Intelligent temperature control method in medicine logistics cold-chain transportation

Abstract

The invention belongs to the technical field of intelligent temperature control, and particularly relates to an intelligent temperature control method in medicine logistics cold-chain transportation. Firstly, medicine temperature requirements and spatial distribution data are obtained, temperature control areas are divided, and temperature data are collected; processing the data, and constructing a dynamic temperature influence factor model to generate an inter-region influence factor; calculating a basic deviation and introducing an influence factor for correction; then dynamically adjusting the PID controller coefficient of each region according to the correction deviation; constructing a temperature prediction model by using the improved long short-term memory network; and finally, the feedforward control compensation amount is calculated in combination with a prediction result, and temperature compound control is achieved by fusing the feedforward control compensation amount with the PID feedback control amount. The problems that traditional PID control cannot respond to complex temperature changes in time, and inter-region heat conduction regulation and control and feedforward control are lacked are solved, and the multi-region cooperative control capability and the system response energy efficiency are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent temperature control, and in particular relates to a method for intelligent temperature control in cold chain transportation of pharmaceutical logistics. Background Art

[0002] With the development of the modern pharmaceutical industry, cold chain logistics plays a vital role in ensuring the quality and efficacy of drugs. Especially in the transportation of highly temperature-sensitive drugs such as vaccines, biologics, and high-end injections, precise control of the temperature in the carriage has become a key factor in ensuring drug safety. Currently, the commonly used cold chain temperature control methods mostly adopt traditional PID control strategies. Although they have certain adjustment capabilities, they are often unable to respond in a timely manner to complex temperature changes in multiple regions and multiple disturbances. Especially under conditions of uneven spatial distribution of drugs, complex carriage structures, and drastic fluctuations in the external environment, local overheating or overcooling is prone to occur, affecting the quality of drugs. In addition, the existing technology lacks a detailed modeling and control mechanism for the heat conduction effect between regions, and fails to effectively combine future temperature trends for feedforward control, resulting in a delayed overall response of the system and low energy efficiency. Summary of the Invention

[0003] In response to the technical problems existing in the above-mentioned background technology, the present invention proposes an intelligent temperature control method in cold chain transportation of pharmaceutical logistics.

[0004] In order to achieve the above object, the technical solution adopted by the present invention includes the following steps:

[0005] S1. Obtain the temperature requirements and spatial distribution data of medicines in the cold chain compartment, divide them into independent temperature control areas, deploy sensors, and collect temperature data within the areas;

[0006] S2. Suppress and normalize the collected temperature data, build a dynamic temperature impact factor model, and generate inter-regional impact factors;

[0007] S3. Calculate the basic deviation based on the target temperature and the real-time temperature average of each temperature control area, introduce the inter-area influencing factor to couple and correct the basic deviation to form a corrected deviation;

[0008] S4. Dynamically adjust the proportional coefficient, integral coefficient, and differential coefficient of the PID controller in each region according to the corrected deviation. The specific implementation is as follows:

[0009] For each temperature control area i, the regional coupling compensation rule is introduced for dynamic adjustment. First, the coupling strength weight is calculated as follows:

[0010] Then for the proportional coefficient, the adjustment formula is: K′ P,i =K P,i ·(1+tanh(γ·|e′i (t)|))·(1+∑ j∈N(i) ω ij (t)·sgn(e′ i (t)·e′ j (t))), where K P,i represents the initial proportional coefficient, γ is the nonlinear factor, e′ j (t) Correction deviation for adjacent areas;

[0011] For the integral coefficient, the adjustment formula is: where K I,i Represents the initial integration coefficient, μ i is the integral attenuation coefficient;

[0012] For the differential coefficient, the adjustment formula is: Among them, K D,i represents the initial differential coefficient, v i is the differential attenuation coefficient, according to the deviation change rate Reduce the differential action, STD(e′ i (t-Δt:t)) represents the standard deviation of the corrected deviation in the most recent Δt period;

[0013] Finally, the PID formula is obtained based on the modified coefficient:

[0014] S5. Use the improved long short-term memory network to build a cold chain compartment temperature prediction model, input the historical temperature data series and the inter-regional influencing factor series, and output the temperature prediction value of each region in the future period;

[0015] S6. Finally, the feedforward control compensation is calculated based on the prediction results and integrated with the PID feedback control to achieve temperature composite control.

[0016] Preferably, the implementation of constructing the dynamic temperature impact factor model in step S2 includes:

[0017] First calculate the heat flux in represent the mean temperature of regions i and j, respectively, and k ij represents the thermal conductivity of the partition, d ij is the area spacing, ε ij (t) represents the random disturbance term;

[0018] Then the first-order inertia link is used to simulate the heat conduction delay: τ ij =τ 0,ij +Δτ ij ·sgn(κ ij (t)), where τ 0,ij is the reference delay, Δτ ijRepresents a dynamically adjusted item.

[0019] As a preferred method, the calculation method for generating the inter-regional impact factor based on the dynamic temperature impact factor model constructed in S2 is: Among them, β represents the correlation weight coefficient, ε represents a constant to prevent the denominator from being zero, and σ represents the standard deviation.

[0020] As a preference, the directionality of temperature conduction between regions is determined by causal test. If it is determined to be i→j, then F ij (t)>0, otherwise F ij (t)=0, for non-adjacent areas, the impact factor is set to 0.

[0021] Preferably, step S3 calculates the basic deviation according to the target temperature of each temperature control area and the real-time temperature average, introduces the inter-area influence factor to couple and correct the basic deviation, and forms the corrected deviation as follows:

[0022] First, for each temperature control area i, calculate the basic deviation Among them, T set,i is the target temperature of region i, represents the real-time mean temperature of area i;

[0023] The temperature influence factor is used to couple the basic deviation of region i. The calculation method is: Where N(i) represents the set of adjacent regions of region i, and sgn is the sign function to ensure that the correction direction is consistent with the heat conduction direction.

[0024] Preferably, the improved long short-term memory network in step S5 introduces a disturbance-aware gating unit into the standard LSTM structure, wherein the gating unit receives the inter-region disturbance feature sequence as a control signal and dynamically adjusts the weight distribution of the input gate and the forgetting gate according to the disturbance intensity.

[0025] Preferably, the step S6 calculates the feedforward control compensation amount in combination with the prediction result, and fuses it with the PID feedback control amount to realize the temperature composite control, which includes:

[0026] First, the temperature forecast values ​​for each temperature-controlled area in the future, predicted by the improved long-short-term memory network, are compared with the target temperature range for drug temperature control. The predicted offset is calculated and converted into an equivalent control response demand, generating the corresponding feedforward control compensation for active pre-regulation.

[0027] Based on the current deviation between the temperature value collected by the sensor in real time and the target temperature, the PID output feedback control amount is used to represent the real-time response compensation for the current deviation;

[0028] The feedforward compensation and PID feedback control are taken as two input sources and fused using a weighted fusion function to generate a fused composite control quantity.

[0029] Convert the composite control quantity into a specific control signal to drive the temperature control equipment in the cold chain compartment.

[0030] Compared with the prior art, the advantages and positive effects of the present invention are:

[0031] 1. By constructing a dynamic temperature influencing factor model, quantifying the inter-regional heat conduction effect, and combining it with the Granger causality test to determine the conduction direction, this solves the problem of the existing technology's lack of precise modeling of inter-regional heat conduction and improves the ability to coordinate temperature control in multiple regions.

[0032] 2. Dynamically adjust the proportional, integral, and differential coefficients of the PID controller based on the corrected deviation, introduce regional coupling compensation rules, and enhance the controller's adaptive adjustment capabilities for multi-region coupled heat conduction and real-time temperature deviations, avoiding the lag defect of traditional PID fixed parameter adjustment.

[0033] 3. Improve the long-short-term memory network and introduce disturbance-aware gating units. Combine historical temperatures with inter-regional influencing factors to predict future temperature trends, generate feedforward control compensation, and integrate it with PID feedback control to achieve compound control of prediction and pre-adjustment plus real-time response, offset the impact of disturbances in advance, and improve system response speed and energy efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following is a brief introduction to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0035] Figure 1 This is a structural flow chart of an intelligent temperature control method in cold chain transportation of pharmaceutical logistics. DETAILED DESCRIPTION

[0036] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described below in conjunction with the accompanying drawings and embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0037] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways than those described herein. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0038] In the cold chain transportation scenario of a certain biopharmaceutical company, the transport carriages need to carry vaccines, biological preparations and chemicals at the same time. The spatial distribution of medicines is uneven and sensitive to temperature. The traditional PID control scheme faces the problem of lag in response to multi-region coupled heat conduction. When the door is opened and closed and causes local thermal shock, the fixed parameter adjustment of the traditional PID is lagging, and it only relies on real-time deviation feedback adjustment, does not predict future temperature trends, and the refrigeration unit frequently starts and stops. For this reason, the present invention proposes a method for intelligent temperature control in cold chain transportation of pharmaceutical logistics. The specific implementation process is as follows: Figure 1 shown.

[0039] Traditional methods ignore the coupling of heat conduction between regions, resulting in the accumulation of control deviations. The present invention quantifies the influencing factors between regions through modeling of heat flux density and temperature gradient, thereby improving the collaborative control capabilities of multiple regions. First, the temperature requirements and spatial distribution data of the medicines in the cold chain compartment are obtained, independent temperature control areas are divided and sensors are deployed to collect temperature data in the area. The collected temperature data is then subjected to noise suppression and normalization. Specifically, noise suppression is first performed, and the real-time temperature series of each temperature control area is processed using a sliding average filter algorithm. The mean is calculated by setting a time window to effectively filter out random noise generated by electromagnetic interference, equipment vibration, etc. during the sensor acquisition process. Normalization is then performed. For each independent temperature control area, the minimum-maximum normalization method is used to first count the minimum and maximum values ​​of the historical temperature data of the area, and then the real-time temperature value is mapped to the [0,1] interval.

[0040] Then we started to build a dynamic temperature influence factor model, first calculating the heat flux density in represent the mean temperature of regions i and j, respectively, and k ij represents the thermal conductivity of the partition, d ij is the area spacing, ε ij (t) represents the random disturbance term; the first-order inertia link is then used to simulate the heat conduction delay: τ ij =τ 0,ij +Δτ ij ·sgn(κ ij (t)), where τ 0,ij is the reference delay, Δτ ijRepresents a dynamic adjustment term. Specifically, the heat flux density is first calculated, and the temperature gradient is reflected by the temperature mean difference between regions i and j, the thermal conductivity coefficient of the partition characterizes the thermal conductivity of the partition material, the regional spacing and the random disturbance term, to construct a heat flux density index. For example, when the temperature mean of region i is higher than that of region j, the heat flux density is positive, indicating that heat is conducted from i to j. The value is proportional to the temperature difference and inversely proportional to the spacing. The random disturbance term is used to fit the subtle effects of uncertain factors such as environmental fluctuations on heat conduction. Subsequently, the heat conduction delay characteristics are simulated, and the first-order inertia link is used to describe the lag effect of heat conduction. Taking the reference delay as the basic time parameter, the delay time is dynamically adjusted in combination with the heat flow direction: the heat flow direction is judged by the sign function. If the heat flow direction is consistent with the reference conduction direction, the dynamic adjustment term shortens the delay time, otherwise it is extended, thereby more realistically reflecting the delay changes caused by direction differences in actual heat conduction.

[0041] After constructing the dynamic temperature impact factor, the inter-regional impact factor is generated based on the model. Specifically, the calculation method is: Among them, β represents the correlation weight coefficient, ε represents the constant to prevent the denominator from being 0, and σ represents the standard deviation. For the directionality of temperature conduction, the Granger causality test is used to determine the directionality of temperature conduction between regions. If it is determined to be i→j, then F ij (t)>0, otherwise F ij (t)=0, for non-adjacent areas, the impact factor is set to 0.

[0042] Then, the basic deviation is calculated based on the target temperature of each temperature control area and the real-time temperature average, and the inter-regional influence factor is introduced to couple the basic deviation to form a corrected deviation. First, for each temperature control area i, the basic deviation is calculated. Among them, T set,i is the target temperature of region i, Represents the real-time temperature average of region i. The temperature influence factor is used to couple the basic deviation of region i with correction, and the calculation method is: Where N(i) represents the set of neighboring regions of region i, and sgn is the sign function, ensuring that the correction direction aligns with the direction of heat conduction. This process quantifies the intensity and direction of heat conduction between regions and incorporates the one-way coupling effect into the deviation calculation. This makes the corrected deviation more accurate to the actual scenario of multi-region temperature linkage, providing more accurate input for subsequent dynamic adjustment of PID controller parameters.

[0043] Then, the proportional coefficient, integral coefficient and differential coefficient of each regional PID controller are dynamically adjusted according to the corrected deviation. For each temperature control area i, the dynamic adjustment introduces the regional coupling compensation rule. First, the coupling strength weight is calculated as follows: Then for the proportional coefficient, the adjustment formula is: K′ P,i =KP,i ·(1+tanh(γ·|e′ i (t)|))·(1+∑ j∈N(i) ω ij (t)·sgn(e′ i (t)·e′ j (t))), where K P,i represents the initial proportional coefficient, γ is the nonlinear factor, e′ j (t) is the correction deviation of adjacent areas. For the integral coefficient, the adjustment formula is: where K I,i Represents the initial integration coefficient, μ i is the integral attenuation coefficient. For the differential coefficient, the adjustment formula is: Among them, K D,i represents the initial differential coefficient, v i is the differential attenuation coefficient, according to the deviation change rate Reduce the differential action, STD(e′ i (t-Δt:t)) represents the standard deviation of the correction deviation in the most recent Δt period. Finally, the PID formula obtained based on the modified coefficient is:

[0044]

[0045] Specifically, first calculate the coupling strength weight of each temperature control area i. The formula accumulates the product of the influence factor of the adjacent area j and the correction deviation, then superimposes a small constant, and finally obtains the weight value through normalization. This weight reflects the comprehensive influence of temperature fluctuations in adjacent areas on the current area: if there are deviations in the same direction in adjacent areas, the weight value increases, indicating that the coordinated regulation between regions needs to be enhanced; if the deviations are in the opposite direction, the weight value decreases, weakening the coupling effect. For the proportional coefficient, the hyperbolic tangent function tanh is introduced to amplify or attenuate the proportional gain in a nonlinear manner. When the absolute value of the correction deviation of the adjacent area is large, the tanh function approaches 1, the proportional coefficient increases significantly, and the coupling deviation is quickly suppressed; when the deviation is small, the function approaches 0, and the proportional coefficient approaches the initial value to avoid over-adjustment. At the same time, the sign function ensures that the adjustment direction is consistent with the heat conduction: if the deviations of i and j have the same sign, it means that heat conduction aggravates the deviation and the proportional effect needs to be enhanced; if they are of different signs, it means that heat conduction alleviates the deviation and the proportional effect is moderately weakened. For the integral coefficient, an exponential attenuation term is used and coupling correction term (1-∑ j∈N(i) ω ij (t)·|e′ i (t)·e′ jThe exponential decay term automatically reduces the integral gain based on the accumulated integral to prevent integral saturation caused by long-term deviations. The coupled correction term adjusts the integral strength by multiplying the deviations of adjacent regions. When opposite deviations exist in adjacent regions, the correction term becomes negative, further attenuating the integral effect and avoiding adjustment delays caused by cross-region interference. The differential coefficient is dynamically attenuated by combining the rate of change of the deviation and the standard deviation (STD) of the corrected deviation. When the rate of change of the deviation increases or the standard deviation rises, the differential attenuation coefficient decreases the differential gain exponentially, suppressing the impact of high-frequency noise on the system and preventing the differential action from amplifying sensor noise or mechanical vibration interference.

[0046] Next, considering that the existing temperature control lacks foresight, this solution predicts future temperature trends by improving LSTM and compensates for the impact of disturbances in advance. The improved LSTM introduces a disturbance-aware gating unit into the standard LSTM structure. The gating unit receives the inter-regional disturbance feature sequence as a control signal and dynamically adjusts the weight distribution of the input gate and the forget gate according to the disturbance intensity. Specifically, a disturbance-aware gating unit is embedded in the input gate and forget gate calculation path of the standard LSTM. The unit takes the disturbance feature as the input signal and first pre-processes the disturbance feature: the disturbance feature sequence is extracted through a sliding window and compressed into a disturbance encoding vector consistent with the dimension of the LSTM hidden layer through a fully connected layer. At each time step, the disturbance encoding vector, the input vector of the LSTM (historical temperature sequence), and the hidden state of the previous moment are jointly input into the gating unit. The gating unit adjusts the gating weight through the following mechanism: for input gate adjustment, the disturbance intensity index rd=tanh(W d ·concat(h t-1 ,c t-1 ,f t )), where W d is the perturbation weight matrix, f t Represents the perturbation feature at the current moment, and concat represents vector concatenation. The perturbation strength rd is then weighted and summed with the standard input gate signal to produce the final input gating signal. Similarly, the perturbation encoding vector interacts with the forget gate input to adjust the forget gate signal in the same manner. Through this mechanism, the improved LSTM dynamically allocates gating resources based on perturbation strength. When strong perturbations occur, the perturbation-aware gating unit significantly improves the input gate's response speed to real-time perturbation features, while suppressing the influence of outdated historical states through the forget gate.

[0047] Finally, the prediction results are combined to calculate a feedforward control compensation and fuse it with the PID feedback control variable to achieve composite temperature control. First, the temperature forecast for each temperature-controlled area for the future period, predicted by the improved long-short-term memory network, is compared with the target temperature range for drug temperature control. A predicted offset is calculated and converted into an equivalent control response requirement to generate the corresponding feedforward control compensation for proactive pre-control. Based on the current deviation between the real-time sensor temperature value and the target temperature, a PID output feedback control variable is used to represent the real-time response compensation for the current deviation. The feedforward compensation and PID feedback control variable are fused as two input sources using a weighted fusion function to generate a fused composite control variable. The composite control variable is then converted into a specific control signal to drive the temperature control equipment in the cold chain compartment. Specifically, the improved LSTM model is used to predict the temperature values ​​for each temperature-controlled area for the future period and compares them with the target temperature range for drug storage to calculate the predicted offset. For example, if the temperature in the vaccine area (target temperature 2-8°C) is predicted to drop to 1°C in one hour, which is 1°C below the lower limit, the predicted offset is -1°C. Through the pre-established control response model, the offset is converted into a feedforward control compensation: for example, for every 1°C drop in temperature, the cooling capacity needs to be reduced by 10% and the heating element needs to be started. The pre-control is triggered 20 minutes in advance, and active compensation is performed by increasing the power of the refrigeration unit to offset the prediction deviation. At the same time, based on the temperature data collected by the sensor in real time, the real-time deviation between the current temperature and the target mean is calculated, and the feedback control quantity is calculated using the dynamically adjusted PID controller parameters to quickly respond to current temperature fluctuations. The feedforward compensation quantity and the feedback control quantity are then dynamically integrated through a weighted fusion function, which not only pre-cools in advance but also fine-tunes in real time. Finally, the composite control quantity is converted into a specific control signal, and the digital signal is converted into an analog voltage / current signal through the analog-to-digital conversion module to drive the refrigeration unit, heating element or damper actuator in the cold chain compartment to achieve precise temperature control.

[0048] The above description is merely a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any person skilled in the art may utilize the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes for application in other fields. However, any simple modification, equivalent change, and modification of the above embodiments made in accordance with the technical essence of the present invention without departing from the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for intelligent temperature control in cold chain transportation of pharmaceutical logistics, characterized in that: The following steps are involved: S1. Obtain the temperature requirements and spatial distribution data of medicines in the cold chain compartment, divide them into independent temperature control areas, deploy sensors, and collect temperature data within the areas; S2. Suppress and normalize the collected temperature data, build a dynamic temperature impact factor model, and generate inter-regional impact factors; S3. Calculate the basic deviation based on the target temperature and the real-time temperature average of each temperature control area, introduce the inter-area influencing factor to couple and correct the basic deviation to form a corrected deviation; S4. Dynamically adjust the proportional coefficient, integral coefficient, and differential coefficient of the PID controller in each region according to the corrected deviation. The specific implementation is as follows: For each temperature control area i, the regional coupling compensation rule is introduced for dynamic adjustment. First, the coupling strength weight is calculated as follows: Then for the proportional coefficient, the adjustment formula is: K′ P,i =K P,i ·(1+tanh(γ·|e′ i (t)|))·(1+∑ j∈N(i) ω ij (t)·sgn(e′ i (t)·e′ j (t))), where K P,i represents the initial proportional coefficient, γ is the nonlinear factor, e′ j (t) Correction deviation for adjacent areas; For the integral coefficient, the adjustment formula is: where K I,i Represents the initial integration coefficient, μ i is the integral attenuation coefficient; For the differential coefficient, the adjustment formula is: Among them, K D,i represents the initial differential coefficient, v i is the differential attenuation coefficient, according to the deviation change rate Reduce differential action, STD(e i ′(t-Δt:t)) represents the standard deviation of the corrected deviation in the most recent Δt period; Finally, the PID formula is obtained based on the modified coefficient: S5. Use the improved long short-term memory network to build a cold chain compartment temperature prediction model, input the historical temperature data series and the inter-regional influencing factor series, and output the temperature prediction value of each region in the future period; S6. Finally, the feedforward control compensation is calculated based on the prediction results and integrated with the PID feedback control to achieve temperature composite control.

2. The method for intelligent temperature control in cold chain transportation of pharmaceutical logistics according to claim 1, characterized in that: The implementation of constructing the dynamic temperature impact factor model in step S2 includes: First calculate the heat flux in represent the mean temperature of regions i and j, respectively, and k ij represents the thermal conductivity of the partition, d ij is the area spacing, ε ij (t) represents the random disturbance term; Then the first-order inertia link is used to simulate the heat conduction delay: τ ij =τ 0,ij +Δτ ij ·sgn(κ ij (t)), where τ 0,ij is the reference delay, Δτ ij Represents a dynamically adjusted item.

3. The method for intelligent temperature control in cold chain transportation of pharmaceutical logistics according to claim 2, characterized in that: The calculation method for generating inter-regional impact factors based on the dynamic temperature impact factor model constructed by S2 is: Among them, β represents the correlation weight coefficient, ε represents a constant to prevent the denominator from being zero, and σ represents the standard deviation.

4. The method for intelligent temperature control in cold chain transportation of pharmaceutical logistics according to claim 3, characterized in that: According to the causal test, the direction of temperature conduction between regions is determined. If it is determined to be i→j, then F ij (t)>0, otherwise F ij (t)=0, for non-adjacent areas, the impact factor is set to 0.

5. The method for intelligent temperature control in cold chain transportation of pharmaceutical logistics according to claim 1, characterized in that: The step S3 calculates the basic deviation based on the target temperature of each temperature control area and the real-time temperature average, introduces the inter-area influence factor to couple and correct the basic deviation, and forms the corrected deviation as follows: First, for each temperature control area i, calculate the basic deviation Among them, T set,i is the target temperature of region i, represents the real-time mean temperature of area i; The temperature influence factor is used to couple the basic deviation of region i. The calculation method is: Where N(i) represents the set of adjacent regions of region i, and sgn is the sign function to ensure that the correction direction is consistent with the heat conduction direction.

6. The method for intelligent temperature control in cold chain transportation of pharmaceutical logistics according to claim 1, characterized in that: The improved long short-term memory network in step S5 introduces a disturbance-aware gating unit into the standard LSTM structure. The gating unit receives the inter-region disturbance feature sequence as a control signal and dynamically adjusts the weight distribution of the input gate and the forget gate according to the disturbance intensity.

7. The method for intelligent temperature control in cold chain transportation of pharmaceutical logistics according to claim 1, characterized in that: The step S6 calculates the feedforward control compensation amount in combination with the prediction result and integrates it with the PID feedback control amount to realize the temperature composite control, which includes: First, the temperature forecast values ​​for each temperature-controlled area in the future, predicted by the improved long-short-term memory network, are compared with the target temperature range for drug temperature control. The predicted offset is calculated and converted into an equivalent control response demand, generating the corresponding feedforward control compensation for active pre-regulation. Based on the current deviation between the temperature value collected by the sensor in real time and the target temperature, the PID output feedback control amount is used to represent the real-time response compensation for the current deviation; The feedforward compensation and PID feedback control are taken as two input sources and fused using a weighted fusion function to generate a fused composite control quantity. Convert the composite control quantity into a specific control signal to drive the temperature control equipment in the cold chain compartment.

Images