Quick-frozen temperature control method and system based on edge computing and digital twin cooperation

Abstract

The application relates to the technical field of computers and discloses a quick-freezing temperature control method and system based on edge computing and digital twin cooperation, aiming to solve the technical problems of temperature control delay, insufficient control precision and high energy consumption caused by the cloud processing mode in the prior art. The method comprises the following steps: deploying an edge computing node in a quick-freezing chamber and running a digital twin model; collecting temperature data in real time through a temperature sensor and sending the temperature data to the edge computing node; performing filtering, calibration and feature extraction preprocessing on the temperature data; updating the digital twin model and performing temperature trend prediction; performing control strategy calculation based on the prediction result to generate an adjustment instruction; and issuing the adjustment instruction to a refrigeration system for execution. The system comprises a quick-freezing chamber, a refrigeration system, a plurality of temperature sensors, an edge computing node and a digital twin model. Through the above scheme, the application can achieve the effects of reducing temperature control delay, improving control precision, enhancing system reliability, reducing network load and reducing energy consumption.

Description

Technical Field

[0001] This invention relates to the field of computer technology, specifically to a quick-freezing temperature control method and system based on edge computing and digital twin collaboration. Background Technology

[0002] Quick-freezing technology is a key technology in modern food processing, cold chain logistics, and biomedicine. By rapidly lowering the temperature of products below freezing, it effectively inhibits microbial growth while maintaining the original structure and nutritional components of food, playing a vital role in ensuring food safety and extending product shelf life. With increasingly stringent requirements for food quality and the rapid development of cold chain logistics systems, the application scope of quick-freezing technology continues to expand, and the demands for temperature control precision and response speed are becoming increasingly stringent.

[0003] In the actual operation of a quick-freezing temperature control system, temperature monitoring and regulation are the core aspects of ensuring quick-freezing quality. Traditional quick-freezing temperature control methods mainly rely on a centralized cloud computing architecture for data processing and decision-making, that is, temperature data analysis, processing, and control strategy distribution are completed on cloud servers. However, this architecture has significant technical bottlenecks when facing the quick-freezing production environment: the temperature changes extremely rapidly inside the quick-freezing chamber, and even a slight delay can lead to temperature overshoot or undershoot, resulting in uneven ice crystal formation and a decline in product quality; in addition, quick-freezing production lines are usually deployed in industrial scenarios with complex network environments, and network fluctuations or interruptions will directly affect the real-time performance and reliability of temperature control commands, while the frequent uploading of a large amount of temperature sensor data can also cause network congestion, further exacerbating the control lag problem.

[0004] Due to the inherent latency of cloud processing in existing technologies, temperature control systems cannot effectively cope with rapid temperature fluctuations during quick-freezing, resulting in problems such as lag in adjustment, insufficient control precision, and high energy consumption. This has become a key obstacle restricting the further development of quick-freezing temperature control technology and urgently needs to be addressed through new technological approaches. Summary of the Invention

[0005] The purpose of this invention is to provide a quick-freezing temperature control method and system based on edge computing and digital twin collaboration, which can effectively solve the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by this invention is: a quick-freezing temperature control method based on edge computing and digital twin collaboration, comprising the following specific steps: S1: Deploy edge computing nodes in the quick-freezing chamber and run a digital twin model in the edge computing nodes; the edge computing nodes are used to process and analyze temperature data locally in the quick-freezing chamber, as well as to calculate control strategies in real time; the digital twin model is used to construct a virtual mapping of the temperature field in the quick-freezing chamber and to predict temperature change trends. S2: Temperature data is collected in real time by multiple temperature sensors deployed in the quick-freezing chamber, and the temperature data is sent to the edge computing node; S3: The collected temperature data is preprocessed in the edge computing node, including data filtering, calibration and feature extraction, to remove noise and outliers and extract key features of temperature changes; S4: Based on the preprocessed temperature data, update the virtual mapping of the temperature field in the digital twin model, and use the updated digital twin model to predict the temperature change trend to obtain the predicted temperature curve. S5: Based on the predicted temperature curve, perform control strategy calculation in the edge computing node to generate adjustment instructions for the refrigeration system; S6: The adjustment command is sent to the refrigeration system of the quick-freezing chamber for execution, so as to realize real-time control of the temperature of the quick-freezing chamber.

[0007] Preferably, in step S1, the edge computing node adopts an industrial-grade embedded computing platform, which has waterproof and dustproof performance, is suitable for the low temperature and humid environment of the quick-freezing room, supports high-frequency floating-point operations, and has data storage capacity that meets preset requirements.

[0008] Preferably, in step S1, the digital twin model is trained based on historical operating data and includes a quick-freezing chamber thermodynamic model, a refrigeration system response model, and an environmental disturbance model. It is constructed using a deep neural network, with the current temperature state and recent temperature sequence as inputs and the temperature prediction value within a preset future time window as outputs.

[0009] Preferably, in step S2, the temperature sensor uses a high-precision thermocouple array, which is distributed to cover the spatial grid of the quick-freezing chamber. The sampling frequency meets the preset requirements, the measurement accuracy meets the high-precision requirements, and the sensor data is transmitted to the edge computing node in real time through the industrial bus.

[0010] Preferably, in step S3, the data preprocessing adopts an adaptive Kalman filter algorithm, and the filter parameters are dynamically adjusted according to the temperature change; the feature extraction includes the statistical features of the current temperature value, the rate of temperature change, the temperature gradient distribution, and the historical temperature curve.

[0011] Preferably, in step S4, the update frequency of the digital twin model is synchronized with the temperature data acquisition frequency, and the model is updated every time temperature data is acquired; the prediction time window covers the temperature changes within a preset time period in the future, and the prediction results are output in the form of a probability distribution.

[0012] Preferably, in step S5, the control strategy calculation adopts a model predictive control algorithm, with the weighted sum of temperature tracking error and energy consumption as the objective function, to solve for the optimal adjustment command under constraints; the refrigeration system adjustment command includes compressor start / stop status, refrigerant flow setpoint and expansion valve opening.

[0013] Preferably, in step S6, the adjustment command is sent to the refrigeration system for execution in real time via the industrial control bus, and the control cycle meets the preset requirements to ensure that the temperature control response speed meets the requirements of the rapid temperature change in the quick-freezing process.

[0014] This invention also proposes a quick-freezing temperature control system based on edge computing and digital twin collaboration.

[0015] According to a second aspect of the present invention, a quick-freezing temperature control system includes: a quick-freezing chamber, a refrigeration system, multiple temperature sensors, an edge computing node, and a digital twin model; the multiple temperature sensors are deployed inside the quick-freezing chamber for real-time temperature data acquisition; the edge computing node is deployed locally in the quick-freezing chamber and is communicatively connected to the temperature sensors and the refrigeration system for local temperature data processing, control strategy calculation, and regulation command issuance; the digital twin model runs on the edge computing node for constructing a virtual mapping of the temperature field in the quick-freezing chamber and predicting temperature change trends.

[0016] Compared with the prior art, the beneficial effects achieved by the present invention are: 1. Reduce temperature control latency. By deploying edge computing nodes locally in the blast freezer, data processing and control strategy calculations are migrated from the cloud to the edge, completely eliminating the time latency caused by network communication. The control response speed meets the requirements for rapid response and effectively avoids temperature overshoot and undershoot.

[0017] II. Improved Control Precision. Based on the temperature prediction function of the digital twin model, the control strategy can detect temperature change trends in advance and actively adjust accordingly, achieving feedforward control. The temperature control precision is significantly improved compared to traditional feedback control methods, resulting in more uniform ice crystal formation and improved product quality.

[0018] Third, enhanced system reliability. Edge computing nodes operate locally without relying on cloud network connections, and can maintain normal temperature control functions even in the event of a network outage, significantly improving system availability.

[0019] Fourth, reduce network load. Temperature data is preprocessed and analyzed at the edge, with only key control results uploaded to the cloud, significantly reducing the amount of data transmitted over the network and effectively avoiding network congestion.

[0020] V. Reducing energy consumption. The operation strategy of the refrigeration system is optimized through the model predictive control algorithm, so as to reduce unnecessary refrigeration energy input on the premise of ensuring temperature control accuracy, and achieve energy saving and consumption reduction. Brief Description of the Drawings

[0021] Figure 1 is a schematic diagram of the overall technical process of the quick-freezing temperature control system based on the collaboration of edge computing and digital twin proposed by the present invention; Figure 2 is a schematic diagram of the core process of the collaborative work of the edge computing node and the digital twin model in the present invention; Figure 3 is a schematic diagram of the logical process of temperature data collection, preprocessing and digital twin model update in the present invention; Figure 4 is a schematic diagram of the process of control strategy calculation and refrigeration system adjustment instruction issuance based on the predicted temperature curve in the present invention. Detailed Description of the Embodiments

[0022] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by a person of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0023] Embodiment 1 This embodiment is applied to the core temperature control link of the quick-frozen food processing production line. In the quick-freezing workshop, precise temperature control of the quick-freezing chamber is required to quickly reduce the core temperature of the food to below minus 18 degrees Celsius, while ensuring the uniformity of temperature distribution and avoiding the degradation of food quality caused by uneven ice crystal formation. By deploying the collaborative work of edge computing nodes and digital twin models, this embodiment achieves millisecond-level temperature control response speed and higher temperature control accuracy, effectively solving the delay problem caused by the traditional cloud processing mode.

[0024] First, a quick-freezing temperature control system including a quick-freezing chamber, a refrigeration system, a plurality of temperature sensors, an edge computing node and a digital twin model is deployed in this scenario. As the core physical space, the quick-freezing chamber has an internal space size of 6 meters in length, 4 meters in width and 3 meters in height, adopts a double-layer thermal insulation structure, and the wall is filled with polyurethane thermal insulation material with a thickness of 150 millimeters to ensure good thermal insulation performance. The door body of the quick-freezing chamber is configured with a fast rolling shutter door, the size of the doorway is 2.5 meters in width and 2.8 meters in height, and the opening speed reaches 1.5 meters per second to reduce the cold loss when the door is opened. A cargo conveying system is arranged inside the quick-freezing chamber, which adopts a three-dimensional shelf structure with a total of three layers of shelves, each layer of shelf is 1.2 meters in height, and the cargo loading capacity reaches 80 kilograms per square meter.

[0025] The refrigeration system includes a refrigeration unit, condenser, evaporator, expansion valve, and connecting piping. The refrigeration unit uses a screw compressor with a rated cooling capacity of 150 kW, equipped with variable frequency speed control, allowing continuous adjustment of output power between 20% and 100%. The condenser is air-cooled, equipped with six axial fans, each with an airflow of 15,000 cubic meters per hour. The evaporator uses a fan-cooled design, suspended from the top of the quick-freezing chamber, with four units installed. Each evaporator has a heat exchange area of ​​80 square meters and is equipped with twelve variable-speed fans, allowing for speed adjustment according to refrigeration load requirements. The expansion valve uses a combination of thermostatic and electronic expansion valves. The thermostatic expansion valve serves as the basic regulating element, while the electronic expansion valve precisely controls the refrigerant flow, achieving fine-tuning of the evaporation temperature. The refrigeration system also includes auxiliary equipment such as an oil separator, gas-liquid separator, liquid receiver, and dryer filter to ensure system reliability.

[0026] Multiple temperature sensors are deployed inside the quick-freezing chamber, forming a spatially gridded temperature monitoring network. The temperature sensors employ a high-precision thermocouple array, specifically K-type thermocouples, with a measurement range of -60°C to +20°C and a measurement accuracy of ±0.1°C. The sensors are spatially distributed throughout the three-dimensional space of the quick-freezing chamber, with three measurement levels along the height, located at 0.5 meters, 1.5 meters, and 2.5 meters above the ground, respectively. Nine measuring points are evenly arranged in a 9x9 grid on each measurement level, totaling twenty-seven temperature sensors. The temperature sensors are connected to the signal acquisition module via compensating wires. These wires are shielded and their length is limited to 30 meters to minimize signal attenuation and interference during transmission. The signal acquisition module has a 16-bit A / D conversion resolution and a sampling frequency of ten times per second, meaning it acquires data from each temperature sensor ten times per second, enabling rapid capture of temperature changes.

[0027] The edge computing nodes are deployed locally within the blast freezer, utilizing an industrial-grade embedded computing platform. The main structure of the edge computing node employs a fully sealed aluminum alloy chassis with an IP65 protection rating, providing waterproof and dustproof performance, enabling stable operation in the low-temperature and humid environment of the blast freezer. Internally, the chassis houses a high-performance multi-core processor, specifically an eight-core ARM Cortex-A72 architecture with a clock speed of 2.5 GHz, supporting millions of floating-point operations per second to meet the demands of real-time data processing and control strategy calculations. The edge computing node is equipped with 8GB of RAM and a 256GB solid-state drive, with a data storage capacity of at least 128GB for storing historical operating data, digital twin model parameters, and intermediate calculation results. The edge computing node also features abundant communication interfaces, including four RS485 serial ports, two Ethernet ports, eight digital input interfaces, and four analog input interfaces, connecting to temperature sensors and the refrigeration system via an industrial bus. The edge computing node runs an embedded Linux operating system, equipped with real-time kernel patches to ensure the timeliness of control command issuance, with a control cycle not exceeding 100 milliseconds.

[0028] The digital twin model runs on edge computing nodes and is trained based on historical operational data. It comprises three sub-models: a blast freezer thermodynamic model, a refrigeration system response model, and an environmental disturbance model. The blast freezer thermodynamic model describes the variation of the internal temperature field, considering factors such as wall heat transfer, cold storage of goods, fan circulation, and heat intrusion from door opening. It is constructed using a deep neural network. The refrigeration system response model establishes the mapping relationship between refrigeration system operating parameters and blast freezer temperature changes. Input parameters include compressor operating frequency, expansion valve opening, and fan speed. The output is the refrigeration capacity response curve of the refrigeration system under different operating conditions. The environmental disturbance model quantifies the impact of external environmental factors on the blast freezer temperature, including ambient temperature changes, heat load fluctuations caused by goods entering the freezer, and cold loss due to door opening. The digital twin model's input consists of the current temperature state and recent temperature sequences. Specifically, the input dimensions are the current values ​​from twenty-seven temperature sensors plus ten historical values ​​sampled every six seconds over the past sixty seconds, totaling a 378-dimensional input vector. The output is the temperature prediction value for one prediction point every two seconds within the next thirty seconds, for a total of fifteen prediction times. The output is in the form of a probability distribution, including the expected temperature value and variance information for each prediction time.

[0029] The workflow of this system is as follows: First, the data acquisition module executes step S2, collecting temperature data in real time through multiple temperature sensors deployed inside the quick-freezing chamber. The temperature sensors continuously collect data at a set sampling frequency, with each sensor generating ten temperature samples per second. Twenty-seven temperature sensors operate simultaneously, generating a total of 270 temperature data points per second. The temperature data is transmitted in real time to the edge computing node via an industrial bus, using the Modbus RTU protocol with a baud rate of 9600 bits per second. After receiving the raw data from the temperature sensors, the data acquisition module performs format parsing and preliminary verification, storing the data in a circular buffer for subsequent processing.

[0030] Subsequently, the data preprocessing module executes step S3, preprocessing the collected temperature data in the edge computing node, including three stages: data filtering, calibration, and feature extraction. Data filtering employs an adaptive Kalman filter algorithm, which automatically adjusts filtering parameters based on the dynamic characteristics of temperature changes. When temperature changes are relatively stable, the filter's process noise covariance matrix is ​​small, resulting in a strong filtering effect and effective suppression of measurement noise. When temperature changes rapidly, the filter's process noise covariance matrix automatically increases, reducing the response delay caused by filtering and ensuring rapid tracking of temperature changes. The calibration stage performs zero-point and range calibration on each temperature sensor. Calibration parameters are stored in the edge computing node's configuration file and updated quarterly. The feature extraction stage extracts key features from the preprocessed temperature data, including the current temperature value, temperature change rate, temperature gradient distribution, and statistical features of historical temperature curves. The current temperature value is the real-time measurement from twenty-seven sensors; the rate of temperature change is the average rate of temperature change over the last six seconds; the temperature gradient distribution is obtained by calculating the standard deviation of temperature values ​​from nine sensors at the same altitude level, used to assess the uniformity of the temperature field; the statistical characteristics of the historical temperature curve include the mean, variance, maximum, and minimum values ​​of the temperature sequence over the past sixty seconds, used to describe the long-term trend of temperature change.

[0031] The model update module executes step S4, updating the virtual temperature field mapping in the digital twin model based on the preprocessed temperature data. The update frequency of the digital twin model is synchronized with the temperature data acquisition frequency; a model update is triggered every time temperature data acquisition is completed. Specifically, each time the data acquisition module generates a new set of temperature data, the model update module immediately inputs this set of data into the digital twin model to update the model's internal state. The model update uses a sliding window method with a window length of sixty seconds. As new data enters the window, the oldest data is removed, maintaining a constant time span for the input data. After the model update is complete, the updated digital twin model is used to predict temperature change trends, obtaining a predicted temperature curve. The prediction time window covers temperature changes within the next thirty seconds, and the prediction results are output in the form of a probability distribution, including the expected temperature value and variance at each prediction time. The prediction results are stored in a prediction result buffer for use by the control strategy calculation module.

[0032] The control strategy calculation module executes step S5, performing control strategy calculations in the edge computing nodes based on the predicted temperature curve to generate adjustment commands for the refrigeration system. The control strategy calculation employs a model predictive control algorithm, which uses a weighted sum of temperature tracking error and energy consumption as the objective function to solve for the optimal adjustment command under constraints. The specific form of the objective function is J equal to weight 1 multiplied by the square of the temperature error, plus weight 2 multiplied by the square of the energy consumption, plus weight 3 multiplied by the square of the control action change. The temperature error is the deviation between the predicted temperature curve and the set temperature; energy consumption is proportional to the compressor operating frequency and fan speed; and the control action change is the change between the current control command and the control command at the previous moment, used to avoid drastic fluctuations in control actions. The weight parameters are set according to actual operating requirements: the temperature error weight is set to 10, the energy consumption weight to 1, and the control action change weight to 5. The constraints include upper and lower limits of compressor operating frequency of 20% and 100%, upper and lower limits of expansion valve opening of 20% and 80%, and upper and lower limits of fan speed of 300 rpm and 1500 rpm, respectively. At the beginning of each control cycle, the control strategy calculation module retrieves the latest predicted temperature curve from the prediction result buffer. Combining this with the current temperature state and system constraints, it solves the aforementioned optimization problem to obtain the optimal control command. The refrigeration system adjustment commands include three parameters: compressor start / stop status, refrigerant flow setpoint, and expansion valve opening. The compressor start / stop status is a binary variable, with one indicating operation and zero indicating shutdown. The refrigerant flow setpoint is expressed as a percentage, ranging from zero to one hundred percent; the expansion valve opening is also expressed as a percentage.

[0033] Finally, the instruction issuing module executes step S6, sending the adjustment instruction to the refrigeration system of the quick-freezing chamber for execution. The adjustment instruction is sent to the refrigeration system in real time via the industrial control bus, with a control cycle not exceeding one hundred milliseconds, ensuring that the temperature control response speed meets the requirements of the rapid temperature changes during the quick-freezing process. Specifically, the edge computing node communicates with the controller of the refrigeration system via an RS485 interface, using the Modbus TCP protocol. Compressor start / stop instructions are sent to the compressor's control contactor via a digital output channel, the refrigerant flow setpoint is sent to the frequency converter via an analog output channel to adjust the compressor's operating frequency, and the expansion valve opening is sent to the drive motor of the electronic expansion valve via a pulse signal. Each control cycle is one hundred milliseconds. At the end of each cycle, the edge computing node calculates a new adjustment instruction and sends it to the refrigeration system via the communication interface. Upon receiving the instruction, the refrigeration system controller immediately executes the corresponding action, achieving real-time control of the quick-freezing chamber temperature.

[0034] Through the above workflow, the quick-freezing temperature control system in this embodiment achieves real-time processing and analysis of temperature data and calculation of control strategies locally within the quick-freezing chamber. Edge computing nodes are deployed near the quick-freezing chamber, with data processing latency controlled to within ten milliseconds. Compared to the latency of hundreds of milliseconds or even seconds in traditional cloud processing models, this represents an order-of-magnitude improvement in response speed. The digital twin model's accurate prediction of temperature changes over the next thirty seconds allows the control strategy to anticipate temperature trends and proactively adjust, achieving feedforward control and effectively avoiding temperature overshoot and undershoot. Temperature control accuracy reaches within ±0.5 degrees Celsius, significantly improving temperature field uniformity, resulting in more uniform ice crystal formation and effectively enhancing the quality of quick-frozen products.

[0035] Example 2 Different filtering algorithms and feature extraction strategies are used in the data preprocessing stage, which are suitable for biopharmaceutical quick-freezing scenarios with higher requirements for temperature control accuracy.

[0036] In the edge computing nodes deployed within the quick-freezing chamber, the data preprocessing module employs a wavelet transform denoising algorithm instead of the adaptive Kalman filter algorithm. The wavelet transform denoising algorithm first performs a three-level wavelet decomposition on the original temperature signal, using the db4 wavelet basis function to decompose the temperature signal into low-frequency approximate components and high-frequency detail components. Then, thresholding is applied to the high-frequency components, with the threshold adaptively determined based on the signal noise level. A hard thresholding function is used to preserve the true signal characteristics while suppressing noise. Finally, the denoised temperature signal is obtained through wavelet reconstruction. The wavelet transform denoising algorithm can effectively distinguish between abrupt changes in temperature signals and measurement noise, and is particularly suitable for signal processing during the rapid temperature drop phase of quick-freezing.

[0037] Building upon Example 1, the feature extraction process adds spatial temperature correlation and periodicity features. Spatial temperature correlation features are obtained by calculating the correlation coefficient between temperatures at different altitudes, reflecting the efficiency of vertical temperature transfer. Periodicity features are extracted by performing Fourier transforms on historical temperature sequences to identify periodic components of temperature changes, used to predict periodic temperature fluctuations. These new features can more comprehensively describe the dynamic characteristics of the temperature field in the freezer, improving the accuracy of the digital twin model's predictions.

[0038] Regarding the control strategy, this embodiment employs a constrained linear quadratic regulator instead of the model predictive control algorithm. The state vector of the linear quadratic regulator includes the temperature error, the temperature error integral, and the control variable. The control gain matrix is ​​obtained by solving the discrete algebraic Riccati equation. This algorithm has a lower computational load, enabling a faster control cycle, reduced to fifty milliseconds, further improving the temperature control response speed. The generation of refrigeration system adjustment commands incorporates an assessment of the evaporator frosting level. When the frosting thickness exceeds a set threshold, a defrosting cycle is automatically triggered to prevent a decrease in refrigeration efficiency due to evaporator frosting.

[0039] This embodiment is applicable to the rapid freezing and preservation of pharmaceutical products that are highly sensitive to temperature, such as vaccines, blood products, and biological samples. The temperature control accuracy reaches ±0.3 degrees Celsius, and the heating rate is controlled within 0.5 degrees Celsius per minute to ensure that the cellular structure of bioactive substances is not damaged.

[0040] Example 3 Replacing centralized edge computing nodes with a distributed edge computing architecture is suitable for large-scale quick-freezing production lines that require simultaneous control of multiple quick-freezing chambers.

[0041] While deploying local edge computing nodes in each blast freezer compartment, regional edge computing servers are also set up throughout the blast freezer workshop. Local edge computing nodes are responsible for temperature data acquisition, preprocessing, and control command issuance within their respective blast freezers, while regional edge computing servers handle coordinated control and global optimization across multiple blast freezers. The regional edge computing servers utilize high-performance industrial servers, configured with dual Xeon processors and 512GB of memory, providing enhanced data storage and computing capabilities.

[0042] The local edge computing node and the regional edge computing server are connected via gigabit industrial Ethernet, with communication latency controlled to within five milliseconds. The regional edge computing server runs a globally optimized digital twin model, which, based on the digital twin model of a single blast freezer compartment, adds a shared resource scheduling model and a load balancing model for the refrigeration system. The shared resource scheduling model considers the scenario where multiple blast freezers share refrigeration units, optimizing the allocation of refrigeration capacity among the compartments and avoiding overloading of the refrigeration units. The load balancing model adjusts the refrigeration strategy in advance based on the operating status of each blast freezer compartment and the goods receiving plan, achieving global optimization of temperature control across multiple blast freezers.

[0043] The control strategy employs a hierarchical architecture. The regional edge computing server calculates the global optimization target every second and distributes temperature setpoint correction parameters for each quick-freezing chamber to each local edge computing node. Based on the received correction parameters and the local state of its quick-freezing chamber, the local edge computing node calculates specific control commands. This hierarchical control architecture ensures the control accuracy of individual quick-freezing chambers while achieving coordinated operation of multiple chambers, thus improving the overall operating efficiency of the quick-freezing production line.

[0044] This embodiment is applicable to quick-freezing production lines in large food processing enterprises with a daily processing capacity of tens of tons. Eight quick-freezing chambers operate simultaneously, and unified temperature management and energy consumption optimization are achieved through a regional edge computing server. Compared to solutions where each quick-freezing chamber is controlled independently, this embodiment improves the overall energy efficiency of the refrigeration system by more than 15%, increases equipment utilization by 20%, and effectively reduces production costs.

[0045] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A quick-freezing temperature control method based on edge computing and digital twin collaboration, characterized in that, Includes the following steps: S1: Deploy edge computing nodes in the quick-freezing chamber and run a digital twin model in the edge computing nodes; the edge computing nodes are used to process and analyze temperature data locally in the quick-freezing chamber, as well as to calculate control strategies in real time; the digital twin model is used to construct a virtual mapping of the temperature field in the quick-freezing chamber and to predict temperature change trends. S2: Temperature data is collected in real time by multiple temperature sensors deployed in the quick-freezing chamber, and the temperature data is sent to the edge computing node; S3: The collected temperature data is preprocessed in the edge computing node, including data filtering, calibration and feature extraction, to remove noise and outliers and extract key features of temperature changes; S4: Based on the preprocessed temperature data, update the virtual mapping of the temperature field in the digital twin model, and use the updated digital twin model to predict the temperature change trend to obtain the predicted temperature curve. S5: Based on the predicted temperature curve, perform control strategy calculation in the edge computing node to generate adjustment instructions for the refrigeration system; S6: The adjustment command is sent to the refrigeration system of the quick-freezing chamber for execution, so as to realize real-time control of the temperature of the quick-freezing chamber.

2. A quick-freezing temperature control system based on edge computing and digital twin collaboration, characterized in that, include: The system includes a quick-freezing chamber, a refrigeration system, multiple temperature sensors, edge computing nodes, and a digital twin model; the multiple temperature sensors are deployed inside the quick-freezing chamber for real-time temperature data acquisition. The edge computing node is deployed locally in the blast freezer and is communicatively connected to the temperature sensor and the refrigeration system. It is used to perform temperature data processing, control strategy calculation, and regulation command issuance locally. The digital twin model runs in the edge computing node and is used to construct a virtual mapping of the temperature field in the blast freezer and predict temperature change trends.

3. The quick-freezing temperature control method as described in claim 1, characterized in that, The edge computing node adopts an industrial-grade embedded computing platform, which is waterproof and dustproof, suitable for the low temperature and humid environment of the quick-freezing room, and its computing power supports high-frequency floating-point operations. The data storage capacity meets the preset requirements.

4. The quick-freezing temperature control method as described in claim 1, characterized in that, The digital twin model is trained based on historical operating data and includes a thermodynamic model of the quick-freezing chamber, a response model of the refrigeration system, and an environmental disturbance model. It is constructed using a deep neural network, with the current temperature state and recent temperature sequence as inputs and the temperature prediction value within a preset future time window as output.

5. The quick-freezing temperature control method as described in claim 1, characterized in that, The temperature sensor uses a high-precision thermocouple array, which is distributed to cover the spatial grid of the quick-freezing chamber. The sampling frequency meets the preset requirements, the measurement accuracy meets the high-precision requirements, and the sensor data is transmitted to the edge computing node in real time through the industrial bus.

6. The quick-freezing temperature control method as described in claim 1, characterized in that, The data preprocessing employs an adaptive Kalman filter algorithm, dynamically adjusting the filter parameters based on temperature changes; feature extraction includes statistical features of the current temperature value, temperature change rate, temperature gradient distribution, and historical temperature curves.

7. The quick-freezing temperature control method as described in claim 1, characterized in that, The update frequency of the digital twin model is synchronized with the temperature data acquisition frequency. The model is updated every time temperature data is acquired. The prediction time window covers the temperature changes within a preset time period in the future, and the prediction results are output in the form of a probability distribution.

8. The quick-freezing temperature control method as described in claim 1, characterized in that, The control strategy calculation adopts a model predictive control algorithm, with the weighted sum of temperature tracking error and energy consumption as the objective function, and solves for the optimal adjustment command under constraints; the refrigeration system adjustment command includes compressor start / stop status, refrigerant flow setpoint and expansion valve opening.

9. The quick-freezing temperature control method as described in claim 1, characterized in that, The adjustment commands are sent to the refrigeration system in real time via the industrial control bus for execution. The control cycle meets the preset requirements, ensuring that the temperature control response speed meets the requirements of the rapid temperature change in the quick-freezing process.

10. The quick-freezing temperature control system as described in claim 2, characterized in that, The quick-freezing chamber adopts a double-layer insulation structure, with the walls filled with insulation material; the refrigeration system includes a refrigeration unit, a condenser, an evaporator, and an expansion valve; the edge computing node communicates with the temperature sensor and the refrigeration system via an industrial bus.

Images