance, lowering the internal temperature to a preset pre-cooling value below the current set point before entering the high-temperature zone. If it predicts traffic congestion causing delivery delays, it issues an energy-saving command to the corresponding container within the acceptable short-term temperature fluctuation range for agricultural product quality, allowing the container to relax its temperature control precision and enter energy-saving mode to reduce energy consumption.

[0035] Specifically, firstly, the coordinating agent generates predicted ambient temperature data along the route based on collected external environmental information. This external environmental information includes, but is not limited to, real-time weather data, meteorological forecasts along the route, geographic information, and measured ambient temperatures reported by other vehicles on the same road segment. The coordinating agent then fuses this multi-source data, combining it with delivery route planning information (including the coordinate sequence on the route and the estimated time to pass through each segment). Using spatial interpolation and time series prediction methods (such as inverse distance weighted interpolation combined with Kalman filtering), it generates predicted ambient temperature values ​​for each segment along the route from the current location to the destination, forming a "path-temperature" sequence with time and location markers. Secondly, based on the aforementioned environmental temperature prediction data along the route and combined with delivery route planning information, the coordinating agent judges in real time the environmental temperature change trend that each delivery box will enter. Specifically, the coordinating agent scans the road sections that the vehicle will pass through in the next period of time (e.g., the next hour), analyzes and predicts the magnitude of temperature rise, fall, rate of change, and whether extreme high or low temperature ranges will occur. For example, if the prediction shows that the vehicle will enter the city center 5 kilometers ahead, and the environmental temperature will rise from 25°C to 35°C within 10 minutes, it is determined that the vehicle will "enter a high-temperature area." If the judgment result is that the vehicle will enter a high-temperature area, the coordinating agent issues a pre-cooling command to the corresponding delivery box-level agent in advance. This pre-cooling command requires the box to actively reduce the internal temperature to a preset pre-cooling value below the current set point before entering the high-temperature area. By storing cold in advance, the delivery box can resist external thermal shock by relying on its own insulation and stored cold energy after entering the high-temperature area, avoiding large fluctuations in internal temperature and thus ensuring the quality of agricultural products. If the assessment indicates that traffic congestion will cause delivery delays, the coordinating agent issues an energy-saving command to the corresponding delivery box. This command requires the box to proactively increase temperature control precision (e.g., expanding the control dead zone from ±0.5℃ to ±2℃) within a short-term temperature fluctuation range acceptable for agricultural product quality, and to put the refrigeration module into energy-saving mode (e.g., reducing compressor operating frequency or increasing start-stop cycles). This significantly reduces energy consumption and extends remaining runtime without compromising agricultural product quality, thus addressing delivery delays caused by congestion.

[0036] In one embodiment, a real-time synchronized digital twin is constructed for each delivery box. The digital twin receives microenvironmental data, equipment status data, and geographic location data, and combines this with a pre-defined knowledge graph of agricultural product temperature control requirements and a delivery box performance and energy consumption model to simulate temperature, humidity, and energy consumption trends under different scheduling strategies, including: A corresponding digital twin instance is created in the cloud for each delivery box, and microenvironmental data, equipment status data, and geographical location data are synchronized to the digital twin instance in real time to ensure that the digital twin and the physical delivery box remain consistent. Based on the agricultural product category information for this delivery task, the digital twin queries a pre-defined knowledge graph of agricultural product temperature control requirements to obtain the suitable temperature and humidity range, chilling injury threshold, ethylene sensitivity, and temperature change curve for the corresponding category, which serve as temperature control constraints for simulation. The digital twin calls the delivery box performance and energy consumption model, which is used to calculate energy consumption and temperature change response under different operating conditions. After completing the loading of the temperature control constraints and the delivery box performance and energy consumption model, the digital twin is used to simulate the temperature, humidity, and energy consumption trends under different scheduling strategies.

[0037] Specifically, firstly, a corresponding digital twin instance is created in the cloud for each physical delivery box. This instance is a virtual digital model that corresponds one-to-one with the physical delivery box. Through IoT communication protocols, real-time micro-environmental data, equipment status data, and geographic location data collected by the sensor array are continuously and in real-time synchronized to this digital twin instance. Through this real-time synchronization mechanism, the digital twin can accurately reflect the current state, historical change trajectory, and dynamic response characteristics of the physical delivery box, ensuring a high degree of consistency between the virtual model and the physical entity, and providing accurate initial state values ​​for subsequent simulations. Secondly, the digital twin queries a pre-defined knowledge graph of agricultural product temperature control requirements based on the product category information loaded in this delivery task. This knowledge graph stores post-harvest physiological characteristic parameters of various agricultural products in the form of a structured database or graph database, including but not limited to: suitable temperature and humidity range (e.g., 0~4℃, 85%~95%RH), chilling injury threshold (e.g., chilling injury occurs when the temperature drops below 10℃ for some tropical fruits), freezing injury threshold (e.g., tissue freezing damage occurs when the temperature drops below -1℃), ethylene sensitivity (high, medium, low), and temperature variation curves for agricultural products with post-ripening characteristics (e.g., mangoes: 12-15℃ in the early stage, 18-20℃ in the later stage). The digital twin uses the above information as the temperature control constraints for subsequent simulations, meaning that all candidate scheduling strategies must ensure that the temperature and humidity inside the box do not exceed these safety boundaries; otherwise, they are deemed infeasible strategies. Then, the digital twin calls the performance and energy consumption model of the delivery box. This model is a mathematical description of the thermodynamic and energy consumption characteristics of the delivery box. It usually includes the thermal insulation performance parameters of the box (such as the comprehensive heat transfer coefficient R or U value), the power-efficiency curve of the cooling / heating module (the relationship between cooling capacity and input power at different set temperatures), the endurance characteristics under different ambient temperature conditions (such as the duration for which an active box can maintain 2~8℃ in an environment of 35℃), and the phase change temperature, latent heat capacity, and release rate curve of the phase change material (such as ice packs, eutectic plates) in the passive insulation box. This model can calculate the response of the internal temperature change and the corresponding cumulative energy consumption value in the future period of time based on the input ambient temperature, initial box temperature, cargo heat load and other operating condition parameters. It is the core computing engine for the digital twin to perform simulation and inference. Finally, after loading the aforementioned temperature control constraints and invoking the performance and energy consumption model of the delivery box, the digital twin officially executes simulations of different scheduling strategies. Specifically, the coordinating intelligence sends a set of candidate scheduling strategies (such as different temperature control setpoint sequences, cooling power allocation schemes, etc.) to the digital twin. Based on the current real-time synchronized initial state (temperature and humidity inside the box, equipment status, geographical location, etc.), and combined with the predicted ambient temperature along the route as external boundary conditions, the digital twin uses the loaded performance and energy consumption model to deduce the dynamic change curve of temperature and humidity inside the box and the corresponding cumulative energy consumption value for each strategy over a future period of time (e.g., remaining delivery time).Simulation results (including whether temperature and humidity remain within safe constraints, temperature fluctuation range, expected energy consumption, and whether the risk of chilling / freezing injury is triggered) are then fed back to the coordinating agent, which selects the optimal scheduling strategy in multi-objective optimization. Through this process, the digital twin achieves a virtual rehearsal of the actual delivery process, enabling temperature control decisions to be based on model prediction rather than post-event feedback, thereby significantly improving the foresight and accuracy of temperature control.

[0038] In one embodiment, after loading the temperature control constraints and the performance and energy consumption model of the delivery box, the temperature, humidity, and energy consumption trends under different scheduling strategies are simulated using a digital twin, including: By fusing real-time synchronized microenvironmental data, equipment status data, geographical location data, and predicted ambient temperature along the route, a dynamic simulation input field is constructed. Based on the dynamic simulation input field, temperature control constraints, and the performance and energy consumption model of the delivery box, the digital twin calls the built-in hybrid driving model to generate multiple candidate scheduling strategies. The hybrid driving model consists of a physical law model and a machine learning prediction model connected in parallel. The physical law model is responsible for the basic thermodynamic calculations, while the machine learning model is responsible for compensating for nonlinear deviations and accelerating the simulation speed. For each candidate scheduling strategy, the digital twin adopts a parallel simulation mechanism, simultaneously inputting multiple candidate scheduling strategies into multiple hybrid driving model instances to deduce the dynamic change curves of temperature and humidity inside the box and the cumulative energy consumption value within a preset time under each strategy, and outputting temperature, humidity, and energy consumption trend data.

[0039] Specifically, firstly, the digital twin fuses real-time synchronized microenvironmental data, equipment status data, geographic location data, and predicted ambient temperature along the delivery route. The fusion process includes timestamp alignment, outlier removal, missing value imputation, and spatial coordinate matching, ultimately constructing a dynamic simulation input field covering the entire delivery process. Secondly, based on the aforementioned dynamic simulation input field, the loaded temperature control constraints, and the delivery box performance and energy consumption model, the digital twin invokes its built-in hybrid driving model. This model consists of two parallel parts: one is a physical law model, based on the first law of thermodynamics, heat transfer equations, and refrigeration cycle principles, calculating the dynamic response of the box's internal temperature to changes in factors such as cold input, heat leakage, and cargo respiration heat from a mechanistic perspective; the other is a machine learning prediction model, using neural networks or Gaussian process regression models trained on historical operating data to compensate in real-time for nonlinear deviations not covered by the physical model (such as the non-ideal characteristics of latent heat release in phase change materials and the non-steady-state response during compressor start-up and shutdown), and accelerating simulation calculations through low-dimensional feature mapping. The two models operate in parallel: the physical model outputs the basic trend, and the machine learning model outputs the correction values. These are then superimposed to obtain high-precision, high-speed predictions of temperature, humidity, and energy consumption. Finally, for various candidate scheduling strategies requiring evaluation (such as different temperature setpoint sequences, cooling power allocation schemes, and temperature change time points), the digital twin employs a parallel simulation mechanism. Specifically, the digital twin simultaneously launches multiple hybrid-driven model instances on a cloud server. Each instance runs independently, corresponding to a candidate strategy. Each instance starts with the same dynamic simulation input field and uses the parallel outputs of the physical law model and the machine learning prediction model to deduce the dynamic change curves of temperature and humidity inside the box and the cumulative energy consumption value within a future preset time (e.g., remaining delivery time) under that strategy. Because multiple instances execute in parallel, the total simulation time is only equivalent to that of a single simulation, significantly improving the efficiency of strategy evaluation. After the simulation is completed, the digital twin outputs the temperature and humidity change curves, energy consumption curves, and indicators of whether temperature control constraints are met for each strategy. This trend data is fed back to the coordinating agent for selecting the optimal scheduling strategy in multi-objective optimization.

[0040] In one embodiment, microenvironment data, device status data, and geographic location data are uploaded to the cloud using an adaptive communication strategy that combines LPWAN and cellular networks. The adaptive communication strategy selects LoRa, NB-IoT, 4G, or 5G transmission channels based on the current signal strength, data urgency, and pricing strategy.

[0041] Specifically, each delivery box integrates a multi-mode communication terminal that simultaneously supports LPWAN (Low Power Wide Area Network) technology (such as LoRa) and cellular network technology (such as NB-IoT, 4G, and 5G). The terminal maintains a communication policy manager that monitors in real time the signal strength (RSSI) of the current location, the urgency level of the data to be uploaded (labeled by the data acquisition module as regular data or abnormal / alarm data), and the operator's pricing policy (e.g., the unit price of data traffic for different network channels). When data needs to be uploaded, the communication policy manager first determines the urgency level of the data. If alarm events such as abnormal temperature fluctuations, equipment failures, or cold chain interruptions are detected, the data is marked as high urgency. In this case, the policy manager prioritizes the cellular network channel with the widest coverage and lowest latency (such as 4G or 5G) to ensure that the data can be delivered to the cloud quickly and reliably to trigger early warnings or emergency dispatch. If the data is regular temperature and humidity, equipment status, or location information (low urgency), the policy manager further evaluates the current signal strength and pricing policy. In warehouse, transit station, or short-distance urban delivery scenarios, if the LoRa gateway signal strength meets the requirements (e.g., RSSI is higher than the threshold) and the operator's pricing strategy favors the use of low-cost LPWAN, then the LoRa channel will be prioritized for data upload. LoRa features strong penetration, low power consumption, and support for large-scale device concurrency, making it suitable for centralized reporting of high-density delivery box clusters, effectively reducing communication costs and battery consumption. When a vehicle enters an area without LoRa coverage (such as highways or remote suburbs) or the LoRa signal quality drops sharply, the policy manager automatically and seamlessly switches to NB-IoT or 4G / 5G cellular networks. NB-IoT is suitable for low-speed, deep-coverage IoT scenarios, ideal for near real-time uploads of regular data; while 4G / 5G is used for scenarios with higher requirements for speed or latency (e.g., uploading high-frequency sampled data or receiving large-capacity scheduling commands). The switching process uses a soft handover mechanism, establishing a connection to the new channel first and then disconnecting the old channel to avoid data loss. Through the aforementioned adaptive communication strategy that combines LPWAN and cellular networks, the optimal transmission channel can be dynamically selected in different environments. While ensuring data real-time performance and reliability, communication power consumption and operating costs can be reduced to the minimum, achieving a balance between communication cost and service quality.

[0042] In one embodiment, such as Figure 3 As shown, a remote temperature control strategy scheduling device for agricultural product delivery boxes is provided, comprising: a data acquisition module 10, a simulation module 20, a decision instruction generation module 30, and an iterative update module 40, wherein: The data acquisition module 10 is used to acquire real-time micro-environmental data, equipment status data, and geographical location data inside the delivery box through a sensor array deployed on the delivery box; The simulation module 20 is used to build a real-time synchronized digital twin for each delivery box. The digital twin receives micro-environment data, equipment status data and geographical location data, and combines them with a preset knowledge graph of agricultural product temperature control requirements and delivery box performance and energy consumption model to simulate the temperature, humidity and energy consumption trends under different scheduling strategies. The decision instruction generation module 30 is used to construct a multi-agent collaborative decision-making framework. The multi-agent collaborative decision-making framework dynamically adjusts the temperature control setpoint and cooling power of each box based on simulation results. At the same time, according to the type of agricultural products being delivered and the mixed packaging situation, it selectively executes a dynamic temperature change strategy or a multi-category zone collaborative temperature control strategy, and generates decision instructions to be sent to the delivery box for execution. The iterative update module 40 is used to iteratively train and update the digital twin and multi-agent collaborative decision-making framework based on the actual execution data after each delivery task is completed.

[0043] In one embodiment, the decision instruction generation module 30 is further configured to deploy a box-level intelligent agent for each delivery box. The box-level intelligent agent acquires the temperature, humidity, and energy consumption trends output by the digital twin of its own box in real time, and monitors the instantaneous microenvironmental changes and equipment status of its own box. A coordination intelligent agent is deployed in the regional dispatch center. The coordination intelligent agent receives the box status data uploaded by all box-level intelligent agents, the simulation results output by the digital twin of each box, external environmental information, and delivery route planning information. Based on the box-level intelligent agents and the coordination intelligent agent, a multi-agent collaborative decision-making framework is constructed. Through the multi-agent collaborative decision-making framework, the temperature control setpoint and cooling power of each box are dynamically adjusted based on the simulation results. At the same time, according to the type of agricultural products being delivered and the mixed packaging situation, a dynamic temperature change strategy or a multi-category zoned collaborative temperature control strategy is selectively executed, and decision instructions are generated and sent to the delivery boxes for execution.

[0044] In one embodiment, the decision instruction generation module 30 is further configured to coordinate the agent to first query a preset knowledge graph of agricultural product temperature control requirements based on the type of agricultural product being delivered, in order to determine whether the current delivery task requires the execution of a dynamic temperature change strategy; if a dynamic temperature change strategy is required, the agent decomposes the temperature change curve of the corresponding agricultural product type into multiple stages of temperature setpoints, and generates temperature switching instructions for each stage in conjunction with the expected delivery time; if the same delivery box contains multiple types of agricultural products, the agent further executes a multi-type zoned collaborative temperature control strategy, generating corresponding temperature control instructions based on the mixed loading situation; and sends the decision instructions formed by the temperature switching instructions and / or temperature control instructions to the delivery box for execution.

[0045] In one embodiment, the decision instruction generation module 30 is further configured to, if the same delivery box contains multiple types of agricultural products, obtain the suitable temperature range and ethylene sensitivity parameters of each type of agricultural product in the same delivery box, and call a preset clustering algorithm to allocate different types of agricultural products to different independent physical partitions in the box according to temperature similarity and ethylene tolerance, generating corresponding temperature and humidity target values ​​for each partition; the coordinating agent calls the fluid dynamics model of the air duct in the box, with the temperature and humidity target values ​​as constraints, and with minimizing thermal interference between partitions and minimizing total cooling energy consumption as dual optimization objectives, iteratively calculating the optimal fan speed combination and damper opening sequence required to maintain independent temperature control of each partition; the coordinating agent combines the calculated optimal fan speed combination and damper opening sequence with the temperature and humidity target values ​​of each partition to generate a temperature control instruction for multi-type partition coordinated temperature control.

[0046] In one embodiment, the remote temperature control strategy scheduling device for agricultural product delivery boxes further includes a prediction and judgment module, which is used to obtain predicted environmental temperature data along the route based on external environmental information, and to judge the environmental temperature change trend that each box is about to enter in real time based on the delivery route planning information and the predicted environmental temperature data along the route; if it is predicted that the box will enter a high-temperature area, a pre-cooling command is issued to the corresponding box in advance, so that the temperature inside the box is reduced to a preset pre-cooling value lower than the current set point before entering the high-temperature area; if it is predicted that traffic congestion will cause delivery delays, an energy-saving command is issued to the corresponding box within the short-term temperature fluctuation range acceptable for the quality of agricultural products, so that the box relaxes the temperature control precision and enters the energy-saving mode to reduce energy consumption.

[0047] In one embodiment, the simulation module 20 is further used to create a corresponding digital twin instance for each delivery box in the cloud, and to synchronize microenvironmental data, equipment status data, and geographical location data to the digital twin instance in real time, so that the digital twin and the physical delivery box maintain consistent status. The digital twin queries a preset knowledge graph of agricultural product temperature control requirements based on the agricultural product category information of this delivery task, and obtains the suitable temperature and humidity range, chilling injury threshold, ethylene sensitivity, and temperature change curve of the corresponding category as temperature control constraints for simulation. The digital twin calls the delivery box performance and energy consumption model, which is used to calculate the energy consumption and temperature change response under different operating conditions. After the temperature control constraints and delivery box performance and energy consumption model are loaded, the temperature, humidity, and energy consumption trends under different scheduling strategies are simulated through the digital twin.

[0048] In one embodiment, the simulation module 20 is further used to fuse real-time synchronized microenvironmental data, equipment status data, geographical location data, and predicted ambient temperature along the route to construct a dynamic simulation input field. Based on the dynamic simulation input field, temperature control constraints, and the performance and energy consumption model of the delivery box, the digital twin calls the built-in hybrid driving model to generate multiple candidate scheduling strategies. The hybrid driving model consists of a physical law model and a machine learning prediction model connected in parallel. The physical law model is responsible for the basic thermodynamic calculations, while the machine learning model is responsible for compensating for nonlinear deviations and accelerating the simulation speed. For each candidate scheduling strategy, the digital twin adopts a parallel simulation mechanism to simultaneously input multiple candidate scheduling strategies into multiple hybrid driving model instances, deduce the dynamic change curve of temperature and humidity inside the box and the cumulative energy consumption value within a preset time under each strategy, and output temperature, humidity, and energy consumption trend data.

[0049] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A remote temperature control strategy scheduling method for agricultural product delivery boxes, characterized in that, include: The sensor array deployed in the delivery box acquires real-time data on the microenvironment inside the box, equipment status data, and geographical location data; A real-time synchronized digital twin is constructed for each delivery box. The digital twin receives the microenvironment data, the equipment status data, and the geographical location data. Combined with a preset knowledge graph of agricultural product temperature control requirements and a delivery box performance and energy consumption model, the temperature, humidity, and energy consumption trends under different scheduling strategies are simulated. A multi-agent collaborative decision-making framework is constructed. The multi-agent collaborative decision-making framework dynamically adjusts the temperature control setpoint and cooling power of each box based on simulation results. At the same time, according to the type of agricultural products being delivered and the mixed packaging situation, it selectively executes dynamic temperature change strategy or multi-category zone collaborative temperature control strategy, and generates decision instructions to be sent to the delivery box for execution. After each delivery task is completed, the digital twin and the multi-agent collaborative decision-making framework are iteratively trained and updated based on the actual execution data.

2. The method for remote temperature control strategy scheduling of agricultural product delivery boxes according to claim 1, characterized in that, The construction of a multi-agent collaborative decision-making framework, which dynamically adjusts the temperature control setpoint and cooling power of each container based on simulation results, and selectively executes a dynamic temperature-changing strategy or a multi-category zone collaborative temperature control strategy according to the type and mixed packaging of the delivered agricultural products, generates decision instructions and issues them to the delivery containers for execution, including: Deploy a box-level intelligent agent for each delivery box. The box-level intelligent agent acquires the temperature, humidity and energy consumption trends output by the digital twin of the box in real time, and monitors the instantaneous microenvironmental changes and equipment status of the box. A coordinating agent is deployed in the regional dispatch center. The coordinating agent receives the status data of the box itself uploaded by all box-level agents, the simulation results output by the digital twins of each box, external environmental information, and delivery route planning information. Based on the box-level intelligent agent and the coordinating intelligent agent, a multi-agent collaborative decision-making framework is constructed. Based on the simulation results, the temperature control setpoint and cooling power of each box are dynamically adjusted through the multi-agent collaborative decision-making framework. At the same time, according to the type of agricultural products being delivered and the mixed packaging situation, dynamic temperature change strategy or multi-category zone collaborative temperature control strategy is selectively executed, and decision instructions are generated and sent to the delivery box for execution.

3. The method for remote temperature control strategy scheduling of agricultural product delivery boxes according to claim 2, characterized in that, The step of selectively implementing a dynamic temperature control strategy or a multi-category zoned collaborative temperature control strategy based on the type and mixed packaging of the delivered agricultural products, and generating decision instructions to be sent to the delivery box for execution, includes: The coordinating agent first queries a preset knowledge graph of agricultural product temperature control requirements based on the type of agricultural product being delivered, in order to determine whether the current delivery task requires the execution of a dynamic temperature control strategy. If a dynamic temperature change strategy needs to be implemented, the coordinating agent will decompose the temperature change curve of the corresponding agricultural product category into multiple stages of temperature setpoints, and generate temperature switching instructions for each stage in combination with the expected delivery time. If the same delivery box contains multiple types of agricultural products, the coordinating agent will further execute a multi-category zoned collaborative temperature control strategy and generate corresponding temperature control instructions based on the mixed loading situation. The decision command formed by the temperature switching command and / or the temperature control command is sent to the delivery box for execution.

4. The method for remote temperature control strategy scheduling of agricultural product delivery boxes according to claim 3, characterized in that, If the same delivery box contains multiple types of agricultural products, the coordinating agent further executes a multi-category zoned collaborative temperature control strategy, generating corresponding temperature control instructions based on the mixed loading situation, including: If the same delivery box contains multiple types of agricultural products, the coordinating agent obtains the suitable temperature range and ethylene sensitivity parameters of each type of agricultural product in the same delivery box, and calls the preset clustering algorithm to allocate different types of agricultural products to different independent physical partitions in the box according to temperature similarity and ethylene tolerance, and generates corresponding temperature and humidity target values ​​for each partition. The coordinating agent invokes the fluid dynamics model of the air duct inside the box, with the target temperature and humidity values ​​as constraints, and with the dual optimization objectives of minimizing thermal interference between zones and minimizing total cooling energy consumption, iteratively calculates the optimal fan speed combination and damper opening sequence required to maintain independent temperature control of each zone. The coordinating agent will combine the calculated optimal fan speed combination and damper opening sequence with the temperature and humidity target values ​​of each zone to generate a temperature control command for multi-category zone collaborative temperature control.

5. The method for remote temperature control strategy scheduling of agricultural product delivery boxes according to claim 2, characterized in that, Also includes: Based on the external environmental information, the predicted environmental temperature along the route is obtained, and based on the delivery route planning information and the predicted environmental temperature along the route, the trend of environmental temperature change that each box will enter in real time is determined. If it is predicted that the chamber will enter a high-temperature zone, a pre-cooling command will be issued to the corresponding chamber in advance, so that the temperature inside the chamber will be reduced to a preset pre-cooling value lower than the current set point before entering the high-temperature zone. If traffic congestion is predicted to cause delivery delays, an energy-saving command will be issued to the corresponding container within the acceptable short-term temperature fluctuation range for agricultural product quality. This will allow the container to relax its temperature control precision and enter energy-saving mode to reduce energy consumption.

6. The method for remote temperature control strategy scheduling of agricultural product delivery boxes according to claim 1, characterized in that, The method involves constructing a real-time synchronized digital twin for each delivery box. This digital twin receives microenvironmental data, equipment status data, and geographic location data. Combined with a pre-defined knowledge graph of agricultural product temperature control requirements and a delivery box performance and energy consumption model, it simulates temperature, humidity, and energy consumption trends under different scheduling strategies, including: A corresponding digital twin instance is created in the cloud for each delivery box, and the microenvironment data, the device status data and the geographical location data are synchronized to the digital twin instance in real time, so that the digital twin and the physical delivery box maintain the same status. The digital twin queries a pre-defined knowledge graph of agricultural product temperature control requirements based on the agricultural product category information of this delivery task, and obtains the appropriate temperature and humidity range, chilling injury threshold, ethylene sensitivity and temperature change curve of the corresponding category as the temperature control constraints for simulation. The digital twin invokes the performance and energy consumption model of the delivery box, which is used to calculate the energy consumption and temperature response under different operating conditions. After loading the temperature control constraints and the performance and energy consumption model of the delivery box, the temperature, humidity and energy consumption trends under different scheduling strategies are simulated using the digital twin.

7. The method for remote temperature control strategy scheduling of agricultural product delivery boxes according to claim 6, characterized in that, After loading the temperature control constraints and the performance and energy consumption model of the delivery box, the process of simulating the temperature, humidity, and energy consumption trends under different scheduling strategies using the digital twin includes: By integrating real-time synchronized microenvironmental data, equipment status data, geographical location data, and predicted ambient temperature along the route, a dynamic simulation input field is constructed. Based on the dynamic simulation input field, the temperature control constraints, and the performance and energy consumption model of the delivery box, the digital twin calls the built-in hybrid driving model to generate a variety of candidate scheduling strategies. The hybrid driving model is composed of a physical law model and a machine learning prediction model in parallel. The physical law model is responsible for the basic thermodynamic calculations, and the machine learning model is responsible for compensating for nonlinear deviations and accelerating the simulation speed. For each candidate scheduling strategy, the digital twin adopts a parallel simulation mechanism to simultaneously input the multiple candidate scheduling strategies into multiple hybrid driving model instances, deduce the dynamic change curve of temperature and humidity inside the box and the cumulative energy consumption value within a preset time under each strategy, and output temperature, humidity and energy consumption trend data.

8. The method for remote temperature control strategy scheduling of agricultural product delivery boxes according to claim 1, characterized in that, The microenvironment data, the device status data, and the geographic location data are uploaded to the cloud using an adaptive communication strategy that combines LPWAN and cellular networks. The adaptive communication strategy selects a LoRa, NB-IoT, 4G, or 5G transmission channel based on the current signal strength, data urgency, and pricing strategy.

9. A remote temperature control strategy scheduling device for agricultural product delivery boxes, characterized in that, include: The data acquisition module is used to acquire real-time data on the microenvironment inside the delivery box, equipment status data, and geographical location data through a sensor array deployed on the delivery box; The simulation module is used to build a real-time synchronized digital twin for each delivery box. The digital twin receives the microenvironment data, the equipment status data, and the geographical location data, and combines them with a preset knowledge graph of agricultural product temperature control requirements and a delivery box performance and energy consumption model to simulate the temperature, humidity, and energy consumption trends under different scheduling strategies. The decision instruction generation module is used to construct a multi-agent collaborative decision-making framework. The multi-agent collaborative decision-making framework dynamically adjusts the temperature control setpoint and cooling power of each box based on simulation results. At the same time, according to the type of agricultural products being delivered and the mixed packaging situation, it selectively executes a dynamic temperature change strategy or a multi-category zone collaborative temperature control strategy, and generates decision instructions to be sent to the delivery box for execution. The iterative update module is used to iteratively train and update the digital twin and the multi-agent collaborative decision-making framework based on the actual execution data after each delivery task is completed.

Images