A refrigeration system parameter optimization method, controller and health management device

Abstract

The application discloses a refrigeration system parameter optimization method, a controller and a health management device, and belongs to the technical field of refrigeration systems. The parameter optimization method comprises the following steps: acquiring training data reflecting the mapping relationship between operating parameters and the energy efficiency ratio COP; training and constructing a COP prediction model; sampling to generate candidate design points in a design space, inputting the model in combination with random working condition combinations, and obtaining performance prediction values of each candidate point; and outputting optimized design parameter combinations, such as the inner diameter, length and refrigerant charge of a capillary tube, according to the prediction values. The COP prediction model trained based on the method can be integrated in an energy-saving controller, used for real-time online search of optimal operating set points, and used for dynamic energy efficiency tracking. The COP prediction model can also be constructed as a health benchmark model and integrated in a health management device, so that early warning of system performance degradation can be realized by comparing the deviation of the actual COP from the health reference value. The whole-process intelligentization from design, operation to maintenance is realized, and the design efficiency and operation energy efficiency are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of refrigeration system technology, and in particular to a method for optimizing refrigeration system parameters, a controller, and a health management device. Background Technology

[0002] In refrigeration systems, especially micro refrigeration systems with extremely high requirements for size, weight and energy efficiency (such as micro chillers, precision air conditioners, etc.), the selection of throttling elements and the determination of refrigerant charge are core design challenges.

[0003] While using capillary tubes as throttling elements offers advantages such as simple structure, low cost, and small size, their design heavily relies on engineers' experience and involves a tedious "design-prototype-testing" iterative process. Because the throttling effect of a capillary tube is determined by its fixed inner diameter and length, and it has a complex nonlinear coupling relationship with the refrigerant charge, traditional methods struggle to quickly and accurately find the globally optimal combination of design parameters. This results in long design cycles, high costs, and the final system energy efficiency may not be optimal.

[0004] Furthermore, the following problems exist in the actual operation and long-term maintenance of refrigeration systems: 1. Difficulty in achieving real-time optimal energy efficiency: For systems using fixed throttling elements, traditional control strategies cannot dynamically adjust the operating state according to real-time changes in operating conditions, causing the system to deviate from the optimal energy efficiency point under partial load or off-design conditions for extended periods, resulting in continuous energy waste. 2. Lack of intelligent assessment of health status: Existing systems mainly rely on threshold alarms or periodic manual inspections, which cannot perform early and accurate predictive diagnosis of system performance degradation based on operating data. Faults are often only discovered after energy efficiency has severely declined or equipment has been damaged, resulting in passive and costly maintenance.

[0005] Therefore, there is an urgent need in this field for an integrated solution that can run through the entire process of design, operation and maintenance, and achieve rapid and accurate optimization of system parameters, real-time energy-saving control and intelligent health management. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a complete intelligent solution for refrigeration systems. Specifically, it aims to: 1. Provide a method for optimizing refrigeration system parameters to efficiently and accurately determine the globally optimal combination of throttling elements and refrigerant charge, solving the optimization problem in the design phase. 2. Provide a refrigeration system controller capable of adaptive energy-saving optimization during operation, solving the problem of non-real-time optimal operating energy efficiency. 3. Provide a refrigeration system health management device capable of real-time monitoring and early warning of system performance, solving the problem of lack of intelligent assessment of health status.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] This invention provides a method for optimizing refrigeration system parameters, comprising: a data acquisition step: acquiring training data, which reflects the mapping relationship between the operating parameters of the refrigeration system and the coefficient of performance (COP); the operating parameters include optimizable design parameters and operating condition parameters reflecting the operating conditions; a model construction step: based on the training data, training a machine learning model with the operating parameters as input and the COP as output to construct an COP prediction model; a simulation evaluation step: setting a design space for the design parameters and setting the variation range for the operating condition parameters; generating multiple candidate design points using uniform sampling within the design space; for each candidate design point, performing multiple random samplings within the variation range of the operating condition parameters to obtain multiple sets of operating condition parameter combinations; inputting each candidate design point and the corresponding multiple sets of operating condition parameter combinations into the COP prediction model to obtain a set of performance prediction values ​​for the candidate design point regarding the COP; and an optimization output step: determining and outputting optimized design parameter combinations from the candidate design points based on the performance prediction values ​​corresponding to each candidate design point.

[0009] This invention also provides an energy-saving controller for a refrigeration system, including a memory and a processor. The memory stores a computer program and a COP prediction model trained based on the aforementioned refrigeration system parameter optimization method. When the processor executes the program, it performs the following: a. Real-time acquisition of the system's operating parameters, combined with fixed throttling element design parameters and refrigerant charge, to form an input feature vector; the operating parameters include ambient temperature and heat load power; b. Using the adjustable operating parameters in the system as decision variables, and the COP prediction model as the objective function, an optimization algorithm is used to search online under the current operating conditions to find the optimal setpoint that maximizes the predicted COP value and the corresponding theoretical optimal COP value; c. Controlling the refrigeration system to operate at the optimal setpoint; d. When the change in the operating parameters exceeds a preset threshold, steps (b) and (c) are repeated to dynamically track and maintain the system's optimal energy efficiency state under the current operating conditions.

[0010] This invention also provides a refrigeration system health management device, including a memory and a processor. The memory stores a computer program and a health benchmark model. The health benchmark model is a machine learning model trained using historical operating data of the refrigeration system under healthy conditions, based on the aforementioned refrigeration system parameter optimization method. It is used to predict the theoretical energy efficiency ratio (COP) value of the system, taking the system's design parameters and operating parameters as inputs. When the processor executes the program, it performs the following: collecting real-time operating parameters of the system and combining them with fixed design parameters to form a real-time feature vector; processing the real-time feature vector using the health benchmark model to obtain a healthy COP reference value; calculating the deviation between the actual COP value of the system and the healthy COP reference value; and generating a corresponding health status warning signal based on the persistence or fluctuation pattern of the deviation exceeding a preset threshold.

[0011] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for optimizing refrigeration system parameters.

[0012] Compared with the prior art, the present invention has the following beneficial effects:

[0013] 1. High Efficiency and Low Cost: Transforms the traditional, experience-based, and lengthy experimental process into a highly efficient, computer-automated optimization process, shortening the design cycle and reducing R&D and prototyping costs. 2. High Precision and Global Optimality: Utilizing the powerful nonlinear fitting capabilities of machine learning models, it can accurately capture the complex coupling relationships between multiple parameters, thereby finding globally optimal or near-optimal design parameters, avoiding the pitfalls of traditional methods that easily get trapped in local optima. 3. Strong Robustness: By optimizing under varying environmental conditions, it ensures that the final design scheme has good adaptability and stability in actual operation, exhibiting robust performance. 4. Strong Design Guidance: By assessing the importance of different operating parameters on the Coefficient of Performance (COP), it provides engineers with a quantitative basis for determining priority design parameters, improving design efficiency and the scientific nature of decision-making. 5. Dynamic Energy Saving Optimization: It can search for the optimal operating setpoint online based on real-time operating conditions, ensuring that the system always maintains at or near the highest energy efficiency state under current conditions, achieving significant energy savings. 6. Strong Adaptability: It automatically re-optimizes when operating conditions change, exhibiting strong adaptability. 7. Predictive Maintenance: By comparing actual performance with the predicted values ​​of the health benchmark model, performance degradation trends can be identified early, enabling predictive maintenance, avoiding sudden failures, and reducing maintenance costs. 8. Intelligent Diagnostics: It can distinguish between different types of performance anomaly patterns (such as continuous degradation or sudden fluctuations), providing more accurate fault warnings. Attached Figure Description

[0014] Figure 1 : Schematic diagram of the refrigeration system parameter optimization method of the present invention.

[0015] Figure 2 : Flowchart of machine learning model training for the refrigeration system parameter optimization method of this invention.

[0016] Figure 3 The flowchart of the iterative optimization method for optimizing the design parameters of the refrigeration system of this invention.

[0017] Figure 4 : Schematic diagram of the refrigeration system controller of this invention.

[0018] Figure 5 : Schematic diagram of the health management device for the refrigeration system of the present invention. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In the following description, suffixes such as "module," "part," or "unit" used to denote elements are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "part," or "unit" may be used interchangeably.

[0020] Example 1: This example provides the most basic technical solution of the present invention, namely a machine learning-based design method for capillary throttling elements in refrigeration systems. Its core lies in using a data-driven model to replace complex physical simulations, quickly and accurately finding the most energy-efficient combination of "capillary structure parameters - refrigerant charge". Reference Figure 1 and Figure 2 This includes the following steps:

[0021] S101, Data acquisition steps: Acquire training data, which includes operating parameters and energy efficiency ratio (COP). Operating parameters include design parameters and operating condition parameters.

[0022] S102, Model training steps: Based on the training data, with the running parameters as input and the energy efficiency ratio (COP) as output, the energy efficiency ratio (COP) prediction model is obtained through training.

[0023] S103, Simulation and Evaluation Steps: Set a design space for the design parameters and set the variation range for the operating parameters; generate multiple candidate design points within the design space using a uniform sampling method; for each candidate design point, perform multiple random samplings within the variation range of the operating parameters to obtain multiple sets of operating parameter combinations; input each candidate design point and the corresponding multiple sets of operating parameter combinations into the COP prediction model to obtain a set of performance prediction values ​​for the COP of the candidate design point.

[0024] S104, Optimization output steps: Based on the performance prediction index values ​​corresponding to each candidate design point, determine and output the optimized candidate design points from the candidate points.

[0025] Specifically, in step S101, the training data is the original dataset obtained through historical experiments. This dataset contains several sets of operating parameter values ​​and corresponding measured and calculated actual COP values. The dataset undergoes preprocessing such as cleaning and feature standardization, and is proportionally divided into a training set for model training and validation and a test set for final performance evaluation. Cross-validation in subsequent step S102 is performed only within the training set. The operating parameters affecting COP are identified, including controllable design parameters and uncontrollable operating condition parameters. Design parameters are the objects to be optimized, including: capillary inner diameter, capillary length, and refrigerant charge. Operating condition parameters simulate the actual operating conditions of the system, including: ambient temperature, compressor speed, and system heat load power. All of these operating parameters affecting COP are directly measurable or directly obtainable data. The COP = Cooling capacity Ø / Compressor power consumption P, and the formula for calculating the system cooling capacity Ø is: Q liq For water flow rate, T in T represents the inlet water temperature. out Let ρ be the outlet water temperature, ρ be the density of water, and C be the specific heat capacity of water. The power consumption P of the compressor can be determined by the input voltage and current. The above data are all easily measured using tools during the operation of the refrigeration system. The energy efficiency ratio (COP) can be calculated based on the above data.

[0026] Specifically, in the S102 model training step, a machine learning model suitable for handling high-dimensional nonlinear relationships is selected, such as random forest, gradient boosting decision tree, or neural network. This embodiment uses the random forest model as an example. During training, cross-validation and grid search strategies are used to optimize the model's hyperparameters. After training, the root mean square error (RMSE) and coefficient of determination (R²) are used as evaluation metrics for the model's final performance, ultimately obtaining a high-precision COP prediction model.

[0027] Specifically, the steps to build a random forest model are as follows:

[0028] Suppose the original training sample set is D={(x1,y1),(x2,y2),…,(x n ,y n By using random sampling with replacement, several subsets D1, D2...D are generated. B , where B represents the number of trees. The generation process for each sample set can be represented as:

[0029]

[0030] For the regression prediction problem involved in this application, when constructing the decision tree, the criterion for selecting splitting features and thresholds for each node is to minimize the variance of the target variable (i.e., energy efficiency ratio COP) of the child nodes, thereby making the predicted values ​​as pure as possible locally. In each sample subset Train a decision tree independently Because each tree has a different sample source, and only a partially randomly selected subset of features is used when each node splits during training, the models naturally exhibit differences. This dual mechanism of sample randomization and feature randomization effectively prevents overfitting and ensures that the model remains highly robust even with noisy data.

[0031] For regression problems, the final prediction result of a random forest is obtained by averaging the prediction outputs of all decision trees, and its ensemble output can be expressed as:

[0032]

[0033] in, This represents the final prediction result for the input sample x, where B is the total number of decision trees. Let represent the predicted value of the b-th tree. Through this averaging mechanism, random forests significantly reduce the errors caused by fluctuations in single models, enabling them to exhibit good stability in complex nonlinear systems.

[0034] In this embodiment, cross-validation specifically refers to a three-fold cross-validation strategy. Specifically, the training set is randomly divided into three groups: two groups are used for training, and one group is used for validation. Training and evaluation are performed alternately three times. The root mean square error (RMSE) and R0 are calculated for the prediction error on each validation set. 2 The average of the three results is taken as the final energy efficiency ratio (COP). The grid search strategy is a systematic method that searches for the optimal configuration by traversing all possible combinations of predefined hyperparameters. It divides the hyperparameter space into a grid, the algorithm scans all combinations, and uses techniques such as cross-validation to evaluate the model performance for each combination, ultimately selecting the best-performing hyperparameter combination as the final model.

[0035] The calculation process for using root mean square error (RMSE) and coefficient of determination (R²) as evaluation metrics for model performance is as follows:

[0036] Suppose there are n samples in the test set, and the true value is y. i The model predicts the value. The formula for calculating the root mean square error (RMSE) is as follows:

[0037]

[0038] The physical meaning of RMSE is the square root of the average of the squared prediction errors. Therefore, it has the same dimensions as the original variables and can intuitively reflect the average deviation of the model's prediction results. The smaller the RMSE, the closer the model's prediction results are to the actual values; the smaller the error, the higher the model accuracy. In the throttle valve performance prediction of this study, RMSE can be used to evaluate the model's accuracy in predicting COP, that is, to measure the model's fitting deviation to the energy efficiency value of the refrigeration system.

[0039] R 2 It is an important indicator for measuring the explanatory power of a model, reflecting how well the model fits the total variance of the target variable. Its calculation formula is as follows:

[0040]

[0041] R² essentially reflects the proportion of variance explained by the model relative to the total variance. The closer the value is to 1, the greater the proportion of variation in the target variable the model can explain, and the better the fit. R² is particularly important in predicting the performance of refrigeration systems. Because the system's coefficient of performance (COP) is affected by various nonlinear factors, such as refrigerant quality, capillary tube size, ambient temperature, and compressor speed, a single variable cannot fully describe the changes in system performance. If a model achieves a high R² on the test set, it means that it can effectively capture the nonlinear coupling relationships between these complex variables and has strong generalization ability.

[0042] To comprehensively evaluate model performance, this embodiment uses RMSE and R² in combination. The former quantifies the absolute value of the prediction error, while the latter measures the goodness of fit. Typically, in experimental results analysis, cross-validation is used to calculate the mean RMSE and R² values ​​of the model at different folds to reduce evaluation bias caused by chance. If the trained model does not meet the performance evaluation criteria of root mean square error (RMSE) and coefficient of determination (R²), the model training process is continued to be corrected until the trained model passes both RMSE and R².

[0043] Specifically, in the S103 simulation and evaluation step: Reasonable physical value ranges are set for each design parameter, forming a three-dimensional design space. Uniform sampling, such as Latin hypercube sampling, is used to generate a large number (e.g., thousands) of candidate design points within this space. For each candidate point, multiple sets (e.g., hundreds) of different combinations of operating parameters are randomly generated to simulate its performance under different operating conditions. All these "design point-operating point" parameter pairs are input into the COP prediction model trained in S102 to obtain the average predicted COP for each candidate design point under varying operating conditions.

[0044] Specifically, in step S104, the optimal design parameters are determined by comparing the average predicted COP of all candidate points and determining the combination of capillary inner diameter, length, and refrigerant charge corresponding to the highest COP as the global optimal design parameters under the system architecture.

[0045] Example 2: Building upon Example 1, the random forest model is further used to evaluate the feature importance of the running parameters. The calculation formula is as follows:

[0046]

[0047] in, In the b-th tree, the feature is used to represent the characteristics. The set of all nodes to be split. This indicates the reduction in variance before and after the node is partitioned. The larger the value, the more significant the contribution of that feature to the model's predictive performance. In throttle valve optimization tasks, this index can quantify the importance of each parameter, thereby guiding subsequent engineering design. For example, when the feature importance of the capillary inner diameter is significantly higher than that of the length, the inner diameter specification can be adjusted first to improve system energy efficiency.

[0048] Example 3: This example, based on Example 1, introduces an iterative optimization mechanism to refine the initial optimization results and obtain more accurate optimal design parameters. (Reference) Figure 3 The specific steps are as follows:

[0049] S301: Perform the first global optimization. Following steps S101 to S104 in Example 1, obtain the first round of globally optimal design parameter combination P1 = (d1, l1, m1), where d1 is the inner diameter of the capillary tube, l1 is the length of the capillary tube, and m1 is the refrigerant charge.

[0050] S302: Construct a refined neighborhood search space. Centered on the optimal parameter P1 from the first round, define a narrowed neighborhood range for each parameter. For example, update the search range as follows: inner diameter ∈ [d1×0.9, d1×1.1], length ∈ [l1×0.9, l1×1.1], refrigerant charge ∈ [m1×0.9, m1×1.1]. The proportion of this neighborhood range (e.g., ±10%) can be adjusted according to actual needs.

[0051] S303: Perform a fine-grained search within the neighborhood. In a new, smaller 3D neighborhood space, employ a higher-density sampling method. For example, increase the number of sampling points by an order of magnitude to generate candidate points. Repeat steps S103 and S104 in Example 1 to generate the next-generation local optimal design parameters P2.

[0052] S304: Iteration and Termination. Determine the latest obtained design parameter P. n The corresponding average COP, and its previous generation parameter P n-1 The corresponding average COP and its change are compared to whether they are less than a preset threshold (e.g., 0.5%). If they are less than the threshold, the iteration terminates and P is output. n As the final optimized design parameter; if it is greater than or equal to the threshold, return to step S302, with P n Build a new neighborhood space around the center and continue to iterate and optimize.

[0053] Example 4: This example transforms the trained machine learning model from a design tool into an online control tool, enabling adaptive energy-saving operation of the refrigeration system. (Reference) Figure 4 The implementation steps are as follows:

[0054] S401: Deploy the online prediction model. The high-precision COP prediction model trained in Example 1 is lightweighted. The lightweighting process must be carried out while ensuring that the decrease in model prediction accuracy is within an acceptable threshold to ensure the reliability of control. The model is then embedded into the firmware of the chiller unit's main controller or edge computing module.

[0055] S402: Real-time Data Acquisition and Input Construction. The controller acquires system operating parameters in real time through sensors, including: current ambient temperature and real-time heat load power. Simultaneously, the controller knows the system's inherent design parameters: capillary inner diameter, capillary length, and refrigerant charge. These real-time data are combined with the fixed design parameters to construct the input feature vector required for the model. It is worth noting that adjustable operating parameters in the system (such as compressor speed) are not used as input feature vectors at this stage; their values ​​will be determined through a search in subsequent steps.

[0056] S403: Online optimization search and setpoint output

[0057] Based on the feature vector (including ambient temperature, real-time heat load power, capillary inner diameter, capillary length, and refrigerant charge) constructed in S402, the controller performs the following steps to find the optimal setpoint under the current operating conditions:

[0058] 1. Fixed known parameters: The real-time collected ambient temperature, heat load power, and the system's inherent capillary inner diameter, length, and refrigerant charge are used as known fixed inputs.

[0059] 2. Define the search problem: Take the compressor speed as the decision variable to be optimized, and set its adjustable range {RPM_min, RPM_max}.

[0060] 3. Perform optimization search: Within the adjustable speed range, using the deployed COP prediction model as the evaluation function, an optimization algorithm is employed to find the speed value S_optimal that maximizes the COP predicted by the model. The essence of this optimization algorithm is to systematically test different candidate values ​​within the speed range and compare their corresponding predicted COPs. As a specific implementation, a uniform sampling comparison method can be used: N candidate speed values ​​are generated uniformly within the speed range; each candidate value is combined with the fixed parameters from step 1 to form a complete input, and the COP prediction model is called to obtain N predicted COP values; the candidate speed corresponding to the maximum predicted COP value is selected, which is S_optimal. The optimization algorithm in this embodiment can be, but is not limited to, one-dimensional optimization algorithms such as the uniform sampling comparison method, the golden section method, and the gradient descent method.

[0061] 4. Output: Record S_optimal and its corresponding theoretical optimal prediction value COP_optimal.

[0062] S404: Adaptive Control and Optimization.

[0063] 1. Setpoint Adjustment: The controller adjusts the adjustable operating parameters of the system to the optimal setpoint S_optimal.

[0064] 2. Status tracking: When the system detects that key operating parameters such as ambient temperature and heat load have changed beyond the set threshold, it automatically returns to step S403 to recalculate and update the optimal setpoint, ensuring that the system always dynamically tracks the optimal energy efficiency state under the current operating conditions.

[0065] Example 5: This example utilizes a trained machine learning model to provide real-time health status assessment and early fault warning for in-service chillers. This method is achieved by deploying the model in the controller and comparing the model's predicted values ​​with actual operating values. (Refer to...) Figure 5 The specific steps are as follows:

[0066] S501: Deploy the health baseline model. After solidifying and lightweighting the high-precision COP prediction model trained in Example 1, it is embedded into the main controller of the chiller unit or a connected edge computing device. The lightweighting process must be performed while ensuring that the decrease in model prediction accuracy remains within an acceptable threshold to guarantee the reliability of control or diagnosis. This model is trained on historical data of system health and normal operation. Its internal parameters are fixed, and it accurately reflects the mapping relationship between the system's operating parameters and the COP under normal conditions, serving as the system's health baseline model.

[0067] S502: Real-time Data Acquisition and Model Prediction. During chiller operation, the controller collects real-time data from system sensors, including operating parameters such as ambient temperature, current heat load power (calculated from inlet and outlet water temperatures and flow rates or measured by a power meter), and compressor speed. This real-time data, along with the system's inherent, immutable design parameters—capillary inner diameter, capillary length, and refrigerant charge—constitutes a real-time feature vector. The controller inputs this feature vector into a deployed health baseline model, which calculates and outputs a "COP value that a healthy system should achieve under the current operating conditions," denoted as COP_Predicted Health. This COP_Predicted Health is the "Healthy COP Reference Value."

[0068] S503: Calculate the energy efficiency ratio deviation. The controller synchronously calculates the current actual operating COP value of the system via the connected power meter and heat meter, denoted as COP_actual. Then, the deviation between the two is calculated, i.e., Deviation = COP_actual - COP_predicted health. When the system is healthy, since the model can simulate the normal state well, this deviation should be close to zero and fluctuate randomly within a very small range.

[0069] S504: Deviation-based fault diagnosis and early warning. The system outputs a corresponding early warning signal based on the calculated deviation, as follows:

[0070] Energy Efficiency Deterioration Warning: If the deviation remains negative and its absolute value exceeds a preset threshold (e.g., the average deviation is below -0.15 for 10 consecutive sampling periods), it indicates that the system's actual energy efficiency is significantly lower than its expected healthy level. This usually foreshadows performance degradation faults such as refrigerant leakage, scale buildup on the heat exchanger surface, internal wear of the compressor, or partial blockage of the throttling element. The controller can issue a warning signal stating "Abnormal decrease in system energy efficiency, inspection recommended."

[0071] Instability warning: If the deviation fluctuates drastically and irregularly within a short period of time, while individual sensor readings show no abnormalities, it may indicate sensor signal interference, electronic expansion valve control malfunction, or unstable flashes within the system. The controller can issue a "system instability warning" signal.

[0072] Furthermore, a continuous health status score (e.g., 0-100 points) can be calculated based on the magnitude and stability of the deviation, allowing maintenance personnel to intuitively understand the equipment degradation trend.

[0073] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0074] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not described in detail or in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0075] In the embodiments provided by this invention, it should be understood that the disclosed apparatus / terminal devices and methods can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0076] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0077] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0078] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0079] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0080] In the description of this specification, references to terms such as "one embodiment," "another embodiment," "other embodiments," or "first embodiment to Xth embodiment," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, method steps, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0081] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments. The above are merely preferred embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for optimizing parameters of a refrigeration system, characterized in that, include: Data acquisition steps: Acquire training data, which reflects the mapping relationship between the operating parameters of the refrigeration system and the Coefficient of Performance (COP). The operating parameters include optimizable design parameters and operating condition parameters that reflect the operating conditions. The design parameters include capillary inner diameter, capillary length, and refrigerant charge. Model building steps: Based on the training data, train a machine learning model with the operating parameters as input and the energy efficiency ratio COP as output to build an energy efficiency ratio COP prediction model; Simulation evaluation steps: Set a design space for the design parameters and set a range of variation for the operating parameters; generate multiple candidate design points within the design space using uniform sampling; for each candidate design point, perform multiple random samplings within the range of variation of the operating parameters to obtain multiple combinations of operating parameters; input each candidate design point and the corresponding multiple combinations of operating parameters into the COP prediction model to obtain a set of performance prediction values ​​for the candidate design point with respect to the COP. Optimization output steps: Based on the performance prediction values ​​corresponding to each candidate design point, an optimized combination of design parameters is determined and output from the candidate design points. The simulation evaluation and optimization output steps are executed multiple times for iterative optimization: After the first execution, the design space is narrowed to form a neighborhood space with the output optimized design parameter combination as the center, and then the simulation evaluation and optimization output steps are executed again in the neighborhood space.

2. A method for optimizing parameters of a refrigeration system according to claim 1, characterized in that, The operating parameters include ambient temperature, compressor speed, and system thermal load power.

3. A method for optimizing parameters of a refrigeration system as set forth in claim 1, wherein It also includes a feature importance assessment step: based on the energy efficiency ratio COP prediction model, assessing the importance of different operating parameters to the energy efficiency ratio COP.

4. The method of claim 1 wherein, In the model building step, a random forest model is used as the machine learning model.

5. An energy saving controller for a refrigeration system, characterized in that, The system includes a memory and a processor, wherein the memory stores a computer program and a COP prediction model trained based on the refrigeration system parameter optimization method according to any one of claims 1 to 4; when the processor executes the program, it is used to perform: a. Real-time acquisition of system operating parameters, combined with fixed throttling element design parameters and refrigerant charge, to form an input feature vector; the operating parameters include ambient temperature and heat load power; b. Using the adjustable operating parameters in the system as decision variables, and the energy efficiency ratio COP prediction model as the objective function, the optimization algorithm searches online under the current operating conditions to find the optimal set point that maximizes the predicted energy efficiency ratio COP value and the corresponding theoretical optimal energy efficiency ratio COP value. c. Control the refrigeration system to operate at the optimal setpoint; d. When the change in the operating parameters is detected to exceed the preset threshold, repeat steps (b) and (c) to dynamically track and maintain the system’s optimal energy efficiency under the current operating conditions.

6. A refrigeration system health management apparatus characterized by comprising: The system includes a memory and a processor. The memory stores a computer program and a health baseline model. The health baseline model is a machine learning model trained using historical operating data of the refrigeration system under healthy conditions, based on the method described in any one of claims 1 to 4. This model is used to predict the theoretical energy efficiency ratio (COP) of the system, taking the optimized design parameters and operating parameters as input. When the processor executes the program, it performs the following: Collect real-time operating parameters of the system and combine them with the optimized design parameters of the system to form a real-time feature vector; The real-time feature vector is processed using the health benchmark model to obtain a health COP reference value; The deviation between the actual COP value of the calculation system and the healthy COP reference value is calculated. Based on the duration or fluctuation pattern of the deviation exceeding the preset threshold, a corresponding health status warning signal is generated.

7. A computer readable storage medium having stored thereon a computer program, characterized in that When the program is executed by the processor, it implements the steps of the refrigeration system parameter optimization method as described in any one of claims 1 to 4.

Images