A real-time monitoring system for fresh product cold chain logistics

Abstract

The application discloses a kind of fresh products cold chain logistics real-time monitoring system, it is related to cold chain logistics monitoring technical field, including comprehensive monitoring center, the comprehensive monitoring center communication connection has following module: identification tracking and temperature zone adaptation module, for by giving each batch of fresh products unique object identifier, collection its temperature trajectory data.The application is by allocating unique encrypted identifier for each batch of fresh products, and combining the data of temperature, humidity, position and the like collected by multi-source sensor in real time, constructs the whole-process state trajectory atlas from packaging, transportation to delivery, dynamically calculates the optimal adaptation path of each batch and target temperature zone based on ant colony algorithm, realizes the intelligent scheduling and dynamic allocation of temperature control resources in multi-batch mixed loading scenario, effectively avoids the local temperature drift problem caused by temperature zone conflict or unreasonable resource allocation in traditional cold chain logistics, and improves the stability and resource utilization of cold chain transportation.

Description

Technical Field

[0001] This invention relates to the field of cold chain logistics monitoring technology, specifically a real-time monitoring system for cold chain logistics of fresh products. Background Technology

[0002] As consumers demand higher levels of freshness and safety, and as the food supply chain becomes more complex, ensuring temperature control during the transportation and storage of fresh products has become a critical factor. Cold chain logistics is an effective means of maintaining the quality of fresh products, involving temperature monitoring and control at multiple stages, including harvesting, processing, transportation, and sales.

[0003] For example, a blockchain-based cold chain logistics management system, published in Chinese Patent Publication No. CN112561250A, proposes and implements a method and technology for multi-physical domain data acquisition and data fusion based on distributed multi-sensors, which can more accurately obtain information about fresh food and better monitor the food status.

[0004] In the real-time monitoring of cold chain logistics for fresh products, existing technologies face the following problems in scenarios involving mixed transport of multiple batches and temperature zones: existing technologies struggle to dynamically identify and track the real-time affiliation and temperature zone adaptation status of each batch of fresh produce, leading to unreasonable allocation of temperature control resources and frequent localized temperature drift. Furthermore, it is difficult to predict the cumulative effect of temperature drift caused by hidden factors such as frequent door openings and equipment aging, making the path of fresh produce quality degradation invisible and hindering early intervention and risk avoidance. To address these issues, a real-time monitoring system for cold chain logistics of fresh products is proposed. Summary of the Invention

[0005] To solve the above-mentioned technical problems, the present invention is implemented through the following technical solution: a real-time monitoring system for cold chain logistics of fresh products, comprising a comprehensive monitoring center, wherein the comprehensive monitoring center is communicatively connected to the following modules:

[0006] The identification tracking and temperature zone adaptation module is used to collect temperature trajectory data in real time by assigning a unique object identifier to each batch of fresh products, and to use the ant colony algorithm to dynamically calculate the optimal temperature zone allocation path for the target temperature zone, so as to carry out intelligent scheduling of temperature control resources in multi-batch mixed transportation, thereby improving the temperature zone matching accuracy and resource utilization.

[0007] The cumulative temperature drift prediction module is used to construct a cold chain knowledge graph that integrates equipment status, operational behavior and environmental factors. Combined with reinforcement learning agents to continuously learn historical and real-time data, it can make advance predictions and risk assessments of the cumulative temperature drift effect caused by hidden factors such as frequent door opening and equipment aging.

[0008] The strategy dynamic optimization module, based on the cumulative temperature drift prediction results, combined with real-time environmental conditions and equipment performance, dynamically adjusts the operating parameters and temperature zone allocation strategy of the refrigeration equipment, adaptively suppresses the temperature drift effect, achieves precise temperature control, reduces energy consumption, and extends equipment life.

[0009] The quality degradation visualization module is used to map the cumulative temperature drift effect into a fresh produce quality degradation curve, display the quality change path of each batch of products in real time, provide managers with intuitive decision support, and facilitate timely intervention and scheduling optimization.

[0010] The intelligent risk avoidance module integrates identification and tracking, temperature drift prediction and control decision data, automatically identifies weak links in the system, proposes risk avoidance suggestions and triggers emergency plans, realizing the transformation from passive response to proactive protection, and comprehensively improving the resilience and safety of cold chain logistics.

[0011] Preferably, the identifier tracking and temperature zone adaptation module includes a dynamic identifier tracking unit and a temperature zone intelligent adaptation unit;

[0012] The dynamic identification tracking unit is used to assign a unique object identifier to each batch of fresh products, collect temperature trajectory data including temperature, humidity and location in real time, realize full-process visual tracking and status identification, ensure the independent traceability of each batch of products, provide accurate data foundation for temperature zone adaptation, realize batch-level full life cycle transparent tracking, and improve traceability accuracy and data credibility.

[0013] The temperature zone intelligent adaptation unit analyzes the temperature zone requirements and real-time environment of each batch of fresh products based on the ant colony algorithm, dynamically calculates the optimal temperature zone allocation path, performs adaptive scheduling of temperature control resources, avoids temperature zone conflicts and local temperature drift, improves refrigeration efficiency and transportation stability, dynamically optimizes temperature zone matching, reduces uneven heating and cooling, and improves energy efficiency.

[0014] Preferably, the dynamic identifier tracking unit performs the following steps:

[0015] Assign a unique object identifier based on an encrypted QR code to each batch of fresh produce, bind it to the product packaging, establish a reliable mapping relationship between the identifier and the physical entity, realize independent and identifiable identity at the batch level, ensure that the batch identity cannot be forged or tampered with, and achieve accurate traceability and anti-counterfeiting authentication;

[0016] By collecting temperature, humidity, and location data of products corresponding to each label in real time through multi-source sensors distributed in the transport vehicle, structured temperature trajectory data is formed and continuously uploaded to the comprehensive monitoring center, realizing continuous collection of environmental data throughout the process and providing real-time and complete data support for intelligent analysis.

[0017] Based on the uploaded temperature trajectory data, a batch-level full-process status trajectory map is constructed in the comprehensive monitoring center to realize real-time visual tracking and automatic identification of abnormal states of products from warehousing, transportation, transfer to delivery. This provides an accurate and dynamic data foundation for temperature zone adaptation, enables automatic detection and early warning of abnormal events, and improves the visualization and control efficiency of the transportation process.

[0018] Preferably, the temperature zone intelligent adaptation unit performs the following steps:

[0019] Receive temperature zone demand data and real-time environmental data from each batch of the dynamic identification tracking unit. Based on the preset temperature sensitivity rule library for fresh produce, initialize the pheromone matrix of the ant colony algorithm and the feasible solution space for temperature zone adaptation, provide basic data support for subsequent optimization, and ensure that the allocation scheme meets the category constraints.

[0020] An improved ant colony algorithm is used to simulate the multi-batch temperature zone allocation path optimization process. Through the pheromone accumulation and volatilization mechanism, the load, cooling efficiency and energy consumption of each temperature zone are dynamically evaluated. The current optimal temperature zone allocation scheme is iteratively calculated to achieve rapid convergence to the global optimal allocation, thereby improving the temperature zone matching accuracy and resource utilization.

[0021] Based on the iterative optimization results, dynamic temperature zone scheduling instructions are generated for multi-batch mixed-load scenarios. This controls the zoned refrigeration equipment to perform adaptive temperature control resource allocation, avoids local temperature drift and temperature zone conflicts, improves the thermal management stability and energy efficiency ratio during transportation, significantly suppresses local temperature fluctuations, and ensures temperature control consistency for each batch of products.

[0022] Preferably, the cumulative temperature drift prediction module includes a knowledge graph construction unit and a reinforcement learning prediction unit;

[0023] The knowledge graph construction unit is used to integrate multi-source data such as cold chain equipment status, operation behavior records, and environmental factors to construct a structured cold chain knowledge graph, characterize the relationship and influence path between various factors, provide multi-dimensional relationship support for temperature drift tracing and prediction, construct an interpretable causal relationship network, and support the tracing of hidden factors and influence path analysis.

[0024] The reinforcement learning prediction unit, based on the cold chain knowledge graph and real-time sensor data, trains reinforcement learning agents to identify temperature drift patterns, predicts the cumulative temperature drift trend and quality degradation risk caused by hidden factors such as frequent door opening and equipment aging, realizes the transformation from post-event alarm to pre-event warning, improves risk perception capabilities, achieves advanced temperature drift prediction, and shortens the warning response time.

[0025] Preferably, the knowledge graph construction unit performs the following steps:

[0026] Collect multi-source heterogeneous data from the entire cold chain logistics chain, including refrigeration equipment operation logs, historical maintenance records, door opening frequency, external environmental temperature and humidity fluctuations, and cargo stacking methods. Clean, standardize, and extract entities from the data to achieve unified integration and high-quality preprocessing of the entire cold chain data, providing a reliable data foundation for knowledge graph construction.

[0027] Based on a predefined entity relationship model, the extracted entities and relationships are structured and modeled to construct a cold chain knowledge graph with equipment-operation-environment-goods as the core nodes. The node attributes include state parameters, timestamps and confidence levels, forming a structured knowledge network that can represent the multi-dimensional relationships of the cold chain system and supports complex semantic queries and relationship inference.

[0028] By utilizing graph embedding technology and a rule-based reasoning engine, the relation weights and path connectivity in the cold chain knowledge graph are dynamically updated. This enables the correlation analysis and interpretability representation of the impact of latent factors (equipment aging trends, frequent door opening behavior) on temperature drift, dynamically revealing the mechanism of action of latent factors on temperature drift and improving the interpretability and traceability accuracy of the prediction model.

[0029] Preferably, the reinforcement learning prediction unit performs the following steps:

[0030] Using the cold chain knowledge graph as the environment representation, a state space, action space and reward function are constructed. A temperature drift tendency agent based on deep reinforcement learning is designed to simulate the formation and evolution of accumulated temperature drift, which greatly improves the dynamic coupling ability of temperature drift simulation and decision-making, and realizes the accurate reproduction of complex temperature drift paths.

[0031] By using historical temperature drift event sequences and real-time sensor data, the temperature drift-oriented agent is trained alternately offline and online. This enables the agent to learn to recognize temperature drift patterns and their evolution caused by latent factors such as frequent door opening and equipment performance degradation. This significantly enhances the agent's ability to identify and generalize latent risk factors and improves prediction accuracy.

[0032] The trained temperature drift trend agent combines the current environmental state with the reasoning results of the knowledge graph to predict the cumulative temperature drift trend of a specific batch in the future and the corresponding probability of quality degradation risk, and outputs an impact range assessment, realizing the leap from passive monitoring to active prediction, and providing a quantitative decision-making basis for quality risk management.

[0033] Preferably, the strategy dynamic optimization module performs the following steps:

[0034] The system receives cumulative temperature drift prediction results, environmental monitoring data, and refrigeration equipment operating status in real time. It constructs a multi-objective optimization function with temperature drift suppression, energy consumption minimization, and equipment life extension as the core optimization objectives. Through multi-objective collaborative optimization, it achieves a balance between temperature drift suppression, energy consumption reduction, and equipment life extension, thereby improving the overall energy efficiency of the system.

[0035] The model predictive control algorithm is adopted to dynamically optimize the combination of operating parameters of the refrigeration equipment, including compressor frequency, fan speed and valve opening, based on the current system state and future temperature drift prediction. It also generates staged control commands to achieve early intervention through predictive control, reduce temperature drift fluctuations, and improve temperature control accuracy and response speed.

[0036] The optimized control commands are sent to the temperature control execution units of each zone to perform dynamic redistribution of temperature zones and adaptive adjustment of refrigeration power, forming a closed-loop control that continuously suppresses temperature drift and maintains stable fresh produce quality. Through closed-loop adaptive adjustment, dynamic matching of temperature zone resources is achieved, which stably maintains fresh produce quality and reduces the risk of spoilage.

[0037] Preferably, the quality degradation visualization module performs the following steps:

[0038] Based on the temperature drift trend data output by the cumulative temperature drift prediction module, combined with the shelf life dynamics model corresponding to fresh food categories, the temperature drift effect is quantified into the cumulative quality degradation amount, generating the real-time quality degradation index of each batch of products, realizing the transformation of quality change from qualitative description to quantitative assessment, and providing numerical basis for precise control.

[0039] The quality degradation index is mapped to a time-quality change curve using a visualization engine, and then displayed in real time on the interactive interface of the integrated monitoring center in the form of a multi-batch overlay heat map. This intuitively presents the real-time evolution of the quality of each batch, improves situational awareness efficiency, and facilitates rapid anomaly location.

[0040] It supports multi-dimensional filtering and drill-down viewing by batch, time interval, and temperature zone, and provides a comparative analysis view of quality change trends. It helps managers to intuitively grasp the quality evolution status, identify high-risk batches and processes, and achieve flexible insight from macro overview to micro details, assisting in efficient decision-making and accurate risk positioning.

[0041] Preferably, the intelligent risk avoidance module performs the following steps:

[0042] By integrating identification and tracking data, temperature drift prediction results, and control strategy execution feedback, a multi-dimensional risk assessment matrix for the cold chain system is constructed. This matrix identifies weak links and potential failure links in the system at four levels: equipment, operation, environment, and goods. This enables comprehensive visualization and identification of system risks, improving the accuracy of problem location and traceability efficiency.

[0043] Based on the risk assessment matrix and preset rule base, it automatically generates targeted risk avoidance suggestions, including equipment pre-maintenance prompts, operating procedure adjustment suggestions and transportation route optimization plans, and triggers early warning notifications to achieve intelligent early warning and proactive suggestion push, assisting managers to respond quickly and reduce operational risks;

[0044] When system anomalies persist without resolution, the pre-set emergency response plan is automatically activated, including dynamically adjusting transportation plans, switching to backup refrigeration equipment, and notifying nearby maintenance personnel to intervene. This improves system resilience from passive response to proactive protection, automates and closes-loops emergency response management, and significantly enhances the system's ability to withstand risks and ensures transportation continuity.

[0045] This invention provides a real-time monitoring system for cold chain logistics of fresh products. It has the following beneficial effects:

[0046] (I) This real-time monitoring system for cold chain logistics of fresh products assigns a unique encrypted identifier to each batch of fresh products and collects data such as temperature, humidity and location in real time by multiple source sensors. It constructs a full-process status trajectory map from packaging and transportation to delivery. Based on the ant colony algorithm, it dynamically calculates the optimal matching path between each batch and the target temperature zone, realizes intelligent scheduling and dynamic allocation of temperature control resources in multi-batch mixed loading scenarios, effectively avoids the local temperature drift problem caused by temperature zone conflict or unreasonable resource allocation in traditional cold chain logistics, and improves the stability and resource utilization of cold chain transportation.

[0047] (II) The real-time monitoring system for cold chain logistics of fresh products adopts a model predictive control algorithm to perform synergistic optimization among multiple objectives such as temperature drift suppression, energy consumption minimization, and equipment life extension. Through rolling optimization and real-time feedback, a closed-loop control is formed, which can dynamically adjust refrigeration parameters according to real-time environmental conditions and equipment performance, and achieve precise on-demand allocation of temperature control resources. While ensuring the stability of fresh product quality, it significantly reduces system energy consumption and extends the service life of key equipment.

[0048] (III) The real-time monitoring system for cold chain logistics of fresh products adopts a model predictive control algorithm to perform collaborative optimization among multiple objectives such as temperature drift suppression, energy consumption minimization and equipment life extension. Through rolling optimization and real-time feedback, a closed-loop control is formed, which can dynamically adjust refrigeration parameters according to real-time environmental conditions and equipment performance, realize precise on-demand allocation of temperature control resources, and significantly reduce system energy consumption and extend the service life of key equipment while ensuring the stability of fresh product quality.

[0049] (iv) The real-time monitoring system for cold chain logistics of fresh products quantifies temperature drift data into a quality degradation index by constructing a quality degradation dynamic model based on the Arrhenius equation, and displays it in real time in the form of heat map, curve comparison, etc., to provide managers with intuitive decision support and facilitate timely intervention and scheduling optimization. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the workflow of a real-time monitoring system for cold chain logistics of fresh products according to the present invention.

[0051] Figure 2 This is a data flow diagram of a real-time monitoring system for cold chain logistics of fresh products according to the present invention. Detailed Implementation

[0052] 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.

[0053] Example 1, please refer to Figure 1 , Figure 2 This invention provides a technical solution: a real-time monitoring system for cold chain logistics of fresh products, including a comprehensive monitoring center, which is connected to the following modules for communication:

[0054] The identification tracking and temperature zone adaptation module is used to assign a unique object identifier to each batch of fresh products, collect their temperature trajectory data in real time, and use the ant colony algorithm to dynamically calculate the optimal temperature zone allocation path for the target temperature zone, so as to carry out intelligent scheduling of temperature control resources in multi-batch mixed transportation, improve temperature zone matching accuracy and resource utilization. The identification tracking and temperature zone adaptation module includes a dynamic identification tracking unit and a temperature zone intelligent adaptation unit.

[0055] The dynamic identification and tracking unit assigns a unique object identifier to each batch of fresh produce, collects real-time temperature trajectory data including temperature, humidity, and location, enabling full-process visual tracking and status identification. This ensures the independent traceability of each batch, provides a precise data foundation for temperature zone adaptation, achieves transparent tracking throughout the entire batch lifecycle, and improves traceability accuracy and data reliability. It assigns a unique object identifier based on an encrypted QR code to each batch of fresh produce, binding it to the product packaging and establishing a reliable mapping relationship between the identifier and the physical entity. This enables independent identification of batch-level identities, ensuring that batch identities cannot be forged or tampered with, achieving accurate traceability and anti-counterfeiting. The system uses multi-source sensors distributed within the transport vehicle to collect real-time temperature, humidity, and location data for each product corresponding to its identification mark. This data forms structured temperature trajectory data and is continuously uploaded to the integrated monitoring center. This enables continuous collection of environmental data throughout the entire process, providing real-time and complete data support for intelligent analysis. Based on the uploaded temperature trajectory data, a batch-level full-process status trajectory map is constructed at the integrated monitoring center. This allows for real-time visual tracking and automatic identification of abnormal states at each stage of the product's journey from warehousing, transportation, and transshipment to delivery. It provides a precise and dynamic data foundation for temperature zone adaptation, enables automatic detection and early warning of abnormal events, and improves the visualization and control efficiency of the transportation process.

[0056] The specific work involves the following steps: Before the system is activated, each batch of fresh produce is assigned a unique object identifier based on the international encryption standard (compliant with AES-256) during the packaging process. This identifier is printed as an encrypted QR code (using the DM code format, supporting fault-tolerant encoding of no less than 300 bytes) on a tear-resistant label that meets waterproof and oil-resistant industrial standards. A professional labeling device with anti-transfer capabilities then firmly affixes this identifier to the outside of the product packaging. The identifier data structure follows the ISO / IEC 15459:2015 standard, including batch number, category code, timestamp, and initial storage condition fields. A signature verification mechanism ensures the immutability of the identifier during generation and binding. Using the entire outer packaging unit of the batch as the smallest tracking granularity, each identifier establishes a one-to-one reliable mapping with all products in that batch, completing the basic initialization work for independent and identifiable identification. During transportation, temperature and humidity sensor nodes and multi-mode positioning modules (supporting GPS / BeiDou dual systems, with positioning accuracy better than 3 meters) integrated into different spatial zones within the vehicle (refrigerated truck / container) collect data on each encrypted QR code identifier in real time. For the corresponding environmental data, the sensor nodes use a low-power wireless transmission protocol to encapsulate the collected data along with the corresponding QR code scanning information (scanning time, operator ID) into a structured message, which is then uploaded to the integrated monitoring center via the vehicle gateway. The message structure includes fields such as data source identifier, collection time, sensor value, location coordinates, and status code, and is serialized in JSON format to ensure data integrity and parsability. At least one full vehicle scan and data synchronization is performed every minute to form continuous and structured three-dimensional temperature trajectory data of temperature, humidity, and location. After receiving the uploaded temperature trajectory data, the integrated monitoring center performs data cleaning and time alignment processing, removes outliers and duplicate records, and clusters them according to batch identifiers. Based on the time axis, combined with geographical location information and operation node records, a batch-level full-process status trajectory map is automatically constructed. This map uses nodes to represent spatiotemporal state points, edges to represent the transition relationship between states, and marks abnormal events that meet preset rules. The map displays the temperature change curves and geographical paths of each batch in the entire process of warehousing, transportation, transfer, and delivery in real time through a visual interface.

[0057] The intelligent temperature zone adaptation unit analyzes the temperature zone requirements and real-time environment of each batch of fresh produce based on the ant colony algorithm, dynamically calculates the optimal temperature zone allocation path, and performs adaptive scheduling of temperature control resources to avoid temperature zone conflicts and local temperature drift, thereby improving refrigeration efficiency and transportation stability. It dynamically optimizes temperature zone matching, reduces uneven heating and cooling, and improves energy efficiency. It receives temperature zone requirement data and real-time environmental data from the dynamic identification tracking unit for each batch, and initializes the pheromone matrix and feasible solution space for temperature zone adaptation based on a pre-set temperature sensitivity rule library for fresh produce categories. This provides foundational data support for subsequent optimization and ensures that the allocation scheme meets category constraints. An improved ant colony algorithm is used to simulate the multi-batch temperature zone allocation path optimization process. Through the pheromone accumulation and volatilization mechanism, the load, cooling efficiency and energy consumption of each temperature zone are dynamically evaluated. The current optimal temperature zone allocation scheme is calculated iteratively to achieve rapid convergence to the global optimal allocation, thereby improving the temperature zone matching accuracy and resource utilization. Based on the iterative optimization results, dynamic temperature zone scheduling instructions are generated for multi-batch mixed loading scenarios. This controls the zoned refrigeration equipment to perform adaptive temperature control resource allocation, avoids local temperature drift and temperature zone conflicts, improves the thermal management stability and energy efficiency ratio during transportation, significantly suppresses local temperature fluctuations, and ensures the temperature control consistency of each batch of products.

[0058] The specific work involves: during the system initialization phase, receiving batch temperature zone requirement data and real-time environmental data from the dynamic identification tracking unit. The temperature zone requirement data includes the category code of each batch of fresh produce and its preset temperature control target range: 0℃ to 4℃ for refrigerated products, -18℃ to -22℃ for frozen products, and 4℃ to 8℃ for fresh products. The real-time environmental data includes the current measured temperature, humidity, and refrigeration equipment operating status of each zone within the vehicle. The input data is preprocessed based on a preset fresh produce category temperature sensitivity rule base, which is stored in a structured format. Constraints were established regarding the permissible temperature fluctuation threshold, maximum exposure time, and cross-contamination sensitivity for products of the same category during transportation. Based on these constraints, the pheromone matrix for the ant colony algorithm was initialized, with the matrix dimensions corresponding to the number of allocable temperature zones and the number of batches to be scheduled. The initial value of each pheromone was set to 0.01. Simultaneously, based on the actual partition layout of the current vehicle, the volume of each partition, and the cooling power limitations, a feasible solution space for temperature zone adaptation was constructed. Each solution represents a possible batch-temperature zone allocation combination. Preliminary screening was performed based on the hard constraints in the rule base, eliminating invalid allocations that did not meet temperature control requirements or had physical space conflicts. In the algorithm iteration and optimization phase, an improved ant colony algorithm is used to simulate the optimization process of multiple batches of temperature zone allocation paths. Each artificial ant represents a complete allocation scheme construction process. Ants allocate batches to a certain temperature zone based on pheromone concentration and heuristic information according to probability. The heuristic information comprehensively considers the real-time load rate of the current temperature zone, the deviation between the target temperature and the measured temperature, and the historical adaptation effect of the species in this temperature zone. In its calculation, the weight of temperature deviation is set to 0.5, the weight of load rate is set to 0.3, and the weight of historical adaptation is set to 0.2. The pheromone update mechanism includes two parts: accumulation and volatilization. After each iteration, for paths traversed by solutions superior to the current global optimum, pheromones are added by the reciprocal of the solution's overall score (with a coefficient of 0.1). Simultaneously, at the end of each iteration, pheromones from all paths are globally evaporated with a volatility coefficient of 0.7. The core indicators for dynamic evaluation by the algorithm include the load balance of each temperature zone (measured by the load variance of each zone, with a target of less than 15%), the overall energy efficiency ratio of the cooling system (with a target of more than 2.8), and the estimated overall energy consumption. Iterative calculations are performed with the goal of maximizing the overall score. In the overall score function, the weights for load balance and energy efficiency are 0.4 and 0 respectively.2. The maximum number of iterations is set to 200. The process terminates early if the optimal solution is not improved after 20 consecutive iterations. Based on the optimal temperature zone allocation scheme obtained through iterative optimization, dynamic temperature zone scheduling instructions are generated. These instructions are a structured control command set, explicitly specifying the target temperature setpoint, fan operating mode, and valve opening for each zone's refrigeration equipment. The instructions are sent to the temperature control execution units of each zone via the vehicle control network. The system monitors the environmental feedback after the instructions are executed in real time and compares it with the predicted state. If the actual monitoring data deviates from the expected value beyond the allowable range, a re-optimization process is triggered, generating adjustment instructions. Through closed-loop adaptive control, precise on-demand allocation of temperature control resources is achieved, effectively suppressing local temperature drift caused by uneven heating or insufficient cooling, avoiding airflow interference and temperature conflicts between different temperature zones, and improving the overall thermal management stability and system energy efficiency ratio during transportation while ensuring the temperature zone adaptation accuracy of each batch of products.

[0059] The cumulative temperature drift prediction module is used to construct a cold chain knowledge graph that integrates equipment status, operational behavior and environmental factors. Combined with the continuous learning history and real-time data of the reinforcement learning agent, it can make advance predictions and risk assessments of the cumulative temperature drift effect caused by the implicit factors of frequent door opening and equipment aging. The cumulative temperature drift prediction module includes a knowledge graph construction unit and a reinforcement learning prediction unit.

[0060] The knowledge graph construction unit integrates multi-source data on cold chain equipment status, operational records, and environmental factors to build a structured cold chain knowledge graph. This graph represents the relationships and influence paths between various factors, providing multi-dimensional relationship support for temperature drift tracing and prediction. It also constructs an interpretable causal relationship network to support the tracing of latent factors and analysis of influence paths. Furthermore, it collects multi-source heterogeneous data from the entire cold chain logistics process, including refrigeration equipment operation logs, historical maintenance records, door opening frequency, external environmental temperature and humidity fluctuations, and cargo stacking methods. The data is then cleaned, standardized, and entity extracted, achieving unified integration and high-quality preprocessing of data across the entire cold chain, providing a reliable data foundation for knowledge graph construction. Based on a predefined entity relationship model, the extracted entities and relationships are structurally modeled to construct a cold chain knowledge graph with equipment-operation-environment-goods as core nodes. The node attributes include state parameters, timestamps, and confidence levels, forming a structured knowledge network that can represent the multi-dimensional relationships of the cold chain system. It supports complex semantic queries and relationship inference. Using graph embedding technology and a rule reasoning engine, the relationship weights and path connectivity in the cold chain knowledge graph are dynamically updated. This enables the correlation analysis and interpretability representation of the impact of latent factors (equipment aging trends, frequent door opening behavior) on temperature drift, dynamically revealing the mechanism of action of latent factors on temperature drift, and improving the interpretability and traceability accuracy of the prediction model.

[0061] The specific work content includes: During the system initialization phase, collecting and preprocessing multi-source heterogeneous data involved in the entire cold chain logistics process. This includes collecting refrigeration equipment operation logs from the control units of various refrigeration equipment models, with a data collection interval of 10 seconds. Log fields include compressor operating current, condenser temperature, evaporator temperature, fan speed, operating mode, and fault codes. The data format follows the JSON standard. Historical maintenance records are extracted from the enterprise maintenance management system, with fields including maintenance time, maintenance type, replaced part model, maintenance personnel, and post-maintenance performance verification data. The frequency of door opening operations is collected using magnetic induction sensors installed on the refrigerator doors, recording the opening time, duration, and door number for each opening. This is then implemented through deployment... Temperature and humidity sensors on the exterior of the vehicle collect ambient temperature and humidity data at a sampling frequency of once per minute. The stacking pattern of the goods is manually entered, recording stacking height, spacing between goods, and ventilation channel width. The collected raw data is first cleaned to remove outliers significantly exceeding physical limits. Missing data is filled using time-series interpolation. All timestamps are uniformly converted to UTC time and accurate to the second. Data standardization uses Z-score normalization to standardize continuous sensor data. In the entity extraction stage, a BERT-based named entity recognition model is used to extract entities such as equipment number, operation type, environmental indicators, and goods batch from the text data. Based on the preprocessed data... Based on multi-source heterogeneous data, a cold chain knowledge graph is constructed with equipment, operations, environment, and goods as core entities. Entity types are predefined into four categories: equipment entities, operation entities, environment entities, and goods entities. Equipment entity attributes include equipment ID, equipment model, installation location, rated power, current operating status, and status confidence (calculated based on sensor data and equipment specifications). Operation entity attributes include operation ID, operation type, operation time, and operation object. Environment entity attributes include environment ID, temperature and humidity values, collection time, and geographic coordinates. Goods entity attributes include batch ID, category code, temperature control requirements, stacking parameters, and current quality status. Relationships between entities are structured according to a predefined entity relationship model. The modeling process includes relationship types such as device-experience-operation, operation-occurrence-environment, environment-impact-goods, and goods-stored-device. Relationship attributes include relationship strength and the effective time range of the relationship. The knowledge graph is stored using the graph database Neo4j, and nodes and relationships are imported in batches using Cypher statements. Graph embedding technology and a rule-based reasoning engine are integrated to dynamically update the knowledge graph and perform relationship reasoning, enabling the correlation analysis and interpretable representation of the influence of latent factors on temperature drift. Specifically, the graph embedding technology of the TransE model is used to map entities and relationships to a 100-dimensional vector space. The distributed representation of entities and relationships is learned by minimizing the loss function, with a training cycle of 100 rounds and a learning rate of 0.001. Every 24 hours, the embedded vector is incrementally updated based on the latest collected data to reflect dynamic information such as equipment performance degradation and changes in operating modes. The rule reasoning engine, using the Drools rule engine, has a built-in rule base containing over 200 business rules. Rule reasoning is triggered every 10 minutes. Based on the latest data and graph embedding results, the relation weights and path connectivity in the graph are dynamically calculated and updated. All update operations are recorded in audit logs to ensure the traceability of graph evolution. Finally, a multi-hop association path starting from equipment aging trends and frequent door opening behavior is output, and the contribution and propagation path of each factor to temperature drift are visualized, providing interpretable structured knowledge support for temperature drift tracing and prediction.

[0062] The reinforcement learning prediction unit, based on the cold chain knowledge graph and real-time sensor data, trains a reinforcement learning agent to identify temperature drift patterns. It predicts the cumulative temperature drift trend and quality degradation risk caused by latent factors such as frequent door opening and equipment aging, shifting from post-event alarms to pre-event warnings. This enhances risk perception capabilities, enables advanced temperature drift prediction, and shortens warning response time. Using the cold chain knowledge graph as the environment representation, it constructs a state space, action space, and reward function. A deep reinforcement learning-based temperature drift tendency agent is designed to simulate the formation and evolution of cumulative temperature drift, significantly improving the dynamic coupling capability between temperature drift simulation and decision-making, and enabling the simulation of complex temperature drift paths. The system accurately reproduces temperature drift events by using historical temperature drift event sequences and real-time sensor data to train the temperature drift trend agent alternately offline and online. This enables the agent to learn and identify temperature drift patterns and their evolution caused by latent factors such as frequent door opening and equipment performance degradation. This significantly enhances the agent's ability to identify and generalize latent risk factors, improving prediction accuracy. The trained temperature drift trend agent, combined with the current environmental state and knowledge graph reasoning results, predicts the cumulative temperature drift trend of a specific batch and the corresponding probability of quality degradation risk in the future, and outputs an impact range assessment. This achieves a leap from passive monitoring to proactive prediction, providing a quantitative decision-making basis for quality risk management.

[0063] The specific work content is as follows: During the system deployment phase, a reinforcement learning environment is established based on the constructed cold chain knowledge graph to build a temperature drift-oriented intelligent agent. The environmental state space consists of multi-dimensional feature vectors, including real-time sensor readings, equipment operating status, operation event markers, and graph embedding vectors. The dimension is set to 128 dimensions. Sensor data is sampled every 10 seconds and retained to 4 decimal places after Z-score normalization. The action space is defined as a discrete control instruction set, including compressor frequency adjustment +5%, compressor frequency adjustment -5%, condenser fan speed increase +10%, condenser fan speed decrease -10%, evaporator fan speed increase +10%, evaporator fan speed decrease -10%, solenoid valve opening increase +15%, solenoid valve opening decrease -15%, start the backup condenser unit (activated when the main condenser efficiency is lower than the set threshold), switch to energy-saving operation mode (compressor frequency limited to 70%, fan speed synchronously reduced to 80%), and start the defrosting program (forced defrosting). The training program includes 12 executable operations, such as frost control for 8 minutes (applicable when the fin frosting coefficient is >0.4) and enabling zone damper adjustment (increasing the opening of the specified zone damper by 20%). Each operation corresponds to a specific adjustment amount of the equipment control parameters. The reward function adopts a piecewise weighted design: when the temperature drift amplitude remains within the threshold, a base reward of +0.1 is given; for each temperature drift exceeding the limit event, a reward of -1.0 is deducted. At the same time, an energy efficiency penalty is introduced: for every 5% increase in cooling power consumption beyond the baseline value, an additional penalty of -0.3 is added. The training cycle is 24 hours as a complete round, and each round contains 8640 decision steps. The temperature drift tendency agent training adopts a two-stage mode of alternating offline pre-training and online fine-tuning. In the offline training stage, 3000 sets of historical temperature drift event sequences recorded in the past 90 days are used as the training set. Each set of sequences contains 72 consecutive hours of environmental state change records. The training adopts a deep Q network architecture, containing 3 fully connected layers with 256, 128, and 64 nodes respectively, and the learning rate is initially set to 0.The value decays to 90% of its original value every 10,000 steps. The experience replay buffer capacity is set to 100,000 entries. The online training phase starts every 6 hours, collecting sensor data and knowledge graph inference results from the most recent 4 hours as incremental training samples. The training batch size is fixed at 32, and each online training session lasts no more than 15 minutes. The training termination condition is set as the reward function fluctuation amplitude being less than 5% within 20 consecutive training cycles or the total training steps reaching 500,000. The trained temperature drift tendency agent is deployed on edge computing nodes and executes a prediction process every 5 minutes. The prediction input is the current environmental state vector and data provided by the cold chain knowledge graph, including equipment aging index, door opening frequency correlation, refrigeration efficiency decay rate, condenser dust accumulation coefficient, refrigerant pressure deviation, airflow organization uniformity index, cargo heat load density, and insulation layer performance degradation. The system utilizes 12 derived features, including the reduction coefficient, cumulative temperature surge from frequent door openings, abnormal equipment vibration index, environmental humidity intrusion risk value, and similarity to historical temperature drift events. The temperature drift trend agent generates predicted temperature drift values ​​every 10 minutes for the next 60 minutes via forward propagation, outputting a 6-bit floating-point array. Simultaneously, a quality degradation dynamics model based on the Arrhenius equation converts the predicted temperature drift sequence into a quality degradation risk probability. The activation energy parameter is set to 65 kJ / mol (refrigerated goods), 80 kJ / mol (frozen goods), and 55 kJ / mol (fresh goods) according to product category. Finally, a structured early warning report is output, including the predicted temperature drift curve, quality degradation risk probability, a list of affected batches, and a priority ranking of recommended intervention measures. All data is transmitted to the integrated monitoring center in JSON format via HTTPS.

[0064] The expression for calculating the probability of quality degradation risk is as follows:

[0065] ;

[0066] ;

[0067] In the formula: This indicates the probability of quality degradation; the higher the value, the higher the risk of quality degradation. Indicates the total predicted duration; Indicates time The constant of the rate of quality degradation at that time; This represents the pre-exponential factor, reflecting the frequency of the reaction, and is set as a fixed constant based on category experience. It represents the activation energy, reflecting the energy barrier of the quality degradation reaction; Represents the gas constant; This represents the baseline absolute temperature, taken as the absolute temperature value corresponding to the standard storage temperature for the product category. (Refrigerated goods:) (Approximately 4°C), Frozen products: (Approximately -18°C), Fresh Product: (Approximately 8°C); Indicates time The temperature drift value is taken from the temperature deviation value at the corresponding time point in the temperature drift prediction sequence. If the temperature drift is negative (temperature is lower than the baseline), then... It is a negative value; This represents the integral of the quality degradation rate over the forecast period, reflecting the cumulative degree of degradation.

[0068] The strategy dynamic optimization module, based on the cumulative temperature drift prediction results and combined with the real-time environmental conditions and equipment performance, dynamically adjusts the operating parameters of the refrigeration equipment and the temperature zone allocation strategy, adaptively suppresses the temperature drift effect, achieves precise temperature control, reduces energy consumption, extends equipment life, and at the same time ensures the stability of fresh food quality. The closed-loop control reduces the temperature drift amplitude and overall energy consumption.

[0069] The quality degradation visualization module is used to map the cumulative temperature drift effect into a fresh produce quality degradation curve, display the quality change path of each batch of products in real time, provide managers with intuitive decision support, facilitate timely intervention and scheduling optimization, visualize the quality degradation path in real time, and help quickly identify high-risk batches.

[0070] The intelligent risk avoidance module integrates identification tracking, temperature drift prediction, and control decision data. It automatically identifies system weaknesses, proposes risk avoidance suggestions, and triggers emergency plans, realizing a shift from passive response to proactive protection. This comprehensively enhances the resilience and safety of cold chain logistics, automatically triggers emergency plans, and shortens system interruption recovery time.

[0071] Example 2, as Figure 1 , Figure 2 As shown, based on Embodiment 1, the present invention provides a technical solution: the strategy dynamic optimization module performs the following steps: receiving the cumulative temperature drift prediction results, environmental monitoring data and the operating status of the refrigeration equipment in real time, constructing a multi-objective optimization function with temperature drift suppression, minimum energy consumption and extended equipment life as the core optimization objectives, achieving a balance between temperature drift suppression, energy consumption reduction and extended equipment life through multi-objective collaborative optimization, improving the overall energy efficiency of the system, adopting a model predictive control algorithm, dynamically optimizing the combination of operating parameters of the refrigeration equipment based on the current system state and future temperature drift prediction, including compressor frequency, fan speed and valve opening, and generating phased control instructions, achieving early intervention through predictive control, reducing temperature drift fluctuations, improving temperature control accuracy and response speed, and sending the optimized control instructions to the temperature control execution units of each zone for dynamic redistribution of temperature zones and adaptive adjustment of refrigeration power, forming a closed-loop control, continuously suppressing the temperature drift effect, maintaining the stability of fresh produce quality, and achieving dynamic matching of temperature zone resources through closed-loop adaptive adjustment, stably maintaining the quality of fresh produce and reducing the risk of loss;

[0072] The specific work involves: during system operation, continuously acquiring the 60-minute temperature drift sequence from the cumulative temperature drift prediction module, collected internal and external temperature and humidity data of the vehicle, and reported real-time operating status. The data is updated once per second and transmitted in JSON format. Based on these inputs, a multi-objective optimization function with three sub-objectives is constructed, encompassing a temperature drift suppression objective function, a minimum energy consumption objective function, and an equipment lifespan extension objective function. The temperature drift suppression objective function... Defined as the weighted sum of the absolute values ​​of temperature drift in each zone during the prediction period, with weighting coefficients set according to the temperature sensitivity category of each batch of goods: 1.0 for refrigerated goods, 1.2 for frozen goods, and 0.8 for fresh goods; the objective function is to minimize energy consumption. Defined as the instantaneous power integral sum of the compressor, condenser fan, evaporator fan, and auxiliary systems, with its reference power value determined based on the equipment model nameplate parameters; the objective function for extending equipment lifespan. Defined as the reciprocal of the cumulative fatigue damage function of key components, this damage model is based on Miner's linear cumulative damage theory and integrates SN curve data provided by equipment manufacturers. The three sub-objectives are integrated into the overall objective function through a weighted summation. Weighting coefficient , and The values ​​are set to 0.5, 0.3, and 0.2 respectively, and can be fine-tuned in the management interface according to the operation strategy, with an adjustment step of 0.05.

[0073] The expression for calculating the overall objective function is as follows:

[0074] ;

[0075] ;

[0076] ;

[0077] ;

[0078] In the formula: This represents the overall objective function value, used to comprehensively evaluate the overall performance of the control system. The smaller the value, the better the overall performance of the system. The weighting coefficients for the temperature drift suppression target; The objective function for suppressing temperature drift reflects the system's ability to control temperature drift. This indicates the total number of temperature zones, i.e., the number of independent temperature control areas divided in the system; This represents the total number of steps in the prediction time domain, corresponding to the number of control cycles within the next 30 minutes, with a step size of 10 seconds, totaling 180 steps; Indicates the first The temperature zone in the first The temperature drift value of each predicted step size reflects the degree to which the temperature in that temperature zone deviates from the set value. Indicates the first The temperature sensitivity weighting coefficient for goods stored in each temperature zone is set according to the category of goods: 1.0 for refrigerated goods, 1.2 for frozen goods, and 0.8 for fresh goods; The weighting coefficients for the goal of minimizing energy consumption; The objective function is to minimize energy consumption, reflecting the energy consumption of the system. The smaller the value, the lower the total energy consumption of the system in the prediction time domain. This represents the current moment, which serves as the starting point for integration. It represents the instantaneous power of the compressor, which is proportional to the cube of the compressor's operating frequency; This represents the instantaneous power of the condenser fan, which is proportional to the cube of the fan speed. This represents the instantaneous power of the evaporator fan, which is proportional to the cube of the fan speed. This indicates the instantaneous power of the auxiliary system, including the energy consumption of auxiliary equipment such as control systems and sensors; The time element is represented by the integral, which represents the accumulation of the total power over the prediction time domain. Weighting coefficients for the goal of extending equipment lifespan; The objective function for extending equipment lifespan is the reciprocal of the cumulative damage to the equipment, used to extend the equipment's service life. The larger the value, the smaller the cumulative damage to the equipment and the longer its lifespan. This represents the cumulative fatigue damage value of critical components (such as compressors, fan bearings, etc.). The damage was calculated using Miner's linear cumulative damage theory. These represent different stress levels, corresponding to different load states during equipment operation. Indicates at stress level The actual number of cycles performed by the device; Indicates at stress level The number of cycles required for the equipment to reach fatigue failure is derived from the SN curve (stress-life curve) provided by the equipment manufacturer.

[0079] A rolling optimization algorithm using model predictive control (MMC) is employed, with the prediction time domain set to the next 30 minutes and the control time domain to the next 10 minutes, and a sampling period of 10 seconds. At the beginning of each control cycle, based on the current system state (including measured temperatures in each zone, equipment operating frequencies, and valve openings) and the predicted future temperature drift sequence, the optimal control sequence is solved under constraints, with the objective of minimizing the overall objective function. These constraints include: compressor frequency adjustment range of 30% to 105% of its rated frequency, with an adjustment resolution of 1 Hz; condenser and evaporator fan speed adjustment range of 40% to 100% of their rated speed, with an adjustment resolution of 10 rpm; and solenoid valve opening adjustment range of 0% (fully closed) to 100% (fully open), with an adjustment resolution of 1%. The optimization solution uses the interior-point method and is implemented in an embedded optimization solver, with each solution time limited to within 200 milliseconds. After the solution is completed, the control sequence for the next 10 minutes is output, which is decomposed into 30-second intervals. The system generates phased control commands; these commands are sent to each zone's temperature control execution unit via an onboard control network conforming to the industrial Ethernet standard. Each command packet contains the target device address, command type, target parameter value, and execution timestamp. After receiving the command, the zone's temperature control execution unit drives the inverter to adjust the compressor frequency, controls the servo motor to adjust the fan speed, and adjusts the valve opening through the proportional valve controller to achieve adaptive and precise adjustment of the cooling power of the temperature zone. Simultaneously, it monitors the environmental feedback after the command is executed, comparing the actual temperature change with the MPC prediction value. If the deviation between the average actual temperature drift and the predicted value exceeds the preset threshold for three consecutive monitoring cycles (30 seconds), a re-optimization process is immediately triggered to generate a new adjustment command. Through the closed-loop control mechanism of monitoring-prediction-optimization-execution-feedback, the system continuously suppresses the temperature drift effect caused by load changes, door opening disturbances, or equipment performance fluctuations, and stably maintains the specified quality status of each batch of fresh products in a dynamic transportation environment.

[0080] The quality degradation visualization module performs the following steps: Based on the temperature drift trend data output by the cumulative temperature drift prediction module, combined with the shelf life dynamics model corresponding to fresh produce, the temperature drift effect is quantified into cumulative quality degradation, generating a real-time quality degradation index for each batch of products. This transforms quality changes from qualitative description to quantitative assessment, providing numerical basis for precise control. The visualization engine maps the quality degradation index to a time-quality change curve, and displays it in real-time on the interactive interface of the comprehensive monitoring center in the form of a multi-batch overlay heat map. This intuitively presents the real-time evolution of the quality of each batch, improving situational awareness efficiency and facilitating rapid anomaly location. It supports multi-dimensional filtering and drill-down viewing by batch, time interval, and temperature zone, providing a comparative analysis view of quality change trends. This helps managers intuitively grasp the quality evolution status, identify high-risk batches and processes, and achieve flexible insight from macro-level overview to micro-level details, assisting in efficient decision-making and precise risk location.

[0081] The specific work involves: calculating the quality degradation index using a shelf-life kinetic model constructed based on the Arrhenius equation; quantifying the predicted temperature drift data into the reaction rate of microbial growth and enzyme activity changes; receiving the temperature drift sequence for the next 60 minutes output by the cumulative temperature drift prediction module. This sequence is a floating-point array of length 6, corresponding to the predicted temperature drift value at 10-minute intervals. During calculation, the corresponding kinetic parameters are called according to the category code in the batch information: the activation energy for refrigerated products is set to 65 kJ / mol, and the pre-exponential factor is 1.2 × 10⁻⁶. 12 min -1 The activation energy of the frozen product is 80 kJ / mol, and the pre-exponential factor is 5.6 × 10⁻⁶. 14 min -1 The activation energy of the fresh product is 55 kJ / mol, and the pre-exponential factor is 3.8 × 10⁻⁶. 10 min -1 Using the standard storage temperature for each product category (refrigerated 4℃, frozen -18℃, fresh 8℃) as the baseline absolute temperature, and combined with the real-time temperature drift value, the quality degradation rate constant k(t) is calculated minute by minute. Then, the cumulative quality degradation amount is calculated by numerical integration of k(t) within the predicted period (default 60 minutes). Simpson's integral rule is adopted with an integration step size of 10 seconds. Then, the quality degradation index in the range of 0-100 is calculated by combining the cumulative quality degradation amount and the cumulative quality degradation threshold, where 0 represents good quality and 100 represents the end of the shelf life. This index is updated every 5 minutes and associated with the unique identifier of the corresponding batch.

[0082] The formula for calculating the quality degradation index is as follows:

[0083] ;

[0084] ;

[0085] In the formula: Indicates the quality decline index; To accumulate the amount of quality degradation; This indicates the cumulative quality degradation threshold corresponding to the end of the shelf life. This threshold is experimentally calibrated and represents the cumulative degradation from the time the product leaves the factory until its quality becomes unacceptable. Total value; This indicates the proportion of the current cumulative decay to the end-of-shelf-life threshold; This indicates that the proportion has been converted to a percentage form; This represents a minimum value function that ensures the exponent does not exceed 100; that is, when the cumulative decline exceeds the threshold, the exponent is fixed at 100. The total length of the predicted period; Represents the constant of instantaneous quality degradation rate From time arrive Perform definite integrals;

[0086] The generated quality degradation index is pushed in real time to the visualization engine of the integrated monitoring center via the WebSocket protocol. The engine uses the ECharts framework to draw time-quality change curves on an interactive dashboard. The curves for each batch are presented as independent layers, and the overall display is in the form of a heatmap. The color gradient from green through yellow to red visually maps the level of the quality degradation index. The horizontal axis is the UTC time axis, with a configurable range, defaulting to the most recent 4 hours; the vertical axis is the quality degradation index. Managers can perform multi-dimensional filtering through graphical controls: precise query by batch ID, supporting batch selection with wildcards; zooming by time interval, with a minimum granularity of 1 minute and a maximum span of 30 days; grouping and viewing by physical temperature zone; clicking on any curve allows drilling down to the details panel for that batch. The panel displays its temperature drift prediction curve, real-time sensor data, and a list of triggered warning events side by side. Simultaneously, the view provides a comparative analysis mode, allowing selection of up to 6 batches for trend synchronization comparison and automatically calculating the Pearson correlation coefficient of their quality degradation index lines to help identify common quality issues. Degradation Mode: Based on real-time updated visual data, an automatic analysis module is built-in to assist in risk identification. It continuously scans the quality degradation index curves of all active batches and automatically marks them as high-risk batches when the following patterns are identified: 1. Slope change, i.e., the average growth rate of the quality degradation index exceeds 15 index points per hour within 3 consecutive calculation periods; 2. Absolute threshold exceedance, i.e. the quality degradation index is above 75 for 2 consecutive minutes; 3. Cross-validation anomaly, i.e. the deviation between the trend of the quality degradation index and the theoretical degradation trend calculated from the real-time monitoring temperature data exceeds 20%. The marked high-risk batches are highlighted and flashed in the interface, and a structured prompt is generated in the warning panel of the sidebar. The content includes batch ID, current quality degradation index, predicted remaining time to reach the end of shelf life, main abnormal links and related temperature drift event IDs. Managers can further view the equipment status, operation records and environmental event chains associated with the batch in the knowledge graph through the link traceability function, and then generate a list of intervention measures suggestions. All intervention measures suggestions are pushed to the comprehensive monitoring center in JSON format via HTTPS protocol.

[0087] The intelligent risk avoidance module performs the following steps: integrating identification and tracking data, temperature drift prediction results, and control strategy execution feedback to construct a multi-dimensional risk assessment matrix for the cold chain system. This identifies weak links and potential failure links in the system at four levels: equipment, operation, environment, and goods, achieving comprehensive and visual identification of system risks, improving the accuracy of problem location and traceability efficiency. Based on the risk assessment matrix and preset rule base, it automatically generates targeted risk avoidance suggestions, including equipment pre-maintenance prompts, operational specification adjustment suggestions, and transportation route optimization plans, and triggers early warning notifications to achieve intelligent early warning and proactive suggestion push, assisting managers in responding quickly and reducing operational risks. When system anomalies persist without relief, it automatically activates preset emergency plan processes, including dynamically adjusting transportation plans, switching backup refrigeration equipment, and notifying nearby maintenance personnel to intervene, thereby improving system resilience from passive response to proactive protection, achieving automated and closed-loop management of emergency response, and significantly improving the system's risk resistance and transportation continuity.

[0088] The specific work involves: in actual operation, continuously collecting and integrating real-time data streams from four dimensions at a frequency of seconds, covering four levels: equipment, operation, environment, and cargo. At the equipment level, data such as operating current, vibration spectrum, cumulative operating time, and historical fault codes of key components like compressors and fans are collected, with a sampling frequency of no less than 10Hz. Deviations of key performance parameters exceeding 5% are marked as potential risks. At the operation level, all door opening and closing events (including timestamps, duration, and operator ID), manual equipment control records, and violation operation markers are recorded. Any operation not performed according to standard operating procedures will be captured in real time and associated with the corresponding equipment. Backup unit; at the environmental level, it integrates internal and external temperature and humidity sensors, light sensors, and positioning data. The external environmental temperature and humidity sampling interval is 30 seconds, and the internal temperature and humidity sampling interval for each zone is 10 seconds. When drastic changes in the external environment or geofence crossing is detected, the risk level is automatically increased. At the cargo level, it is associated with the quality degradation index of each batch, real-time temperature drift value, and stacking parameters. The quality degradation index is calculated every 5 minutes. When the index slope exceeds a preset threshold, cargo status abnormality is triggered. All dimension data are normalized and then weighted according to preset weight coefficients (equipment layer weight 0.3, operation layer 0.2, environmental layer 0.25, cargo layer 0).25) A weighted fusion is performed to construct a real-time updated fourth-order risk assessment matrix. This matrix automatically identifies cross-level associated failure links through eigenvalue decomposition and cluster analysis. The overall health score of the entire matrix is ​​calculated every 30 minutes. When the score is below 70 points (out of 100), it is automatically marked as a weak link requiring priority intervention. Based on the output of the real-time risk assessment matrix, a preset expert rule base is invoked to automatically generate targeted avoidance suggestions. The rule base contains more than 300 IF-THEN production rules defined by domain experts, covering three major categories: equipment pre-maintenance, operation specifications, and path optimization. All generated suggestions are accompanied by a priority score (level 1-5, calculated based on the probability of risk occurrence and the degree of impact). Suggestions with a score ≥4 will automatically trigger an audible and visual warning notification and generate a continuously flashing warning card on the main interface of the integrated monitoring center until relevant personnel confirm and handle the situation. When a high-risk abnormal state is detected and continues to be unresolved, i.e., one of the following conditions is met: the performance indicators of key equipment continue to deteriorate within 15 minutes and deviate from the benchmark value by more than 15%; the local temperature drift value exceeds the safety threshold for three consecutive monitoring cycles and the control strategy is ineffective; or If the quality degradation index accelerates to above 75 within 30 minutes, the pre-set emergency plan is automatically activated. Once activated, based on real-time traffic and weather API data, the transportation route is dynamically replanned within 3 minutes to avoid congested or inclement weather sections. Route replanning prioritizes ensuring that the total travel time does not increase by more than 20% of the original plan. Simultaneously, if the main refrigeration system's efficiency falls below 60% of its rated value, a hard-wired control signal automatically switches to the backup refrigeration unit. The switching process ensures that the temperature control interruption time does not exceed 90 seconds. The backup unit's activation parameters are calculated and set in real-time based on the current cargo heat load. Furthermore, a work order is automatically dispatched via the integrated communication module (supporting 4G / 5G and satellite communication dual links) to the nearest on-site maintenance personnel (within a 50km radius based on real-time GPS positioning). The work order includes an event summary, location coordinates, on-site sensor snapshots, and preliminary diagnostic conclusions. Personnel are required to respond and confirm within 30 minutes. The entire emergency plan execution process is fully recorded and an emergency event report is generated for subsequent effectiveness evaluation and iterative optimization, achieving proactive protection against systemic risks and enhancing resilience.

[0089] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0090] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A real-time monitoring system for cold chain logistics of fresh produce, comprising a comprehensive monitoring center, characterized in that, The integrated monitoring center has the following communication modules: The identification tracking and temperature zone adaptation module is used to collect temperature trajectory data by assigning a unique object identifier to each batch of fresh products, and to use the ant colony algorithm to calculate the optimal temperature zone allocation path for the target temperature zone, so as to carry out intelligent scheduling of temperature control resources in multi-batch mixed transportation. The cumulative temperature drift prediction module is used to construct a cold chain knowledge graph that integrates equipment status, operational behavior and environmental factors. Combined with reinforcement learning agents to continuously learn historical and real-time data, it can make advance predictions and risk assessments of the cumulative temperature drift effect caused by hidden factors such as frequent door opening and equipment aging. The strategy dynamic optimization module dynamically adjusts the operating parameters and temperature zone allocation strategy of the refrigeration equipment based on the cumulative temperature drift prediction results and combined with the real-time environmental conditions and equipment performance, so as to adaptively suppress the temperature drift effect. The quality degradation visualization module is used to map the cumulative temperature drift effect into a fresh produce quality degradation curve, and to display the quality change path of each batch of products in real time. The intelligent risk avoidance module integrates identification tracking, temperature drift prediction, and control decision data to automatically identify system weaknesses, propose risk avoidance suggestions, and trigger emergency plans.

2. The real-time monitoring system for cold chain logistics of fresh products according to claim 1, characterized in that: The identifier tracking and temperature zone adaptation module includes a dynamic identifier tracking unit and a temperature zone intelligent adaptation unit. The dynamic identification tracking unit is used to assign a unique object identifier to each batch of fresh products and to collect temperature trajectory data including temperature, humidity and location in real time. The temperature zone intelligent adaptation unit analyzes the temperature zone requirements and real-time environment of each batch of fresh products based on the ant colony algorithm, dynamically calculates the optimal temperature zone allocation path, and performs adaptive scheduling of temperature control resources.

3. The real-time monitoring system for cold chain logistics of fresh products according to claim 2, characterized in that: The dynamic identifier tracking unit performs the following steps: Assign a unique object identifier based on an encrypted QR code to each batch of fresh produce, bind it to the product packaging, and establish a reliable mapping relationship between the identifier and the physical entity; By collecting temperature, humidity, and location data of each product corresponding to each label in real time through multi-source sensors distributed in the transport vehicle, structured temperature trajectory data is formed and continuously uploaded to the comprehensive monitoring center. Based on the uploaded temperature trajectory data, a batch-level full-process status trajectory map is constructed in the comprehensive monitoring center to realize real-time visual tracking and automatic identification of abnormal states of products from warehousing, transportation, transfer to delivery.

4. The real-time monitoring system for cold chain logistics of fresh products according to claim 2, characterized in that: The temperature zone intelligent adaptation unit performs the following steps: Receive temperature zone demand data and real-time environmental data for each batch from the dynamic identification tracking unit, and initialize the pheromone matrix and temperature zone adaptation feasible solution space of the ant colony algorithm based on the preset temperature sensitivity rule library for fresh food categories. An improved ant colony algorithm is used to simulate the multi-batch temperature zone allocation path optimization process. Through the pheromone accumulation and volatilization mechanism, the load, cooling efficiency and energy consumption of each temperature zone are dynamically evaluated, and the current optimal temperature zone allocation scheme is iteratively calculated. Based on the iterative optimization results, dynamic temperature zone scheduling instructions are generated for multiple batches of mixed-load scenarios to control the zoned refrigeration equipment to perform adaptive temperature control resource allocation.

5. The real-time monitoring system for cold chain logistics of fresh products according to claim 2, characterized in that: The cumulative temperature drift prediction module includes a knowledge graph construction unit and a reinforcement learning prediction unit; The knowledge graph construction unit is used to integrate multi-source data such as cold chain equipment status, operation behavior records, and environmental factors to construct a structured cold chain knowledge graph that represents the relationship and influence path between various factors. The reinforcement learning prediction unit, based on the cold chain knowledge graph and real-time sensor data, trains a reinforcement learning agent to identify temperature drift patterns and predicts the cumulative temperature drift trend and quality degradation risk caused by latent factors such as frequent door opening and equipment aging.

6. The real-time monitoring system for cold chain logistics of fresh products according to claim 5, characterized in that: The knowledge graph construction unit performs the following steps: Collect multi-source heterogeneous data from the entire cold chain logistics process, including refrigeration equipment operation logs, historical maintenance records, door opening frequency, external environmental temperature and humidity fluctuations, and cargo stacking methods; and clean, standardize, and extract entities from the data. Based on a predefined entity relationship model, the extracted entities and relationships are structured and modeled to construct a cold chain knowledge graph with equipment-operation-environment-goods as the core nodes. The node attributes include state parameters, timestamps, and confidence levels. By utilizing graph embedding technology and a rule-based reasoning engine, the relation weights and path connectivity in the cold chain knowledge graph are dynamically updated, enabling the correlation analysis and interpretable representation of the impact of latent factors on temperature drift.

7. The real-time monitoring system for cold chain logistics of fresh products according to claim 5, characterized in that: The reinforcement learning prediction unit performs the following steps: Using a cold chain knowledge graph as an environment representation, a state space, action space, and reward function are constructed. A temperature drift tendency agent based on deep reinforcement learning is designed to simulate the formation and evolution of cumulative temperature drift. By using historical temperature drift event sequences and real-time sensor data, the temperature drift-oriented agent is trained alternately offline and online, enabling it to learn to recognize temperature drift patterns and their evolution caused by latent factors such as frequent door opening and equipment performance degradation. The trained temperature drift trend agent combines the current environmental state with the reasoning results of the knowledge graph to predict the cumulative temperature drift trend of a specific batch in the future and the corresponding probability of quality degradation risk, and outputs an assessment of the scope of impact.

8. The real-time monitoring system for cold chain logistics of fresh products according to claim 7, characterized in that: The strategy dynamic optimization module performs the following steps: Real-time data collection of cumulative temperature drift prediction results, environmental monitoring data and refrigeration equipment operating status; construction of multi-objective optimization function with temperature drift suppression, energy consumption minimization and equipment life extension as core optimization objectives. The model predictive control algorithm is adopted to dynamically optimize the combination of operating parameters of the refrigeration equipment, including compressor frequency, fan speed and valve opening, based on the current system state and future temperature drift prediction, and to generate staged control commands. The optimized control commands are sent to the temperature control execution units of each zone to dynamically redistribute the temperature zones and adaptively adjust the cooling power, forming a closed-loop control and continuously suppressing the temperature drift effect.

9. A real-time monitoring system for cold chain logistics of fresh produce according to claim 8, characterized in that: The quality degradation visualization module performs the following steps: Based on the temperature drift trend data output by the cumulative temperature drift prediction module, and combined with the shelf life dynamics model corresponding to fresh food categories, the temperature drift effect is quantified into the cumulative quality degradation amount, and a real-time quality degradation index for each batch of products is generated. The quality degradation index is mapped to a time-quality change curve using a visualization engine, and then displayed in real time on the interactive interface of the comprehensive monitoring center in the form of multi-batch overlay heat maps. It supports multi-dimensional filtering and drill-down viewing by batch, time interval, and temperature zone, and provides a comparative analysis view of quality change trends to help managers intuitively grasp the quality evolution status and identify high-risk batches and processes.

10. A real-time monitoring system for cold chain logistics of fresh produce according to claim 9, characterized in that: The intelligent risk avoidance module performs the following steps: By integrating identification and tracking data, temperature drift prediction results, and feedback on the execution of control strategies, a multi-dimensional risk assessment matrix for the cold chain system is constructed to identify weak links and potential failure links in the system at four levels: equipment, operation, environment, and goods. Based on the risk assessment matrix and the preset rule base, it automatically generates targeted risk avoidance suggestions, including equipment pre-maintenance prompts, operating procedure adjustment suggestions and transportation route optimization plans, and triggers early warning notifications; If the system anomaly persists without resolution, the pre-set emergency response plan will be automatically activated, including dynamically adjusting the transportation plan, switching to backup refrigeration equipment, and notifying nearby maintenance personnel to intervene.

Images