Air compressor vehicle-mounted refrigeration and refrigeration house collaborative temperature control system based on AI scheduling

Abstract

The invention relates to the technical field of cold-chain logistics transportation and storage temperature control, in particular to an air compressor vehicle-mounted refrigeration and refrigeration house collaborative temperature control system based on AI scheduling, which comprises a multi-dimensional state sensing module used for collecting real-time temperature field data, environmental meteorological parameters, compressor operation states and goods category information in a refrigerator car and a refrigeration house, and sending the real-time temperature field data to a server; a structured multi-dimensional state vector is output; and the thermodynamic evolution prediction module is used for setting a prediction time domain with a finite length based on the multi-dimensional state vector, and calling and constructing a mathematical model of a heat balance relationship between the refrigerating capacity generated by the operation frequency of the compressor and the environment and the cargo heat load. According to the method, the comprehensive cost function fusing the energy consumption, the cargo quality loss and the equipment wear is constructed, and the hidden quality attenuation is converted into the dominant economic cost by using the Arrhenius model, so that the global operation benefit maximization of the system on the premise of ensuring the cargo safety is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of cold chain logistics transportation and storage temperature control technology, and more particularly to a vehicle-mounted air compressor refrigeration and cold storage collaborative temperature control system based on AI scheduling. BACKGROUND

[0002] In the field of cold chain logistics transportation and storage, the operation control of the refrigeration system is the core link to ensure the quality and safety of perishable goods (such as food, medicine, etc.) and control the operating cost. Among them, the vehicle-mounted air compressor and the cold storage compressor unit are the main energy-consuming equipment, and their running frequency and start-stop state directly determine the energy consumption level and the temperature stability of the refrigeration space. The temperature fluctuation is closely related to the biochemical reaction rate of the goods and the remaining shelf life. Therefore, developing a reasonable control strategy to maintain a suitable temperature environment while reducing energy consumption has always been the research focus in the field of cold chain control technology.

[0003] However, the existing cold chain control technology mostly only sets rigid upper and lower threshold constraints for temperature, lacks quantitative mathematical description of the dynamic decay process of goods quality with temperature and time, cannot convert the implicit shelf life loss into explicit economic indicators using Arrhenius kinetics model, and makes it difficult for the control system to perceive the potential quality depreciation cost caused by the accumulation of small temperature fluctuations. At the same time, when pursuing energy saving optimization, the existing scheme simply focuses on reducing compressor power consumption, ignoring the mechanical wear and tear cost caused by fast control action or frequent start-stop of the equipment, so that the system cannot find a balance point between energy expenditure, goods quality preservation and equipment maintenance life, and it is difficult to achieve the global minimum of the overall operation cost. SUMMARY

[0004] In order to make up for the above shortcomings, the present application provides a vehicle-mounted air compressor refrigeration and cold storage collaborative temperature control system based on AI scheduling, which aims to improve the problem that the existing technology only controls by rigid temperature threshold, lacks quantitative evaluation of goods quality decay and equipment wear and tear, and cannot achieve global cost optimization of energy consumption, quality and life.

[0005] The present application provides the following technical solutions: a vehicle-mounted air compressor refrigeration and cold storage collaborative temperature control system based on AI scheduling, comprising: A multi-dimensional state perception module is used to collect real-time temperature field data, environmental meteorological parameters, compressor operating states and goods category information in the refrigerated truck and the cold storage, and output a structured multi-dimensional state vector; A thermodynamic evolution prediction module is based on the multi-dimensional state vector, sets a limited length of the prediction time domain, and calls a mathematical model that builds the heat balance relationship between the refrigeration capacity generated by the compressor operating frequency and the environmental and goods thermal load, calculates the system state evolution with time step advancing in the prediction time domain, and outputs a prediction state sequence containing temperature change trajectory and energy consumption trend; a comprehensive cost construction module configured to construct, based on the predicted state sequence and the cargo category information, a comprehensive operation cost function including a cargo quality loss calculation item, a device mechanical wear calculation item, and an energy consumption calculation item by a weighted manner; a robust optimization solving module configured to, based on the comprehensive operation cost function, set a temperature safety boundary as a soft constraint condition and introduce a nonlinear penalty item, solve, by an iterative calculation, a control amount that minimizes the comprehensive operation cost function, and output a target compressor frequency sequence; a cooperative execution and feedback module configured to, based on the target frequency sequence, execute frequency conversion adjustment on the vehicle-mounted air compressor and the cold storage compressor group respectively, and in response to a vehicle predicted arrival time being less than a preset threshold, generate a precooling instruction to trigger precooling operation of a cold storage platform, and simultaneously collect state data after execution and feed back to the multi-dimensional state perception module.

[0006] Preferably, the output structured multi-dimensional state vector specifically includes the following steps: obtaining sensor time series data, external weather interface data, and cargo category metadata of the vehicle-mounted end and the cold storage end to form a multi-source heterogeneous original data stream; performing clock synchronization calibration on the original data stream, and using a moving average filtering algorithm to remove noise data to generate an aligned physical parameter sequence; performing dimensionless normalization processing on continuous numerical data in the physical parameter sequence, and vectorizing discrete category data; splicing the processed data sequence and the environmental disturbance variable at the current time to construct the multi-dimensional state vector composed of state variables, disturbance variables, and attribute variables.

[0007] Preferably, the calculation of the system state evolution advancing with time steps in the prediction time domain specifically includes the following steps: taking the multi-dimensional state vector at the current time as an initial state point of the prediction time domain, and establishing a discretized time step and a prediction step number; obtaining corresponding environmental weather forecast data in the prediction time domain and estimated respiration heat data based on the cargo category to construct an external disturbance sequence varying with time steps; for each time step in the prediction time domain, calculating a heat enthalpy change amount of the system based on a discretized heat balance equation, the heat enthalpy change amount being jointly determined by a heat transfer increment, an internal energy increment, and a refrigeration decrement corresponding to the compressor operation frequency; calculating a temperature state value at the next time step using the heat enthalpy change amount and a total heat capacity parameter of the system, and taking the state value as an input to recursively execute the step of calculating the heat enthalpy change amount until a temperature change trajectory of the whole time domain is generated.

[0008] Preferably, the constructing the comprehensive operation cost function comprising the goods quality loss calculation item, the equipment mechanical wear calculation item and the energy consumption calculation item by weighting specifically comprises the following steps: Obtaining the energy unit price information, combining the power model corresponding to the compressor operating frequency, calculating the cumulative energy consumption cost in the prediction time domain as the energy consumption calculation item; Accessing the activation energy parameters corresponding to the goods category, processing the temperature change trajectory in the prediction time domain by using the Arrhenius metamorphic kinetics model, calculating the decay cumulative value of the remaining shelf life of the goods and converting it into the quality decay value, as the goods quality loss calculation item; Calculating the square of the difference of the compressor operating frequency between adjacent time steps in the prediction time domain, and taking the cumulative result of the square value as the equipment mechanical wear calculation item; Respectively assigning the preset weight coefficients to the energy consumption calculation item, the goods quality loss calculation item and the equipment mechanical wear calculation item, and accumulating the weighted values of each item to generate the comprehensive operation cost function.

[0009] Preferably, the setting the temperature safety boundary as a soft constraint condition and introducing a nonlinear penalty term specifically comprises the following steps: Obtaining the boundary threshold of the temperature safety interval, and setting the relaxation penalty coefficient for representing the tolerance of temperature violation; Traversing the temperature prediction value in the prediction state sequence, calculating the deviation amount of the temperature prediction value from the temperature safety interval, wherein the deviation amount of the temperature prediction value in the safety interval is recorded as zero; Carrying out square operation on the deviation amount to construct a nonlinear relationship, and multiplying the squared result by the relaxation penalty coefficient to obtain the penalty value corresponding to this time step; Accumulating the penalty values of all time steps in the prediction time domain, and adding them to the comprehensive operation cost function as a soft constraint cost item.

[0010] Preferably, the solving the control amount that minimizes the comprehensive operation cost function by iterative calculation specifically comprises the following steps: Setting the physical constraint set of the compressor frequency converter, the physical constraint set including the operating frequency allowed range and the frequency change rate gradient limit; Converting the comprehensive operation cost function into a numerical optimization model, and taking the multi-dimensional state vector at the current time as the initial condition; In the search space defined by the physical constraint set, iteratively optimizing the candidate frequency sequence in the prediction time domain until the preset convergence termination condition is met, obtaining the full-time domain optimized frequency sequence; Extract the frequency value corresponding to the first time step in the full-time domain optimization frequency sequence as the output value of the current time, and discard the remaining prediction control amount in the sequence.

[0011] Preferably, the frequency regulation of the vehicle-mounted air compressor and the cold storage compressor group comprises the following steps: Analyze the target frequency sequence, and convert it into a bottom layer control signal according to the communication protocol corresponding to the control object identifier; The bottom layer control signal is subjected to safety check, and the protection logic is triggered when the device operating parameter is monitored to exceed the physical limit value, and the execution is issued when the check is passed; For the vehicle-mounted air compressor, the control signal is sent to the vehicle-mounted frequency converter to adjust the motor speed; For the cold storage compressor group, the control signal is sent to the cold storage controller to adjust the refrigeration output parameter of the compressor group.

[0012] Preferably, the precooling instruction is generated to trigger the precooling operation of the cold storage platform in response to the vehicle predicted arrival time being less than the preset threshold, and the precooling instruction comprises the following steps: Track the position of the refrigerated truck and calculate the predicted remaining time of the vehicle to arrive at the specified platform in combination with the road condition information; Determine whether the predicted remaining time is less than the cooperative precooling time threshold based on the platform heat capacity characteristic calibration; When it is determined that the condition is met, the target storage temperature of the goods in the multi-dimensional state vector is read and set as the precooling target temperature of the platform area; The precooling instruction is generated and sent to the cold storage control system, and the refrigeration equipment of the platform area is started to cool down to establish a low-temperature buffer zone.

[0013] The present application has the following beneficial effects: 1、In the present application, by constructing a comprehensive cost function integrating energy consumption, goods quality loss and equipment wear and tear, the implicit quality decay is converted into explicit economic cost by using the Arrhenius model, and the system global operation benefit maximization under the premise of guaranteeing the safety of goods is realized.

[0014] 2、In the present application, the temperature soft constraint mechanism with a nonlinear penalty term is used to allow flexible adjustment of the boundary in the prediction control to cope with environmental disturbances, avoiding frequent start-stop and severe vibration of the equipment, and significantly enhancing the robustness and control stability of the system operation.

[0015] 3、In the present application, the vehicle arrival time is used to realize the cooperative precooling of the vehicle-mounted end and the cold storage end, the low-temperature buffer zone of the platform is established in advance to eliminate the temperature control blind area of the unloading link, effectively preventing the damage of goods quality caused by thermal shock and improving the seamless connection level of the whole cold chain. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The framework diagram of the air compressor vehicle-mounted refrigeration and cold storage collaborative temperature control system based on AI scheduling is provided for the application. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the application.

[0018] The application provides an air compressor vehicle-mounted refrigeration and cold storage collaborative temperature control system based on AI scheduling, as shown in the accompanying drawings, comprising: Figure 1 A multi-dimensional state perception module is configured to collect real-time temperature field data, environmental meteorological parameters, compressor operating states and cargo category information in the refrigerated truck and the cold storage, and output a structured multi-dimensional state vector. Further, the output of the structured multi-dimensional state vector comprises the following steps: Obtain sensor time series data, external meteorological interface data and cargo category metadata at the vehicle-mounted end and the cold storage end to form a multi-source heterogeneous raw data stream. Clock synchronization calibration is performed on the raw data stream, and a sliding average filtering algorithm is used to remove noise data to generate an aligned physical parameter sequence. The continuous numerical data in the physical parameter sequence is subjected to dimensionless normalization processing, and the discrete category data is subjected to vectorization encoding. The processed data sequence and the current time environmental disturbance variable are spliced to construct a multi-dimensional state vector composed of state variables, disturbance variables and attribute variables.

[0019] Specifically, the multi-dimensional state perception module obtains real-time data during operation through vehicle-mounted and cold storage end sensors, compressor controller frequency feedback interfaces and external meteorological interfaces, and uniformly processes these data into a structured multi-dimensional state vector that can be directly called by a temperature field prediction model and an energy consumption optimization model. The system first receives temperature time series data output by an array of temperature sensors inside the vehicle-mounted box, which represents the temperature distribution of different spatial positions of the vehicle-mounted box over time. The vehicle-mounted temperature data is represented as: ; Wherein is the sampling time index of the system, representing the discretized time point of the temperature field data, is the spatial position index of the vehicle-mounted temperature sensor, corresponding to different sensor points inside the vehicle box.

[0020] ​The temperature time series data collected by the temperature sensor network at the cold storage end is represented as: ; wherein is the unified sampling time index of the system, is the sensor position index arranged inside the cold storage, reflecting the temperature distribution of different storage areas, and the external weather interface provides real-time weather data including ambient temperature, ambient humidity, solar radiation intensity, and wind speed, represented as: ; wherein is the ambient temperature at time , is the ambient humidity, is the solar radiation intensity, is the wind speed.

[0021] The device operating state is obtained through the frequency feedback of the compressor inverter, and the vehicle-mounted compressor frequency is represented as: ; wherein is the operating frequency of the vehicle-mounted variable frequency compressor at time , and the cold storage compressor frequency is represented as : wherein is the variable frequency operating frequency of the cold storage compressor unit at time .

[0022] The cargo category metadata includes information such as cargo name, respiratory heat generation coefficient, and temperature-sensitive quality decay interval, represented by the set: ; , wherein correspond to the attribute parameters of different categories of goods, respectively.

[0023] Since the above data comes from different devices with different original time bases, the system uses a unified master clock timestamp to resample and time-align all data. For any data source, the observation value at the original sampling time is converted to the corresponding value at the master time point , and linear interpolation is used to complete the synchronization. The interpolation calculation formula is: ; wherein and are the observation values at the original sampling time points and , respectively, and satisfy , through the processing mode, temperature, weather data and compressor frequency data are mapped to a unified time axis .

[0024] The synchronized data may contain high-frequency noise caused by electromagnetic interference or sensor jitter, so a moving average filtering process is performed on the aligned sequence, and the aligned data sequence is , the filtered data is represented as: ; where is the length of the filtering window, controlling the number of consecutive sampling points participating in the average, is the smoothed data at time , used to improve the stability of the sensor data.

[0025] The filtered continuous physical parameter data needs to be dimensionless normalized due to the difference in dimensions, so that it remains numerically consistent in model input. The normalization of any continuous physical quantity is defined as ; where and represent the minimum and maximum values that the physical quantity can take under engineering planning conditions, and the normalized output is located in the interval zero to one, which can improve the training and inference stability of the subsequent model.

[0026] Cargo category metadata is a discrete categorical variable and cannot be directly input into the model, so it needs to be vectorized encoded. The present embodiment uses one-hot encoding to map the cargo category to a vector: ; where the only one value in the vector corresponds to the number position of the cargo category , and the other positions are zero. This encoding method ensures that different cargo categories have mutual exclusivity in the input structure.

[0027] In order to reflect the influence of short-term changes in external environment on the system thermal load, the system generates an environmental disturbance variable according to the short-term prediction results of the weather service, represented as: ; where is the rate of change of the ambient temperature in the future short time, is the rate of change of the ambient humidity, is the rate of change of the wind speed, each of which reflects the dynamic disturbance trend of the external conditions.

[0028] After all the above processing steps, the normalized vehicle temperature, cold storage temperature, weather data, compressor operating frequency data: ; And the goods category one-hot encoding: ; And environmental disturbance variables ; Spliced into a structured multi-dimensional state vector: ; Where is the final state vector at time , which can fully express the thermal state, equipment operating state, goods attribute information and external environmental disturbance characteristics of the system.

[0029] Thus, the structured fusion of multi-source, large-span, dimensionally inconsistent and different sampling frequency data under the unified time reference is realized, so that the system can provide input to the prediction model and optimization model in a standardized form, thereby significantly improving the temperature prediction accuracy, energy consumption inference ability and effectiveness of the compressor scheduling strategy.

[0030] The thermodynamic evolution prediction module, based on the multi-dimensional state vector, sets a finite length of the prediction time domain, and calls the mathematical model that builds the heat balance relationship between the refrigeration capacity generated by the compressor operating frequency and the environmental and cargo thermal load, calculates the system state evolution advancing with time steps in the prediction time domain, and outputs the prediction state sequence containing the temperature change trajectory and energy consumption trend; Further, the calculation of the system state evolution advancing with time steps in the prediction time domain includes the following steps: The multi-dimensional state vector at the current time is taken as the initial state point of the prediction time domain, and the time step and the prediction step number after discretization are established; Obtain the corresponding environmental weather forecast data in the prediction time domain and the estimated respiration heat data based on the goods category, and construct the external disturbance sequence varying with time steps; For each time step in the prediction time domain, calculate the enthalpy change of the system based on the discretized heat balance equation, which is determined by the heat transfer increment, the internal energy increment and the refrigeration decrement corresponding to the compressor operating frequency; Use the enthalpy change and the total heat capacity parameter of the system to calculate the temperature state value of the next time step, and take this state value as input to recursively execute the above step of calculating the enthalpy change until the temperature change trajectory of the whole time domain is generated.

[0031] Specifically, the thermodynamic evolution prediction module calculates the temperature change trajectory and energy consumption trend of the system in the prediction time domain based on the constructed mathematical model of the refrigeration system's thermal balance by continuously extrapolating the multidimensional state vector over time. The system first uses the multidimensional state vector obtained at the current moment as the initial state point for the prediction operation. This vector contains the vehicle body temperature, cold storage temperature, meteorological data, compressor operating frequency, cargo category code, and external disturbance information. Based on this initial state point, the system determines the discretized time step and the number of prediction steps, thereby forming a prediction time domain of finite length.

[0032] After the prediction time domain is established, the system acquires the meteorological forecast data corresponding to that time domain, including the ambient temperature, humidity, radiation intensity and wind speed at future time points, and estimates the respiratory heat that changes over time based on the metadata of the cargo category. The resulting external disturbance sequence participates in the subsequent heat balance calculation. At each discrete time step, the system calculates the dynamic balance relationship between cooling capacity and heat load, which constitutes the core of the thermodynamic evolution prediction process.

[0033] The system first calculates the enthalpy change at that time step. The enthalpy change represents the total heat change of the system during that time step, denoted as: ; in For heat transfer increment, For the increase of internal energy, For cooling capacity, The heat transfer increment is generated by the temperature difference between the cold storage end temperature and the vehicle body temperature, calculated according to the heat conduction formula, and expressed as: ; in The thermal conductivity coefficient, For heat transfer surface area, The temperature at the cold storage end. This refers to the temperature of the vehicle's enclosure.

[0034] The increase in internal energy is caused by the cargo itself and the internal heat load of the system, and is expressed as: ; in For the quality of goods, For the specific heat capacity of the goods, The temperature change of the system before and after this time step is also included in this item, with the breathing heat coefficient stored in the cargo category metadata also added as a source of heat load that increases over time.

[0035] The cooling capacity is determined by the compressor's operating frequency. The compressor's frequency directly affects the cooling capacity per unit time, expressed as: ; in for the refrigeration efficiency of the compressor, for the discrete time point of the compressor operating frequency, derived from the compressor feedback data in the multi-dimensional state vector.

[0036] The system sums the above three items to obtain the change in enthalpy, and after obtaining the change in enthalpy, the system calculates the temperature at the next discrete time step using the total heat capacity parameter, which represents the system's ability to absorb or release heat, composed of the vehicle-mounted box, the cargo body, and the refrigeration system structure, represented as: ; wherein is the heat capacity of the vehicle-mounted box, is the heat capacity of the cargo, is the heat capacity of the refrigeration system structure.

[0037] According to the relationship between heat capacity and enthalpy, the change in temperature at the next time step is represented as: ; wherein is the change in temperature, is the calculated change in enthalpy, is the total heat capacity of the system, from which the temperature state value at the next time step can be calculated, and the above heat balance calculation is recursively performed as new input state until the derivation of all prediction steps is completed.

[0038] Throughout the prediction time domain, the system records the temperature change value and the corresponding refrigeration demand at each discrete time step, forming a prediction state sequence of temperature change trajectory and energy consumption trend, which reflects the evolution law of system temperature and energy consumption over time within the set future time domain.

[0039] This thermodynamic evolution prediction module can accurately predict future temperature changes and energy consumption trends based on physical mechanisms and actual engineering data, providing reliable basis for subsequent compressor operation optimization and energy consumption regulation strategies.

[0040] The comprehensive cost construction module, based on the prediction state sequence and the cargo category information, constructs a comprehensive operating cost function including the cargo quality loss calculation item, the equipment mechanical wear calculation item, and the energy consumption calculation item through a weighting method; Further, the comprehensive operating cost function including the cargo quality loss calculation item, the equipment mechanical wear calculation item, and the energy consumption calculation item through a weighting method specifically includes the following steps: Obtain the energy unit price information, combine the power model corresponding to the compressor operating frequency, and calculate the cumulative energy consumption cost in the prediction time domain as the energy consumption calculation item; The activation energy parameter corresponding to the cargo category is called, the temperature change trajectory in the prediction time domain is processed by using the Arrhenius metamorphic kinetics model, the decay cumulative value of the remaining shelf life of the cargo is calculated and converted into a quality decay equivalent value, which is used as a cargo quality loss calculation item; The square of the difference value of the compressor operating frequency between adjacent time steps in the prediction time domain is calculated, and the cumulative result of the square value is used as a device mechanical wear calculation item; The preset weight coefficients are allocated to the energy consumption calculation item, the cargo quality loss calculation item and the device mechanical wear calculation item, and the weighted values of each item are added to generate a comprehensive operation cost function.

[0041] Specifically, the comprehensive cost construction module quantifies the energy consumption, cargo quality loss and device mechanical wear based on the predicted state sequence output by the thermodynamic evolution prediction module and the cargo category information, and constructs a comprehensive operation cost function system through a weighting method. First, the temperature change trajectory, compressor operating frequency sequence and time step information in the future time domain are obtained from the prediction results, and the activation energy parameter, pre-exponential factor and value parameter are read from the cargo category metadata.

[0042] When the system starts to calculate the energy consumption cost, first, the energy unit price is obtained according to the business configuration, and the energy unit price is denoted as , which represents the price of unit electric energy. The system calls the compressor power model to convert the compressor frequency into power, and the power of the compressor at discrete time points is expressed as: ; Wherein is the power consumption of the compressor at time point , is the compressor power conversion coefficient, which represents the power change value corresponding to the change of 1 Hz of compressor frequency, is the operating frequency of the compressor at time point in the predicted state sequence. The system calculates the total energy consumption according to the time step , and then multiplies the energy unit price to obtain the energy consumption cost, which is denoted as: ; Wherein is the total number of time steps in the prediction time domain, is the energy consumption calculation item.

[0043] The system then calculates the cargo quality loss cost. To evaluate the impact of temperature change on cargo quality, the system calls the activation energy parameter and the pre-exponential factor defined in the cargo category, and calculates the cargo quality decay rate based on the Arrhenius model. The quality decay rate is expressed as: ; wherein is the quality decay rate of the goods at time point , is the pre-exponential factor related to the goods category, is the goods quality decay kinetic activation energy, is the universal gas constant, is the goods temperature at time point in the predicted temperature trajectory, 273.15 is used to convert Celsius temperature to thermodynamic temperature The system accumulates the decay rate according to the time step to form the shelf life decay amount, which is represented as: ; wherein is the decay accumulation amount of the remaining shelf life of the goods The system converts the shelf life decay into economic loss according to the goods value coefficient to obtain the quality loss cost, which is represented as: ; wherein is the goods quality loss calculation item: The system continues to calculate the equipment mechanical wear cost The compressor frequency change will cause the wear of mechanical parts, and the system evaluates according to the frequency change amplitude of adjacent time steps, taking the square of the frequency difference as the calculation basis of the wear amount, which is represented as: ; wherein is the cumulative mechanical wear degree of the compressor in the predicted time domain The system converts the wear degree into mechanical wear cost according to the wear economic cost coefficient , which is represented as: ; wherein is the equipment mechanical wear calculation item.

[0044] The system finally sets a weight coefficient to the three types of costs and performs a weighted sum to obtain a comprehensive operation cost function, which is represented as: ; wherein is the comprehensive operation cost result, , , are the weight coefficients corresponding to the energy consumption, goods quality loss and equipment mechanical wear respectively, which are used to reflect the emphasis degree of different cost indicators in the actual operation strategy.

[0045] Through the above implementation process, the comprehensive cost construction module can convert the temperature change, energy consumption and equipment load information contained in the predicted state sequence into quantifiable and optimizable cost metrics, so that the subsequent operation strategy design can be solved with the goal of optimizing the comprehensive benefits.

[0046] The robust optimization solving module sets the temperature safety boundary as a soft constraint condition and introduces a nonlinear penalty term based on the comprehensive operation cost function, solves the control quantity that minimizes the comprehensive operation cost function through iterative calculation, and outputs the target compressor frequency sequence; Further, setting the temperature safety boundary as a soft constraint condition and introducing a nonlinear penalty term specifically includes the following steps: Obtain the boundary threshold of the temperature safety interval, and set a relaxation penalty coefficient for representing the tolerance of temperature violation; Iterate through the temperature prediction values in the predicted state sequence, and calculate the deviation of the temperature prediction value from the temperature safety interval, wherein the deviation of the temperature prediction value within the safety interval is recorded as zero; Square the deviation to construct a nonlinear relationship, and multiply the squared result by the relaxation penalty coefficient to obtain the penalty value corresponding to the time step; Accumulate the penalty values of all time steps in the prediction time domain, and add them as a soft constraint cost item to the comprehensive operation cost function.

[0047] Further, solving the control quantity that minimizes the comprehensive operation cost function through iterative calculation specifically includes the following steps: Set the physical constraint set of the compressor frequency converter, which includes the allowed range of operating frequency and the gradient limit of frequency change rate; Convert the comprehensive operation cost function into a numerical optimization model, and use the multi-dimensional state vector at the current time as the initial condition; Within the search space defined by the physical constraint set, iteratively optimize the candidate frequency sequence in the prediction time domain until the preset convergence termination condition is met, and obtain the full-time domain optimized frequency sequence; Extract the frequency value corresponding to the first time step in the full-time domain optimized frequency sequence as the output value at the current time, and discard the remaining predicted control quantities in the sequence.

[0048] Specifically, the robust optimization solving module constructs an optimization solving process containing a soft constraint based on the comprehensive operation cost function and the requirement of the refrigeration system for the temperature stability of the goods, to obtain the target compressor frequency sequence that minimizes the total operation cost. The system first obtains the upper and lower boundaries of the temperature safety interval from the device operation configuration, denoted as and , which are defined as the lower and upper bounds of the temperature range where the quality of the goods will not be damaged, and the relaxation penalty coefficient of the temperature violation penalty, denoted as , which represents the amplification weight of the unit temperature deviation in the cost function after the temperature boundary is violated, and is used to control the sensitivity of the over-temperature penalty.

[0049] The system traverses the temperature prediction value of each future time in the predicted state sequence, denoted as , where represents the discrete time step, and the system calculates the temperature deviation amount , which is defined as the boundary amplitude of the temperature deviating from the safety interval, and the mathematical form of the boundary is: ; In the formula , the over-regulation amount of the temperature exceeding the upper limit is , the under-regulation amount of the temperature being lower than the lower limit is, and the deviation amount is zero when the temperature is in the safety interval. In order to increase the constraint strength of the temperature boundary, the system constructs the deviation amount into a nonlinear penalty form, and establishes a gain relationship of being more sensitive to large boundary through squaring, and the penalty value of the generation value is: ; , where is the temperature boundary penalty result of the time step , and is the relaxation penalty coefficient, which is used to convert the temperature boundary into the corresponding economic cost. The system adds all the penalty generation values in the whole time domain to obtain the temperature soft constraint cost item, which is denoted as: ; , where is the total number of discrete time steps in the prediction time domain, is the total penalty value of the temperature boundary in the whole time domain, and the system adds this item to the comprehensive operation cost function to obtain the optimization objective function containing the soft constraint, which is denoted as: ; , where is the final solution objective value, is the comprehensive cost result obtained by the comprehensive operation cost construction module.

[0050] The system constructs a physical constraint set of the compressor frequency converter for the feasibility constraint of the control variable. The set first includes the frequency allowed range constraint, which is denoted as: ; , where is the target frequency of the compressor in the time step , , and respectively represent the minimum and maximum operating frequency allowed by the device hardware to output, the system further sets the gradient limit of the compressor frequency variation rate to avoid mechanical fatigue caused by frequency jumping, the limit is represented as: ; wherein is the maximum frequency variation amplitude acceptable in unit time, used to limit the compressor adjustment speed.

[0051] The system constructs a numerical optimization model according to the optimization objective function and the physical constraint set, and takes the built-in state vector as the initial condition of the current rolling period, the system generates a candidate frequency sequence within the feasible region defined by the constraint set, and iteratively optimizes the compressor frequency distribution in the prediction time domain through a gradient descent method or a model predictive control-based solving strategy. In the iteration process, the system adjusts the candidate frequency sequence according to the gradient direction or cost decrease trend of the objective function each time, so that the objective function gradually decreases. When the system judges that the iteration has reached the convergence termination condition, such as the change of the objective function being less than a threshold or the maximum number of iterations being reached, the optimal frequency sequence in the full time domain is output, denoted as: ; The system extracts the optimal frequency value of the first time step from the sequence as the control instruction issued to the compressor frequency converter at the current time, and then discards the rest of the sequence, so that the next period can be optimized based on the latest predicted state, thereby realizing rolling robust control.

[0052] Through the above steps, the robust optimization solving module can minimize the comprehensive operating cost in the future time domain while ensuring temperature safety and allowing a certain flexibility, and output the real-time updated optimal control frequency of the compressor, realizing the balance optimization among the safety, energy efficiency and service life of the refrigeration system.

[0053] The cooperative execution and feedback module executes frequency regulation for the vehicle-mounted air compressor and the cold storage compressor group based on the target frequency sequence, and generates a pre-cooling instruction to trigger the pre-cooling operation of the cold storage platform in response to the vehicle arrival time being less than a preset threshold, and collects the state data after execution and feeds back to the multi-dimensional state perception module.

[0054] Further, the frequency regulation for the vehicle-mounted air compressor and the cold storage compressor group includes the following steps: Parse the target frequency sequence and convert it into underlying control signals of the corresponding communication protocol according to the control object identifier; Perform safety check on the underlying control signals, trigger protection logic when monitoring that the device operating parameters exceed the physical limit, and issue execution when the check is passed; For the vehicle-mounted air compressor, the control signal is sent to the vehicle-mounted frequency converter driver to adjust the motor speed; For the cold storage compressor unit, the control signal is sent to the cold storage controller to adjust the refrigeration output parameter of the compressor unit.

[0055] Further, in response to the vehicle estimated arrival time being less than the preset threshold, a pre-cooling instruction is generated to trigger the pre-cooling operation of the cold storage platform, which specifically includes the following steps: Track the position of the refrigerated truck and calculate the estimated remaining time for the vehicle to arrive at the specified platform in combination with the road condition information; Determine whether the estimated remaining time is less than the cooperative pre-cooling time threshold based on the platform thermal capacity characteristic calibration; When it is determined that the condition is met, the target storage temperature of the goods in the multi-dimensional state vector is read and set as the pre-cooling target temperature of the platform area; A pre-cooling instruction is generated and sent to the cold storage control system to start the refrigeration equipment in the platform area to cool down and establish a low-temperature buffer zone.

[0056] Specifically, the cooperative execution and feedback module performs frequency conversion adjustment on the vehicle-mounted air compressor and the cold storage compressor unit based on the target frequency sequence, and generates a pre-cooling instruction when the vehicle estimated arrival time is less than the preset threshold to trigger the pre-cooling operation of the cold storage platform. During execution, the system analyzes the frequency sequence, converts the control signal, performs frequency conversion adjustment, and collects real-time state data feedback to the multi-dimensional state perception module.

[0057] The system first receives the target frequency sequence obtained by the robust optimization solving module, which is composed of optimized frequency values in the future time domain, denoted as: ; Where represents the discrete time step The target frequency value needs to be issued to the compressor, and the time step sequence is obtained by dividing the prediction control time domain. The system converts the above frequency value into the control signal required by the bottom layer communication protocol according to the device identification of the control object, denoted as: ; Where is the time step The corresponding control signal, represents the mapping function of converting the frequency value into the bottom layer communication format recognizable by the device driver, which will be specifically implemented as MODBUS or CAN communication frame coding rules in practice. The system then performs safety check on the generated bottom layer control signal, and the check content includes whether the current, voltage, coil temperature rise and other key physical operating parameters related to the compressor are out of limits. This check can be represented as: ; Where represents the value of the real-time monitored equipment operating parameter, represents the maximum threshold value allowed for the safe operation of the equipment, if is exceeded , the system immediately triggers the protection logic to block the execution flow, ensuring that the hardware equipment is not damaged.

[0058] Under the condition of passing the verification, the system sends a control signal to the vehicle-mounted air compressor or the cold storage compressor unit. For the vehicle-mounted air compressor, the control signal is sent to the vehicle-mounted frequency drive, which adjusts the motor speed driving the vehicle-mounted compressor. The motor speed can be represented as: ; where is the target speed of the motor at time step , and represents the conversion factor between the frequency and the synchronous speed of the motor. On the side of the cold storage compressor unit, the control signal is sent to the cold storage controller, which adjusts the output capacity of the refrigeration system according to , including key operating parameters such as compressor load, evaporation temperature, and condensing pressure, ultimately changing the actual refrigeration effect of the cold storage.

[0059] While implementing equipment adjustment, the system calculates the estimated remaining time for the transport vehicle to reach the cold storage platform according to the real-time location and path information of the vehicle. The remaining arrival time of the vehicle is represented as: ; where is the estimated remaining time, is the remaining driving distance of the vehicle from the unloading platform, is the average driving speed of the vehicle on the current section. The system compares the estimated remaining time with the cooperative pre-cooling threshold value, which is represented as: ; where is used to represent the minimum effective pre-cooling time required by the platform area according to its wall heat capacity, heat exchange conditions and refrigeration capacity. When the condition is met: ; , the system starts the pre-cooling process.

[0060] After entering the pre-cooling process, the system reads the target storage temperature of the goods in the multi-dimensional state vector, which is represented by the symbol: ; where represents the optimal storage temperature required by the goods carried by the vehicle, and the system sets this temperature as the target temperature for pre-cooling the platform area, which is represented by the symbol: ; Subsequently, the system generates pre-cooling instructions and sends them to the cold storage control system, so that the platform area refrigeration equipment operates according to the target temperature to establish a heat buffer zone in advance to avoid temperature rapid rise during the unloading process after the vehicle arrives, which leads to unstable temperature control of the goods.

[0061] After the system completes the above execution actions, the real-time motor speed of the on-board air compressor, the refrigeration output state of the cold storage compressor unit, and the measured temperature of the platform area are collected, and are fed back to the multi-dimensional state perception module through the communication interface, so that the next round of optimization control cycle can be calculated based on the latest state.

[0062] The cooperative execution and feedback module can ensure the safety and reliability of the control signal, synchronize the adjustment of the on-board equipment and the cold storage equipment, dynamically trigger the pre-cooling operation based on the vehicle arrival time, and provide a temperature-stable buffer environment for the goods handover, thereby significantly improving the temperature control safety and overall operation efficiency of the cold chain transportation process.

[0063] Finally, it should be noted that: the above only describes the preferred embodiments of the present application and does not limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or replace some technical features with equivalent ones, as long as they are within the spirit and principles of the present application. Any modification, equivalent replacement, improvement, etc. shall be included in the protection scope of the present application.

Claims

1. An air compressor vehicle-mounted refrigeration and cold storage collaborative temperature control system based on AI scheduling, characterized in that, The application relates to a refrigeration system state prediction and control method, which comprises the following steps: a multi-dimensional state perception module is used to collect real-time temperature field data, environmental meteorological parameters, compressor operation states and goods category information in a refrigerated truck and a cold storage room, and outputs a structured multi-dimensional state vector; a thermodynamic evolution prediction module is used to set a limited length prediction time domain based on the multi-dimensional state vector, to call a mathematical model for establishing a heat balance relationship between refrigerating capacity generated by compressor operation frequency and environmental and goods heat load, to calculate system state evolution advancing with time steps in the prediction time domain, and to output a prediction state sequence containing temperature change trajectory and energy consumption trend; a comprehensive cost construction module is used to construct a comprehensive operation cost function containing goods quality loss calculation item, equipment mechanical wear calculation item and energy consumption calculation item through a weighting method based on the prediction state sequence and goods category information; a robust optimization solving module is used to set a temperature safety boundary as a soft constraint condition and introduce a nonlinear penalty term based on the comprehensive operation cost function, to solve control quantity minimizing the comprehensive operation cost function through iterative calculation, and to output a target compressor frequency sequence; a cooperative execution and feedback module is used to execute frequency regulation on a vehicle-mounted air compressor and a cold storage compressor group based on the target frequency sequence, to generate a precooling instruction to trigger precooling operation of a cold storage platform in response to vehicle predicted arrival time being less than a preset threshold, and to collect state data after execution and feed back to the multi-dimensional state perception module.

2. The air compressor on-truck refrigeration and cold storage collaborative temperature control system based on AI scheduling according to claim 1, characterized in that, The output structured multi-dimensional state vector specifically comprises the following steps: obtaining sensor time sequence data, external meteorological interface data and goods category metadata of a vehicle-mounted end and a cold storage end to form a multi-source heterogeneous original data stream; carrying out clock synchronization calibration on the original data stream, and removing noise data by using a sliding average filtering algorithm to generate aligned physical parameter sequences; carrying out dimensionless normalization processing on continuous numerical data in the physical parameter sequences, and vectorizing discrete category data; splicing the processed data sequence and environmental disturbance variables at the current moment to construct the multi-dimensional state vector composed of state variables, disturbance variables and attribute variables.

3. The AI-based scheduling air compressor on-board refrigeration and cold storage cooperative temperature control system according to claim 1, characterized in that, The calculation of system state evolution advancing with time steps in the prediction time domain specifically comprises the following steps: taking the multi-dimensional state vector at the current moment as an initial state point of the prediction time domain, and establishing a time step after discretization processing and a prediction step number; obtaining corresponding environmental meteorological forecast data in the prediction time domain and respiratory heat data estimated based on goods category to construct an external disturbance sequence varying with time steps; for each time step in the prediction time domain, calculating the heat enthalpy change amount of the system based on a discretized heat balance equation, wherein the heat enthalpy change amount is jointly determined by heat transfer increment, internal energy increment and refrigeration decrement corresponding to the compressor operation frequency; calculating the temperature state value at the next time step by using the heat enthalpy change amount and total heat capacity parameters of the system, and taking the state value as input to recursively execute the step of calculating the heat enthalpy change amount until the temperature change trajectory of the whole time domain is generated.

4. The air compressor on-truck refrigeration and cold storage collaborative temperature control system based on AI scheduling according to claim 1, characterized in that, The step of constructing the comprehensive operation cost function by weighting includes the following steps: Obtain the unit price of energy, combine the power model corresponding to the compressor operating frequency, and calculate the cumulative energy consumption cost in the prediction time domain as the energy consumption calculation item; Retrieve the activation energy parameters corresponding to the cargo category, process the temperature change trajectory in the prediction time domain using the Arrhenius metamorphic kinetics model, calculate the decay cumulative value of the remaining shelf life of the cargo, and convert it into a quality decay equivalent value as the cargo quality loss calculation item; Calculate the square of the difference in compressor operating frequency between adjacent time steps in the prediction time domain, and take the cumulative result of the square value as the equipment mechanical wear calculation item; Assign a preset weight coefficient to each of the energy consumption calculation item, the cargo quality loss calculation item, and the equipment mechanical wear calculation item, and accumulate the weighted values of each item to generate the comprehensive operation cost function.

5. The AI-based scheduling air compressor on-board refrigeration and cold storage cooperative temperature control system according to claim 1, characterized in that, The step of setting the temperature safety boundary as a soft constraint condition and introducing a nonlinear penalty term includes the following steps: Obtain the boundary threshold of the temperature safety interval and set a relaxation penalty coefficient to represent the tolerance for temperature violations; Iterate through the temperature prediction values in the prediction state sequence, calculate the deviation of the temperature prediction value from the temperature safety interval, and set the deviation of the temperature prediction value within the safety interval to zero; Square the deviation to build a nonlinear relationship, multiply the squared result by the relaxation penalty coefficient, and obtain the penalty equivalent value corresponding to this time step; Accumulate the penalty equivalent values of all time steps in the prediction time domain and add them as a soft constraint cost item to the comprehensive operation cost function.

6. The AI-based scheduling air compressor on-board refrigeration and cold storage cooperative temperature control system according to claim 1, characterized in that, The step of solving the control amount that minimizes the comprehensive operation cost function through iterative calculation includes the following steps: Set the physical constraint set of the compressor frequency converter, which includes the allowed range of operating frequency and the gradient limit of frequency change rate; Convert the comprehensive operation cost function into a numerical optimization model and use the multi-dimensional state vector at the current time as the initial condition; Within the search space defined by the physical constraint set, iteratively optimize the candidate frequency sequence in the prediction time domain until the preset convergence termination condition is met, and obtain the full-time domain optimized frequency sequence; Extract the frequency value corresponding to the first time step in the full-time domain optimized frequency sequence as the output value at the current time, and discard the remaining prediction control amounts in the sequence.

7. The AI-based scheduling air compressor on-board refrigeration and cold storage cooperative temperature control system according to claim 1, characterized in that, The step of performing frequency regulation on the vehicle-mounted air compressor and the cold storage compressor set includes the following steps: Analyze the target frequency sequence and convert it into a bottom-layer control signal according to the communication protocol corresponding to the control object identifier; Perform safety verification on the bottom-layer control signal, trigger protection logic when the device operating parameters exceed the physical limit, and issue execution when the verification is passed; For the vehicle-mounted air compressor, send the control signal to the vehicle-mounted frequency drive to adjust the motor speed; For the cold storage compressor set, send the control signal to the cold storage controller to adjust the refrigeration output parameters of the compressor set.

8. The AI-based scheduling air compressor on-board refrigeration and cold storage cooperative temperature control system according to claim 1, characterized in that, The generating a pre-cooling instruction to trigger a pre-cooling operation of the cold storage platform in response to the vehicle estimated arrival time being less than a preset threshold specifically comprises the following steps: Tracking the position of the refrigerated vehicle and calculating the estimated remaining time for the vehicle to arrive at the designated platform in combination with road condition information; Determining whether the estimated remaining time is less than a cooperative pre-cooling time threshold calibrated based on the thermal capacity characteristics of the platform; When it is determined that the condition is met, reading the target storage temperature of the goods in the multi-dimensional state vector and setting it as the pre-cooling target temperature of the platform area; Generating a pre-cooling instruction and sending it to the cold storage control system to start the refrigeration equipment of the platform area to cool down and establish a low-temperature buffer zone.

Images