A method and system for temperature control in cold chain logistics transportation of agricultural products

Abstract

This invention provides a method and system for temperature control in cold chain logistics transportation of agricultural products, belonging to the field of cold chain temperature control technology. The method includes: acquiring operating condition data characterizing the thermal state and airflow state during cold chain transportation; processing the operating condition data to characterize dynamic temperature features, obtaining feature parameters reflecting these features, and training a temperature rebound prediction model based on these feature parameters; generating a temperature rebound prediction curve based on the temperature rebound prediction model, and applying a preset correction rule to the basic target temperature curve corresponding to the transportation task based on the temperature rebound prediction curve to generate a corrected target temperature curve; generating target control parameters for driving the refrigeration unit and circulating fan based on the corrected target temperature curve and the operating condition data, and performing temperature regulation based on the target control parameters. This invention enables cold chain transportation temperature regulation to better match real-time heat load changes and maintain more stable temperature control performance.

Description

Technical Field

[0001] This invention relates to the field of cold chain temperature control technology, specifically to a method for controlling the temperature of agricultural product cold chain logistics transportation and a control system for agricultural product cold chain logistics transportation. Background Technology

[0002] Cold chain transportation is often accompanied by a complex and constantly changing thermal environment. The temperature inside the truck compartment is not a simple steady-state quantity, but exhibits obvious time-varying characteristics under the combined effects of factors such as the metabolic heat of the cargo, external heat intrusion, and impeded air circulation. Different agricultural products show significant differences in their thermal sensitivity during transportation; some release high metabolic heat during the initial cooling stage, while others are more sensitive to temperature fluctuations during long-distance transportation. These factors cause the compartment temperature to exhibit periodic rebounds, localized warming, or delayed changes, and these changes are often not synchronized with the operating rhythm of the refrigeration unit and circulating fans.

[0003] Existing cold chain temperature control methods mostly employ a single temperature difference control strategy, adjusting the cooling capacity or fan speed based on the difference between the set target temperature and the real-time compartment temperature. This approach relies on steady-state temperature difference judgment, ignoring the dynamic changes of internal heat sources and failing to reflect localized heat accumulation when airflow is obstructed. Some studies have attempted to incorporate ambient temperature or cargo type as adjustment criteria, but most approaches remain at the static setting level, lacking characterization of the temperature evolution process. In actual transportation, the rate of change in cargo metabolic activity often fluctuates on an hourly scale. If adjustments are made solely based on instantaneous temperature differences, problems such as cooling overshoot, delayed reheating, or unresponsive airflow can easily occur.

[0004] On the other hand, the air circulation efficiency in the carriage also exhibits nonlinear variation. The operating status of the circulating fan, the stacking method, and local blockages all affect airflow organization, thereby altering the temperature distribution. Traditional control strategies struggle to infer the air circulation state from a single-point temperature signal and are also difficult to respond to localized heat accumulation under artificially set target curves. As transportation distances increase, such cumulative deviations are often only discovered after delivery, posing significant challenges to the timeliness and accuracy of temperature control.

[0005] Existing methods generally lack the ability to predict the rebound behavior of vehicle compartment temperature. Temperature typically experiences a certain degree of rise, slow fall, or stagnation after external disturbances or changes in cargo metabolism, and these changes cannot be inferred from simple temperature difference relationships. For transportation tasks, a more reasonable approach is to predict future temperature trends in advance and adjust the target temperature curve accordingly, making the adjustment actions proactive. However, related technologies still rely heavily on empirical corrections or rule settings, making it difficult to consider the dynamic coupling characteristics of multi-source operating condition data.

[0006] In summary, temperature control in cold chain transportation faces two long-standing challenges: firstly, it is difficult to extract features reflecting dynamic temperature behavior from multi-source operating data; secondly, there is a lack of models capable of predicting temperature rebound trends, resulting in a lack of foresight in temperature regulation actions regarding future thermal conditions. These shortcomings limit the adaptability of temperature control strategies and make it difficult for the coordinated regulation of refrigeration units and circulating fans to maintain stable performance across different transportation stages. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for temperature control in cold chain logistics transportation of agricultural products, so as to at least solve the problem that the temperature regulation in existing cold chain transportation relies on a single temperature difference judgment and cannot reflect the temperature rebound behavior caused by changes in the metabolic heat of goods and air circulation, which is difficult to identify in advance.

[0008] To achieve the above objectives, a first aspect of the present invention provides a method for temperature control in cold chain logistics transportation of agricultural products. The method includes: acquiring operating condition data characterizing the thermal state and airflow state during cold chain transportation; performing processing on the operating condition data to characterize dynamic temperature features, thereby obtaining feature parameters reflecting the dynamic temperature features, and training a temperature rebound prediction model based on the feature parameters; generating a temperature rebound prediction curve based on the temperature rebound prediction model, and performing a preset correction rule on a basic target temperature curve corresponding to the transportation task based on the temperature rebound prediction curve to generate a corrected target temperature curve; generating target control parameters for driving the refrigeration unit and the circulating fan based on the corrected target temperature curve and the operating condition data, and performing temperature regulation based on the target control parameters.

[0009] Optionally, acquiring operating condition data to characterize the thermal state and airflow state during cold chain transportation includes: collecting temperature information from multiple locations inside the vehicle compartment and ambient temperature information outside the vehicle to form a temperature information set reflecting the temperature distribution inside the vehicle compartment and the external heat load; collecting operating status information of the circulating fan during transportation to form a fan operation information set reflecting changes in air circulation; collecting trace gas concentration information reflecting the metabolic intensity of the cargo to form a metabolism-related information set reflecting changes in internal heat sources; and aggregating the temperature information set, the fan operation information set, and the metabolism-related information set to form the operating condition data.

[0010] Optionally, the operating data is processed to characterize the dynamic characteristics of temperature to obtain feature parameters reflecting the dynamic characteristics of temperature, including: constructing a temperature change sequence over time based on the temperature information set, and extracting time-series feature parameters from the temperature change sequence to reflect the hysteresis response characteristics of the compartment temperature; calculating the concentration change per unit time based on the metabolism-related information set, and extracting metabolic feature parameters reflecting the change level of internal heat sources of the cargo according to a preset metabolic characteristic model; determining the load change of the circulating air path based on the fan operation information set, and extracting flow feature parameters reflecting the degree of air circulation obstruction according to a preset flow resistance model; and fusing the time-series feature parameters, the metabolic feature parameters, and the flow feature parameters into feature parameters reflecting the dynamic characteristics of temperature.

[0011] Optionally, the concentration change per unit time is calculated based on the metabolism-related information set, and metabolic characteristic parameters reflecting the level of change of internal heat sources in the cargo are extracted according to a preset metabolic characteristic model. This includes: performing time difference operation on the trace gas concentration data in the metabolism-related information set to obtain the concentration change value per unit time; inputting the concentration change value into a preset metabolic response model to determine the instantaneous metabolic quantity reflecting the metabolic intensity of the cargo; wherein, the preset metabolic response model is a model used to characterize the relationship of change in the metabolic intensity of the cargo, trained based on the trace gas release patterns of different cargo categories under closed transportation conditions; performing curve fitting operation on the instantaneous metabolic quantity to obtain a metabolic change curve characterizing the metabolic change trend; and extracting metabolic characteristic parameters reflecting the level of change of internal heat sources in the cargo based on the metabolic change curve.

[0012] Optionally, the load change of the circulating air path is determined based on the fan operation information set, and flow characteristic parameters reflecting the degree of air circulation obstruction are extracted according to a preset flow resistance model. This includes: performing joint differential analysis on the operating current and speed data in the fan operation information set to obtain a load change value characterizing the instantaneous load change of the circulating air path; inputting the load change value into the preset flow resistance model to determine the flow resistance amount reflecting the degree of air circulation obstruction; wherein, the preset flow resistance model is a model characterizing the resistance change law of the circulating air path, trained based on the electrical dynamic correspondence of the circulating fan under different load conditions; performing smoothing processing on the flow resistance amount to obtain a flow resistance change curve reflecting the air circulation change trend; and extracting flow characteristic parameters reflecting the degree of air circulation obstruction based on the flow resistance change curve.

[0013] Optionally, training a temperature rebound prediction model based on the feature parameters includes: performing feature normalization on the feature parameters to form a normalized feature vector for model training; performing model training calculations based on the normalized feature vector to obtain training results that reflect the correspondence between dynamic temperature features and temperature evolution laws; performing time series reconstruction on the training results to form a temperature response sequence that describes the temperature rebound process; and generating a temperature rebound prediction model based on the temperature response sequence to characterize the temperature rebound law under transportation conditions.

[0014] Optionally, a temperature rebound prediction curve is generated based on the temperature rebound prediction model, and a preset correction rule is applied to the basic target temperature curve corresponding to the transportation task based on the temperature rebound prediction curve to generate a corrected target temperature curve. This includes: performing time series interpolation processing on the temperature response sequence output by the temperature rebound prediction model to form a temperature rebound prediction curve describing the temperature rebound change process; extracting an offset parameter reflecting the temperature offset amplitude based on the difference between the temperature rebound prediction curve and the basic target temperature curve; applying the offset parameter to the basic target temperature curve according to the preset correction rule to obtain an intermediate correction curve characterizing the temperature adjustment trend; and generating the corrected target temperature curve based on the intermediate correction curve.

[0015] Optionally, extracting offset parameters reflecting the temperature shift amplitude based on the difference between the temperature rebound prediction curve and the baseline target temperature curve includes: performing time synchronization calculations on the differences between the temperature rebound prediction curve and the baseline target temperature curve at each time node to form a difference sequence describing the instantaneous shift level; determining corresponding stage sensitivity coefficients based on the stage attributes of the transportation task, and quantifying the stage sensitivity coefficients into sensitivity factors characterizing the degree of thermal sensitivity response in different transportation stages; performing offset enhancement calculations on the difference sequence according to the sensitivity factors to form a weighted offset sequence reflecting the stage-specific thermal sensitivity impact; and extracting offset parameters reflecting the temperature shift amplitude based on the weighted offset sequence.

[0016] Optionally, generating target control parameters for driving the chiller unit and the circulating fan based on the corrected target temperature curve and the operating condition data, and performing temperature regulation based on the target control parameters, includes: performing joint analytical operations on the corrected target temperature curve and the operating condition data to obtain heat load parameters describing the current heat load change level; determining the cooling intensity adjustment amount of the chiller unit based on the heat load parameters, and converting the cooling intensity adjustment amount into a cooling control command for driving the chiller unit; determining the air circulation demand of the circulating air path based on the heat load parameters, and converting the air circulation demand into a fan control command for driving the circulating fan; and using the cooling control command and the fan control command as the target control parameters and applying them to the chiller unit and the circulating fan to perform temperature regulation.

[0017] A second aspect of the present invention provides a temperature control system for cold chain logistics transportation of agricultural products. The system includes: an acquisition unit for acquiring operating condition data characterizing the thermal state and airflow state during cold chain transportation; a processing unit for processing the operating condition data to characterize temperature dynamics, thereby obtaining feature parameters reflecting the temperature dynamics, and training a temperature rebound prediction model based on the feature parameters; a prediction unit for generating a temperature rebound prediction curve based on the temperature rebound prediction model, and executing a preset correction rule on a basic target temperature curve corresponding to the transportation task based on the temperature rebound prediction curve to generate a corrected target temperature curve; and an execution unit for generating target control parameters for driving the refrigeration unit and the circulating fan based on the corrected target temperature curve and the operating condition data, and performing temperature regulation based on the target control parameters.

[0018] Through the above technical solution, this invention collects and processes operating data on the thermal state and airflow state during cold chain transportation, enabling the construction of a dynamic temperature characteristic expression that conforms to the actual transportation environment. This allows for a more accurate characterization of temperature lag, rebound magnitude, and localized heat accumulation caused by obstructed air circulation. Based on this, a temperature rebound prediction model obtained through feature parameter training can provide the temperature evolution trend over a future period, allowing the control process to no longer rely on instantaneous temperature differences but possess a forward-looking judgment of rebound behavior. Correcting the basic target temperature curve using the prediction curve makes the target curve closer to the thermosensitive state of goods at different transportation stages, avoiding a disconnect between adjustment actions and actual heat load changes. Finally, combining the corrected target temperature curve and the target control parameters generated from the operating data allows the adjustment actions of the refrigeration unit and circulating fan to better match the dynamic heat load changes during transportation, thus making the temperature adjustment process more aligned with the needs of the transportation task in terms of timing and magnitude.

[0019] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings:

[0021] Figure 1 This is a flowchart of the steps of a method for controlling the temperature of cold chain logistics transportation of agricultural products according to one embodiment of the present invention;

[0022] Figure 2 This is a detailed flowchart of step S2 of the temperature control method for cold chain logistics transportation of agricultural products provided in one embodiment of the present invention;

[0023] Figure 3 This is a system structure diagram of a temperature control system for cold chain logistics transportation of agricultural products provided in one embodiment of the present invention. Detailed Implementation

[0024] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0025] like Figure 1 As shown, this invention provides a method for temperature control in cold chain logistics transportation of agricultural products, the method comprising:

[0026] Step S1: Obtain operating condition data to characterize the thermal state and airflow state of the cold chain transportation process.

[0027] Specifically, temperature information from multiple locations inside the carriage and ambient temperature information outside the carriage are collected to form a temperature information set reflecting the temperature distribution inside the carriage and the external heat load; the operating status information of the circulating fan during transportation is collected to form a fan operation information set reflecting changes in air circulation; trace gas concentration information reflecting the metabolic intensity of the cargo is collected to form a metabolism-related information set reflecting changes in internal heat sources; the temperature information set, the fan operation information set, and the metabolism-related information set are then combined to form the operating condition data.

[0028] In this embodiment of the invention, the basic data acquisition process is first established around the temperature field inside the cold chain truck compartment. Multiple sets of temperature sensors are installed at different heights and positions within the compartment, covering at least the areas near the air ducts, the center of the cargo stack, and the areas near the doors. Each temperature sensor collects temperature data according to a uniform sampling period, and this data is aligned with the sampling time of the external ambient temperature acquisition device to form a time-stamped temperature information set. This deployment method records not only the overall temperature level but also the spatial distribution and gradient changes near the air inlets and outlets, facilitating subsequent determination of the area where temperature rebound first occurs.

[0029] To assess airflow conditions, the operating status of the circulating fan is used as the observation entry point. Specifically, this involves collecting information on the fan's speed, operating current, and start / stop status during operation. If necessary, fan speed signals or duty cycle parameters can also be recorded. This operating status information is compiled into a unified fan operating information set and synchronized with the temperature information set according to timestamps, ensuring a one-to-one correspondence between temperature changes and fan operating status within the same time window. In this way, the fan operating information set can reflect the strength of air circulation, potential blockage trends in the airflow path, and prolonged high-load operation conditions, which can then be used to infer the impact of changes in air circulation on localized temperature accumulation.

[0030] To characterize the changes in the heat source of the cargo itself, trace gas concentration was introduced as an indirect measure of metabolic intensity. Trace gas sensors were deployed inside the carriage near the cargo stacking area to collect gas concentration data related to the cargo's respiratory activity, such as carbon dioxide concentration or other gaseous components that can serve as metabolic indicators. A uniform sampling period was used during the data collection process, and abnormal sampling points that significantly deviated from the normal range were removed or marked, forming a metabolism-related information set. After aligning the metabolism-related information set with the temperature information set and the fan operation information set on the time axis, it can reflect the changing rhythm of the cargo's metabolic intensity at different transportation stages, providing data support from the internal heat source side for subsequent construction of dynamic temperature characteristics.

[0031] After completing the collection and alignment of the aforementioned multi-source data, the temperature information set, fan operation information set, and metabolism-related information set are aggregated to form operating condition data characterizing the thermal and airflow states of the cold chain transportation process. Structurally, the operating condition data can be viewed as a time-series collection of various physical quantities, with each time slice simultaneously containing the temperature distribution characteristics of the vehicle compartment, the strength of air circulation, and the intensity of cargo metabolism. Through this aggregation process, the temperature field, wind field, and internal heat sources are no longer treated in isolation but are organized into a unified operating condition description, providing a complete input foundation for subsequent dynamic extraction of temperature features and training of temperature rebound prediction models. Overall, this data acquisition and aggregation method enables the temperature control process to be analyzed and adjusted based on operating condition information that more closely reflects the actual transportation conditions.

[0032] Step S2: Perform processing on the operating data to characterize the dynamic characteristics of temperature, so as to obtain feature parameters that reflect the dynamic characteristics of temperature, and train a temperature rebound prediction model based on the feature parameters.

[0033] Specifically, the operating data, after multi-dimensional processing, is reorganized into a dynamic feature set that describes the temperature change pattern. Temperature information, fan status, and metabolic intensity are synchronized on the time axis, and then three key features—temperature lag, changes in internal heat sources, and obstructed air circulation—are extracted using rules such as differencing, smoothing, or weighting. The processed feature parameters retain the main driving forces behind the evolution of the carriage temperature, clearly expressing the inertia, rebound magnitude, and stage inflection points of temperature changes. Based on this, a temperature rebound prediction model is built through a training process, enabling the model to provide an estimated temperature curve for a future period based on the feature parameters. This approach gives the temperature control strategy a forward-looking judgment capability, no longer relying on temperature difference information at a single moment. Specifically, such as... Figure 2 Step S2 includes the following steps:

[0034] Step S21: Construct a temperature change sequence based on the temperature information set, and extract time-series feature parameters from the temperature change sequence to reflect the hysteresis response characteristics of the carriage temperature.

[0035] Specifically, a temperature change sequence over time is constructed based on the temperature information set, and temporal characteristic parameters reflecting the hysteresis response characteristics of the carriage temperature are extracted from this temperature change sequence. Specifically, multi-location temperature samples with timestamps are rearranged on the time axis according to a preset sampling step size to form a temperature change sequence covering the entire transportation period. For time points with missing samples or abnormal jumps, adjacent time-time interpolation and median filtering are used for correction, ensuring the temperature change sequence remains continuous and resolvable in the time dimension. Through this arrangement, temperature changes are no longer discrete, scattered points, but a continuously evolving time series, facilitating subsequent dynamic analysis of temperature response behavior.

[0036] After obtaining the temperature change sequence, time-series analysis rules are introduced to extract hysteresis features. Analysis windows can be defined before and after each cooling output adjustment event to statistically analyze the slope of the temperature change caused by the control action, the time required to reach a new stable range, and the maximum magnitude of temperature drop. Furthermore, statistics of the first and second differences of temperature changes can be calculated across multiple consecutive windows to characterize the speed of the temperature response and the presence of significant hysteresis. These quantification results are summarized into a set of time-series characteristic parameters, including at least a time constant parameter reflecting the response time, a recovery time parameter reflecting the stabilization speed, and a fluctuation amplitude parameter reflecting the degree of temperature fluctuation.

[0037] In some implementations, the temperature change sequences can be grouped by location. Independent temperature change subsequences are constructed for the top of the carriage, the middle of the cargo, and the area near the air duct. The aforementioned temporal feature extraction process is repeated in each subsequence. By comparing the time constants and recovery times at different locations, the regions with the slowest temperature response can be identified, providing more representative hysteresis features for subsequent model training. Finally, the temporal feature parameters obtained from each location or window are uniformly organized into a feature parameter set, which is input into the temperature rebound prediction model training process as an important component of the temperature dynamic features. Overall, this processing step transforms the raw temperature sampling data into temporal features that can directly describe the relationship between cooling action and temperature evolution, which is beneficial for the model to capture the hysteresis response pattern of the carriage temperature.

[0038] Step S22: Calculate the concentration change per unit time based on the metabolism-related information set, and extract metabolic characteristic parameters that reflect the change level of internal heat sources of the cargo according to the preset metabolic characteristic model.

[0039] Specifically, time difference calculation is performed on the trace gas concentration data in the metabolism-related information set to obtain the concentration change value per unit time; the concentration change value is input into a preset metabolic response model to determine the instantaneous metabolic quantity used to reflect the metabolic intensity of the cargo; wherein, the preset metabolic response model is a model used to characterize the relationship of changes in the metabolic intensity of the cargo, trained based on the trace gas release patterns of different cargo categories under closed transportation conditions; curve fitting calculation is performed on the instantaneous metabolic quantity to obtain a metabolic change curve used to characterize the metabolic change trend; metabolic characteristic parameters used to reflect the change level of internal heat sources of the cargo are extracted based on the metabolic change curve.

[0040] Specifically, the construction of metabolic characteristic parameters usually starts with the temporal changes of concentration sequences. During transportation, the concentration of trace gases is not a stable quantity but fluctuates slowly with the respiration intensity of the cargo. Therefore, the original concentration sequence must first be processed into incremental information suitable for analysis. A common practice is to perform temporal differencing on the sequence to obtain the concentration change per unit time. For ease of expression, a first-order difference expression is used here, defined as:

[0041]

[0042] in, Indicates the first The trace gas concentration values ​​at each sampling time. This represents the concentration value at the previous sampling time. Indicates the time interval between two adjacent samples. This represents the change in concentration per unit time. This difference result directly reflects the local increase or decrease trend of concentration over time and is more sensitive to metabolic changes than the original concentration sequence.

[0043] Concentration changes do not directly describe metabolic intensity; therefore, a mapping model is needed to express the metabolic response of goods. Metabolic response models are typically trained on extensive experimental data, including respiration experiments of different types of goods in a closed environment. These models take the concentration change per unit time as input and output the instantaneous metabolic rate. To clearly express this mapping relationship, the following formula can be used:

[0044]

[0045] in, Indicates the first Instantaneous metabolic rate at each sampling time. This represents a metabolic response model trained based on the patterns of cargo gas release. The model is not limited to a specific form and can be a linear regression model, an exponential response model, or even a more complex nonlinear learning structure. Its task is to map local concentration changes to a metabolic intensity order that more closely approximates the actual heat source output.

[0046] After obtaining the instantaneous metabolic rate sequence, curve fitting is usually performed on the sequence to more stably characterize the metabolic trend over time. Various fitting methods can be used, such as polynomial fitting or spline curve fitting. The key to selection is to reduce the interference of instantaneous noise while maintaining the overall shape of the trend. The metabolic change curve obtained through fitting allows observation of metabolic peaks, plateaus, and decay phases, providing a more compact feature representation for subsequent modeling.

[0047] When extracting metabolic characteristic parameters from metabolic change curves, several typical indicators are generally considered. For example, peak metabolic rate and its timing can be extracted to reflect maximum heat load; the slope of the rising segment can be extracted to describe the phase of rapid metabolic enhancement; or the average metabolic rate over the entire time window can be calculated as a representative of steady-state metabolic level. Fluctuation measures, such as standard deviation or coefficient of variation, can also be introduced to characterize metabolic stability. These parameters ultimately constitute the metabolic feature vector, an important component of temperature dynamics.

[0048] Through this processing pathway from concentration sequences to metabolic characteristics, changes in internal heat sources are no longer limited to simple concentration records, but are refined into a set of explicit parameters that can participate in the training of temperature prediction models. This transformation makes it easier for temperature rebound prediction models to capture the impact of cargo metabolic behavior on temperature evolution, making the overall prediction structure more closely aligned with actual transportation environments.

[0049] Step S23: Determine the load change of the circulating air path based on the fan operation information set, and extract flow characteristic parameters to reflect the degree of air circulation obstruction according to the preset flow resistance model.

[0050] Specifically, joint differential analysis is performed on the operating current and speed data in the fan operation information set to obtain a load change value that characterizes the instantaneous load change in the circulating air path; the load change value is input into a preset flow resistance model to determine the flow resistance quantity that reflects the degree of air circulation obstruction; wherein, the preset flow resistance model is a model that characterizes the resistance change law of the circulating air path based on the electrical dynamic correspondence of the circulating fan under different load conditions; the flow resistance quantity is smoothed to obtain a flow resistance change curve that reflects the air circulation change trend; and flow characteristic parameters that reflect the degree of air circulation obstruction are extracted based on the flow resistance change curve.

[0051] Specifically, in characterizing air circulation constraints, the electrical and mechanical quantities on the fan side need to be organized into a description that reflects changes in the airflow load. The fan operation information set typically includes operating current, speed, and start / stop status markers sampled over time. The combination relationships of these quantities vary significantly under different operating conditions. To avoid interference from random fluctuations in single-point readings, the current and speed sequences on the same time axis are usually denoised and interpolated to fill in missing points and suppress isolated spikes. After this processing, the fan operation information set is more suitable for analyzing the load variation levels of the circulating airflow at different stages.

[0052] In the calculation of load changes, a joint differential method can be used to compress the changes in current and speed into a unified index. Taking the k-th sampling time as an example, the changes in current and speed between two adjacent sampling times are normalized to a unit time, constructing a first-order differential component. The joint differential index can be written in the form of equation (1):

[0053] (1)

[0054] in, Indicates the first The operating current of the wind turbine at each sampling time. This represents the current at the previous sampling time. Indicates the first Fan speed at each sampling time This indicates the rotational speed at the previous sampling time. The time interval between two adjacent samples. This represents the load change value within the corresponding time interval. and These are weighting coefficients determined based on the fan's rated parameters and calibration tests. Through this joint differential calculation, the torque change reflected on the current side and the aerodynamic condition change reflected on the speed side are superimposed in the same evaluation quantity.

[0055] The load change value obtained by the combined differential equation only reflects the electrical side. It is still necessary to use a pre-defined flow resistance model to link this change to the resistance of the circulating air path. The pre-defined flow resistance model is derived from the electro-aerodynamic relationship of the circulating fan under different load conditions. It can be understood as the fitting result of the relationship between current, speed, air volume, and pressure head under steady-state or quasi-steady-state conditions. To express this mapping, the flow resistance can be written in the form of equation (2):

[0056] (2)

[0057] in, Indicates the first The flow resistance quantity at each sampling time point is used to characterize the degree of obstruction to air circulation. This represents a flow resistance model trained based on calibration data. The input to this model is the load variation value. The output is the equivalent resistance level at the corresponding time, which can be in the form of a piecewise linear model, an empirical multinomial model, or other nonlinear regression structures suitable for this type of wind turbine.

[0058] After obtaining the flow resistance over time, a smoothing and trend extraction step is usually performed. A simple approach is to... The sequence is subjected to moving average or low-pass filtering to suppress local sampling noise and preserve the drag trend that evolves slowly over time. The smoothed sequence can be regarded as a flow drag change curve, and the curve shape will show characteristics such as steps, gradual rises or falls when the wind path is blocked, the loading mode changes, or frost adheres. Those skilled in the art can select an appropriate smoothing window length according to the actual scenario of shipping or land transportation, so that the change curve neither loses key inflection points nor excessively amplifies transient disturbances.

[0059] When extracting flow characteristic parameters from flow resistance variation curves, the focus is typically on several aspects. One type of parameter describes the overall resistance level, such as the average or median resistance within a certain time window. Another type of parameter emphasizes the pattern of change, such as the rising or falling slope of the resistance curve over a local time period, and the amplitude of fluctuations over several consecutive sampling periods. Indicators representing the accumulation of anomalies can also be introduced, such as the duration or frequency of exceeding a preset resistance threshold. These quantities are combined according to pre-designed rules and organized into a flow characteristic parameter vector, used to reflect the degree of air circulation obstruction and its evolution trend.

[0060] Through the above processing flow, the wind turbine operating information set is no longer just a simple record of current and speed, but is rewritten into flow characteristic parameters that can indirectly reflect the wind path resistance state. In this embodiment of the invention, a joint differential and flow resistance model is introduced at this stage, enabling temperature rebound prediction and subsequent control stages to sense changes in air circulation conditions, providing more specific quantitative basis for the risk of localized thermal buildup.

[0061] Step S24: The time-series characteristic parameters, the metabolic characteristic parameters, and the flow characteristic parameters are fused into characteristic parameters that reflect the dynamic characteristics of temperature.

[0062] Specifically, the three types of features are first normalized or standardized to ensure that different physical quantities are on comparable scales, avoiding excessive weighting of a single dimension in subsequent training. Then, according to a pre-defined feature arrangement rule, the time-series feature parameters related to temperature lag, the metabolic feature parameters related to internal heat source intensity, and the flow feature parameters related to air circulation obstruction are sequentially combined into a one-dimensional feature vector. If necessary, feature weight coefficients can be introduced before fusion, assigning higher weights to parameters more sensitive to temperature dynamics and lower weights to auxiliary parameters, ensuring that the fusion result retains key information without excessive redundancy. After completing the above steps, the resulting feature vector serves as the feature parameters reflecting temperature dynamics, inputting into the subsequent model training stage, allowing temperature rebound prediction to be learned based on the same fused operating condition representation.

[0063] Step S25: Train and obtain a temperature rebound prediction model based on the feature parameters.

[0064] Specifically, feature normalization is performed on the feature parameters to form a normalized feature vector for model training; model training calculations are performed based on the normalized feature vector to obtain training results that reflect the correspondence between temperature dynamics and temperature evolution; time series reconstruction is performed on the training results to form a temperature response sequence that describes the temperature rebound process; and a temperature rebound prediction model is generated based on the temperature response sequence to characterize the temperature rebound law under transportation conditions.

[0065] In this embodiment of the invention, the training of the temperature rebound prediction model revolves around the previously constructed feature parameters. These feature parameters have already integrated time-series, metabolic, and flow features into a unified feature vector. Directly using this vector for training is not suitable; a numerical "shaping" process is required first. A common practice is to statistically analyze the mean and standard deviation of the training samples along each feature dimension, performing standardization on each feature to bring physical quantities from different sources to a similar numerical range. This results in a set of normalized feature vectors that preserve the relative differences between dimensions while avoiding the excessive influence of a single large-scale quantity on the training process.

[0066] After preparing the normalized feature vectors, it's necessary to define the object the model will learn. The target here isn't the temperature at a specific moment, but rather the evolution of temperature within a predefined prediction window. More simply, each training sample consists of a "current operating condition feature vector" and a "future temperature sequence," and the model learns the mapping relationship between them. To this end, temperature sequences are extracted from historical cold chain transportation data using a sliding window approach. The normalized feature vector corresponding to the starting moment of the window is used as input, and the temperature value sequence from that starting point to several subsequent sampling moments is used as the output label, forming a multi-step time-series prediction dataset.

[0067] To describe this training objective, a relatively complete form of the loss function can be given. Assume there are a total of... There are 10 training samples, each corresponding to a normalized feature vector. and length is True temperature sequence The model is given parameters Time output prediction sequence Then the objective function can be constructed as shown in equation (3):

[0068] (3)

[0069] in, For the overall loss function, This is the parameter set for the temperature rebound prediction model. Indicates the first Each sample in the prediction step size Predicted temperature at the location, This corresponds to the actual temperature. For the first The time weights of each prediction moment are used to give greater attention to errors in the early stages of the rebound or specific phases when needed. The regularization coefficient is . The squared L2 norm of the parameters is used to constrain model complexity. By constraining the objective function, the model maintains a relatively balanced parameter size while fitting temperature time series data.

[0070] Based on the above loss function, gradient descent-like optimization methods can be used to optimize the parameters. The process involves iterative updates. In each iteration, a batch of normalized feature vectors and corresponding real temperature sequences are extracted from the training set. The predicted sequence is calculated, and then substituted into the objective function to obtain the loss value and gradient. The parameters are updated according to a preset learning rate. After multiple iterations, the loss function gradually converges, and the model develops a relatively stable intrinsic expression of the temperature evolution patterns of different types of working conditions. Those skilled in the art can choose different training strategies, such as batch gradient, stochastic gradient, or mini-batch gradient, depending on the sample size and model structure.

[0071] After training convergence, the model's direct output is still a window-based predicted temperature segment. To better reflect real-world applications, these predictions need to be reconstructed into a time series. Specifically, based on the actual transportation timeline, the predicted segments obtained at different starting points can be aligned by time, weighted averaged according to overlapping intervals, or segments with higher confidence can be selected and spliced ​​into a continuous temperature response sequence. The reconstructed temperature response sequence is used to check whether the model's predictive behavior during the temperature rebound phase is reasonable. For example, after adjusting the cooling capacity, does the temperature show the expected rapid drop and slow rise, and are the local fluctuations roughly consistent with the previously extracted operating condition characteristics?

[0072] After completing the above training and reconstruction process, the mapping structure from the normalized feature vector to the temperature response sequence can be viewed as the temperature rebound prediction model. This model receives the fused feature parameters at any given time, normalizes them, and uses them as input, outputting the temperature rebound response within the corresponding prediction window. Overall, this training process links cross-source operating condition features with temperature time series, enabling the temperature rebound prediction model to provide a relatively reasonable rebound trend under given transportation conditions, providing a reusable dynamic tool for subsequent target temperature curve correction and control parameter determination.

[0073] Step S3: Generate a temperature rebound prediction curve based on the temperature rebound prediction model, and execute a preset correction rule on the basic target temperature curve corresponding to the transportation task based on the temperature rebound prediction curve to generate a corrected target temperature curve.

[0074] Specifically, time series interpolation is performed on the temperature response sequence output by the temperature rebound prediction model to form a temperature rebound prediction curve describing the temperature rebound change process; an offset parameter reflecting the temperature offset magnitude is extracted based on the difference between the temperature rebound prediction curve and the base target temperature curve; the offset parameter is applied to the base target temperature curve according to a preset correction rule to obtain an intermediate correction curve characterizing the temperature adjustment trend; and a corrected target temperature curve is generated based on the intermediate correction curve.

[0075] Furthermore, based on the difference between the temperature rebound prediction curve and the baseline target temperature curve, an offset parameter reflecting the temperature offset amplitude is extracted, including: performing time synchronization calculations on the differences between the temperature rebound prediction curve and the baseline target temperature curve at each time node to form a difference sequence describing the instantaneous offset level; determining the corresponding stage sensitivity coefficient based on the stage attributes of the transportation task, and quantifying the stage sensitivity coefficient into a sensitivity factor characterizing the degree of thermal sensitivity response in different transportation stages; performing an offset enhancement operation on the difference sequence according to the sensitivity factor to form a weighted offset sequence reflecting the stage-specific thermal sensitivity impact; and extracting an offset parameter reflecting the temperature offset amplitude based on the weighted offset sequence.

[0076] In this embodiment of the invention, the output of the temperature rebound prediction model is first organized into a continuous prediction curve. After receiving the characteristic parameters, the model typically outputs a temperature response sequence in discrete time steps, with a fixed sampling interval between each time step. To facilitate alignment with the baseline target temperature curve, time series interpolation is performed on this discrete sequence, resampling the prediction results onto a time grid consistent with the control period. The interpolation method can employ piecewise linear interpolation, spline interpolation, or other smoothing interpolation methods, which can be selected by those skilled in the art based on data smoothing requirements. The temperature rebound prediction curve obtained after interpolation can continuously describe the process of temperature decline and subsequent rebound on the time axis.

[0077] After obtaining the temperature rebound prediction curve, the prediction result needs to be compared with the pre-set baseline target temperature curve for the transportation task. The baseline target temperature curve is usually determined by the type of goods, transportation time, and destination requirements, and is often a piecewise constant or slowly changing curve. On the time axis, performing synchronous calculations on the temperature values ​​of the two curves at each discrete time point directly yields the instantaneous offset sequence. The offset includes both low-temperature offsets caused by excessive cooling and high-temperature offsets caused by insufficient cooling or increased heat load; therefore, it is more suitable to retain it in a signed difference form. Through this difference sequence, the degree to which temperature control deviates from the target during a certain transportation stage can be identified more intuitively.

[0078] Given the varying sensitivities to temperature deviations across different transportation stages, introducing a stage sensitivity coefficient to weight the deviation is a reasonable approach. Transportation tasks can be divided into three phases based on the operational flow: the initial temperature drop phase after loading, the stable operation phase, and the transition phase near arrival. The tolerance for temperature fluctuations often differs across these stages. For example, a certain degree of short-term deviation is acceptable during the temperature drop phase, while sustained deviation is more critical during the long-term stable operation phase. To reflect this difference, a corresponding stage sensitivity coefficient is configured for each stage, and further quantified as a sensitivity factor acting on specific time points. This factor is used to adjust the weight of the deviation at each time point in the overall deviation parameter.

[0079] When constructing the offset parameters, a stage-weighted offset metric can be used. Let the temperature rebound prediction curve be at the [missing information - likely a specific point or stage]. The predicted value at each time point is The target value of the basic target temperature curve at the same time point is The difference between the two is The transportation process is divided into several stages, using indexes. The identification phase, the phase sensitivity factor is denoted as This factor is assigned to the time nodes belonging to this stage. One implementation of the comprehensive offset parameter can be written as equation (4):

[0080] (4)

[0081] in, The offset parameter is used to reflect the overall temperature shift magnitude. Representation phase The corresponding set of time points, For the stage Sensitivity factor, For the first Temperature offset at each time point This is an exponential parameter greater than or equal to 1, used to adjust the penalty for large offsets. When The time bias is towards the mean absolute shift, when The emphasis is placed on moments with larger offsets. Through the weighted structure of Equation (4), the offset parameter numerically integrates the offset levels of each stage, while also reflecting the thermal response weights of different stages.

[0082] After obtaining the weighted offset sequence and the overall offset parameters, the preset correction rules come into play. The goal of these rules is not to completely redraw the base target temperature curve, but rather to make limited adjustments to the curve's amplitude and shape based on the offset amount and parameters. One possible strategy is to superimpose a correction term, obtained by scaling the weighted offset sequence, onto the base target temperature curve. The scaling factor can be derived from the comparison between the offset parameters and a preset offset threshold. For example, when the overall offset parameters exceed the upper limit of a certain allowable range, the proportion of the correction term is increased; conversely, a smaller correction magnitude is maintained. In this way, a clear correspondence is established between the offset parameters and the correction intensity.

[0083] The intermediate corrected curve is obtained by superimposing correction terms onto the base target temperature curve. The intermediate corrected curve maintains a high degree of similarity to the base target temperature curve numerically, but it more closely approximates the temperature rebound trend predicted by the model in terms of time position and amplitude. To avoid introducing excessively rapid target changes during the correction process, smoothing constraints can be added to the curve correction stage. For example, limiting the maximum gradient of target temperature changes between adjacent time periods, or using low-pass filtering to weaken the high-frequency components of the correction terms. Those skilled in the art can set these constraints based on transportation safety boundaries and equipment response speed.

[0084] After completing the above correction steps, a corrected target temperature curve is generated based on the intermediate correction curve. This corrected target temperature curve serves as the direct basis for subsequent control parameter calculations and, compared to the original base target temperature curve, better reflects the actual temperature evolution requirements under current operating conditions. This curve inherits the constraints of the original target setting on cargo quality requirements and transportation specifications, while also making appropriate fine-tuning in certain parts of the curve, giving the control strategy a certain degree of foresight regarding temperature rebound behavior. Overall, this processing flow from the predicted curve to the offset parameters, and then to the corrected target temperature curve, introduces dynamic adaptive capability into the temperature control process while maintaining the executability of the target curve.

[0085] In another possible implementation, a micro-area temperature perturbation correction method based on dynamic thermal imaging residual fields is introduced to further improve the sensitivity to local thermal anomalies during temperature rebound prediction. This method does not rely on additional sensors but utilizes existing low-resolution thermal imaging modules inside the carriage to extract weak local temperature difference perturbations from continuous thermal images. Specifically, spatial registration is first performed on adjacent thermal imaging frames to ensure that the same cargo surface area is aligned on the time axis; then, a pixel-level differential residual field is constructed from the registered image sequence, and a threshold-adaptive connected component extraction rule is used to identify local temperature rise areas. Local temperature rises in the residual field often correspond to small-scale heat source changes such as obstructed ventilation, overly dense stacking, or sudden increases in metabolism in individual containers, changes that are difficult to capture in data from traditional temperature sampling points.

[0086] To enable this method to participate in temperature prediction, the area, maximum temperature rise, and rate of rise of local residual regions can be summarized to form new micro-region thermal disturbance characteristic parameters. These parameters, along with existing time-series, metabolic, and flow characteristic parameters, are incorporated into the temperature rebound prediction model, giving the model a stronger ability to identify local thermal accumulation risks. Through this additional image residual analysis path, the temperature control method can respond more sensitively to atypical thermal behaviors in cold chain transportation, such as extreme loading patterns and localized blockages.

[0087] Step S4: Generate target control parameters for driving the refrigeration unit and the circulating fan based on the corrected target temperature curve and the operating condition data, and perform temperature regulation based on the target control parameters.

[0088] Specifically, a joint analytical operation is performed on the corrected target temperature curve and the operating condition data to obtain heat load parameters describing the current heat load change level; based on the heat load parameters, the cooling intensity adjustment amount of the refrigeration unit is determined, and the cooling intensity adjustment amount is converted into a cooling control command for driving the refrigeration unit; based on the heat load parameters, the air circulation demand of the circulating air path is determined, and the air circulation demand is converted into a fan control command for driving the circulating fan; the cooling control command and the fan control command are used as the target control parameters and applied to the refrigeration unit and the circulating fan to perform temperature regulation.

[0089] In this embodiment of the invention, the temperature regulation process no longer relies solely on a single target temperature, but rather on a combination of a modified target temperature curve and operating condition data. The modified target temperature curve provides the desired temperature trajectory under the current transportation task and temperature rebound prediction, while the operating condition data includes information such as the temperature field, metabolic intensity, and air circulation status. Combining these two factors allows for the depiction of the thermal balance within the carriage from both the "required temperature level" and the "current thermal environment state." Thus, the control quantity is no longer purely driven by temperature difference, but rather a quantified result linked to real-time heat load.

[0090] Constructing the heat load parameters is one of the core steps in this process. Specifically, this involves performing joint analytical operations on the corrected target temperature curve and the operating data, comparing the expected temperature, actual temperature of the carriage, ambient temperature, cargo metabolic characteristic parameters, and flow characteristic parameters on the same time axis. A weighted superposition method can be used to construct the estimated heat load. For example, in one embodiment, equation (5) is used to construct the instantaneous heat load parameters:

[0091] (5)

[0092] in, For a moment Heat load parameters, The current temperature of the carriage. To correct the target temperature curve at time 10:00 Target temperature The equivalent metabolic intensity is calculated from the aforementioned metabolic characteristic parameters. These are the flow characteristic parameters corresponding to the degree of air circulation obstruction. The weighting coefficients are determined based on calibration tests. Through equation (5), temperature deviation, internal heat sources of the cargo, and airflow resistance are combined to form a heat load description that can be directly used for control calculations.

[0093] After obtaining the heat load parameters, the control requirements for the chiller unit and the circulating fan need to be determined separately. For the chiller unit, the required cooling intensity adjustment can be calculated based on the magnitude and sign of the heat load parameters. For example, when... When the value is positive and exceeds the preset buffer zone, it indicates that the current overall heat input is too high, and the cooling output needs to be increased; when... When the temperature is close to zero or falls within the allowable fluctuation range, the current cooling level can be maintained. The adjustment amount can be obtained from the heat load parameters through piecewise functions or proportional-integral control laws, and further quantified into control quantities such as compressor speed, expansion valve opening, or refrigerant flow rate, and then organized into cooling control commands. In this way, the operation of the refrigeration unit is closely centered on the overall heat load change, rather than simply chasing the instantaneous temperature difference.

[0094] For circulating fans, the control logic focuses more on estimating the air circulation demand. This demand is related to both the spatial uniformity requirements of the target temperature curve and the airflow resistance level reflected by the flow characteristic parameters. When resistance is low and temperature distribution is relatively uniform, lower fan speeds can be used to reduce unnecessary energy consumption. However, when flow characteristic parameters indicate localized circulation obstruction or significant temperature stratification, it is necessary to increase the fan speed or adjust the fan operating mode to enhance mixing and ventilation. The air circulation demand can be calculated based on a combination of heat load parameters and flow characteristic parameters, then mapped to fan speed levels or duty cycle parameters to form specific fan control commands.

[0095] Once the refrigeration control commands and fan control commands are determined, these two types of commands together constitute the target control parameters and act on the refrigeration unit and the circulating fan. In actual execution, a discrete control cycle can be used, comparing the execution result of the previous cycle with the heat load parameters of the current cycle, and adjusting the control quantity within the new cycle, forming a control closed loop with feedforward components and local feedback. To avoid frequent adjustments causing mechanical shock to equipment and energy waste, a minimum adjustment step size and a minimum hold time can be set in the command generation stage to appropriately suppress small-amplitude, short-cycle heat load fluctuations.

[0096] Through the aforementioned control path, the operating states of the refrigeration unit and the circulating fan are no longer adjusted in isolation, but rather coordinated and allocated around the same heat load parameter. Correcting the target temperature curve ensures that the temperature control direction aligns with the transportation task requirements, while operating data ensures that the control variables are consistent with the actual thermal environment. The heat load parameter acts as a bridge between the two.

[0097] like Figure 3 As shown, this invention provides a temperature control system for cold chain logistics transportation of agricultural products. The system includes: an acquisition unit for acquiring operating condition data characterizing the thermal state and airflow state during cold chain transportation; a processing unit for processing the operating condition data to characterize temperature dynamics, obtaining feature parameters reflecting these dynamics, and training a temperature rebound prediction model based on the feature parameters; a prediction unit for generating a temperature rebound prediction curve based on the temperature rebound prediction model, and executing a preset correction rule on the basic target temperature curve corresponding to the transportation task based on the temperature rebound prediction curve to generate a corrected target temperature curve; and an execution unit for generating target control parameters for driving the refrigeration unit and the circulating fan based on the corrected target temperature curve and the operating condition data, and performing temperature regulation based on the target control parameters.

[0098] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a microcontroller, chip, or processor to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0099] The optional embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details described above. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention. It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not further describe the various possible combinations.

[0100] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the embodiments of the present invention, they should also be regarded as the content disclosed by the embodiments of the present invention.

Claims

1. A method for temperature control of agricultural product cold chain logistics transportation, characterized in that, The method includes: Acquire operating condition data to characterize the thermal and airflow states during cold chain transportation; The operating data is processed to characterize the dynamic characteristics of temperature, so as to obtain feature parameters that reflect the dynamic characteristics of temperature, and a temperature rebound prediction model is trained based on the feature parameters. A temperature rebound prediction curve is generated based on the temperature rebound prediction model, and a preset correction rule is applied to the basic target temperature curve corresponding to the transportation task based on the temperature rebound prediction curve to generate a corrected target temperature curve; including: Time series interpolation is performed on the temperature response sequence output by the temperature rebound prediction model to form a temperature rebound prediction curve describing the temperature rebound change process; an offset parameter reflecting the temperature offset magnitude is extracted based on the difference between the temperature rebound prediction curve and the base target temperature curve; the offset parameter is applied to the base target temperature curve according to a preset correction rule to obtain an intermediate correction curve characterizing the temperature adjustment trend; and a corrected target temperature curve is generated based on the intermediate correction curve. Extracting offset parameters reflecting the temperature shift amplitude based on the difference between the temperature rebound prediction curve and the baseline target temperature curve includes: performing time synchronization calculations on the differences between the temperature rebound prediction curve and the baseline target temperature curve at each time node to form a difference sequence describing the instantaneous shift level; determining corresponding stage sensitivity coefficients based on the stage attributes of the transportation task, and quantifying the stage sensitivity coefficients into sensitivity factors characterizing the degree of thermal sensitivity response in different transportation stages; performing offset enhancement calculations on the difference sequence according to the sensitivity factors to form a weighted offset sequence reflecting the stage-specific thermal sensitivity impact; and extracting offset parameters reflecting the temperature shift amplitude based on the weighted offset sequence. Based on the corrected target temperature curve and the operating condition data, target control parameters for driving the refrigeration unit and the circulating fan are generated, and temperature regulation is performed based on the target control parameters.

2. The produce cold chain logistics transport temperature control method of claim 1, wherein, Acquire operating condition data to characterize the thermal and airflow states during cold chain transportation, including: Temperature information from multiple locations inside the carriage and ambient temperature information outside the carriage are collected to form a temperature information set that reflects the temperature distribution inside the carriage and the external heat load. Collect operating status information of circulating fans during transportation to form a fan operation information set that reflects changes in air circulation; Collect trace gas concentration information to reflect the metabolic intensity of cargo, so as to form a metabolic-related information set to reflect changes in internal heat sources; The temperature information set, the fan operation information set, and the metabolism-related information set are aggregated to form the operating condition data.

3. The produce cold chain logistics transport temperature control method of claim 2, wherein, The operating condition data is processed to characterize the dynamic characteristics of temperature, thereby obtaining characteristic parameters that reflect the dynamic characteristics of temperature, including: Based on the temperature information set, a temperature change sequence is constructed to reflect the temperature change over time, and time-series feature parameters reflecting the hysteresis response characteristics of the carriage temperature are extracted from the temperature change sequence. The concentration change per unit time is calculated based on the metabolism-related information set, and metabolic characteristic parameters reflecting the change level of internal heat sources of the cargo are extracted according to the preset metabolic characteristic model. The load change of the circulating air path is determined based on the fan operation information set, and flow characteristic parameters reflecting the degree of air circulation obstruction are extracted according to the preset flow resistance model. The time-series characteristic parameters, the metabolic characteristic parameters, and the flow characteristic parameters are fused into characteristic parameters that reflect the dynamic characteristics of temperature.

4. The produce cold chain logistics transport temperature control method of claim 3, wherein, The concentration change per unit time is calculated based on the aforementioned metabolic information set, and metabolic characteristic parameters reflecting the level of change of internal heat sources in the cargo are extracted according to a preset metabolic characteristic model, including: Perform time difference operation on the trace gas concentration data in the metabolism-related information set to obtain the concentration change value per unit time. The concentration change value is input into a preset metabolic response model to determine the instantaneous metabolic amount that reflects the metabolic intensity of the cargo; wherein, the preset metabolic response model is a model used to characterize the relationship of changes in the metabolic intensity of the cargo, which is trained based on the trace gas release patterns of different cargo categories under closed transportation conditions. Perform curve fitting on the instantaneous metabolic rate to obtain a metabolic change curve that characterizes the trend of metabolic change; Based on the metabolic change curve, metabolic characteristic parameters reflecting the level of change of internal heat sources in the cargo are extracted.

5. The produce cold chain logistics transport temperature control method of claim 3, wherein, Based on the fan operation information set, the load change of the circulating air path is determined, and flow characteristic parameters reflecting the degree of air circulation obstruction are extracted according to a preset flow resistance model, including: A joint differential analysis is performed on the operating current and speed data in the wind turbine operation information set to obtain load change values ​​that characterize the instantaneous load change in the circulating air path; The load change value is input into a preset flow resistance model to determine the flow resistance amount that reflects the degree of air circulation obstruction; wherein, the preset flow resistance model is a model used to characterize the change law of circulation air resistance, which is trained based on the electrical dynamic correspondence of the circulating fan under different load conditions. The flow resistance is smoothed to obtain a flow resistance change curve that reflects the changing trend of air circulation. Based on the flow resistance variation curve, flow characteristic parameters reflecting the degree of air circulation obstruction are extracted.

6. The produce cold chain logistics transport temperature control method of claim 3, wherein, A temperature rebound prediction model is obtained by training based on the aforementioned feature parameters, including: The feature parameters are subjected to feature normalization to form a normalized feature vector for model training; The model training calculation is performed based on the normalized feature vector to obtain training results that reflect the correspondence between temperature dynamics and temperature evolution. The training results are reconstructed over time to form a temperature response sequence that describes the temperature rebound process. A temperature rebound prediction model is generated based on the temperature response sequence to characterize the temperature rebound law under transportation conditions.

7. The produce cold chain logistics transport temperature control method of claim 1, wherein, Based on the corrected target temperature curve and the operating condition data, target control parameters for driving the chiller unit and circulating fan are generated, and temperature regulation is performed based on the target control parameters, including: Perform joint analytical operations on the corrected target temperature curve and the operating condition data to obtain heat load parameters that describe the current level of heat load change; Based on the heat load parameters, the cooling intensity adjustment amount of the refrigeration unit is determined, and the cooling intensity adjustment amount is converted into a cooling control command for driving the refrigeration unit. The air circulation demand of the circulating air path is determined based on the heat load parameters, and the air circulation demand is converted into fan control commands for driving the circulating fan. The refrigeration control command and the fan control command are used as the target control parameters and applied to the refrigeration unit and the circulating fan to perform temperature regulation.

8. A temperature control system for cold chain logistics transportation of agricultural products, characterized in that, The system is used to execute the temperature control method for cold chain logistics transportation of agricultural products according to any one of claims 1-7, and the system includes: The acquisition unit is used to acquire operating condition data that characterizes the thermal state and airflow state during the cold chain transportation process. The processing unit is used to perform processing on the operating condition data to characterize the dynamic characteristics of temperature, so as to obtain feature parameters that reflect the dynamic characteristics of temperature, and to train a temperature rebound prediction model based on the feature parameters. The prediction unit is used to generate a temperature rebound prediction curve based on the temperature rebound prediction model, and to execute a preset correction rule on the basic target temperature curve corresponding to the transportation task based on the temperature rebound prediction curve to generate a corrected target temperature curve. The execution unit is used to generate target control parameters for driving the refrigeration unit and the circulating fan based on the corrected target temperature curve and the operating condition data, and to perform temperature regulation based on the target control parameters.

Images