An internet-of-things-based cold-chain logistics temperature and humidity monitoring method and system

Abstract

The present application relates to cold chain logistics temperature and humidity monitoring technical field, disclose a kind of cold chain logistics temperature and humidity monitoring method and system based on Internet of Things, the method includes obtaining the initial thermal state set of car initial temperature and return air inlet initial humidity contained cold bridge, analysis obtains cargo load fluctuation value;Collecting cold bridge real-time temperature and return air inlet real-time humidity, calculate temperature and humidity deviation;Fusion above-mentioned parameter and precooling balance, predict cooling demand estimate value;Mapping obtains initial compression speed and fan air volume, determine target parameter by filtering and fitting;Generation frequency conversion and air duct regulation instruction, analysis obtains start-stop delay length, finally complete the collaborative matching of cooling supply and demand.This method can realize the space-time accurate matching of cooling supply and demand, effectively solve the problem of local temperature anomaly of cold bridge.

Description

Technical Field

[0001] This invention relates to the field of temperature and humidity monitoring technology in cold chain logistics, and in particular to a method and system for monitoring temperature and humidity in cold chain logistics based on the Internet of Things. Background Technology

[0002] Currently, in the cold chain transportation of fresh food and biological products, the internal thermal state of the vehicle is constrained by both external solar radiation and ambient temperature fluctuations, as well as the respiratory heat of the goods and disturbances from door openings. The nonlinear coupling of multiple heat loads leads to spatiotemporal heterogeneity in the temperature and humidity field. With the popularization of IoT sensing technology, intelligent temperature control methods based on industrial data processing are gradually becoming a core means to improve cold chain quality. However, existing solutions still struggle to achieve a precise match between cold energy supply and spatial requirements.

[0003] In one existing technology, the electric vehicle air conditioning control method receives target operating parameters via a network, collects single-point temperature values ​​inside the vehicle at fixed intervals, calculates the rate of temperature difference change, and inputs it into the compressor speed calculation model. From the perspective of thermal control mechanism, this scheme only extracts feedback signals from a single temperature dimension, without decoupling analysis of multi-source heat loads such as solar radiation, cold bridge heat leakage, and latent heat of cargo, nor does it introduce the contribution of humidity changes to cooling demand. Furthermore, the linear speed mapping model it uses cannot adapt to the nonlinear fluctuation characteristics of the heat load, and the fan airflow adjustment is completely disconnected from the spatial distribution demand for cooling, resulting in the inability to compensate for heat intrusion at cold bridge locations in a timely manner, and the continuous accumulation of local temperature deviations over transportation time.

[0004] In summary, existing technologies lack a mechanism for the coordinated decoupling of multi-source heat loads and precise control of spatial cold distribution, resulting in problems such as delayed cold response and continuously widening temperature deviations at cold bridge locations under complex operating conditions. Summary of the Invention

[0005] This invention provides a method and system for monitoring temperature and humidity in cold chain logistics based on the Internet of Things, so as to realize real-time cold supply and cold bridge temperature control under complex working conditions.

[0006] Firstly, to address the aforementioned technical problems, this invention provides a method for monitoring temperature and humidity in cold chain logistics based on the Internet of Things, comprising:

[0007] The initial temperature at the cold bridge and the initial humidity at the return air inlet are collected, and the parameters are fused based on the initial temperature and the initial humidity to obtain the initial thermal state set.

[0008] Load analysis is performed based on the initial thermal state set to obtain the cargo load fluctuation value;

[0009] The real-time temperature at the cold bridge and the real-time humidity at the return air vent are obtained. The deviations between the real-time temperature and the initial temperature, and between the real-time humidity and the initial humidity are calculated to obtain the temperature deviation and humidity deviation.

[0010] Based on the temperature deviation, humidity deviation, cargo load fluctuation value, and pre-acquired pre-cooling margin, a multi-factor fusion prediction is performed to obtain an estimated cooling demand.

[0011] Based on the estimated cooling demand, the initial compression speed and initial fan air volume are obtained by mapping. The initial compression speed is then used to calculate the smoothing slope to obtain the speed ramp slope. The speed ramp slope is then integrated to obtain the target compression speed. Finally, the initial fan air volume is fitted to obtain the target fan air volume.

[0012] A first adjustment command is generated based on the target compression speed and the speed ramp rate, and a second adjustment command is generated based on the target fan air volume.

[0013] The first and second adjustment commands are parsed to obtain the start-stop delay duration, and the cooling supply and cooling demand are matched in a coordinated manner based on the start-stop delay duration.

[0014] Secondly, the present invention provides a cold chain logistics temperature and humidity monitoring system based on the Internet of Things, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described above.

[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] (1) This invention collects solar radiation intensity, cargo breathing heat and cold bridge temperature and humidity data, constructs an initial heat source distribution matrix and introduces a condensation risk index for weight correction, and obtains an initial thermal state set through clustering. This solves the problem that the existing technology does not distinguish the coupling characteristics of multi-source heat load and ignores the thermal and humidity coupling effect of cold bridge, and realizes a comprehensive and accurate assessment of the thermal state of the carriage, providing a reliable data basis for subsequent cold demand prediction.

[0018] (2) This invention integrates cargo load fluctuation value, temperature and humidity deviation and pre-cooling margin, and uses a random forest model to predict multivariate cold load. At the same time, it superimposes the instantaneous disturbance of door opening to correct the deviation. This solves the problem of low cold load prediction accuracy and slow response caused by single temperature feedback in the prior art. It significantly improves the accuracy of cold load prediction and system response speed, and can adapt to complex and ever-changing transportation conditions in a timely manner.

[0019] (3) This invention constructs a spatial distribution field of cold energy to perceive local differences in cold energy, uses a long short-term memory network model to predict the cold energy demand at the location of the cold bridge, and adjusts the local guide plate to achieve precise supplemental cooling. This solves the problem of the disconnect between global cold energy regulation and spatial distribution demand in the prior art, effectively controls the temperature deviation at the location of the cold bridge, and at the same time reduces the overall energy consumption of the refrigeration system and ensures the quality of cargo transportation. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of a method for monitoring temperature and humidity in cold chain logistics based on the Internet of Things, provided in the first embodiment of the present invention. Detailed Implementation

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

[0022] Reference Figure 1 The first embodiment of the present invention provides a method for monitoring temperature and humidity in cold chain logistics based on the Internet of Things, including the following steps:

[0023] S11, collect the initial temperature at the cold bridge and the initial humidity at the return air vent, and perform parameter fusion based on the initial temperature and the initial humidity to obtain the initial thermal state set;

[0024] S12, perform load analysis based on the initial thermal state set to obtain the cargo load fluctuation value;

[0025] S13, obtain the real-time temperature at the cold bridge and the real-time humidity at the return air vent, and calculate the deviation between the real-time temperature and the initial temperature, and the deviation between the real-time humidity and the initial humidity, respectively, to obtain the temperature deviation and humidity deviation;

[0026] S14. Based on the temperature deviation, humidity deviation, cargo load fluctuation value and pre-acquired pre-cooling margin, a multi-factor fusion prediction is performed to obtain a cooling demand estimate.

[0027] S15, Map the estimated cooling demand to obtain the initial compression speed and initial fan air volume, calculate the smoothing slope based on the initial compression speed to obtain the speed ramp slope, integrate the speed ramp slope to obtain the target compression speed, and fit the initial fan air volume to obtain the target fan air volume.

[0028] S16, a frequency conversion command is generated based on the target compression speed and the speed climb slope to obtain a first adjustment command, and an air duct command is generated based on the target fan air volume to obtain a second adjustment command;

[0029] S17, the first adjustment command and the second adjustment command are parsed to obtain the start-stop delay time, and the coordinated matching of cooling supply and cooling demand is completed according to the start-stop delay time.

[0030] In step S11, the initial temperature at the cold bridge and the initial humidity at the return air vent are collected. Based on the initial temperature and initial humidity, parameters are fused to obtain an initial thermal state set, including:

[0031] Acquire the solar radiation intensity and initial heat source data of the carriage, and perform spatial mapping on the initial heat source data based on the solar radiation intensity to obtain the initial heat source distribution;

[0032] Based on the initial heat source distribution, the initial temperature at the location of the cold bridge and the initial humidity at the return air inlet are collected, and the initial temperature and initial humidity are mapped to obtain a heat-humidity coupling vector.

[0033] Based on the thermal-humidity coupling vector, the initial heat source distribution is corrected to obtain the corrected heat load distribution;

[0034] The modified heat load distribution is clustered to obtain the initial thermal state set.

[0035] It should be noted that solar radiation intensity data is collected using a silicon photovoltaic cell-type photoelectric sensor installed in the center of the refrigerated truck's roof. The sampling frequency is fixed at 1 Hz, allowing real-time capture of solar radiation variations at different times and geographical locations. The unit of measurement is watts per square meter (W / m²), with a measurement accuracy controlled within ±10 W / m². The sensor's spectral response range covers 300 nm to 1100 nm, enabling accurate measurement of solar radiation across the entire wavelength band. Initial heat source data is collected by heat flow sensors arranged at multiple points inside the truck. These sensors are embedded in pre-embedded slots in the truck's inner walls, with measurement points evenly distributed at different heights within the cargo stacking area. The spacing between measurement points is 1 meter, with a total of 40 measurement points. The installation positions of each measurement point avoid the contact surface after standard cargo stacking. The sensor probes are covered with breathable and waterproof protective covers, maintaining an air gap of at least 50 mm between them and the cargo. This ensures complete coverage of the effective heat exchange area of ​​all cargo space inside the truck, preventing cargo obstruction from interfering with heat flow detection. Spatial mapping is achieved using an inverse distance weighted three-dimensional interpolation algorithm. With solar radiation intensity data as the boundary condition, the interior of the carriage is divided into a uniform grid of 0.5m × 0.5m × 0.5m. During the interpolation process, the power exponent is set to 2 and the search radius is 1m. Each grid node corresponds to an initial heat value, and finally an initial heat source distribution matrix with dimensions of 20×10×8 is generated. This matrix is ​​updated every 10 minutes and can accurately locate the high heat load area inside the carriage that is most affected by external solar radiation.

[0036] It is worth noting that cold bridge locations specifically refer to weak points in the insulation structure, such as door frame seals, metal connectors, and the junction of the floor and side walls. A total of 12 temperature monitoring points are set up, each with a sampling frequency of 1 Hz and a measurement accuracy of ±0.1 degrees Celsius. The sensors are waterproof and dustproof, adaptable to the high humidity environment inside the carriage. A high-precision capacitive humidity sensor is installed at the return air vent, with a sampling frequency of 1 Hz, a measurement accuracy of ±1% relative humidity, and a response time of less than 5 seconds. The thermal-humidity coupling vector includes two key dimensions: temperature gradient and water vapor partial pressure difference. The temperature gradient is obtained by calculating the temperature difference between the cold bridge location and the adjacent normal area, while the water vapor partial pressure difference is calculated based on the Magnus saturated vapor pressure formula. The dew point threshold is dynamically calculated using the Magnus saturated vapor pressure formula based on the real-time relative humidity collected at the return air vent and the real-time temperature at the cold bridge location. The calculation formula is as follows:

[0037]

[0038]

[0039] In the formula, Take 17.27, Using 237.7 degrees Celsius, the fitting error between this set of parameters and the measured water vapor pressure data is less than 0.1% within a temperature range of 0 to 60 degrees Celsius, making it suitable for the temperature and humidity environment inside cold chain transport vehicles. This represents the real-time temperature at the location of the cold bridge. The relative humidity is measured in real time at the return air vent, expressed as a percentage. As an intermediate calculation variable, it is dimensionless and represents the value at temperature. and relative humidity Under the given conditions, the logarithmic transformation relationship between the saturated vapor pressure and the current vapor pressure is established. The dew point threshold is dynamically updated according to the humidity changes inside the carriage, with an update frequency of 1 Hz, consistent with the sampling frequency of the temperature and humidity sensors.

[0040] When the temperature characteristic value in the thermal-humidity coupling vector is lower than the dew point threshold, a fuzzy evaluation algorithm is used to calculate the cold bridge condensation risk index. The algorithm uses a triangular membership function, and the coordinates of the three vertices of the function are respectively ( ,1),( ,0.5) and ( ,0), where This is the dew point threshold. The formula for calculating the condensation risk index is:

[0041]

[0042] Where R is the cold bridge condensation risk index, and T is the temperature characteristic value of the cold bridge location. The index ranges from 0 to 1, and the higher the value, the greater the condensation risk.

[0043] It is worth further explaining that the initial heat source distribution is corrected through matrix multiplication. The cold bridge condensation risk index is used as a weighted penalty term and multiplied by the corresponding element of the initial heat source distribution matrix to amplify the heat load weight of the cold bridge area, with a maximum amplification factor of 10 times. Clustering is performed using the k-means clustering algorithm, with a cluster size of 3, Euclidean distance as the distance metric, a maximum of 100 iterations, and a convergence threshold of 0.001. The initial cluster centers are determined using the k-means++ algorithm, avoiding local optima. The algorithm divides the interior space of the carriage into three clusters: a high-risk condensation zone, a temperature fluctuation zone, and a stable refrigeration zone. Each cluster corresponds to different thermal characteristics and temperature control priorities. Verified through 1000 sets of standard operating condition tests, the initial thermal state set generated by this method accurately reflects the actual thermal state inside the carriage, with the heat load assessment error controlled within ±5%, providing reliable data support for subsequent precise temperature control.

[0044] In step S12, load analysis is performed based on the initial thermal state set to obtain the cargo load fluctuation value, including:

[0045] Based on the initial thermal state set, heat conduction calculations are performed to obtain the radiative heat penetration and the overall heat leakage flux.

[0046] Numerical analysis was performed on the radiative heat penetration and the overall heat leakage flux to obtain the sensible heat change gradient.

[0047] The latent heat conversion ratio is obtained by matching the latent heat conversion based on the sensible heat change gradient.

[0048] Regression calculations are performed on the sensible heat change gradient and the latent heat conversion ratio to obtain the cargo load fluctuation value.

[0049] It should be noted that the heat conduction calculation is divided into two independent parts: the calculation of radiative heat penetration and the calculation of the overall heat leakage flux. The radiative heat penetration is obtained by solving the one-dimensional heat conduction differential equation, which is as follows:

[0050]

[0051] In the formula, The density of the insulation layer is expressed in kilograms per cubic meter. The specific heat capacity of the insulation layer is expressed in joules per kilogram (Kelvin). The thermal conductivity of the insulation layer is expressed in watts per meter Kelvin. The internal temperature of the insulation layer is expressed in Kelvin. Time is measured in seconds. Let be the spatial coordinates along the thickness of the insulation layer, with dimensions in meters. The boundary conditions for solving the equations are set as follows: [Equation missing - likely referring to the outer surface of the carriage]. The temperature at the point of contact is equal to the equivalent temperature of solar radiation, obtained by multiplying the solar radiation intensity by the absorptivity of the outer surface of the carriage and then dividing by the convective heat transfer coefficient of the outer surface; the temperature at the inner surface of the carriage is equal to the equivalent temperature of solar radiation, obtained by multiplying the solar radiation intensity by the absorptivity of the outer surface of the carriage and then dividing by the convective heat transfer coefficient of the outer surface of the carriage. The temperature at the point of contact is equal to the temperature of the air inside the carriage. The insulation layer thickness is taken as 100 mm. A fully implicit finite difference method is used for the solution, with a spatial step of 10 mm and a time step of 1 second. The initial condition is that the temperature inside the insulation layer is uniformly distributed and equal to the air temperature inside the carriage. After the numerical solution is completed, the radiative heat penetration is... Calculated using Fourier's law of thermal conductivity:

[0052]

[0053] In the formula, The area is the interior surface area of ​​the carriage, measured in square meters. This refers to the heat penetrating through radiation, measured in watts, which reflects the heat generated by solar radiation penetrating the insulation layer and entering the interior of the carriage.

[0054] The overall heat leakage flux is obtained by linearly weighting the heat leakage from the box walls and the heat leakage from the door gaps, with the weight of the heat leakage from the box walls being 70% and the weight of the heat leakage from the door gaps being 30%. This weighting ratio was statistically calibrated using test data from 500 sets of different wind speeds and temperature differences, and can accurately reflect the overall heat leakage characteristics of the carriage. (Box wall heat leakage) Calculated using the following formula:

[0055]

[0056] In the formula, The overall heat transfer coefficient of the carriage wall is taken as 0.35 Kelvin per square meter; This represents the total area of ​​the carriage walls, measured in square meters. The temperature difference between the inside and outside of the carriage is expressed in degrees Celsius.

[0057] Heat from air leakage through the gaps in the carriage doors Calculated using the following formula:

[0058]

[0059] In the formula, The density of air is taken as 1.2 kg per cubic meter; The specific heat capacity of air at constant pressure is taken as 1005 joules per kilogram Kelvin. The volumetric flow rate of air leakage through the door gap is expressed in cubic meters per second. It is calculated using Bernoulli's equation based on the door gap area and the pressure difference between the inside and outside of the carriage. The door gap area is dynamically assessed based on the aging of the door seals and the closing pressure, and is taken as 0.0005 to 0.002 square meters.

[0060] It is worth noting that the sensible heat gradient was calculated using a finite volume method numerical analysis model. This model divides the cargo area inside the carriage into a fine grid of 0.2m × 0.2m × 0.2m, containing a total of 12,500 grid cells. Radiation penetration heat and total heat leakage flux were used as heat source inputs, while cargo specific heat capacity and cargo density were used as thermal inertial drag terms. The model boundary conditions were that the heat flux density on the carriage walls equaled the total heat leakage flux, and the heat flux density on the cargo surface equaled the cargo respiration heat flux. The SIMPLE algorithm was used to solve the pressure-velocity coupling problem, with a convergence residual set to 1e-6. The rate of change of the temperature difference between adjacent grid nodes over time was calculated. The cargo specific heat capacity was obtained by matching the cargo category information from a pre-set cargo thermal property database after reading the cargo category information via RFID tags. The database contains thermal property parameters for over 200 common cold chain goods, including vegetables, fruits, meat, and dairy products. For example, the specific heat capacity of chilled meat is 3800 joules per kilogram Kelvin. The cargo density is evaluated using an onboard vision sensor combined with an image recognition algorithm. The algorithm uses the YOLOv5 target detection model to identify the cargo stack outline and calculate the ratio of stack volume to total cargo mass. The evaluation accuracy is controlled within ±20 kg per cubic meter.

[0061] It is worth further explaining that the latent heat conversion ratio is closely related to the type of goods, temperature, and humidity, and is obtained by querying a pre-set database of goods' thermal properties. The latent heat conversion ratio ranges from 0.2 to 0.6 for different goods; for example, the latent heat conversion ratio for leafy vegetables is approximately 0.5, for frozen meat it is approximately 0.2, and for dairy products it is approximately 0.35. During the matching process, a basic latent heat conversion ratio is first determined based on the goods category, and then corrected according to the current temperature and humidity inside the compartment. For every 1 degree Celsius increase in temperature, the latent heat conversion ratio increases by 0.02; for every 10% decrease in relative humidity, the latent heat conversion ratio increases by 0.05. The corrected latent heat conversion ratio maintains the same dimension as the sensible heat change gradient matrix, with each grid cell corresponding to an independent latent heat conversion ratio value, accurately reflecting the latent heat change characteristics of goods at different locations.

[0062] Subsequently, a second-order polynomial nonlinear regression algorithm was employed for regression calculation. This algorithm uses the quadratic and linear terms of the sensible heat change gradient and the latent heat conversion ratio as independent variables, and the cargo load fluctuation value as the dependent variable. The fitting calculation is completed using pre-calibrated regression coefficients. The calibration of the regression coefficients was performed in a programmable constant temperature and humidity chamber, simulating environmental temperature, solar radiation intensity, and humidity conditions in different regions and seasons across the country. Ten typical cargo types and five cargo density gradients were set up, and a total of 1000 sets of control experiments were conducted. In each experiment, the actual cargo load fluctuation value was collected as the true value using heat flow sensors placed at 10 different depths inside the cargo. The sensible heat change gradient and latent heat conversion ratio under the corresponding operating conditions were used as input features, and the optimal regression coefficients were solved using the least squares method. After calibration, the generalization ability was verified using 300 independent validation sets to ensure that the calculation error of the regression coefficients was less than ±5% across the entire operating condition range. Using the sensible heat change gradient matrix and latent heat conversion ratio matrix as input, the fitting calculation was performed cell by cell to obtain the cargo load fluctuation value for each cell, ultimately forming a three-dimensional cargo load fluctuation matrix. Each element of this matrix represents the additional cooling power required by the cargo within the corresponding grid cell due to changes in its state at the current moment, measured in watts. Multiple batches of tests have verified that the deviation between the cargo load fluctuation value calculated by this method and the actual measured value is less than ±8%, providing a crucial basis for subsequent cooling demand forecasting.

[0063] In step S13, the real-time temperature at the cold bridge and the real-time humidity at the return air vent are obtained. The deviations between the real-time temperature and the initial temperature, and between the real-time humidity and the initial humidity are calculated to obtain the temperature deviation and humidity deviation, including:

[0064] The real-time temperature at the cold bridge and the real-time humidity at the return air vent are collected by sensors.

[0065] The initial temperature deviation and the initial humidity deviation are obtained by subtracting the real-time temperature from the initial temperature and the real-time humidity from the initial humidity.

[0066] The instantaneous disturbance is obtained by calculating the state disturbance based on the pre-acquired door opening frequency;

[0067] The initial temperature deviation and the initial humidity deviation are corrected based on the instantaneous disturbance to obtain the temperature deviation and humidity deviation.

[0068] It should be noted that the real-time temperature at the cold bridge is collected by PT100 platinum resistance temperature sensors located at 12 metal connectors on the carriage wall, while the real-time humidity at the return air vent is collected by an SHT30 high-precision humidity sensor installed at the center of the return air vent. Both types of sensors have a sampling frequency of 1 Hz, consistent with the initial temperature and humidity data collection frequency, ensuring data time synchronization. The analog signals output by the sensors are converted into digital signals by a 16-bit analog-to-digital converter and transmitted to the vehicle controller for further processing. Data transmission uses the CAN bus protocol with a transmission rate of 250 kilobits per second, a transmission delay of less than 10 milliseconds, and a data packet loss rate controlled within 0.01%. Each data frame includes a CRC checksum to effectively detect transmission errors.

[0069] The initial temperature deviation is the arithmetic difference between the real-time temperature and the initial temperature, and the initial humidity deviation is the arithmetic difference between the real-time humidity and the initial humidity. The dimensions are kept consistent throughout the calculation: the temperature deviation is measured in degrees Celsius, and the humidity deviation is measured in relative humidity percentage. The initial temperature and initial humidity correspond one-to-one with the location of the corresponding sensor measurement point. Data is aligned with timestamps before difference calculation to avoid errors caused by timing misalignment. For example, when the initial temperature at the cold bridge location is -18 degrees Celsius, and the actual measured temperature is -15 degrees Celsius, the calculated initial temperature deviation is 3 degrees Celsius; when the initial relative humidity at the return air vent is 85%, and the actual measured humidity is 90%, the calculated initial humidity deviation is 5%.

[0070] It is worth noting that the state perturbation calculation is based on pre-acquired door opening frequency data, which is recorded by the carriage door control sensor. The door control sensor is magnetic induction type, installed on the carriage door frame, with a response time of less than 10 milliseconds, and can accurately record the start and end times of each door opening. The state perturbation calculation uses a pre-trained multilayer perceptron neural network model. The model input layer contains two neurons: door opening frequency and single door opening duration. The output is a door frame sealing state decay value, ranging from 0 to 1, where 0 indicates that the sealing state is intact with no decay, and 1 indicates that the sealing state is completely ineffective. The hidden layer is set to 3 layers, each containing 20 neurons, using the ReLU activation function. The output layer uses the Sigmoid activation function. The training process uses the Adam optimizer with a learning rate of 0.001, 100 training epochs, and a batch size of 32. The training dataset contains 2000 sets of measured sealing state decay data from different regions and seasons across the country in cold chain transportation. Each set of data includes a label for door opening frequency, single door opening duration, and the corresponding door frame sealing state decay value. The label value is calibrated by measuring the actual air leakage of the truck door frame using the tracer gas method after each transport mission. The ratio of the leakage to the baseline leakage when the vehicle leaves the factory is normalized to the range of 0 to 1, and this value is used as the label for the sealing state attenuation value corresponding to that transport mission. The model prediction accuracy can reach over 95%, and the output door frame sealing state attenuation value ranges from 0 to 1. The instantaneous disturbance is obtained by multiplying the door frame sealing state attenuation value by a preset limit disturbance coefficient, which is fixed at 10. This coefficient is calibrated using real vehicle test data under 300 sets of different door opening frequencies, different ambient temperatures, and different temperature differences between the inside and outside of the truck. When the attenuation value is 1, it corresponds to the maximum temperature and humidity impact under extreme door opening conditions, namely a 10-degree Celsius temperature impact and a 20% relative humidity impact. The aforementioned temperature limit disturbance coefficient of 10 degrees Celsius and humidity limit disturbance coefficient of 20% relative humidity are derived from real vehicle tests in which the truck door is fully opened for 30 seconds under extreme conditions. This experiment simulated the worst-case scenario of door opening during transportation, where the temperature and humidity difference between the inside and outside of the carriage reaches its maximum, and the temperature and humidity fluctuation caused by door opening also reaches its maximum. Therefore, the temperature change at the cold bridge location and the humidity change at the return air vent measured in the experiment are the maximum temperature and humidity shock values ​​under extreme door opening conditions. After conservative rounding, these values ​​are used as the temperature limit disturbance coefficient of 10 degrees Celsius and the humidity limit disturbance coefficient of 20%, respectively. At the same time, different disturbance correction coefficients were set for different types of goods: 1.2 for leafy vegetables, 0.8 for frozen meat, and 1.0 for dairy products, which can more accurately reflect the degree of impact of door opening disturbance on different goods.

[0071] It is worth further explaining that the corrections for initial temperature and humidity deviations are achieved using a weighted summation method. The correction weight for temperature deviation is 0.6, and the correction weight for humidity deviation is 0.4. This weighting ratio was statistically calibrated based on 500 sets of experimental data. A higher weight was assigned because temperature changes have a greater impact on cargo quality. The instantaneous disturbance is multiplied by its corresponding weight and cargo correction coefficient, and then summed to the corresponding initial deviation value to obtain the final temperature and humidity deviations. Verified through multiple batches of real-vehicle tests, this correction method effectively eliminates temperature and humidity measurement errors caused by door opening operations, achieving a deviation correction accuracy of ±0.2 degrees Celsius and ±2% relative humidity. The corrected temperature and humidity deviations are stored in real-time in the vehicle controller's cache, providing core input data for subsequent threshold judgments and cooling demand predictions.

[0072] In step S14, a multi-factor fusion prediction is performed based on the temperature deviation, the humidity deviation, the cargo load fluctuation value, and the pre-acquired pre-cooling margin to obtain an estimated cooling demand, including:

[0073] Obtain the compressor's operating parameters, perform thermodynamic calculations based on the operating parameters, and obtain the precooling margin;

[0074] If the temperature deviation exceeds a preset temperature threshold, or the humidity deviation exceeds a preset humidity threshold, the cargo load fluctuation value, the temperature deviation, the humidity deviation, and the precooling margin are input into a preset random forest model to obtain a cooling demand estimate.

[0075] If neither the temperature deviation nor the humidity deviation exceeds the threshold, the current cooling state remains unchanged.

[0076] It should be noted that the current cooling capacity is obtained based on the thermodynamic calculation formula of the vapor compression refrigeration cycle. The calculation formula is as follows:

[0077]

[0078] in This is the current cooling capacity, measured in watts. The compressor's volumetric efficiency was calibrated through 300 bench tests under different speed and pressure ratio conditions, with values ​​ranging from 0.7 to 0.95. During real-time operation, the specific volumetric efficiency value under the current operating condition was obtained by consulting a pre-calibrated speed-pressure ratio MAP table. This MAP table covers the entire compressor speed range of 0 to 4000 rpm and the entire pressure ratio range of 1.5 to 8. After 30 minutes of stable operation at each operating point, the volumetric efficiency was calculated by measuring the actual and theoretical intake volumes of the compressor. A total of 300 sets of valid data were collected. For operating points not directly calibrated, linear interpolation was used for calculation, and the lookup accuracy was controlled within ±0.02. The real-time speed of the compressor is collected by the compressor's built-in Hall effect speed sensor, and the unit of measurement is revolutions per minute. This is the theoretical displacement of the compressor, a fixed value specified by the manufacturer when the equipment is manufactured, and the unit of measurement is cubic meters per revolution. The density of the refrigerant in the suction state is obtained by consulting the international standard physical property database of refrigerant R404A based on the suction pressure and suction temperature. The unit of measurement is kilograms per cubic meter. The refrigeration capacity per unit mass is calculated based on the isobaric heat absorption of the refrigerant in the evaporator using the suction and discharge pressures. Specifically, the calculation process involves consulting the international standard property database for refrigerant R404A using the suction pressure and temperature to obtain the enthalpy of the refrigerant at the evaporator outlet; and then consulting the enthalpy of the refrigerant at the condenser outlet using the discharge pressure and condensing temperature. The throttling process is an isenthalpic process, and since the outlet of the throttling valve is directly connected to the evaporator inlet, the enthalpy of the refrigerant at the evaporator inlet is equal to the enthalpy of the refrigerant at the condenser outlet. The refrigeration capacity per unit mass is the enthalpy difference between the evaporator inlet and outlet, measured in joules per kilogram.

[0079] The maximum cooling capacity is the compressor's maximum cooling output capability at rated speed, rated suction pressure, and rated discharge pressure, a fixed value calibrated at the factory. The evaporator heat transfer efficiency loss coefficient reflects the attenuation effect of frost on the evaporator's heat transfer capacity. It is obtained through bench testing. The calibration experiment was conducted in a programmable constant temperature and humidity chamber, using an evaporator test bench identical to the actual vehicle, with the ambient temperature controlled at 25 degrees Celsius and relative humidity at 60%, simulating normal operating conditions of the refrigeration system. The frost thickness was controlled by adjusting the frost time on the evaporator surface, and an ultrasonic thickness gauge was used to measure the frost thickness in real time with a measurement accuracy of ±0.1 mm. After stabilizing for 20 minutes at each frost thickness operating point, the actual heat transfer and theoretical heat transfer of the evaporator were measured, and the corresponding heat transfer efficiency loss coefficient was calculated. A total of 20 sets of test data with different frost thicknesses were collected. The relationship between the loss coefficient and frost thickness was obtained through linear fitting, with a coefficient of determination R0. 2 A value greater than 0.98 ensures calibration accuracy. The coefficient is 1 when the frost thickness is 0 mm, and decreases by 0.03 for every 1 mm increase in thickness, down to a minimum of 0.5.

[0080] The pre-cooling margin is calculated by subtracting the ratio of the current cooling capacity to the maximum cooling capacity from 1, and then multiplying by the evaporator heat exchange efficiency loss coefficient. The pre-cooling margin ranges from 0 to 1, with a larger value indicating stronger remaining cooling capacity of the refrigeration system. When the compressor's rated speed is 4000 rpm, the current speed is 3400 rpm, the volumetric efficiency is 0.9, and the evaporator frost thickness is 5 mm, the calculated current cooling capacity is 85% of the maximum cooling capacity, and the pre-cooling margin is approximately 0.13. At this point, the system will automatically trigger a defrost warning, prompting the execution of the defrost procedure.

[0081] It is worth noting that the preset temperature threshold is fixed at 2 degrees Celsius, and the humidity threshold is fixed at 8%. These two thresholds were calibrated using test data from 1000 sets of standard cold chain transportation conditions. Test results show that when the temperature deviation exceeds 2 degrees Celsius or the humidity deviation exceeds 8%, the risk of product quality deterioration increases significantly, with leafy vegetables experiencing a spoilage rate increase of over 15% and frozen meat experiencing a juice loss rate increase of over 8%. However, within the thresholds, product quality remains stable, with a quality deterioration rate of less than 5% within 72 hours. The threshold comparison process is performed on a sampling cycle basis, with the comparison frequency consistent with the temperature and humidity data acquisition frequency of 1 Hz. Whenever either the temperature deviation is greater than 2 degrees Celsius or the humidity deviation is greater than 8%, the system automatically generates a cold load prediction trigger command, activating the pre-deployed random forest regression model.

[0082] It is worth further explaining that the random forest model contains 100 decision trees, each with a maximum depth of 10 and a minimum number of splits of 5, using the Gini impurity criterion for node splitting. The model's training dataset contains 5000 sets of measured cold chain transportation data under different operating conditions. Each set of data includes cargo load fluctuations, temperature deviations, humidity deviations, pre-cooling margins, and the corresponding actual cooling demand. The data covers transportation scenarios in different regions, seasons, and cargo types across the country. Five-fold cross-validation was used during training to avoid overfitting, and the final model's prediction accuracy reached over 92%. Feature importance analysis showed that temperature deviation accounted for 35% of the feature importance, pre-cooling margins for 28%, cargo load fluctuations for 22%, and humidity deviations for 15%, indicating that temperature changes and the system's remaining cooling capacity have the most significant impact on cooling demand.

[0083] If both temperature and humidity deviations do not exceed their respective thresholds, the system determines that the thermal state inside the carriage is stable, maintains the current refrigeration parameters, and continues to monitor temperature and humidity changes in real time. When the pre-cooling margin is below 10%, the system automatically activates the energy-saving protection mechanism, prioritizing the reduction of cooling supply to non-critical areas to ensure the cooling demand of cold bridge locations and core cargo areas, extending the continuous operation time of the refrigeration system. If any deviation exceeds the threshold, the system combines cargo load fluctuation, temperature deviation, humidity deviation, and pre-cooling margin into a multi-dimensional feature vector and inputs it into a random forest model. By arithmetically averaging the prediction results of the leaf nodes of all decision trees, the final estimated cooling demand is obtained, with the unit of measurement being watts. The estimated cooling demand is updated every minute, responding in real time to dynamic changes in the heat load inside the carriage. For example, when the input cargo load fluctuation is 500 watts, the temperature deviation is 3 degrees Celsius, the humidity deviation is 5%, and the pre-cooling margin is 0.15, the model outputs an estimated cooling demand of 5200 watts. After multiple rounds of transportation tests, the prediction error of this cooling demand forecasting method is controlled within ±10%, and the response time is less than 1 second. It can accurately capture the dynamic changes in the heat load inside the carriage.

[0084] In step S15, the initial compression speed and initial fan air volume are obtained by mapping based on the estimated cooling demand. A smoothing slope is calculated based on the initial compression speed to obtain the speed ramp slope. The speed ramp slope is integrated to obtain the target compression speed. The target fan air volume is obtained by fitting the initial fan air volume. This includes:

[0085] The estimated cooling demand is indexed, and the initial compression speed and initial fan air volume are obtained by querying the preset cooling capacity mapping table.

[0086] Obtain the current compression speed and calculate the speed difference between the initial compression speed and the current compression speed;

[0087] If the speed difference is less than the preset speed difference threshold, the initial compression speed is taken as the target compression speed, and the preset speed ramp slope is taken as the speed ramp slope.

[0088] If the speed difference is greater than the preset speed difference threshold, the speed difference is subjected to first-order low-pass filtering. The dynamic speed rise slope is calculated based on the filtered speed difference. The dynamic speed rise slope is then integrated cycle by cycle to obtain the target compression speed.

[0089] The initial fan air volume is converted by pulse width modulation to obtain the target fan air volume.

[0090] It should be noted that the cooling capacity mapping table is a two-dimensional data structure using estimated cooling demand and the temperature difference between the inside and outside of the vehicle compartment as coordinate axes. The data in the table has been calibrated through bench tests under 500 sets of different cooling demand and ambient temperature conditions, covering a cooling demand range of 0 to 10,000 watts and a temperature difference range of -40 to 40 degrees Celsius between the inside and outside of the vehicle compartment. The cooling capacity mapping table is calibrated in a programmable constant temperature and humidity chamber. During the calibration process, a refrigeration system test bench identical to that of the actual vehicle is built. The ambient temperature outside the vehicle compartment is precisely controlled through the chamber, and electric heating rods are used to simulate different levels of heat load inside the vehicle compartment. After each operating point has been running stably for 30 minutes, the average temperature inside the vehicle compartment, compressor speed, and fan airflow data are continuously collected for 10 minutes. The measurements are repeated 3 times, and the arithmetic mean is used as the basic data for the mapping table. For operating points that are not directly measured, the corresponding parameter values ​​are calculated using a bilinear interpolation algorithm. The cooling capacity mapping table uses a quarterly update mechanism, adjusting parameter weights based on ambient temperature and solar radiation intensity in different seasons. In summer, the proportion of fan airflow is increased, while in winter, the compressor base speed is reduced, further reducing system energy consumption by approximately 8%. The query process uses the estimated cooling demand and the real-time temperature difference between the inside and outside of the vehicle compartment as a joint index, employing a bilinear interpolation algorithm to address and calculate the corresponding initial compressor speed and initial fan airflow in the mapping table. When the estimated cooling demand is 5200 watts and the temperature difference between the inside and outside of the vehicle compartment is 15 degrees Celsius, the queried initial compressor speed is 3500 rpm, and the initial fan airflow is 1200 cubic meters per hour.

[0091] Subsequently, the current compression speed is acquired by the compressor's built-in Hall effect speed sensor at a sampling frequency of 1 Hz and a measurement accuracy of ±10 rpm. The speed difference is the arithmetic difference between the initial compression speed and the current compression speed. The preset speed difference threshold is fixed at 500 rpm. This threshold has been calibrated using real-vehicle test data under 300 sets of different operating conditions to balance the refrigeration system's response speed and operational stability, avoiding increased energy consumption and equipment wear caused by frequent compressor start-stop. If the speed difference is less than or equal to 500 rpm, the initial compression speed is directly determined as the target compression speed, and the preset speed ramp slope is directly used as the speed ramp slope, uniformly set at 30 rpm. This slope has been calibrated using bench tests under 100 sets of different load conditions to ensure a smooth increase in compressor speed, with the operating current not exceeding 1.1 times the rated current, while preventing compressor mechanical vibration from exceeding 0.5g, thus balancing equipment operational reliability and cooling response speed.

[0092] It is worth noting that if the speed difference is greater than 500 rpm, a first-order low-pass filter is applied to the speed difference, with a filter time constant set to 2 seconds, to obtain a smoothed speed difference. The dynamic speed ramp-up slope is then calculated based on the smoothed speed difference, using the formula: smoothed speed difference divided by the preset maximum adjustment time of 30 seconds. Subsequently, the dynamic speed ramp-up slope is integrated cycle by cycle, with each adjustment cycle being 1 second. The single-cycle speed increment is superimposed on the current compression speed to obtain the target compression speed at each moment, until the initial compression speed is reached. When the initial compression speed is 3500 rpm and the current compression speed is 2000 rpm, after the above smoothing and integration processing, the compression speed is increased to 3500 rpm within 30 seconds at a ramp-up slope of 50 rpm. During the increase, if the cooling demand is met ahead of schedule, the integration can be terminated early, for example, stopping acceleration when the target compression speed reaches 3200 rpm.

[0093] The initial fan airflow is converted using pulse width modulation (PWM) to obtain the target fan airflow. The relationship between fan airflow and motor drive duty cycle is determined using a third-order polynomial fitting algorithm. The fitting coefficients are calibrated using 100 sets of airflow measurement data at different duty cycles, and the coefficient of determination of the fitting model reaches over 0.99, accurately realizing the conversion between airflow and duty cycle. The PWM frequency is fixed at 20 kHz, which is outside the range of human hearing and avoids generating operating noise. When the initial fan airflow is 1200 cubic meters per hour, the converted fan duty cycle is 75%, and the corresponding target fan airflow is consistent with the initial fan airflow. Multiple batches of tests have verified that the control accuracy of this parameter mapping method can reach ±50 revolutions per minute and ±50 cubic meters per hour, meeting the precise adjustment requirements of the refrigeration system.

[0094] In step S16, a frequency converter command is generated based on the target compression speed and the speed ramp rate to obtain a first adjustment command, and an air duct command is generated based on the target fan air volume to obtain a second adjustment command, including:

[0095] The compressor's real-time operating current is collected. If the target compression speed is greater than the current compression speed and the real-time operating current is less than a preset current threshold, a first adjustment command is generated based on the speed ramp rate.

[0096] If the real-time operating current is greater than the current threshold, the current compression speed remains unchanged;

[0097] Obtain the current fan air volume, calculate the air volume deviation between the current fan air volume and the target fan air volume, correct the duct deviation based on the air volume deviation, and obtain the second adjustment command.

[0098] It is worth noting that the compressor's real-time operating current is collected by an open-circuit current transformer installed in the three-phase power supply circuit of the stator coil. The sampling frequency is 1 Hz, and the measurement accuracy is ±0.1 amperes, allowing for real-time monitoring of the compressor's load status. The preset current threshold is fixed at 15 amperes, which is 1.2 times the compressor's rated operating current of 12.5 amperes. This setting has been verified through 200 bench tests under different ambient temperatures and load conditions, maximizing the utilization of its cooling capacity while protecting the compressor from overload damage. The speed comparison process is performed cycle by cycle. If the target compression speed is greater than the current compression speed and the real-time operating current is less than 15 amperes, the system determines that it is safe to increase the compressor speed and generates the first adjustment command according to the speed ramp rate. The first adjustment command contains two core parameters: the target compression speed and the speed ramp rate. It is transmitted using the CAN 2.0B protocol, identified by 0x123, and has a data length of 8 bytes. The first 4 bytes store the target compression speed, and the last 4 bytes store the speed ramp rate. Each data frame includes a 16-bit CRC checksum to effectively detect transmission errors, with a transmission delay of less than 10 milliseconds. For example, when the target compression speed is 4000 rpm, the current compression speed is 3000 rpm, and the real-time operating current is 12 amps, a first adjustment command with a ramp rate of 30 rpm per second is generated, driving the compressor to smoothly increase to the target speed within approximately 33 seconds.

[0099] If the real-time operating current is greater than or equal to 15 amps, the system determines that the compressor is overloaded and immediately maintains the current compressor speed to prevent the windings from overheating and burning out. Simultaneously, it automatically records the occurrence time, duration, corresponding ambient temperature, and cabin temperature and humidity data of the overload event, and uploads this data to the cloud monitoring platform for subsequent fault analysis and equipment maintenance. When the operating current drops below 12 amps for 30 seconds, the system automatically deactivates the overload protection and restores normal speed regulation. The system has a tiered warning mechanism for different levels of overload. When the current reaches 14 amps, a yellow warning message is displayed on the vehicle screen; when the current reaches 15 amps, a red warning message is displayed and a buzzer alarm is triggered. This protection logic has been verified through long-term real-vehicle operation and can effectively extend the compressor's lifespan and reduce the equipment failure rate by approximately 15%.

[0100] Subsequently, the current fan airflow is measured by a hot-wire anemometer installed at the center of the fan outlet, with a sampling frequency of 0.5 Hz and a measurement accuracy of ±10 cubic meters per hour. This sensor accurately measures the actual output airflow regardless of air temperature and humidity. The airflow deviation is the arithmetic difference between the target fan airflow and the current fan airflow. If the deviation is not zero, the system calculates the duct valve opening based on the airflow deviation and the return air pressure. The return air pressure is collected by a differential pressure sensor installed inside the return air filter, with a measurement accuracy of ±1 Pascal, accurately reflecting changes in the duct system's resistance. The duct valve opening is calculated using a feedforward neural network model. The model contains two hidden layers and was trained based on 1000 sets of measured data under different airflow and pressure conditions, with a prediction error controlled within ±5%.

[0101] Finally, a second adjustment command is generated based on the calculated duct valve opening. This command includes two parameters: the fan duty cycle and the duct valve opening. It is transmitted using the CAN 2.0B protocol, with an identifier of 0x124 and a data length of 8 bytes. If the airflow deviation is zero, the system maintains the current fan and duct parameters unchanged. In another optional implementation, when the compressor current approaches a threshold, the system prioritizes reducing the fan airflow in non-critical areas to reduce the overall system load and prevent compressor overload shutdown. Multiple batches of cross-seasonal real-vehicle transportation tests have verified that this command generation method can achieve coordinated adjustment of the compressor and fan, with a cooling capacity adjustment response time of less than 5 seconds, a temperature control accuracy of ±0.5 degrees Celsius inside the vehicle, and an overall system energy consumption reduction of approximately 12% compared to traditional single-parameter feedback control methods.

[0102] In step S17, the first adjustment command and the second adjustment command are parsed to obtain the start-stop delay duration. Based on the start-stop delay duration, the coordinated matching of cooling supply and cooling demand is completed, including:

[0103] Analyze the execution timestamps of the first and second adjustment commands to calculate the start-stop delay duration;

[0104] Drive the compressor to reach the target compression speed according to the start-stop delay time to generate the initial cooling capacity supply;

[0105] The initial cooling capacity is queried from a preset pulse width modulation mapping table to determine the corresponding fan duty cycle. The fan is driven according to the fan duty cycle to construct a cooling capacity spatial distribution field.

[0106] Based on the spatial distribution of cold energy, supplemental cold energy is supplied to the local cold bridge locations.

[0107] It should be noted that both the first and second adjustment commands include precise execution timestamps with an accuracy of 1 millisecond, generated by the high-precision real-time clock of the vehicle controller, with a daily clock error of less than 1 second. The start-stop delay is the arithmetic difference between the execution timestamps of the first and second adjustment commands. The execution timestamp of the first adjustment command is the moment the compressor start command is issued, and the execution timestamp of the second adjustment command is the moment the fan start command is issued. When the first adjustment command requires a response within 2 seconds and the mechanical action of the second adjustment command requires 5 seconds to complete, the algorithm outputs a start-stop delay of 3 seconds. This delay setting ensures that the compressor starts first to generate cooling capacity, and the fan starts later to deliver cooling capacity, avoiding energy waste caused by fan idling, and preventing premature frosting of the evaporator.

[0108] It is worth noting that after the compressor starts according to the start-stop delay time, the initial cooling capacity is generated by mapping the collected exhaust temperature and condensing pressure, with the unit being watts. The exhaust temperature is collected by a PT100 temperature sensor installed at the compressor exhaust port, with a measurement accuracy of ±0.5 degrees Celsius; the condensing pressure is collected by a piezoresistive pressure sensor installed at the condenser outlet, with a measurement accuracy of ±0.1 bar. When the initial cooling capacity is 5000 watts, the system automatically records the compressor operating parameters at that moment, including speed, current, and exhaust pressure, and stores them in a local database for subsequent analysis. During compressor startup, the system monitors the operating current in real time. If abnormal current fluctuations occur, the system immediately suspends speed increase and triggers a fault warning to ensure equipment operation safety.

[0109] Based on the initial cooling capacity supply, a preset pulse width modulation (PWM) mapping table is consulted to extract the corresponding fan duty cycle. This mapping table was calibrated during bench testing. During calibration, the compressor's operating parameters were fixed to ensure a stable cooling capacity supply. The fan's PWM duty cycle was gradually adjusted until the average temperature inside the compartment stabilized at the set refrigeration temperature. The optimal fan duty cycle corresponding to this cooling capacity supply was recorded. The calibration covered a cooling capacity range of 0 to 10000 watts, with each cooling capacity gradient representing 200 watts. Each operating point was measured three times, and the average value was taken to establish a one-to-one correspondence between the cooling capacity supply and the fan duty cycle. After calibration, the airflow control accuracy of the mapping table reached ±30 cubic meters per hour. The fan operated according to this duty cycle, while the evaporator airflow angle was adjusted to a preset angle. Temperature gradient data in three-dimensional space was collected through a multi-point temperature sensor array at the top and bottom of the compartment. The temperature sensor array comprises 32 measuring points, evenly distributed across eight height levels within the carriage, with four measuring points per level. The sampling frequency is 0.2 Hz, providing complete coverage of all spatial areas within the carriage. Kriging interpolation is used to spatially smooth the discrete temperature data. The algorithm employs a Gaussian semi-variogram model, with parameters calibrated using 500 sets of experimental data on temperature distribution within the carriage. The nugget value is 0.02, the pedestal value is 0.5, and the range is 1.5 meters. During interpolation, the search neighborhood is set to the eight nearest measuring points, and ordinary Kriging is used for calculation. The average absolute error of the interpolation result is less than 0.2 degrees Celsius, accurately reproducing the spatial distribution characteristics of cold energy within the carriage. A cold energy spatial distribution field with a resolution of 0.5 m × 0.5 m × 0.5 m is constructed to visually reflect the cold energy density distribution at different locations within the carriage.

[0110] Finally, the actual cold density at the corresponding coordinates in the cold energy spatial distribution field of the cold bridge location is extracted, and a Long Short-Term Memory (LSTM) network model is used to calculate the predicted cold energy demand at the cold bridge location. The model's input features include ambient temperature, vehicle speed, and historical temperature data from the previous 10 time steps at the cold bridge location. Each time step corresponds to a 1-second sampling period, i.e., a 10-second historical window, and the sampling frequency is consistent with the acquisition frequency (1 Hz) of the cold bridge temperature sensor. The training dataset contains 3000 sets of measured data on cold bridge temperature and cold energy demand under different seasons and driving conditions, covering major cold chain transportation routes nationwide. The model consists of two LSTM layers, each with 64 hidden units, using the tanh activation function, and an output layer with one linear neuron. The training process uses the Adam optimizer with a learning rate of 0.001, a batch size of 32, and 50 training epochs, employing an early stopping mechanism to prevent overfitting. After 5-fold cross-validation, the model's average prediction error is controlled within ±50 watts, and the prediction accuracy reaches 93%. When the actual cooling supply is 600 watts and the predicted cooling demand is 800 watts, a cooling gap of 200 watts is calculated, generating a deflection angle command for the local deflector. The adjustment range of the local deflector is 0 to 90 degrees, with an adjustment accuracy of ±1 degree. Each 1-degree deflection increases the cooling supply at the cold bridge location by approximately 10 watts, and a 20-degree deflection is sufficient to fill the cooling gap. If the temperature at the cold bridge location still does not meet the target, the system will feed back the cooling gap to the cooling forecasting stage to recalculate the overall cooling demand. Multiple batches of cross-seasonal real-vehicle transportation tests have verified that this method can control the temperature deviation at the cold bridge location within ±0.3 degrees Celsius, extending the shelf life of goods by approximately 20% compared to traditional control methods.

[0111] In summary, this invention discloses an IoT-based method for monitoring temperature and humidity in cold chain logistics. The method includes collecting initial temperature at the cold bridge and initial humidity at the return air vent; fusing parameters based on the initial temperature and initial humidity to obtain an initial thermal state set; performing load analysis based on the initial thermal state set to obtain cargo load fluctuation values; acquiring real-time temperature at the cold bridge and real-time humidity at the return air vent; calculating the deviations between the real-time temperature and the initial temperature, and between the real-time humidity and the initial humidity, respectively, to obtain temperature deviation and humidity deviation; and performing multi-factor fusion prediction based on the temperature deviation, humidity deviation, cargo load fluctuation values, and pre-acquired pre-cooling margin to obtain the required cooling capacity. The process involves: estimating the required cooling capacity; mapping the estimated cooling capacity to obtain the initial compression speed and initial fan airflow; calculating the smoothing slope based on the initial compression speed to obtain the speed ramp slope; integrating the speed ramp slope to obtain the target compression speed; fitting the initial fan airflow to obtain the target fan airflow; generating a frequency converter command based on the target compression speed and the speed ramp slope to obtain a first adjustment command; generating a duct command based on the target fan airflow to obtain a second adjustment command; parsing the first and second adjustment commands to obtain the start-stop delay duration; and achieving coordinated matching between cooling capacity supply and demand based on the start-stop delay duration. This invention achieves precise spatiotemporal matching of cooling capacity supply and demand through multi-source thermal data fusion and industrial data processing technology, effectively solving the problem of localized temperature anomalies due to cold bridges, ensuring the quality of cold chain goods, and reducing the energy consumption of the refrigeration system.

[0112] The second embodiment of the present invention provides a cold chain logistics temperature and humidity monitoring system based on the Internet of Things, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of the method described above.

[0113] It should be noted that the IoT-based cold chain logistics temperature and humidity monitoring system provided in this embodiment of the invention is used to execute all the process steps of the IoT-based cold chain logistics temperature and humidity monitoring method in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0114] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0115] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for monitoring temperature and humidity in cold chain logistics based on the Internet of Things, characterized in that, include: The initial temperature at the cold bridge and the initial humidity at the return air inlet are collected, and the parameters are fused based on the initial temperature and the initial humidity to obtain the initial thermal state set. Load analysis is performed based on the initial thermal state set to obtain the cargo load fluctuation value; The real-time temperature at the cold bridge and the real-time humidity at the return air vent are obtained. The deviations between the real-time temperature and the initial temperature, and between the real-time humidity and the initial humidity are calculated to obtain the temperature deviation and humidity deviation. Based on the temperature deviation, humidity deviation, cargo load fluctuation value, and pre-acquired pre-cooling margin, a multi-factor fusion prediction is performed to obtain an estimated cooling demand. Based on the estimated cooling demand, the initial compression speed and initial fan air volume are obtained by mapping. The initial compression speed is then used to calculate the smoothing slope to obtain the speed ramp slope. The speed ramp slope is then integrated to obtain the target compression speed. Finally, the initial fan air volume is fitted to obtain the target fan air volume. A first adjustment command is generated based on the target compression speed and the speed ramp rate, and a second adjustment command is generated based on the target fan air volume. The first and second adjustment commands are parsed to obtain the start-stop delay duration, and the cooling supply and cooling demand are matched in a coordinated manner based on the start-stop delay duration.

2. The method for monitoring temperature and humidity in cold chain logistics based on the Internet of Things according to claim 1, characterized in that, The initial temperature at the cold bridge and the initial humidity at the return air vent are collected. Based on these initial temperatures and humidity values, parameters are fused to obtain an initial thermal state set, including: Acquire the solar radiation intensity and initial heat source data of the carriage, and perform spatial mapping on the initial heat source data based on the solar radiation intensity to obtain the initial heat source distribution; Based on the initial heat source distribution, the initial temperature at the location of the cold bridge and the initial humidity at the return air vent are collected, and the initial temperature and initial humidity are mapped to obtain a heat-humidity coupling vector. Based on the thermal-humidity coupling vector, the initial heat source distribution is corrected to obtain the corrected heat load distribution; The modified heat load distribution is clustered to obtain the initial thermal state set.

3. The method for monitoring temperature and humidity in cold chain logistics based on the Internet of Things according to claim 1, characterized in that, The step of performing load analysis based on the initial thermal state set to obtain the cargo load fluctuation value includes: Based on the initial thermal state set, heat conduction calculations are performed to obtain the radiative heat penetration and the overall heat leakage flux. Numerical analysis was performed on the radiative heat penetration and the overall heat leakage flux to obtain the sensible heat change gradient. The latent heat conversion ratio is obtained by matching the latent heat conversion based on the sensible heat change gradient. Regression calculations are performed on the sensible heat change gradient and the latent heat conversion ratio to obtain the cargo load fluctuation value.

4. The method for monitoring temperature and humidity in cold chain logistics based on the Internet of Things according to claim 1, characterized in that, The process of obtaining the real-time temperature at the cold bridge and the real-time humidity at the return air vent, and calculating the deviations between the real-time temperature and the initial temperature, and between the real-time humidity and the initial humidity, respectively, to obtain the temperature deviation and humidity deviation, includes: The real-time temperature at the cold bridge and the real-time humidity at the return air vent are collected by sensors. The initial temperature deviation and the initial humidity deviation are obtained by subtracting the real-time temperature from the initial temperature and the real-time humidity from the initial humidity. The instantaneous disturbance is obtained by calculating the state disturbance based on the pre-acquired door opening frequency; The initial temperature deviation and the initial humidity deviation are corrected based on the instantaneous disturbance to obtain the temperature deviation and humidity deviation.

5. The method for monitoring temperature and humidity in cold chain logistics based on the Internet of Things according to claim 1, characterized in that, The step of performing multi-factor fusion prediction based on the temperature deviation, humidity deviation, cargo load fluctuation value, and pre-acquired pre-cooling margin to obtain a cooling demand estimate includes: Obtain the compressor's operating parameters, perform thermodynamic calculations based on the operating parameters, and obtain the precooling margin; If the temperature deviation exceeds a preset temperature threshold, or the humidity deviation exceeds a preset humidity threshold, the cargo load fluctuation value, the temperature deviation, the humidity deviation, and the precooling margin are input into a preset random forest model to obtain a cooling demand estimate. If neither the temperature deviation nor the humidity deviation exceeds the threshold, the current cooling state remains unchanged.

6. The method for monitoring temperature and humidity in cold chain logistics based on the Internet of Things according to claim 1, characterized in that, The process involves mapping the estimated cooling demand to obtain initial compression speed and initial fan airflow, calculating a smoothing slope based on the initial compression speed to obtain a speed ramp slope, integrating the speed ramp slope to obtain the target compression speed, and fitting the initial fan airflow to obtain the target fan airflow. This includes: The estimated cooling demand is indexed, and the initial compression speed and initial fan air volume are obtained by querying the preset cooling capacity mapping table. Obtain the current compression speed and calculate the speed difference between the initial compression speed and the current compression speed; If the speed difference is less than the preset speed difference threshold, the initial compression speed is taken as the target compression speed, and the preset speed ramp slope is taken as the speed ramp slope. If the speed difference is greater than the preset speed difference threshold, the speed difference is subjected to first-order low-pass filtering. The dynamic speed rise slope is calculated based on the filtered speed difference. The dynamic speed rise slope is then integrated cycle by cycle to obtain the target compression speed. The initial fan air volume is converted by pulse width modulation to obtain the target fan air volume.

7. The method for monitoring temperature and humidity in cold chain logistics based on the Internet of Things according to claim 6, characterized in that, The process of generating a frequency conversion command based on the target compression speed and the speed rise slope to obtain a first adjustment command, and generating a duct command based on the target fan air volume to obtain a second adjustment command, includes: The compressor's real-time operating current is collected. If the target compression speed is greater than the current compression speed and the real-time operating current is less than a preset current threshold, a first adjustment command is generated based on the speed ramp rate. If the real-time operating current is greater than the current threshold, the current compression speed remains unchanged; Obtain the current fan air volume, calculate the air volume deviation between the current fan air volume and the target fan air volume, correct the duct deviation based on the air volume deviation, and obtain the second adjustment command.

8. The method for monitoring temperature and humidity in cold chain logistics based on the Internet of Things according to claim 1, characterized in that, The step of parsing the first and second adjustment commands to obtain the start-stop delay duration, and then performing coordinated matching of cooling supply and cooling demand based on the start-stop delay duration, includes: Analyze the execution timestamps of the first and second adjustment commands to calculate the start-stop delay duration; Drive the compressor to reach the target compression speed according to the start-stop delay time to generate the initial cooling capacity supply; The initial cooling capacity is queried from a preset pulse width modulation mapping table to determine the corresponding fan duty cycle. The fan is driven according to the fan duty cycle to construct a cooling capacity spatial distribution field. Based on the spatial distribution of cold energy, supplemental cold energy is supplied to the local cold bridge locations.

9. A cold chain logistics temperature and humidity monitoring system based on the Internet of Things, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that, When the processor executes the program, it implements the steps of the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method described in any one of claims 1 to 8.

Images