Agricultural product cold chain compartment temperature control method based on space equalization

Abstract

The invention discloses an agricultural product cold chain carriage temperature control method based on space equalization, and relates to the technical field of intelligent cold chain logistics, and the method comprises the steps: collecting carriage three-dimensional point cloud data and agricultural product electronic tag information, building a carriage loading three-dimensional model, carrying out the space grid division, and generating a partition topological feature vector; collecting temperature, humidity and carbon dioxide concentration data, calculating average temperature and average gas concentration observation values, and generating a breathing heat estimation value in combination with the partition topological feature vector; inputting the breathing heat estimation value into a dynamic prediction model, generating a subarea cold load curve and a priority weight by combining the subarea temperature deviation, and outputting a subarea cold load sequence; and according to the partition cold load sequence, generating a local execution proposal, constructing a global optimization problem taking the space temperature balance degree as an optimization target, and generating a global control plan. According to the method, global temperature equalization and energy consumption collaborative optimization are realized through network reasoning and dynamic optimization of the meta-reinforcement learning strategy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent cold chain logistics, and in particular to a temperature control method for a cold chain carriage of agricultural products based on spatial balance. BACKGROUND

[0002] With the rapid development of the cold chain logistics industry, the quality assurance of agricultural products during transportation is increasingly valued. As a core link, the technology of cold chain carriage temperature control has evolved from mechanical temperature control to electronic PID control, and then to modern intelligent control. In recent years, the integration of Internet of Things technology and artificial intelligence has brought new breakthroughs to cold chain temperature control, and multi-sensor data acquisition, model predictive control (MPC) and distributed optimization algorithm have gradually become research hotspots. In particular, in the aspect of spatial temperature field regulation, the combination of computational fluid dynamics (CFD) simulation and real-time data assimilation technology has improved the accuracy of temperature prediction and control.

[0003] Despite the obvious progress made by existing technologies, there are still two key deficiencies: first, traditional methods lack quantitative characterization and dynamic optimization of the thermal coupling relationship in the carriage space. Most studies use fixed partitioning or simple weighted average strategies, which cannot adapt to the changes in cargo respiratory heat and the complexity of air flow organization, resulting in frequent local overheating or overcooling; second, existing control strategies are mostly based on static models or single-objective optimization, and fail to deeply integrate biological thermodynamic characteristics, multi-modal sensor data and spatio-temporal correlation analysis, resulting in lagging control response and insufficient balance. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a temperature control method for a cold chain carriage of agricultural products based on spatial balance to solve the problems of uneven spatial temperature distribution and lagging control response.

[0006] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides a space balance-based temperature control method for agricultural product cold-chain compartments, which comprises: collecting three-dimensional point cloud data of the compartment and electronic tag information of the agricultural products, constructing a three-dimensional model of the compartment load and performing space grid division to generate a partition topological feature vector; collecting temperature, humidity and carbon dioxide concentration data, calculating average temperature and average gas concentration observation values, and combining the partition topological feature vector to generate a respiration heat estimate value; inputting the respiration heat estimate value into a dynamic prediction model, combining the partition temperature deviation to generate a partition cold load curve and priority weight and outputting a partition cold load sequence; generating a local execution proposal according to the partition cold load sequence, and constructing a global optimization problem with the space temperature balance degree as the optimization objective to generate a global control plan; performing dynamic adjustment in the refrigeration unit, the main fan and the partition air guide actuator according to the global control plan to generate a high-precision temperature field and a structured feedback data stream; performing multi-modal spatio-temporal data coding and feature fusion according to the high-precision temperature field and the structured feedback data stream to generate a dynamic thermal coupling correlation weight, and generating an executable space balance configuration strategy through meta-reinforcement learning strategy network reasoning and dynamic optimization.

[0007] As a preferred scheme of the space balance-based temperature control method for agricultural product cold-chain compartments, the method comprises the following steps of collecting the compartment geometric parameters and the agricultural product batch information, constructing a three-dimensional model of the compartment load and performing space grid division, specifically as follows, Collecting three-dimensional point cloud data of the compartment and electronic tag information of the agricultural products, and obtaining the optimal storage temperature and the respiration heat dynamics coefficient; Collecting the cargo position information, combining the three-dimensional point cloud data of the compartment, constructing a three-dimensional model of the compartment load and performing space grid division.

[0008] As a preferred scheme of the space balance-based temperature control method for agricultural product cold-chain compartments, the method comprises the following steps of generating the partition topological feature vector, specifically as follows, Based on the optimal storage temperature and the respiration heat dynamics coefficient, obtaining the biological thermodynamic attribute feature vector of each region in the space grid and performing partition division to generate a dynamic partition identification set and a space adjacent relationship; Constructing a compartment partition topological graph according to the dynamic partition identification set and the space adjacent relationship to generate the partition topological feature vector.

[0009] As a preferred scheme of the space balance-based temperature control method for agricultural product cold-chain compartments, the method comprises the following steps of collecting the temperature, humidity and carbon dioxide concentration data, calculating the average temperature and the average gas concentration observation values, specifically as follows, Collecting the temperature, humidity and carbon dioxide concentration data of each partition, and performing filtering processing; Based on the filtered temperature and humidity and carbon dioxide concentration data, the average temperature and average gas concentration observation value of each partition are calculated.

[0010] As a preferred scheme of the space balance-based agricultural product cold chain vehicle compartment temperature control method, wherein: the respiration heat estimation value is generated by combining the partition topology feature vector, and the specific steps are as follows, According to the partition topology feature vector, a nonlinear relationship between the respiration heat, the average temperature and the average gas concentration observation value is established. Based on the nonlinear relationship, a prediction step of ensemble Kalman filtering is performed to obtain the predicted state of the respiration heat. Using the average temperature and average carbon dioxide concentration observation value, the predicted state of the respiration heat is assimilated and corrected by performing an update step of ensemble Kalman filtering, and the respiration heat estimation value is generated.

[0011] As a preferred scheme of the space balance-based agricultural product cold chain vehicle compartment temperature control method, wherein: the respiration heat estimation value is input into a dynamic prediction model, and the partition cold load curve and priority weight are generated by combining the partition temperature deviation, and the partition cold load sequence is output, and the specific steps are as follows, The respiration heat estimation value is input into a dynamic prediction model, and the future respiration heat estimation value sequence and the future temperature deviation sequence at the future time point are predicted by combining the partition temperature deviation; According to the future respiration heat estimation value sequence and the future temperature deviation sequence, the cold load curve of each partition is obtained through a manifold learning algorithm. According to the peak absolute value of the cold load curve, the priority weight of each partition is distributed; Based on the priority weight of each partition, the cold load curve is weighted and packaged to generate the partition cold load sequence.

[0012] As a preferred scheme of the space balance-based agricultural product cold chain vehicle compartment temperature control method, wherein: according to the partition cold load sequence, a local execution proposal is generated, and a global optimization problem with spatial temperature balance degree as the optimization objective is constructed to generate a global control plan, and the specific steps are as follows, Based on the partition cold load sequence, a local execution proposal is generated by combining the local device constraint condition through an optimization algorithm; According to the local execution proposal, a global optimization strategy with spatial temperature balance degree as the optimization objective is constructed; A multi-agent collaborative evolution algorithm is used to analyze the global optimization problem to obtain a global optimal control sequence; The global optimal control sequence is encoded into a global control plan through space-time instruction coding.

[0013] As a preferred scheme of the space balance-based agricultural product cold chain vehicle compartment temperature control method, wherein: the dynamic adjustment is performed in the refrigeration unit, the main fan and the partitioned air guide actuator according to the global control plan, a high-precision temperature field and a structured feedback data stream are generated, and the specific steps are as follows The global control plan is compiled into driving instructions of the refrigeration unit, the main fan and the partitioned air guide actuator; The driving instructions are used to set simulation boundary conditions, and a predicted temperature field distribution is obtained through real-time fluid dynamics simulation; The temperature and humidity and carbon dioxide concentration data are numerically assimilated with the predicted temperature field distribution to generate a high-precision temperature field; Based on the high-precision temperature field and the operation state data of the refrigeration unit, the main fan and the partitioned air guide actuator, a structured feedback data stream is generated through multi-source data standardization coding.

[0014] As a preferred scheme of the space balance-based agricultural product cold chain vehicle compartment temperature control method, wherein: the dynamic thermal coupling correlation weight is generated through multi-modal spatio-temporal data coding and feature fusion based on the high-precision temperature field and the structured feedback data stream, and the specific steps are as follows, The multi-modal spatio-temporal data coding and feature fusion are performed based on the high-precision temperature field and the structured feedback data stream to generate fused spatio-temporal features; Based on the fused spatio-temporal features, the dynamic thermal coupling correlation weight between spaces is calculated through a spatio-temporal attention mechanism.

[0015] As a preferred scheme of the space balance-based agricultural product cold chain vehicle compartment temperature control method, wherein: the executable space balance configuration strategy is generated through meta-reinforcement learning strategy network reasoning and dynamic optimization, and the specific steps are as follows, Based on the dynamic thermal coupling correlation weight, the control action parameters are generated through meta-reinforcement learning strategy network reasoning and optimization; The control action parameters are decoded into space balance regulation and control instructions to generate an executable space balance configuration strategy.

[0016] The present application has the advantages that: through the prediction step and the update step of the ensemble Kalman filter, the respiratory heat estimation value is generated, the high-precision dynamic inversion of the respiratory heat of the agricultural products in the vehicle compartment, which is a key state that cannot be directly measured, is realized, and through the meta-reinforcement learning strategy network reasoning and dynamic optimization, the global temperature balance and energy consumption collaborative optimization are realized. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0018] Fig. 1 Flowchart of the space balance-based agricultural product cold chain compartment temperature control method.

[0019] Fig. 2 Flowchart of the compartment loading three-dimensional model construction and space grid division.

[0020] Fig. 3 Flowchart of generating the respiration heat estimation value and the cold load prediction.

[0021] Fig. 4 Flowchart of generating the partitioned cold load sequence. DETAILED DESCRIPTION

[0022] In order to make the above-mentioned objects, features and advantages of the present application more apparent and comprehensible, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0023] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the concept of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0024] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0025] Reference Figs. 1-4 For one embodiment of the present application, the embodiment provides a space balance-based agricultural product cold chain compartment temperature control method, comprising the following steps: S1, collecting compartment three-dimensional point cloud data and agricultural product electronic tag information, constructing a compartment loading three-dimensional model and performing space grid division, and generating a partitioned topological feature vector.

[0026] Collecting compartment three-dimensional point cloud data and agricultural product electronic tag information, and obtaining the best storage temperature and respiration heat dynamics coefficient.

[0027] It should be noted that the three-dimensional point cloud data of the carriage is used to obtain the geometric structure characteristics of the internal space of the carriage, and the electronic tag information of the agricultural products is collected to obtain the storage requirements related to the agricultural products. In this process, the temperature-related parameters are extracted based on the agricultural product characteristic data recorded in the agricultural product electronic tag information, so as to obtain the optimal storage temperature of the agricultural products, and further, the respiration heat dynamics coefficient is obtained according to the respiration characteristic parameters of the agricultural products recorded in the agricultural product electronic tag information, and finally the optimal storage temperature and the respiration heat dynamics coefficient are obtained.

[0028] The three-dimensional point cloud data of the carriage includes a set of spatial coordinates of the inner wall of the carriage, the surface of the goods and the outline of the equipment, which is usually collected by scanning with a laser radar or a depth camera.

[0029] The electronic tag information of the agricultural products includes variety identification, optimal storage temperature threshold and respiration heat dynamics coefficient, which is collected by near field communication or ultra-high frequency radio frequency identification technology.

[0030] The goods position information is collected, combined with the three-dimensional point cloud data of the carriage, to construct a three-dimensional model of the carriage loading and perform spatial grid division.

[0031] The specific process includes collecting the goods position information to determine the specific placement position of the agricultural products in the carriage, matching the goods position information with the geometric characteristics of the carriage space based on the internal space structure of the carriage described by the three-dimensional point cloud data of the carriage, so as to accurately map the distribution of the goods in the carriage space, and on this basis, constructing a three-dimensional model of the carriage loading and performing spatial grid division within the spatial range of the three-dimensional model of the carriage loading.

[0032] Further, the training process of the three-dimensional model of the carriage loading is based on the geometric structure characteristics of the internal space of the carriage obtained from the three-dimensional point cloud data of the carriage, and the loading distribution representation of the agricultural products in the carriage space is established based on the goods position information. In this process, a large number of three-dimensional model sample data of the carriage loading are used for training, and through the feature extraction and matching of different carriage geometric structures and different goods placement modes, the three-dimensional model of the carriage loading can quickly generate stable three-dimensional model of the carriage loading when new three-dimensional point cloud data of the carriage and goods position information are input.

[0033] Based on the optimal storage temperature and the respiration heat dynamics coefficient, the biological thermodynamic attribute feature vectors of each region in the spatial grid are obtained and partitioned, and a dynamic partition identification set and a spatial adjacent relationship are generated.

[0034] The specific process includes using the optimal storage temperature and respiratory thermodynamic coefficient to generate a biothermodynamic property feature vector for each cell within the spatial grid. The biothermodynamic property feature vector includes temperature sensitivity and respiratory heat intensity parameters. A clustering algorithm is used to group adjacent spatial grid cells with similar biothermodynamic property feature vectors into the same partition and assign a unique partition identifier to form a dynamic partition identifier set. At the same time, the spatial adjacency relationship between each partition is recorded.

[0035] A carriage partition topology map is constructed based on the dynamic partition identifier set and spatial adjacency relationship, and partition topology feature vectors are generated.

[0036] The specific process includes: a dynamic partition identifier set provides unique identifiers for all partitions; spatial adjacency defines the physical connection state between partitions; each partition is abstracted as a topological graph node and undirected edges are established between nodes based on spatial adjacency; the topological graph nodes are mapped into low-dimensional dense vector representations through a graph embedding algorithm; and the resulting partition topological feature vector can simultaneously retain the spatial location relationship and biothermodynamic property association features of the partitions.

[0037] S2. Collect temperature, humidity and carbon dioxide concentration data, calculate the average temperature and average gas concentration observations, and generate respiratory heat estimates by combining the regional topological feature vectors.

[0038] Temperature, humidity, and carbon dioxide concentration data for each zone were collected and then filtered.

[0039] It should be noted that the temperature, humidity, and carbon dioxide concentration data are collected to obtain real-time information on the internal environment of the carriage. After the data is collected, the temperature, humidity, and carbon dioxide concentration data are filtered to remove environmental interference and noise components, so as to keep the temperature, humidity, and carbon dioxide concentration data stable and continuous, thereby obtaining temperature, humidity, and carbon dioxide concentration data that can reflect the real environmental conditions.

[0040] Based on the filtered temperature, humidity, and carbon dioxide concentration data, the average temperature and average gas concentration observations for each zone are calculated using the following expressions: ; ; in, Indicates the partition number. Indicates the first Average temperature of each zone Indicates the first The total number of effective sensor nodes deployed within each partition. This represents the traversal index of each sensor node. This represents the temperature data value measured and reported by the i-th sensor node. Indicates the first average gas concentration observation value of the zone, represents the gas concentration value measured and reported by the i-th sensor node.

[0041] The specific process includes that the calculation method of the average temperature and the average gas concentration observation value is that the temperature values and the gas concentration values measured and reported by multiple sensor nodes in each zone are collected, the measurement results of all sensor nodes in the zone are summarized, and the overall balance level is calculated based on this, so that the average temperature and the average gas concentration observation value capable of representing the overall environment state of the zone are obtained.

[0042] According to the zone topology feature vector, a nonlinear relationship between the respiratory heat, the average temperature, and the average gas concentration observation value is established.

[0043] The specific process includes that according to the zone topology feature vector, a multivariate nonlinear regression or an artificial neural network is used, the zone topology feature vector and the average temperature and the average gas concentration observation value in the zone are taken as input variables, and a nonlinear relationship between the respiratory heat and the average temperature and the average gas concentration observation value is established through parameter fitting and training for inferring the estimated value of the respiratory heat.

[0044] Based on the nonlinear relationship, a prediction step of the ensemble Kalman filter is performed to obtain the predicted state of the respiratory heat.

[0045] The specific process includes that based on the nonlinear relationship, the state of the respiratory heat is deduced by using the prediction step of the ensemble Kalman filter, and in the prediction process, the change trend of the zone topology feature vector and the average temperature and the average gas concentration observation value is taken as an input condition, and the predicted state of the respiratory heat (for example, a respiratory heat state set containing a mean value and a covariance is generated) is obtained through state transition and uncertainty propagation.

[0046] The average temperature and the average carbon dioxide concentration observation value are used to perform an update step of the ensemble Kalman filter to assimilate and correct the predicted state of the respiratory heat, and generate the estimated value of the respiratory heat.

[0047] The specific process includes that the average temperature and the average carbon dioxide concentration observation value are used as observation inputs, the predicted state of the respiratory heat is corrected by using the update step of the ensemble Kalman filter, in the correction process, the observation inputs are compared with the predicted state of the respiratory heat, and the predicted state of the respiratory heat is assimilated by combining the update mechanism of the ensemble Kalman filter, so as to generate the estimated value of the respiratory heat.

[0048] S3, inputting the estimated value of the respiratory heat into a dynamic prediction model, combining a zone temperature deviation, generating a zone cooling load curve and a priority weight, and outputting a zone cooling load sequence.

[0049] The respiratory heat estimation value is input into a dynamic prediction model, and a future respiratory heat estimation value sequence and a future temperature deviation sequence at a future time point are predicted in combination with the zoned temperature deviation.

[0050] The specific process includes that after the respiratory heat estimation value is input into the dynamic prediction model, the dynamic prediction model analyzes based on the current state and historical change trend to generate the future respiratory heat estimation value sequence at the future time point. In combination with the zoned temperature deviation, the dynamic prediction model predicts the future temperature deviation sequence at the future time point by using the current value and historical information of the zoned temperature deviation. Finally, the future respiratory heat estimation value sequence and the future temperature deviation sequence are output as the prediction result.

[0051] Further, the construction process of the dynamic prediction model is to establish the dynamic relationship between the respiratory heat and the temperature deviation according to the historical respiratory heat estimation value sequence and the zoned temperature deviation data by using a time series analysis method, to describe the evolution law of the respiratory heat with time in a state space equation form, and to determine the parameters of the state transition matrix and the observation matrix by using a system identification technology, and finally to form a dynamic prediction model capable of predicting the respiratory heat state and the temperature deviation at a future time point.

[0052] Further, the training process of the dynamic prediction model first uses the historical respiratory heat estimation value sequence and the corresponding historical zoned temperature deviation sequence as training data. By adjusting the internal parameters of the dynamic prediction model, the dynamic prediction model can learn the time sequence dependence relationship and dynamic change law between the respiratory heat estimation value and the zoned temperature deviation. In the training process, the dynamic prediction model continuously optimizes the prediction ability to minimize the difference between the predicted future respiratory heat estimation value sequence and the future temperature deviation sequence and the real sequence. Finally, the dynamic prediction model capable of accurately performing multi-step prediction is obtained.

[0053] According to the future respiratory heat estimation value sequence and the future temperature deviation sequence, a cold load curve of each zone is obtained by using a manifold learning algorithm.

[0054] The specific process includes that according to the future respiratory heat estimation value sequence and the future temperature deviation sequence, a manifold learning algorithm is used to perform dimensionality reduction processing on high-dimensional time sequence features, the nonlinear relationship between the future respiratory heat estimation value sequence and the future temperature deviation sequence and the difference features between zones are retained in the dimensionality reduction process, so as to extract low-dimensional representations capable of representing the change trend of the cold load, and to generate the cold load curve of each zone.

[0055] The difference features between zones mainly reflect the spatial heterogeneity of the biological thermodynamic attribute feature vectors, and the difference features between zones are obtained by using a clustering algorithm to measure the similarity and group the biological thermodynamic attribute feature vectors of the spatial grid cells.

[0056] It should be noted that in the process of obtaining the cooling load curve of each subzone according to the future breathing heat estimation value sequence and the future temperature deviation sequence, the role of the manifold learning algorithm is to map the high-dimensional time sequence characteristics contained in the future breathing heat estimation value sequence and the future temperature deviation sequence to a low-dimensional space. Through this mapping, the nonlinear relationship and overall structural characteristics existing in the sequence can be preserved, so that the complex change pattern can be clearly expressed in the low-dimensional representation, so that the cooling load change trend of each subzone at different time points can be intuitively presented in the form of a cooling load curve.

[0057] The priority weight of each subzone is allocated according to the peak absolute value of the cooling load curve.

[0058] The specific process includes: based on the peak absolute value of the cooling load curve, by comparing the change amplitudes of the cooling load curves of each subzone at different time points, the interval with the strongest temperature regulation demand in the cooling load curve is identified, and the importance of the subzone in the overall regulation is dynamically determined according to the size of the change amplitude, thereby generating the priority weight of each subzone.

[0059] Based on the priority weight of each subzone, the cooling load curve is weighted and packaged to generate a subzone cooling load sequence.

[0060] The specific process includes: the priority weight of each subzone is applied to the time sequence value of the cooling load curve item by item, and a weighted aggregation method is used to package the cooling load curve, thereby generating a subzone cooling load sequence reflecting the priority difference and representing the cooling load demand intensity of each subzone at different time points.

[0061] S4, according to the subzone cooling load sequence, generating a local execution proposal, and constructing a global optimization problem with spatial temperature uniformity as the optimization objective, generating a global control plan.

[0062] Based on the subzone cooling load sequence, combined with the local device constraint condition, a local execution proposal is generated through an optimization algorithm.

[0063] The specific process includes: based on the future refrigeration demand information provided by the subzone cooling load sequence, combined with the provisions about the device operating range and performance limit in the local device constraint condition, the optimization algorithm identifies the device operation strategy under the premise of meeting all local device constraint conditions, and finally generates a local execution proposal.

[0064] It should be noted that the local device constraint condition is determined by the physical characteristics and safe operation requirements of the device, and directly comes from the technical specifications and field operation manual of the device.

[0065] According to the local execution proposal, a global optimization strategy with spatial temperature uniformity as the optimization objective is constructed.

[0066] The specific process includes: collecting the local execution proposals generated by each partition agent to a global coordinator, the global coordinator integrates all local execution proposals and defines a space temperature uniformity index as the core optimization target, the space temperature uniformity index quantifies the uniformity of the temperature distribution in each partition in the carriage to represent the global thermal environment consistency, and the global coordinator constructs a constrained optimization problem based on the space temperature uniformity index, which aims to find a set of global control instructions to make the space temperature uniformity optimal.

[0067] The multi-agent collaborative evolution algorithm is used to analyze the global optimization problem and obtain the global optimal control sequence.

[0068] The specific process includes: using a multi-agent collaborative evolution algorithm to iteratively analyze the global optimization problem, and in the analysis process, multiple agents exchange candidate solutions and evaluation results through collaborative evolution to gradually improve the quality of global solutions, and finally converge to a global optimal control sequence in the search space that can meet the optimization target, which is determined according to the matching degree of the partition cooling load sequence and the future temperature deviation sequence.

[0069] It should be noted that the multi-agent collaborative evolution algorithm improves the analysis efficiency of the global optimization problem through parallel search and interactive evolution mechanism, thereby accelerating the generation of the global optimal control sequence.

[0070] The global optimal control sequence is encoded into a global control plan through space-time instruction coding.

[0071] The specific process includes: structurally analyzing the control actions corresponding to different time points and different space partitions in the global optimal control sequence, and then sequentially arranging and coding the control actions according to the time sequence and the difference characteristics between partitions, thereby forming a space-time instruction set that can express both time and space dimensions. In this process, the space-time instruction coding not only preserves the control logic in the global optimal control sequence, but also converts it into an executable global control plan, so that the refrigeration unit, main fan and partition air guide executor can be operated collaboratively under a unified time frame and space frame, thereby realizing the coordinated regulation of the overall environment.

[0072] S5, according to the global control plan, executing dynamic adjustment in the refrigeration unit, main fan and partition air guide executor, generating high-precision temperature field and structured feedback data stream.

[0073] The global control plan is compiled into driving instructions for the refrigeration unit, main fan and partition air guide executor.

[0074] The specific process includes resolving each control action in the global control plan for each time point and partition, and translating it according to the operating characteristics and execution rules of the refrigeration unit, main fan and partition air guide actuator, so as to convert the abstract global control plan into driving instructions that can directly drive the operation of the equipment. In this process, the control action is refined into a power adjustment signal of the refrigeration unit, a wind volume control signal of the main fan, and an angle adjustment signal of the partition air guide actuator, ensuring that different equipment can work according to the unified global control logic, so as to realize dynamic adjustment and precise control of the overall environment.

[0075] The execution rule refers to the physical operation constraints and logical control rules of the refrigeration unit, main fan and partition air guide actuator, which are summarized through the technical parameter manual provided by the equipment manufacturer and the engineering configuration in actual deployment.

[0076] Based on the driving instruction, the simulation boundary condition is set, and the predicted temperature field distribution is obtained through real-time fluid dynamics simulation.

[0077] The specific process includes converting the operating state of the refrigeration unit, main fan and partition air guide actuator into simulation boundary conditions of air flow, temperature exchange and gas diffusion, and applying corresponding energy input and fluid boundary in the spatial range. Then, through real-time fluid dynamics simulation, the air flow field and temperature transmission process are numerically identified, and the motion trajectory of air flow and heat distribution law are dynamically simulated, so as to obtain the predicted temperature field distribution that can reflect the future environmental change trend.

[0078] The temperature, humidity and carbon dioxide concentration data are numerically assimilated with the predicted temperature field distribution to generate a high-precision temperature field.

[0079] The specific process includes fusing the temperature, humidity and carbon dioxide concentration data with the predicted temperature field distribution in a unified framework through numerical assimilation method, so that the predicted temperature field distribution can be closer to the actual situation in spatial distribution. The specific method is to use the humidity and carbon dioxide concentration data to correct the deviation of the predicted temperature field distribution, so that the predicted temperature field distribution can reflect the actual environmental characteristics in time sequence change and spatial details, thereby generating a more detailed and reliable high-precision temperature field.

[0080] Based on the high-precision temperature field and the operating state data of the refrigeration unit, main fan and partition air guide actuator, through multi-source data standardization coding, a structured feedback data stream is generated.

[0081] The specific process includes: the high-precision temperature field and the operation state data of the refrigeration unit, the main fan and the partition air guide executor are dimensionally unified and scaled, so that the various types of data are consistent in numerical range and expression form, thereby eliminating the heterogeneous differences between various types of data. After standardization processing, all data are converted into a unified coding format and organized according to time sequence and spatial position, and finally a structured feedback data stream is formed, so that the high-precision temperature field and the equipment operation state information can be fused in the same data framework.

[0082] The various types of data include temperature distribution data in the high-precision temperature field, operation power and refrigerant flow data of the refrigeration unit, rotation speed and air pressure data of the main fan, and opening angle and guiding position data of the partition air guide executor.

[0083] S6, according to the high-precision temperature field and the structured feedback data stream, multi-modal spatio-temporal data coding and feature fusion are carried out, dynamic thermal coupling correlation weights are generated, and through meta-reinforcement learning strategy network reasoning and dynamic optimization, an executable spatial balance configuration strategy is generated.

[0084] Based on the high-precision temperature field and the structured feedback data, multi-modal spatio-temporal data coding and feature fusion are carried out to generate fused spatio-temporal features.

[0085] The specific process includes: first, the high-precision temperature field and the equipment operation state data are aligned in the time dimension, and mapped in the spatial dimension according to the corresponding regions, so as to realize the synchronous expression of the high-precision temperature field and the equipment operation state data. Then, the numerical standardization and feature extraction are carried out on the data of different modalities, the temperature change features and the equipment operation features are converted into unified spatio-temporal representation form, and the multi-modal features are combined into continuous whole feature sequence through fusion method, and finally the fused spatio-temporal features are generated, which are used to reflect the dynamic correlation between temperature distribution and equipment operation state.

[0086] Based on the fused spatio-temporal features, the dynamic thermal coupling correlation weights between the positions in space are calculated through the spatio-temporal attention mechanism, and the expression is: ; Among them, represents the dynamic thermal coupling correlation weight between the space position and the space position , a represents the identification number of a specific position in space, represents the identification number of another specific position in space, represents the natural exponential function, represents the fused spatio-temporal feature vector of the space position , and represents the fused spatio-temporal feature vector of the space position The fusion of spatiotemporal feature vectors, This represents the total number of points in space. An index identifier representing a spatial location. Indicates spatial location The fusion of spatiotemporal feature vectors, Indicates based on spatial location eigenvectors and spatial location eigenvectors Calculated attention score, Indicates based on spatial location eigenvectors and spatial location eigenvectors Calculated attention score.

[0087] The specific process involves matching feature vectors of spatial locations with each other, and measuring the correlation between spatial locations by calculating attention scores among the feature vectors of different spatial locations. Each spatial location's feature vector is compared with the feature vectors of other spatial locations, and a corresponding weight distribution is obtained through a natural exponential function calculation. This weight is then normalized to reflect the relative importance of different spatial locations. The resulting dynamic thermodynamic coupling correlation weights can describe the dynamic relationship between spatial locations in terms of heat transfer and mutual influence.

[0088] Based on dynamic thermal coupling associated weights, the control action parameters are generated through reasoning and optimization via a meta-reinforcement learning policy network.

[0089] The specific process involves feeding dynamic thermal coupling weights as input into a meta-reinforcement learning policy network. During this process, the network continuously updates its policy to adaptively adjust the dynamic relationships between spatial locations, thereby extracting the most advantageous feature representations for control during temporal inference. Subsequently, the network utilizes policy optimization methods from reinforcement learning to evaluate and screen candidate control methods. After balancing the dynamic thermal coupling effect with target constraints, it generates control action parameters that ensure high execution effectiveness while adapting to complex spatiotemporal changes.

[0090] Further, the training process of the meta-reinforcement learning strategy network is to repeatedly train in a large number of data scenes containing different temperature field distributions, gas concentration changes and equipment running states, so that the meta-reinforcement learning strategy network can quickly extract spatio-temporal features and learn the corresponding relationship between different control action parameters and results in a diversified environment. In the training process, the meta-reinforcement learning strategy network continuously iteratively optimizes the strategy parameters to improve the adaptability and generalization ability of the dynamic thermal coupling correlation weight in unobserved scenes, and finally forms a strategy structure that can efficiently infer and optimize the control action parameters.

[0091] The control action parameters are decoded into spatial balance regulation instructions to generate an executable spatial balance configuration strategy.

[0092] The specific process includes: the control action parameters generated by the meta-reinforcement learning strategy network are parsed one by one into operation regulation signals of the refrigeration unit, the main fan and the partition air guide actuator, and then the regulation amplitude and timing arrangement of different spatial positions are determined according to the dynamic thermal coupling correlation weight, so that the temperature and airflow distribution of each region in space are coordinated and adjusted in the execution process, thereby realizing balanced control of the air environment while maintaining overall energy efficiency, and finally forming an executable spatial balance configuration strategy.

[0093] The embodiment also provides a computer device suitable for the case of the agricultural product cold chain vehicle compartment temperature control method based on spatial balance, which comprises a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the agricultural product cold chain vehicle compartment temperature control method based on spatial balance proposed in the above embodiment.

[0094] The computer device can be a terminal, and the computer device comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. In addition, the input device can be an external keyboard, touchpad or mouse, etc.

[0095] The embodiment also provides a storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method for controlling the temperature of the cold chain compartment of the agricultural product based on the spatial balance as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disk.

[0096] In summary, the present application achieves high-precision dynamic inversion of the respiratory heat of the agricultural product in the compartment, which is a key state that cannot be directly measured, by performing the prediction step and the update step of the ensemble Kalman filter to generate the respiratory heat estimation value, and achieves global temperature balance and energy consumption collaborative optimization through meta-reinforcement learning strategy network reasoning and dynamic optimization.

[0097] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A spatially balanced agricultural product cold chain trailer temperature control method, comprising: The application relates to a method for dynamically optimizing the temperature field of a refrigerated vehicle. The method comprises the following steps: Collecting three-dimensional point cloud data of a vehicle compartment and electronic tag information of agricultural products, constructing a three-dimensional model of the vehicle compartment and performing spatial grid division to generate a partition topological feature vector; Collecting temperature and humidity and carbon dioxide concentration data, calculating average temperature and average gas concentration observation values, and combining the partition topological feature vector to generate a respiration heat estimation value; Inputting the respiration heat estimation value into a dynamic prediction model, combining a partition temperature deviation to generate a partition cold load curve and a priority weight, and outputting a partition cold load sequence; According to the partition cold load sequence, a local execution proposal is generated, and a global optimization problem with spatial temperature uniformity as an optimization objective is constructed to generate a global control plan; According to the global control plan, dynamic adjustment is performed in the refrigeration unit, the main fan and the partition air guide actuator to generate a high-precision temperature field and a structured feedback data stream; 2. The spatially-balanced agricultural cold-chain trailer temperature control method of claim 1, wherein: According to the high-precision temperature field and the structured feedback data stream, multi-modal space-time data coding and feature fusion are performed to generate a dynamic thermal coupling correlation weight, and an executable space balance configuration strategy is generated through meta-reinforcement learning strategy network reasoning and dynamic optimization. The method for dynamically optimizing the temperature field of the refrigerated vehicle comprises the following steps of collecting three-dimensional point cloud data of a vehicle compartment and electronic tag information of agricultural products, constructing a three-dimensional model of the vehicle compartment and performing spatial grid division, Collecting three-dimensional point cloud data of a vehicle compartment and electronic tag information of agricultural products, and obtaining a best storage temperature and a respiration heat dynamics coefficient; 3. The spatially-balanced agricultural cold-chain trailer temperature control method of claim 2, wherein: Collecting cargo position information, combining the three-dimensional point cloud data of the vehicle compartment, constructing a three-dimensional model of the vehicle compartment and performing spatial grid division. The method for dynamically optimizing the temperature field of the refrigerated vehicle comprises the following steps of collecting three-dimensional point cloud data of a vehicle compartment and electronic tag information of agricultural products, constructing a three-dimensional model of the vehicle compartment and performing spatial grid division, Based on the best storage temperature and the respiration heat dynamics coefficient, the biological thermodynamic attribute feature vector of each region in the spatial grid is obtained and partitioned to generate a dynamic partition identification set and a spatial adjacent relationship; 4. The spatially-balanced agricultural cold-chain trailer temperature control method of claim 3, wherein: According to the dynamic partition identification set and the spatial adjacent relationship, a vehicle compartment partition topological graph is constructed to generate a partition topological feature vector. The method for dynamically optimizing the temperature field of the refrigerated vehicle comprises the following steps of collecting three-dimensional point cloud data of a vehicle compartment and electronic tag information of agricultural products, constructing a three-dimensional model of the vehicle compartment and performing spatial grid division, Collecting temperature and humidity and carbon dioxide concentration data of each partition and performing filtering processing; 5. The spatially-balanced agricultural cold-chain trailer temperature control method of claim 4, wherein: Based on the filtered temperature and humidity and carbon dioxide concentration data, the average temperature and average gas concentration observation values of each partition are calculated. The method for dynamically optimizing the temperature field of the refrigerated vehicle comprises the following steps of collecting three-dimensional point cloud data of a vehicle compartment and electronic tag information of agricultural products, constructing a three-dimensional model of the vehicle compartment and performing spatial grid division, According to the partition topological feature vector, a nonlinear relationship among the respiration heat, the average temperature and the average gas concentration observation value is established; Based on the nonlinear relationship, a prediction step of ensemble Kalman filtering is performed to obtain a prediction state of the respiration heat; 6. The spatially-balanced agricultural cold-chain trailer temperature control method of claim 5, wherein: Using the average temperature and the average gas concentration observation value, an update step of ensemble Kalman filtering is performed to assimilate and correct the prediction state of the respiration heat to generate a respiration heat estimation value. The method for dynamically optimizing the temperature field of the refrigerated vehicle comprises the following steps of collecting three-dimensional point cloud data of a vehicle compartment and electronic tag information of agricultural products, constructing a three-dimensional model of the vehicle compartment and performing spatial grid division, The respiration heat estimation value is input into a dynamic prediction model, and combined with a partition temperature deviation, a future respiration heat estimation value sequence and a future temperature deviation sequence at a future time point are predicted. According to the future respiratory heat estimation value sequence and the future temperature deviation sequence, the cold load curve of each partition is obtained through a manifold learning algorithm; According to the peak absolute value of the cold load curve, the priority weight of each partition is allocated; Based on the priority weight of each partition, the cold load curve is weighted and packaged to generate a partition cold load sequence.

7. The spatially-balanced agricultural cold-chain trailer temperature control method of claim 6, wherein: According to the partition cold load sequence, a local execution proposal is generated, and a global optimization problem with spatial temperature uniformity as the optimization objective is constructed to generate a global control plan, the specific steps are as follows, Based on the partition cold load sequence, combined with the local device constraint condition, a local execution proposal is generated through an optimization algorithm; According to the local execution proposal, a global optimization strategy with spatial temperature uniformity as the optimization objective is constructed; A multi-agent collaborative evolution algorithm is used to analyze the global optimization strategy to obtain a global optimal control sequence; The global optimal control sequence is encoded into a global control plan through space-time instruction coding.

8. The spatially-balanced agricultural cold-chain trailer temperature control method of claim 7, wherein: According to the global control plan, dynamic adjustment is performed in the refrigeration unit, main fan and partition air guide executor to generate a high-precision temperature field and structured feedback data stream, the specific steps are as follows The global control plan is compiled into driving instructions for the refrigeration unit, main fan and partition air guide executor; Based on the driving instructions, the simulation boundary conditions are set, and the predicted temperature field distribution is obtained through real-time fluid dynamics simulation; The temperature and humidity and carbon dioxide concentration data are numerically assimilated with the predicted temperature field distribution to generate a high-precision temperature field; Based on the high-precision temperature field and the operation state data of the refrigeration unit, main fan and partition air guide executor, multi-source data standardization coding is performed to generate a structured feedback data stream.

9. The spatially-balanced agricultural cold-chain trailer temperature control method of claim 8, wherein: According to the high-precision temperature field and the structured feedback data stream, multi-modal spatio-temporal data coding and feature fusion are performed to generate dynamic thermal coupling correlation weights, the specific steps are as follows, Based on the high-precision temperature field and the structured feedback data stream, multi-modal spatio-temporal data coding and feature fusion are performed to generate fused spatio-temporal features; Based on the fused spatio-temporal features, the dynamic thermal coupling correlation weights between different positions in space are calculated through a spatio-temporal attention mechanism.

10. The spatially-balanced agricultural cold-chain trailer temperature control method of claim 9, wherein: Through meta-reinforcement learning strategy network reasoning and dynamic optimization, an executable space balance configuration strategy is generated, the specific steps are as follows, Based on the dynamic thermal coupling correlation weights, the meta-reinforcement learning strategy network is used for reasoning and optimization to generate control action parameters; The control action parameters are decoded into space balance regulation instructions to generate an executable space balance configuration strategy.

Images