Control method and device of refrigerating system and refrigerating system

Abstract

The invention provides a control method and device of a refrigerating system and the refrigerating system, and relates to the field of refrigerating control. The method comprises the following steps: acquiring multidimensional operation parameters including thermodynamic, electrical and mechanical vibration parameters in real time; dominant state analysis is carried out according to thermodynamic and electrical parameters to obtain dominant state indexes, and meanwhile, recessive state analysis is carried out according to mechanical vibration parameters to diagnose compressor valve plate leakage and refrigerant oil mixing states; in the digital twinborn model, taking the lowest comprehensive operation cost of the system in a future optimization time domain as a target, constructing and solving an optimization problem including energy consumption, performance degradation and reliability risk cost, and obtaining a comprehensive optimization instruction of the opening degree of a liquid supply valve and the rotating speed of a compressor; and adjusting the valve opening and the rotating speed of the compressor according to the instruction. According to the method, dynamic collaborative optimization of liquid supply and operation of the refrigerating system is achieved, and the energy efficiency and economical efficiency of the system are remarkably improved while the reliability is guaranteed.

Description

Technical Field

[0001] This application relates to the field of refrigeration control, and more specifically, to a control method, apparatus and refrigeration system for a refrigeration system. Background Technology

[0002] In refrigeration systems, precise control of the refrigerant supply is crucial for ensuring efficient, stable, and safe operation. Appropriate refrigerant flow rate and superheat control directly affect the system's energy efficiency, equipment lifespan, and operational reliability.

[0003] Currently, traditional refrigeration systems generally use a liquid supply method based on a fixed circulation rate. This method typically relies on preset, static control logic and cannot make dynamic and precise adjustments based on real-time changes in system load, environmental conditions, and the health status of the equipment itself. Summary of the Invention

[0004] The purpose of this application is to provide a control method, apparatus, and system for a refrigeration system, which dynamically adjusts the forward-looking range of optimization planning based on the real-time operating status of the system. This method enables the system to perform long-term economic optimization under stable operating conditions and quickly switch to short-term robust regulation when anomalies occur, thereby balancing energy efficiency, safety, and equipment lifespan.

[0005] In a first aspect, embodiments of this application provide a control method for a refrigeration system, the control method comprising: Step S1: Real-time acquisition of multi-dimensional operating parameters of the refrigeration system; wherein, the operating parameters include at least thermodynamic parameters, electrical parameters, and mechanical vibration parameters; Step S2: Based on the thermodynamic parameters and electrical parameters, performing explicit state analysis on the refrigeration system to obtain explicit state indices; performing implicit state analysis on the refrigeration system based on the mechanical vibration parameters to obtain implicit state analysis results; wherein, the implicit state analysis includes diagnosis of compressor valve leakage state and refrigerant mixing with lubricating oil state; Step S3: In a digital twin model, with the overall objective of minimizing the comprehensive operating cost of the system within a future optimization time domain, constructing and solving an optimization problem including equipment energy consumption cost, performance degradation cost, and reliability risk cost, to obtain a comprehensive optimization instruction within the current control cycle; wherein, the decision variables of the optimization problem include at least the opening sequence of the liquid supply valve and the speed sequence of the compressor; Step S4: Based on the comprehensive optimization instruction, adjusting the opening of the liquid supply valve to the target opening and adjusting the speed of the compressor to the target speed.

[0006] In the above implementation process, the refrigeration system control method provided in this application achieves comprehensive state perception of the refrigeration system by integrating multi-dimensional parameters such as thermodynamics, electrical, and mechanical vibration. Based on this, it not only performs explicit performance analysis but also introduces the diagnosis of implicit health conditions such as compressor valve leakage and refrigerant mixing. Furthermore, using a digital twin model, with the goal of minimizing the comprehensive operating cost including energy consumption, equipment degradation, and risks, it continuously solves for and executes the optimal coordinated control commands for the liquid supply valve and compressor. This enables integrated intelligent control of system health status and economic operation, moving from passive response to proactive optimization, thereby significantly improving the overall energy efficiency and economy of the refrigeration system while ensuring long-term reliable operation of the equipment.

[0007] Optionally, in this embodiment, step S2, which involves performing latent state analysis based on mechanical vibration parameters, includes: performing wavelet transform on the mechanical vibration parameters to decompose them into multiple sub-signals of different frequencies; extracting the energy features of each sub-signal to form a feature vector; inputting the feature vector into a pre-trained multi-layer neural network model to obtain the output of the neural network model, including the classification results of compressor valve leakage state and refrigerant mixed with lubricating oil state; wherein, the output layer of the neural network model is configured to perform joint multi-label classification of compressor valve leakage state and refrigerant mixed with lubricating oil state; compressor valve leakage state includes three categories: normal, slight leakage, and severe leakage, and refrigerant mixed with lubricating oil state includes three categories: no mixing, small amount mixing, and large amount mixing.

[0008] In the above implementation process, the control method for the refrigeration system provided in this application combines wavelet transform with deep learning neural networks to achieve automatic, precise, and joint diagnosis of two types of latent faults: valve plate leakage in the refrigeration compressor and refrigerant mixing. It can automatically extract time-frequency features strongly correlated with the fault from complex vibration signals and utilize the powerful nonlinear fitting capability of neural networks for high-precision classification. This overcomes the shortcomings of traditional diagnostic methods, such as reliance on human experience, single feature extraction, and high false alarm rates, greatly improving the intelligence level of condition monitoring and early fault detection capabilities, and laying a solid data foundation for predictive maintenance and reliability-based optimization control.

[0009] Optionally, in this embodiment, step S2, which involves performing implicit state analysis based on mechanical vibration parameters, includes: fuzzifying the mechanical vibration parameters to obtain an input fuzzy set; inputting the input fuzzy set into a preset fuzzy inference system; wherein the fuzzy inference system includes a fuzzy rule base determined based on historical fault data and vibration characteristics; performing inference through the fuzzy inference system to output fuzzy results regarding the possibility of compressor valve plate leakage and refrigerant mixing with lubricating oil; and defuzzifying the fuzzy results to obtain implicit state analysis results expressed as percentages.

[0010] In the above implementation process, the refrigeration system control method provided in this application is based on a fuzzy logic-based implicit state analysis method. It does not rely on a precise mathematical model of the fault, but instead simulates the experience-based judgment process of human experts, utilizing a fuzzy rule base to handle the uncertainties and nonlinear characteristics in vibration signals. This enables effective identification and probability assessment of early implicit faults such as compressor valve plate leakage and refrigerant mixing, even with limited data samples or complex fault mechanisms, providing crucial state inputs for subsequent risk cost calculation and optimized control.

[0011] Optionally, in this embodiment of the application, the performance degradation cost in step S3 is calculated in the following way: based on the compressor operating state in the digital twin model, the compressor life micro-loss increment caused by the current control decision is calculated according to the preset life prediction model; the life micro-loss increment is multiplied by the preset unit replacement cost coefficient to obtain the performance degradation cost.

[0012] In the above implementation process, the refrigeration system control method provided in this application transforms the physical equipment wear that is difficult to compare directly into quantifiable and comparable economic costs, enabling the digital twin optimization model to endogenously consider the long-term impact of control decisions, thereby achieving true full life cycle cost optimization and avoiding short-sighted behavior of excessively damaging equipment lifespan in pursuit of short-term energy efficiency.

[0013] Optionally, in this embodiment of the application, the reliability risk cost in step S3 is calculated in the following way: based on the implicit state analysis results obtained in step S2, the development rate of potential failure modes is evaluated; combined with historical maintenance data, the unplanned downtime and maintenance costs that may be caused by each failure mode are predicted; and the predicted expected economic loss is used as the reliability risk cost.

[0014] In the above implementation process, the control method for the refrigeration system provided in this application transforms the elusive equipment health risks into quantifiable economic indicators that can be directly added to and compared with current energy consumption costs and performance degradation costs. This enables the digital twin model to rationally weigh the pros and cons of aggressive operation to obtain higher energy efficiency and conservative operation to reduce failure risks when performing multi-objective optimization, thereby generating comprehensive optimization instructions that are both efficient and robust, achieving optimal management of the refrigeration system's entire life cycle cost.

[0015] Optionally, in this embodiment, step S3 further includes: dynamically determining the operating state of the refrigeration system based on the explicit state indicators and implicit state analysis results obtained in step S2; increasing the length of the optimization time domain when the operating state of the refrigeration system is stable; and shortening the length of the optimization time domain when the operating state of the refrigeration system is experiencing a fault warning or drastic parameter fluctuations; wherein the length of the optimization time domain is used to define the future time span covered by the liquid supply valve opening sequence and the compressor speed sequence in the optimization problem.

[0016] In the above implementation process, the control method of the refrigeration system in this application embodiment can perform long-term and economical global planning when the system is stable, and switch to rapid and robust emergency response when the system is unstable, thereby maintaining the effectiveness and safety of the control strategy in a complex and ever-changing working environment.

[0017] Optionally, in this embodiment, step S3, solving an optimization problem that includes equipment energy consumption cost, performance degradation cost, and reliability risk cost to obtain a comprehensive optimization instruction for the current control cycle, includes: encoding the opening sequence of the liquid supply valve and the rotational speed sequence of the compressor frequency converter to generate an initial population; constructing a fitness function with the sum of equipment energy consumption cost, performance degradation cost, and reliability risk cost as the evaluation index; performing selection, crossover, and mutation operations on the population for iterative evolution until a preset termination condition is met; and decoding the individual with the best fitness after iteration as the comprehensive optimization instruction for the current control cycle.

[0018] In the above implementation process, a genetic algorithm is used to solve the optimization problem with multiple cost constraints, transforming the complex nonlinear programming into an efficient parallel search process. This scheme can avoid getting trapped in local optima and find the cooperative control command that brings the long-term comprehensive operating cost (energy consumption, equipment wear and tear, and risk) close to the global optimum within a large decision space, thereby significantly improving the intelligence level and economic optimization potential of the system control.

[0019] Secondly, embodiments of this application provide a control device for a refrigeration system. The refrigeration system control device is used to execute the method described in the first aspect. The refrigeration system control device includes: a multimodal sensor group for collecting thermodynamic parameters, electrical parameters, and mechanical vibration parameters; an edge computing unit configured to execute steps S2 to S4, which integrates a digital twin model; an actuator group including a liquid supply valve and a compressor frequency converter driver; and a communication unit for realizing data interaction between the refrigeration system control device and a cloud platform.

[0020] Thirdly, embodiments of this application provide a refrigeration system, including a compressor, a condenser, an evaporator, a throttling device, and a refrigeration system control device as described in the second aspect above.

[0021] Fourthly, embodiments of this application provide an electronic device, which includes a memory and a processor. The memory stores program instructions, and when the processor reads and runs the program instructions, it executes the steps in any of the implementations of the first aspect described above.

[0022] Fifthly, embodiments of this application also provide a computer-readable storage medium storing computer program instructions, which are read and executed by a processor to perform the steps in any of the above implementations. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A flowchart of a control method for a refrigeration system provided in an embodiment of this application;

[0025] Figure 2 A flowchart illustrating a first implementation of implicit state analysis provided in an embodiment of this application;

[0026] Figure 3 A flowchart illustrating a second implementation of implicit state analysis provided in this application embodiment;

[0027] Figure 4 A flowchart for performance degradation calculation provided in the embodiments of this application;

[0028] Figure 5 A flowchart for calculating reliability risk costs provided in this application embodiment;

[0029] Figure 6 A solution flowchart provided for embodiments of this application;

[0030] Figure 7 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application; Detailed Implementation

[0031] The technical solutions of the embodiments of this application will now be described with reference to the accompanying drawings. For example, the flowcharts and block diagrams in the drawings illustrate the architecture, functions, and operations of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, program segment, or part of code, which contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions. In addition, the functional modules in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0032] The control of a refrigeration system refers to the technical process of optimizing energy efficiency, reliability, and economy by adjusting the opening of the liquid supply valve, the speed of the compressor, and other actuators to meet the refrigeration demand.

[0033] Currently, the mainstream refrigeration system control methods can be mainly divided into two categories:

[0034] One approach is open-loop control based on fixed logic or empirical formulas, such as using a fixed circulation ratio or a pre-set liquid supply curve based on nominal operating conditions. This method relies entirely on assumptions made during the design phase and cannot respond to fluctuations in key parameters such as actual load and ambient temperature during operation.

[0035] The second approach is closed-loop control based on univariate feedback. The most common method uses a proportional-integral-derivative (PID) controller to adjust the liquid supply valve, maintaining the evaporator outlet superheat around a fixed setpoint. While this method can handle certain disturbances, its control logic is relatively simple: a sensor measures the current superheat, the controller calculates the deviation from the setpoint, and outputs valve opening adjustment commands based on fixed PID parameters. Existing technologies typically rely only on limited "explicit" thermodynamic parameters such as temperature and pressure, have a single control objective (usually stable superheat), and the controller parameters rarely change once tuned.

[0036] The inventors discovered that, firstly, relying solely on parameters such as superheat and pressure is insufficient to discern the equipment's operating status, leaving the system operating unknowingly with inherent defects, thus creating a potential for sudden shutdowns. Secondly, even if a decision-making model exists, it is simplistic and static; neither fixed logic nor PID control considers the changes in equipment performance under complex and variable environments.

[0037] Based on this, this application proposes a control method, device, and system for a refrigeration system. This method integrates thermodynamic, electrical, and mechanical vibration parameters for comprehensive state perception and diagnosis. Utilizing a digital twin model, it dynamically optimizes the time domain to achieve the optimal coordinated control command for the liquid supply valve and compressor, aiming to minimize overall operating costs (including energy consumption, equipment degradation, and risk costs). This method enables integrated intelligent control, from early warning of latent faults and multi-objective dynamic optimization to coordinated regulation of actuators. Therefore, while ensuring long-term reliable system operation, it significantly improves the system's economic efficiency and energy efficiency throughout its entire lifecycle.

[0038] Please refer to Figure 1 , Figure 1 This application provides a flowchart of a control method for a refrigeration system according to an embodiment of the present application; the present application provides a control method for a refrigeration system that can... Figure 7 The electronic devices execute.

[0039] The control method includes the following steps:

[0040] Step S1: Real-time acquisition of multi-dimensional operating parameters of the refrigeration system.

[0041] In step S1 above, various sensors deployed on the physical entity of the refrigeration system synchronously collect multi-dimensional and heterogeneous operating parameters at a fixed sampling frequency, forming a data foundation that reflects the instantaneous state of the system.

[0042] Operating parameters include at least thermodynamic parameters, electrical parameters, and mechanical vibration parameters.

[0043] In the embodiments of this application, thermodynamic parameters directly describe the thermodynamic state and cycle process of the refrigerant. These typically include: the temperature and pressure at the evaporator inlet and outlet, the temperature and pressure at the condenser outlet, the compressor suction temperature and pressure, and the discharge temperature and pressure. For example, by installing a PT100 temperature sensor and pressure transmitter on the evaporator outlet pipeline, the temperature (e.g., 5°C) and pressure (e.g., 4.5 bar) of the refrigerant leaving the evaporator can be measured in real time; these are crucial inputs for calculating superheat.

[0044] In this embodiment, electrical parameters reflect the power input of the system's main energy-consuming components. These typically include the instantaneous power (which can be measured using a power meter) and operating current of devices such as the compressor drive motor, refrigerant pump, and cooling fan. For example, if the compressor's current input power is measured to be 7.5 kW, its operating energy efficiency can be preliminarily assessed by combining this with thermodynamic parameters.

[0045] In this embodiment, mechanical vibration parameters are acquired through a vibration acceleration sensor and used to monitor the mechanical operating status of rotating mechanical components such as compressors and pumps. The sensor is typically installed on the compressor housing, base, or adjacent pipelines to collect broadband vibration signals generated during equipment operation. This signal contains information about the internal mechanical state of the equipment (such as rotor balance, valve plate movement, and component friction).

[0046] Step S2: Based on thermodynamic and electrical parameters, perform explicit state analysis on the refrigeration system to obtain explicit state indices; perform implicit state analysis on the refrigeration system based on mechanical vibration parameters to obtain implicit state analysis results. The implicit state analysis includes diagnosing compressor valve leakage and refrigerant mixing with lubricating oil.

[0047] In step S2 above, the multidimensional parameters collected in step S1 are analyzed in depth to evaluate the system state from two levels: explicit state analysis and implicit state analysis.

[0048] Explicit state analysis, based on thermodynamic and electrical parameters, uses classical physical models or empirical formulas to calculate a series of indicators that can directly characterize the current external performance of the system, and can analyze the current operating status of the refrigeration system.

[0049] For example, the superheat is calculated based on the difference between the evaporator outlet temperature and the saturation temperature at the corresponding pressure. The superheat is a key indicator for measuring the heat exchange efficiency of the evaporator and preventing "liquid slugging".

[0050] For example, the system energy efficiency ratio (COP) can be calculated based on the system cooling capacity (which can be estimated from the heat exchange on the evaporator side) and the total input power of the main energy-consuming equipment such as the compressor, which can directly reflect the system's energy utilization efficiency.

[0051] For example, the volumetric efficiency of a compressor is evaluated by comparing its actual gas delivery volume with its theoretical gas delivery volume, reflecting its mechanical performance conditions such as internal leakage.

[0052] Latent condition analysis, specifically targeting mechanical vibration parameters, aims to diagnose the internal health of equipment that, while not yet significantly affecting thermodynamic performance, indicates potential faults. This includes diagnosing compressor valve leakage and refrigerant contamination with lubricating oil. Minor valve leakage or excessive refrigerant contamination with lubricating oil may not immediately lead to significant changes in superheat or COP, but it will alter the internal vibration characteristics of the compressor. The system analyzes the time-domain, frequency-domain, or time-frequency-domain characteristics of real-time vibration signals and compares them with baseline characteristics under normal conditions or processes them using specific analytical models to determine the presence of such latent faults and their severity trends. For example, analysis revealing an abnormally high energy increase in the vibration signal at a specific high frequency band may indicate a minor gas leak caused by a loose valve closure.

[0053] Step S3: In the digital twin model, with the overall goal of minimizing the comprehensive operating cost of the system in the next optimization time domain, construct and solve an optimization problem that includes equipment energy consumption cost, performance degradation cost, and reliability risk cost, and obtain the comprehensive optimization instructions for the current control cycle.

[0054] The decision variables for the optimization problem include at least the opening sequence of the liquid supply valve and the speed sequence of the compressor;

[0055] In step S3 above, a digital twin model that runs synchronously with the physical refrigeration system is used to virtually map the physical system, integrating the equipment's physical characteristics, control logic, and cost model.

[0056] In each control cycle, taking 5 minutes as an example, the optimization process is as follows:

[0057] The first step is to set optimization goals and problems. The overall goal is to minimize the system's overall operating cost within a future optimization timeframe, for example, within the next 30 minutes. This overall cost is not simply energy consumption, but includes equipment energy consumption costs, performance degradation costs, and reliability risk costs. Among these, equipment energy consumption costs refer to the electrical energy consumed during operation; performance degradation costs refer to the lifespan loss costs calculated due to wear and aging of critical equipment such as compressors caused by current operating strategies (such as high speed and large pressure differential); and reliability risk costs refer to the expected economic losses, such as unplanned downtime and maintenance, that may result from potential failures, as assessed based on the current state, especially the results of implicit state analysis.

[0058] The second step is to define the decision variables: the object of the optimization problem, i.e., the sequence of control variables that need to be decided. In this embodiment, the decision variables include at least: the opening sequence of the liquid supply valve and the speed sequence of the compressor. The opening sequence of the liquid supply valve refers to the valve opening setpoint for each control step (e.g., per minute) within the future optimization time domain, such as 30%, 35%, 40%, etc. The speed sequence of the compressor refers to the compressor speed setpoint for each control step within the future optimization time domain, such as 2900 rpm, 2950 rpm, 3000 rpm, etc.

[0059] The third step involves mathematically solving the aforementioned optimization problem, which includes multiple cost terms, within the digital twin model based on the current system state. The solution process aims to find an optimal sequence of liquid supply valve opening degrees and compressor speed sequences that minimizes the predicted overall operating cost. Finally, the instructions corresponding to the current control cycle are extracted from the solution results and used as comprehensive optimization instructions, such as a target opening degree of 35% and a target speed of 2950 rpm.

[0060] Step S4: According to the comprehensive optimization instructions, adjust the opening degree of the liquid supply valve to the target opening degree and adjust the speed of the compressor to the target speed.

[0061] In step S4 above, the comprehensive optimization command generated in step S3 is translated into specific physical actions. A control signal is sent to the liquid supply valve, i.e., the electronic expansion valve, to drive its valve core to move and precisely adjust the actual opening degree to the target opening degree required by the command. Additionally, a control signal is sent to the compressor's variable frequency drive to adjust the compressor's actual operating speed to the target speed required by the command.

[0062] Pass Figure 1 As can be seen, the control method for the refrigeration system provided in this application achieves comprehensive state perception of the refrigeration system by integrating multi-dimensional parameters such as thermodynamics, electrical, and mechanical vibration. Based on this, it not only performs explicit performance analysis but also introduces the diagnosis of implicit health conditions such as compressor valve leakage and refrigerant mixing. Furthermore, using a digital twin model, with the goal of minimizing the comprehensive operating cost including energy consumption, equipment degradation, and risks, it continuously solves for and executes the optimal coordinated control commands for the liquid supply valve and compressor. This enables integrated intelligent control of system health status and economic operation, moving from passive response to proactive optimization, thereby significantly improving the overall energy efficiency and economy of the refrigeration system while ensuring long-term reliable operation of the equipment.

[0063] In an alternative embodiment, please refer to... Figure 2 , Figure 2 The flowchart illustrates a first implementation of latent state analysis provided in this application embodiment; step S2, which involves performing latent state analysis based on mechanical vibration parameters, can be achieved through the following steps:

[0064] Step S211: Perform wavelet transform on the mechanical vibration parameters to decompose them into multiple sub-signals of different frequencies.

[0065] In step S211 above, the one-dimensional mechanical vibration acceleration signal acquired in real time is processed by wavelet transform. Unlike the traditional Fourier transform, which only provides global frequency information, the wavelet transform has good time-frequency localization characteristics, making it very suitable for analyzing non-stationary vibration signals. Specifically, a suitable wavelet basis function, such as the Daubechies wavelet, is selected to perform multi-resolution analysis on the original vibration signal, decomposing it into multiple sub-signals with different frequency components. For example, after 5 layers of wavelet decomposition, one low-frequency approximate sub-signal (a5) and five high-frequency detail sub-signals (d1, d2, d3, d4, d5) can be obtained. These sub-signals represent the projections of the original signal onto different frequency bands, and can more effectively reveal the characteristic vibration modes existing in specific frequency bands caused by specific faults such as valve leakage or refrigerant mixing.

[0066] Step S212: Extract the energy features of each sub-signal to form a feature vector.

[0067] In step S212 above, features characterizing the energy distribution of each sub-signal obtained after wavelet decomposition are extracted. The energy of each sub-signal is calculated, and the energy E of the j-th layer sub-signal is... j This can be obtained by calculating the sum of squares of all data points of the sub-signal. Then, the energy values ​​of all sub-signals are arranged in order to form an eigenvector T = [E]. a5 E d5 E d4 E d3 E d2 E d1 This feature vector encapsulates the energy distribution information of the original vibration signal in different frequency bands. The occurrence of faults often changes this distribution. For example, valve plate leakage may cause a significant increase in energy in a specific high-frequency band. Therefore, this vector can serve as an effective characterization of the fault.

[0068] Step S213: Input the feature vector into the pre-trained multilayer neural network model to obtain the output of the neural network model, including the classification results of the compressor valve plate leakage state and the refrigerant mixed with the lubricating oil state.

[0069] The output layer of the neural network model is configured to perform joint multi-label classification of compressor valve leakage status and refrigerant mixing with lubricating oil status. The compressor valve leakage status includes three categories: normal, slight leakage, and severe leakage. The refrigerant mixing with lubricating oil status includes three categories: no mixing, small amount mixing, and large amount mixing.

[0070] In step S213 above, the constructed feature vector T is input into a pre-trained multi-layer neural network model, such as a feedforward neural network or a convolutional neural network. This model is configured as a joint multi-label classifier, outputting two diagnostic results: compressor valve leakage status and refrigerant contamination with lubricating oil status. The compressor valve leakage status is divided into three mutually exclusive categories: normal, minor leakage, and severe leakage; the refrigerant contamination with lubricating oil status is divided into three mutually exclusive categories: no contamination, minor contamination, and significant contamination.

[0071] The network is trained on a large amount of labeled data—specifically, vibration signals and their corresponding wavelet energy feature vectors collected under known valve plate conditions and refrigerant mixing conditions—learning a complex mapping relationship from feature vectors to these two fault states. During the inference phase, after receiving real-time feature vectors, the network performs forward propagation calculations and ultimately provides a specific combination of classification results at the output layer. This transforms the raw vibration signals, which are difficult to interpret directly, into clear and quantifiable diagnostic conclusions about the key latent fault states of the compressor.

[0072] pass Figure 2 As can be seen, the control method for the refrigeration system provided in this application combines wavelet transform with deep learning neural networks to achieve automatic, precise, and joint diagnosis of two types of latent faults: valve plate leakage in the refrigeration compressor and refrigerant mixing. It can automatically extract time-frequency features strongly correlated with the fault from complex vibration signals and utilize the powerful nonlinear fitting capability of neural networks for high-precision classification. This overcomes the shortcomings of traditional diagnostic methods, such as reliance on human experience, single feature extraction, and high false alarm rates, greatly improving the intelligence level of condition monitoring and early fault detection capabilities, and laying a solid data foundation for predictive maintenance and reliability-based optimization control.

[0073] In an alternative embodiment, please refer to... Figure 3 , Figure 3 The flowchart illustrates a second implementation of the latent state analysis provided in this application embodiment; the latent state analysis based on mechanical vibration parameters in step S2 can also be achieved through the following steps:

[0074] Step S221: Fuzzify the mechanical vibration parameters to obtain the input fuzzy set.

[0075] In step S221 above, feature extraction is performed on the real-time acquired mechanical vibration parameters to obtain key feature values ​​such as the dominant vibration frequency (unit: Hz) and the effective vibration amplitude (unit: m / s²). These precise numerical values ​​are then converted into fuzzy language values ​​that can be processed by the fuzzy logic system.

[0076] Specifically, several fuzzy subsets and their membership functions are defined for each input feature. For example, for the input "dominant vibration frequency," three fuzzy subsets can be defined: "low frequency," "mid frequency," and "high frequency." For the "effective vibration amplitude," three fuzzy subsets can be defined: "small," "medium," and "violent." By querying the pre-defined membership functions, a specific frequency measurement value, such as 1250 Hz, will be converted into a fuzzy set description similar to "belonging to 'mid frequency' with a degree of 0.7 and belonging to 'high frequency' with a degree of 0.3." The amplitude is processed in the same way to obtain the input fuzzy set.

[0077] Step S222: Input the input fuzzy set into the preset fuzzy inference system.

[0078] The fuzzy inference system includes a fuzzy rule base determined based on historical fault data and vibration characteristics.

[0079] In step S222 above, the input fuzzy set obtained in step S221 is fed into a preset fuzzy inference system. It should be noted that the preset fuzzy inference system is a fuzzy rule base constructed based on a large amount of historical fault data, expert experience, and vibration characteristic analysis. Each rule associates the fuzzy state of the input features with the fuzzy state of the output (i.e., the probability of a fault). For example, the rule base may contain the following typical rules:

[0080] Rule 1: If the vibration frequency is "high frequency" and the amplitude is "medium", then the probability of compressor valve plate leakage is "medium".

[0081] Rule 2: If the vibration frequency is "low frequency" and the amplitude is "intense", then the possibility of refrigerant mixing with lubricating oil is "high".

[0082] This rule base encapsulates the complex, nonlinear empirical relationships between vibration characteristics and latent faults.

[0083] Step S223: Perform inference through a fuzzy inference system and output fuzzy results regarding the possibility of compressor valve plate leakage and refrigerant mixing with lubricating oil.

[0084] In step S223 above, the fuzzy inference engine performs inference calculations based on all activated rules (i.e., the antecedent and the current input fuzzy set have a non-zero matching degree). It comprehensively evaluates the consequents (conclusions) of all relevant rules and generates fuzzy results for the two output variables, "probability of compressor valve plate leakage" and "probability of refrigerant mixing with lubricating oil," through a "synthesis" operation. This result itself is still a fuzzy set; for example, the output "probability of compressor valve plate leakage" might be described as "0.4 for a 'high' probability and 0.6 for a 'medium' probability."

[0085] Step S224: Defuzzify the fuzzy results to obtain the latent state analysis results expressed as percentages.

[0086] In step S224 above, the fuzzy output set obtained in step S223 is converted into a definite, precise value that can be used for subsequent decision-making through defuzzification. Specifically, the abscissa value corresponding to the center of the area under the membership function curve of the output fuzzy set is calculated. Through this processing, the fuzzy "probability" is quantified into a specific value between 0% and 100%. For example, the final output is "Compressor valve plate leakage status: probability 62%" and "Refrigerant mixed with lubricating oil status: probability 18%". The percentage analysis results are the implicit state analysis results provided by this embodiment, which intuitively reflect the confidence level of the system in assessing the occurrence of various potential faults.

[0087] pass Figure 3 As can be seen, the control method for the refrigeration system provided in this application is based on the implicit state analysis method of fuzzy logic. It does not rely on a precise mathematical model of the fault, but rather simulates the experience-based judgment process of human experts, utilizing a fuzzy rule base to handle the uncertainties and nonlinear characteristics in vibration signals. This allows for the effective identification and probability assessment of early implicit faults such as compressor valve plate leakage and refrigerant mixing, even with limited data samples or complex fault mechanisms, providing crucial state inputs for subsequent risk cost calculation and optimized control.

[0088] In an alternative embodiment, please refer to Figure 4 , Figure 4 The performance degradation calculation flowchart provided in this application embodiment; the performance degradation cost in step S3 is calculated in the following way:

[0089] In step S3 above, when constructing an optimization problem within the digital twin model with the goal of minimizing overall operating costs, the calculation of performance degradation costs is a key component. This aims to quantify the physical impact of control decisions on the long-term lifespan of equipment into immediate economic costs, thereby enabling the optimization model to make an economically optimal trade-off between saving current energy consumption and extending equipment lifespan.

[0090] Step S311: Based on the compressor operating status in the digital twin model, calculate the incremental compressor life loss caused by the current control decision according to the preset life prediction model.

[0091] In step S311 above, based on the real-time operating status of the compressor in the digital twin model, the additional loss of equipment life caused by adopting the current candidate control decision, such as a specific set of speeds and loads, is evaluated.

[0092] Key state parameters, both current and under candidate control strategies, are extracted from a digital twin compressor model synchronized with the physical entity. These parameters include at least: compressor speed, discharge temperature, average pressure ratio (discharge pressure / intake pressure), and cumulative operating time.

[0093] The aforementioned state parameters are input into a preset lifetime prediction model. The preset lifetime prediction model in this application embodiment is typically established based on accelerated life theory or cumulative damage model in reliability engineering.

[0094] For example, a simplified model can be expressed as the rate of compressor life loss being an increasing function of its rotational speed and exhaust temperature. The model quantifies the number of life loss units caused by operating for a unit of time (e.g., 1 hour) under specific operating conditions (rotational speed N, exhaust temperature T_d).

[0095] The formula for calculating the lifetime micro-loss increment ΔL can be conceptually expressed as:

[0096]

[0097] Where f is the lifetime decay rate function, and It represents the state under the candidate control strategy. and This is a reference, relatively mild baseline operating condition, such as the state at economical speed. ΔL represents the additional life loss per unit time caused by using the current candidate strategy instead of the baseline strategy.

[0098] Step S312: Multiply the lifetime micro-loss increment by the preset unit replacement cost coefficient to obtain the performance degradation cost.

[0099] In step S312 above, after obtaining the lifetime micro-loss increment ΔL, it is converted into economic cost so that it can be measured in a unified manner with energy consumption cost and other costs.

[0100] To calculate the cost of performance degradation, we first need to determine the unit replacement cost coefficient C. unit This coefficient is a preset economic parameter representing the total cost of a compressor reaching the end of its overall lifespan (i.e., requiring replacement with a new unit), amortized across unit lifespan loss. Its calculation is based on: the compressor's purchase cost, installation and commissioning fees, minus its residual value, divided by the compressor's design lifespan under standard operating conditions, expressed as the total number of consuming "life units," i.e., C. unit = (Procurement cost + Installation cost - Residual value) / Total design life units.

[0101] Calculate the cost of performance degradation C degradationMultiplying the incremental lifetime loss by the unit replacement cost coefficient yields the performance degradation cost caused by the current control decision over a future optimization time domain (e.g., the next 30 minutes), i.e., C. degradation = ΔL × C unit .

[0102] For example, suppose the life prediction model of a certain type of screw compressor indicates that the life loss after one hour of operation at 3000 rpm and 90°C is 0.01 life units, while the loss at the economical speed of 2800 rpm and 85°C is 0.008 life units. If the current candidate control strategy is to use 3000 rpm, then ΔL = 0.01 - 0.008 = 0.002 life units / hour. If the total replacement cost of this compressor is equivalent to 100,000 yuan, and its designed total life is 100,000 life units, then C unit = 100,000 yuan / 100,000 = 1 yuan per lifespan unit. Therefore, the performance degradation cost C for running for 1 hour under this strategy is... degradation = 0.002 × 1 = 0.002 yuan. This cost will be included in the overall operating cost of this candidate strategy for evaluation.

[0103] Therefore, the refrigeration system control method provided in this application transforms the physical wear and tear of equipment, which is difficult to compare directly, into quantifiable and comparable economic costs. This enables the digital twin optimization model to endogenously consider the long-term impact of control decisions, thereby achieving true life-cycle cost optimization and avoiding short-sighted behavior of excessively damaging equipment lifespan in pursuit of short-term energy efficiency.

[0104] In an optional implementation, please refer to Figure 5 , Figure 5 The flowchart for calculating reliability risk cost provided in this application embodiment; the reliability risk cost in step S3 above is calculated in the following way:

[0105] Step S321: Based on the latent state analysis results obtained in step S2, assess the development rate of potential failure modes.

[0106] In step S321 above, based on the latent state analysis results obtained in step S2, a quantitative risk assessment is performed on the identified potential failure modes. Specifically, each diagnosed latent state, such as a slight leak in the compressor valve plate or a small amount of lubricating oil mixed in the refrigerant, is mapped to one or more specific potential failure modes; for example, valve plate fatigue fracture leading to complete failure, or lubricating oil dilution leading to accelerated wear of the compressor bearings.

[0107] For each failure mode, its rate of progression is assessed based on its currently diagnosed severity level and the trends in historical status data. The assessment model can be a prediction based on a physical failure model or a regression model based on statistical learning. For example, for valve plate leakage, the rate of increase in leakage indicators (such as energy values ​​in a specific frequency band) in vibration characteristics over a past period, such as 24 hours, combined with the known material fatigue characteristics of the valve plate for that compressor model, is estimated to predict the possible timeframe for the leakage area to expand, thus predicting the time it may take to develop into a severe leak or even fracture. This rate of progression is typically characterized by the rate of change of failure severity over time, or the expected time required to reach the next severity level.

[0108] Step S322: Based on historical maintenance data, predict the unplanned downtime and maintenance costs that may be caused by each failure mode.

[0109] In step S322 above, a pre-built and continuously updated historical maintenance database is used to quantitatively predict the potential consequences of each failure mode assessed in step S321. This database records the handling records of similar or identical failures that have occurred in the history of similar equipment, including at least: failure type, handling measures, required spare parts, maintenance man-hours, total equipment downtime, and total costs incurred.

[0110] For predicting unplanned downtime, the average repair time is calculated based on historical data of similar faults, and adjusted for factors such as current production schedule, spare parts inventory, and maintenance team availability. For example, historical data shows that replacing a compressor valve plate takes an average of 8 hours (including cooling, disassembly, replacement, vacuuming, refrigerant recharging, and testing). If spare parts are currently in stock and maintenance shifts are normal, the predicted downtime is 8 hours.

[0111] For the prediction of maintenance costs, historical data is also used to calculate the average direct cost of handling the fault, including spare parts cost, labor cost, and auxiliary material cost. The calculation is then updated based on the current spare parts price and labor rate. For example, the historical average maintenance cost is 5,000 yuan (of which valve plate costs 2,000 yuan and labor costs 3,000 yuan).

[0112] Step S323: Use the predicted expected economic loss as the reliability risk cost.

[0113] In step S323 above, typically, risk cost = probability of failure × economic loss caused by failure.

[0114] The probability of failure occurrence, based on the current failure mode's evolution rate, predicts the likelihood that a failure will worsen to the point of requiring downtime for maintenance within the optimization time domain or a foreseeable period thereafter. For example, according to the evolution model, the probability that a minor leak will develop into a serious leak requiring immediate repair within the next week is 10%.

[0115] The economic loss from a single failure is calculated as the total cost of unplanned downtime and repairs that would result if the failure were to occur. For example, the production loss caused by the downtime is estimated at 8,000 yuan, plus 5,000 yuan for repairs, for a total of 13,000 yuan.

[0116] For example, if the probability of a certain failure occurring within a relevant period is predicted to be P=10%, and the corresponding expected economic loss per failure is C=13000 yuan, then under the current decision, the reliability risk cost contributed by this failure mode can be calculated as R. risk = P × C = 0.1 × 13000 = 1300 yuan. If there are multiple independent potential failure modes, the total reliability risk cost is the sum of the risk costs of each mode.

[0117] Therefore, the control method for the refrigeration system provided in this application transforms the elusive equipment health risks into quantifiable economic indicators that can be directly added to and compared with current energy consumption costs and performance degradation costs. This enables the digital twin model to rationally weigh the pros and cons of aggressive operation to achieve higher energy efficiency versus conservative operation to reduce failure risks when performing multi-objective optimization, thereby generating comprehensive optimization instructions that are both efficient and robust, achieving optimal management of the refrigeration system's entire lifecycle cost.

[0118] In an optional embodiment, before constructing and solving the optimization problem in step S3, a pre-adaptive stage is introduced to match the predictability of the optimization control with the real-time dynamic characteristics of the system. Step S3 further includes:

[0119] Based on the explicit state indicators and implicit state analysis results obtained in step S2, the operating status of the refrigeration system is dynamically determined.

[0120] In the above implementation process, explicit state indicators, such as superheat, system energy efficiency ratio, and the standard deviation of key operating parameters over a period of time, are comprehensively analyzed. If these indicators are all within the preset safe and efficient ranges and have very small fluctuations, it indicates that the system is in a stable state, such as smooth operation under normal load. In addition, the results of implicit state analysis are also considered in depth. For example, if the possibility of compressor valve leakage output by the diagnostic model exceeds the warning threshold, or if vibration signal analysis indicates signs of sudden mechanical impact, the system is judged to have entered a fault warning state. Furthermore, if thermodynamic or electrical parameters experience drastic changes beyond the normal range in a short period of time, such as a sharp drop in evaporation pressure, it is judged to be a state of drastic parameter fluctuation.

[0121] When the refrigeration system is operating stably, the length of the optimization time domain is increased. When the refrigeration system is operating under fault warning conditions or experiencing drastic parameter fluctuations, the length of the optimization time domain is shortened. The length of the optimization time domain defines the future time span covered by the liquid supply valve opening sequence and the compressor speed sequence in the optimization problem.

[0122] In the above implementation process, when the system is determined to be in a steady state, the control system has sufficient confidence to predict and plan for the more distant future. At this time, the length of the optimization time domain is proactively increased, for example, from the usual 20 minutes to 40 minutes or longer. This allows the optimization solution to be coordinated and scheduled within a wider time window, potentially accepting small short-term fluctuations temporarily in order to achieve long-term overall energy efficiency optimization, thereby achieving a deeper level of economic optimization.

[0123] Conversely, when the system is identified as experiencing a fault warning or severe parameter fluctuations, the primary task is to quickly mitigate these fluctuations, prevent the situation from worsening, or adapt to rapidly changing conditions. At this point, the system's predictability decreases, and long-term predictions become unreliable. Therefore, the system will decisively shorten the optimization time domain, for example, reducing it to 5 minutes or less. This allows the optimization solution to focus on the most pressing near future, resulting in more frequent and agile generation and adjustments of control commands to quickly respond to transient changes, prioritizing the system's instantaneous safety and stability, and creating conditions for subsequent handling of faults or anomalies.

[0124] Therefore, the control method of the refrigeration system in this application embodiment can perform long-term and economical global planning when the system is stable, and switch to rapid and robust emergency response when the system is unstable, thereby maintaining the effectiveness and safety of the control strategy in a complex and ever-changing working environment.

[0125] In an alternative embodiment, please refer to Figure 6 , Figure 6The flowchart provided in this application embodiment shows the solution process. Step S3 involves solving an optimization problem that includes equipment energy consumption cost, performance degradation cost, and reliability risk cost to obtain a comprehensive optimization instruction for the current control cycle. This can be achieved through the following steps:

[0126] Step S331: Encode the opening sequence of the liquid supply valve and the rotational speed sequence of the compressor inverter driver to generate an initial population.

[0127] In step S331 above, the control sequence is encoded and an initial population is generated. First, the control variables to be optimized, i.e., the opening sequence of the liquid supply valve in the future optimization time domain, and the rotational speed sequence of the compressor variable frequency drive, are combined into a complete decision vector. Then, this vector is encoded using binary encoding or real number encoding, representing it as a chromosome that can be processed by the genetic algorithm. Next, an initial population containing multiple such chromosomes is created by random generation or generation based on historical best solutions, where each individual in the population represents a potential control scheme.

[0128] Step S332: Construct a fitness function with the sum of equipment energy consumption cost, performance degradation cost and reliability risk cost as the evaluation index.

[0129] In step S332 above, a comprehensive cost fitness function is constructed. To evaluate the quality of each individual in the population, a quantitative evaluation standard, namely the fitness function, needs to be built. This function directly corresponds to the overall objective of the optimization problem. Its input is the control sequence obtained after decoding an individual, and its output is a scalar value. This scalar value is the sum of the predicted equipment energy consumption cost, performance degradation cost, and reliability risk cost under this control sequence. In genetic algorithms, the lower the fitness value, the lower the total cost, indicating that the individual is of better quality and more likely to be preserved in evolution.

[0130] Step S333: Perform selection, crossover, and mutation operations on the population to iteratively evolve until the preset termination condition is met.

[0131] In step S333 above, the algorithm begins its iterative evolution process. In each generation, a selection operation is first performed based on the fitness value of individuals, selecting superior individuals to enter the next generation. Subsequently, a crossover operation is performed on the selected individuals, that is, exchanging some genetic information to generate new individuals that possess characteristics of their parents. Afterward, a mutation operation is performed on some of the new individuals with a low probability, that is, randomly changing some of their genetic values, to introduce new search possibilities and prevent the algorithm from getting trapped in local optima too early. This selection-crossover-mutation cycle is repeated continuously, and the quality of the population continues to improve until a preset termination condition is reached, such as reaching the maximum number of iterations or the optimal fitness value of the population no longer significantly improves over several generations.

[0132] Step S334: Decode the individual with the best fitness after iteration and use it as the comprehensive optimization instruction for the current control cycle.

[0133] In step S334 above, after the iteration terminates, the individual with the best fitness is selected from the final generation population, representing the control scheme with the lowest overall operating cost. The chromosome of this individual is decoded to restore the specific liquid supply valve opening sequence and compressor speed sequence. The first control variable corresponding to the current control cycle is extracted from this sequence, thus obtaining the comprehensive optimization command for the current control cycle, including the target opening degree and target speed.

[0134] pass Figure 6 It is evident that by employing genetic algorithms to solve optimization problems with multiple cost constraints, complex nonlinear programming is transformed into an efficient parallel search process. This scheme can avoid getting trapped in local optima and find cooperative control commands within a vast decision space that bring the long-term comprehensive operating costs (energy consumption, equipment wear and tear, and risk) close to the global optimum, thereby significantly improving the intelligence level and economic optimization potential of the system control.

[0135] This application provides a control device for a refrigeration system, which is used to execute the above-described control method for a refrigeration system. The refrigeration system control device includes:

[0136] A multimodal sensor array is used to collect thermodynamic parameters, electrical parameters, and mechanical vibration parameters.

[0137] The edge computing unit is configured to execute steps S2 to S4 and integrates a digital twin model.

[0138] The actuator assembly includes a liquid supply valve and a compressor variable frequency drive.

[0139] The communication unit is used to enable data interaction between the refrigeration system control device and the cloud platform.

[0140] This application provides a refrigeration system, which includes a compressor, a condenser, an evaporator, a throttling device, and a refrigeration system control device as described above.

[0141] Please see Figure 7 , Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. An electronic device 300 provided in this application includes: a processor 301 and a memory 302. The memory 302 stores machine-readable instructions executable by the processor 301. When the machine-readable instructions are executed by the processor 301, the method described above is performed.

[0142] Based on the same inventive concept, embodiments of this application also provide a computer program product including a computer program / instruction, which, when executed by a processor, performs steps in any implementation of the control method for the above-described refrigeration system.

[0143] Based on the same inventive concept, embodiments of this application also provide a computer-readable storage medium storing computer program instructions. When the computer program instructions are read and executed by a processor, they perform the steps in any implementation of the control method of the above-described refrigeration system.

[0144] Computer-readable storage media can be any medium capable of storing program code, such as Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), and Electrically Erasable Programmable Read-Only Memory (EEPROM).

[0145] In this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, without necessarily requiring or implying any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0146] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A control method of a refrigeration system, characterized by, The control method comprises: Step S1: collecting multi-dimensional operating parameters of the refrigeration system in real time; wherein the operating parameters at least include thermodynamic parameters, electrical parameters and mechanical vibration parameters; Step S2: performing explicit state analysis on the refrigeration system according to the thermodynamic parameters and the electrical parameters to obtain an explicit state index; performing implicit state analysis on the refrigeration system according to the mechanical vibration parameters to obtain an implicit state analysis result; wherein the implicit state analysis includes diagnosing the leakage state of the compressor valve and the state of refrigerant mixing into lubricating oil; Step S3: in the digital twin model, taking the minimum system comprehensive operating cost in a future optimization time domain as the total target, constructing and solving an optimization problem containing equipment energy consumption cost, performance degradation cost and reliability risk cost to obtain comprehensive optimization instructions in the current control period; wherein the decision variables of the optimization problem at least include the opening sequence of the liquid supply valve and the speed sequence of the compressor; Step S4: adjusting the opening of the liquid supply valve to the target opening and adjusting the speed of the compressor to the target speed according to the comprehensive optimization instructions.

2. The method of claim 1, wherein, In the step S2, the implicit state analysis according to the mechanical vibration parameters comprises: performing wavelet transform on the mechanical vibration parameters to decompose them into a plurality of sub-signals of different frequencies; extracting energy features of each of the sub-signals to form a feature vector; inputting the feature vector into a pre-trained multi-layer neural network model to obtain an output of the neural network model including a classification result of the leakage state of the compressor valve and the state of refrigerant mixing into lubricating oil; wherein the output layer of the neural network model is configured to perform joint multi-label classification on the leakage state of the compressor valve and the state of refrigerant mixing into lubricating oil; the leakage state of the compressor valve includes three categories of normal, slight leakage and serious leakage, and the state of refrigerant mixing into lubricating oil includes three categories of no mixing, small amount of mixing and large amount of mixing.

3. The method of claim 1, wherein, In the step S2, the implicit state analysis according to the mechanical vibration parameters comprises: performing fuzzy processing on the mechanical vibration parameters to obtain an input fuzzy set; inputting the input fuzzy set into a pre-set fuzzy reasoning system; wherein the fuzzy reasoning system contains a fuzzy rule base determined based on historical fault data and vibration features; outputting a fuzzy result about the possibility of compressor valve leakage and refrigerant mixing into lubricating oil by reasoning through the fuzzy reasoning system; de-fuzzifying the fuzzy result to obtain an implicit state analysis result represented in percentage form.

4. The method of claim 1, wherein, In the step S3, the performance degradation cost is calculated by: based on the running state of the compressor in the digital twin model, calculating the increment of compressor life micro-loss caused by the current control decision according to a preset life prediction model; multiplying the increment of compressor life micro-loss by a preset unit replacement cost coefficient to obtain the performance degradation cost.

5. The method of claim 1, wherein, In the step S3, the reliability risk cost is calculated by: based on the implicit state analysis result obtained in the step S2, evaluating the development rate of potential failure modes; In combination with historical maintenance data, predict the unplanned downtime and maintenance cost that each failure mode may cause; Use the predicted expected economic loss as the reliability risk cost.

6. The method of claim 1, wherein, The step S3 further comprises: Dynamically determine the operating state of the refrigeration system according to the explicit state indicators and the implicit state analysis results obtained in the step S2; In the case that the operating state of the refrigeration system is stable, increase the length of the optimization time domain; In the case that the operating state of the refrigeration system is failure warning or parameter dramatic fluctuation, shorten the length of the optimization time domain; The length of the optimization time domain is used to define the future time span covered by the sequence of the liquid supply valve opening degree and the sequence of the compressor speed in the optimization problem.

7. The method of claim 1, wherein, The step S3 comprises solving the optimization problem containing the equipment energy consumption cost, performance degradation cost and reliability risk cost to obtain the comprehensive optimization instruction in the current control period, including: Encode the sequence of the liquid supply valve opening degree and the sequence of the compressor variable frequency drive speed to generate an initial population; Construct a fitness function taking the sum of the equipment energy consumption cost, performance degradation cost and reliability risk cost as the evaluation index; Iteratively evolve the population through selection, crossover and mutation operations until the preset termination condition is met; Decode the individual with the optimal fitness after iteration as the comprehensive optimization instruction in the current control period.

8. A control device for a refrigeration system, characterized by The refrigeration system control device is configured to execute the method of any one of claims 1-7, and comprises: A multi-modal sensor group for collecting thermodynamic parameters, electrical parameters and mechanical vibration parameters; An edge computing unit configured to execute steps S2-S4, and the digital twin model is integrated therein; An actuator group including a liquid supply valve and a compressor variable frequency drive; A communication unit for realizing data interaction between the refrigeration system control device and a cloud platform.

9. A refrigeration system characterized by, A refrigeration system comprising a compressor, a condenser, an evaporator, a throttling device and a control device as claimed in claim 8.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer program instructions, and the computer program instructions are run by a processor to execute the steps in the method of any one of claims 1-7.

Images