Dynamic refrigerant filling process based on complete machine simulation operation

Abstract

The invention discloses a refrigerant dynamic charging process based on complete machine simulation operation, and belongs to the technical field of air conditioners and heat pumps. The problem that existing refrigerant charging depends on a static target value and cannot adapt to system individual differences and working condition changes in a self-adaption mode is solved. According to the process, a digital twinborn simulation model of a to-be-charged air conditioning system is constructed, the system is started to operate under a preset simulation working condition, and a first operation parameter is collected to initialize and calibrate the model; dynamically filling the refrigerant according to a preset step length, continuously collecting a real-time second operation parameter, and feeding the real-time second operation parameter back to the model; on the basis of real-time data, the model performs online fitting on a changing curve of parameters such as the supercooling degree and the superheat degree along with the charging amount, and predicts the refrigerant amount when the energy efficiency ratio of the system reaches the optimal as the optimal charging target; and when the parameters collected in real time and the extreme value region predicted by the model converge, it is judged that filling is completed. According to the system and the method, 'tailoring 'type dynamic accurate charging aiming at each specific air conditioning system is realized, and the charging precision and the system operation energy efficiency are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of air conditioning and heat pump technology, and in particular to a refrigerant dynamic charging process based on whole-machine simulation operation. Specifically, it relates to a refrigerant charging method, system and medium based on whole-machine simulation operation and digital twin technology for dynamic optimization. Background Technology

[0002] The performance, efficiency, and reliability of air conditioning and heat pump systems (hereinafter referred to as air conditioning systems) largely depend on the amount of refrigerant (refrigerant) charged into the system. Insufficient refrigerant charge will lead to a decrease in the system's cooling / heating capacity, excessively high exhaust temperature, poor compressor lubrication, and in severe cases, even compressor burnout. Excessive refrigerant charge will lead to increased condensing pressure, increased compressor power consumption, reduced energy efficiency ratio, and may cause liquid slugging risk, while also increasing greenhouse gas emissions. Therefore, achieving accurate refrigerant charging is a key technology in the manufacturing, installation, and maintenance of air conditioning systems.

[0003] Furthermore, with the increasing complexity of air conditioning system design, such as the widespread adoption of inverter technology, the application of electronic expansion valves, and the rapid development of multi-split systems, the system's sensitivity to refrigerant charge has become increasingly higher. A tiny charge deviation can lead to significant fluctuations in system energy efficiency and even trigger control logic malfunctions. Therefore, the industry's requirements for charge accuracy have been raised from the initial "gram-level" to "lean gram-level".

[0004] Currently, the common refrigerant charging methods in the industry can be mainly divided into several categories: One method is the quantitative charging method, which involves directly weighing and charging the air conditioner according to the rated charge amount indicated on its nameplate using an electronic scale. However, this method cannot adapt to changes in the system's internal volume caused by factors such as differences in the length of connecting pipes at the installation site and the height difference between the indoor and outdoor units, often resulting in poor actual operating performance.

[0005] Another type is the parameter monitoring method. For example, Chinese patent application CN104990320A discloses a control method and system for automatic refrigerant charging, including step S1, detecting the outdoor ambient temperature T1 and the air conditioning pressure value P1; step S2, looking up the minimum pressure value Ps corresponding to the outdoor ambient temperature T1 according to the minimum pressure correspondence table; step S3, determining whether Ps≥P1. If yes, the solenoid valve of the refrigerant storage device is opened to add refrigerant, and step S1 is executed again; if no, the compressor of the refrigerant storage device is turned on, and after a cooling operation time t, step S4 is executed; step S4, detecting the air conditioner condenser outlet temperature T2; step S5, determining whether ΔT≥Ts. If yes, the solenoid valve of the refrigerant storage device is opened to add refrigerant, and step S4 is executed again; if no, the refrigerant is determined to be sufficient, and refrigerant charging is stopped. This invention can ensure accurate refrigerant charging amount after installation or maintenance of the air conditioning system, reduce the refrigerant calculation process during installation and maintenance, and ensure high reliability of the air conditioning system. It determines whether charging is needed by detecting the outdoor ambient temperature and system pressure and comparing them with a preset table. Although this method takes environmental factors into account, its control logic is simple and relies on static, universal empirical data, making it unable to make precise adjustments for the individual characteristics of each specific air conditioning system (such as differences in compressor efficiency and heat exchanger performance deviations).

[0006] Another type of charging method is based on subcooling or superheating. For example, Chinese Patent Publication No. CN102893096B describes a method for charging an HVAC system by determining the relationship between the liquid line temperature, the suction line pressure, and the outdoor ambient temperature. Another method involves charging an HVAC system by adjusting the amount of refrigerant to approach a target minimum liquid line temperature. Yet another method involves charging an HVAC system by testing it according to at least three sets of test parameters, two of which include testing the system at approximately the same outdoor ambient temperature, and at least one of the remaining sets including testing the system at different outdoor ambient temperatures. This method adjusts the refrigerant dosage to achieve a target subcooling. This target subcooling is usually pre-calibrated through numerous experiments under standard operating conditions. However, in practical applications, air conditioning operating conditions vary greatly, and system components have manufacturing tolerances and performance degradation after long-term use. A fixed target value cannot guarantee that the system will operate optimally in all situations.

[0007] The limitation of this method is that its target subcooling is obtained by testing a prototype with a specific configuration (such as standard connecting pipe length) under specific operating conditions (such as standard refrigeration conditions). When faced with individuals with different installation environments, different connecting pipe lengths, or manufacturing deviations, the actual charge quantity corresponding to the fixed target subcooling is not the energy-efficient charge quantity for that individual. In other words, existing technologies pursue "achieving the subcooling target" rather than "optimal system energy efficiency".

[0008] Furthermore, Chinese patent application CN106198043A discloses a method for calibrating the refrigerant charge of an automotive air conditioning system. This method first provides a precise environmental simulation chamber to simulate the actual operating conditions of a vehicle, a refrigerant recovery and charging machine for recovering and charging refrigerant, and a data acquisition instrument for collecting temperature and pressure data. The vehicle is fixed inside the precise environmental simulation chamber and operated under preset conditions. A predetermined amount of refrigerant is charged at predetermined intervals. The data acquisition instrument collects the condenser outlet pressure, condenser outlet temperature, the temperature of each air outlet on the vehicle's interior surface, and the external circulation inlet temperature. The subcooling of the condenser outlet is calculated, and a curve showing the relationship between the condenser outlet subcooling, condenser outlet pressure, and refrigerant charge is plotted. Finally, the optimal refrigerant charge for the air conditioning system is calculated. Although this method utilizes an environmental simulation chamber for precise calibration, it is primarily suitable for the research and development stage. The process is complex and time-consuming, making it unsuitable for rapid charging requirements in production lines or on-site maintenance.

[0009] Some existing technologies attempt to dynamically adjust the refrigerant circulation volume during system operation, such as by setting up a liquid receiver or using complex control algorithms to cope with changes in operating conditions. However, these methods are all "passive compensation" or "dynamic adjustment" based on the system's existing initial charge volume. Their purpose is not to solve the problem of accurately determining the initial charge volume, and their adjustment range and effectiveness are limited by the hardware structure.

[0010] In summary, existing technologies generally lack a dynamic and precise refrigerant charging process capable of adapting to individual system differences and real-time operating condition changes. The fundamental problem lies in the fact that the charging process is open-loop or semi-open-loop, preventing the air conditioning system from "actively expressing" its optimal performance characteristics during the charging process. In other words, existing methods determine when to stop charging based on preset, static targets (such as weight, pressure, and subcooling values), which cannot reflect the actual refrigerant requirements of each specific air conditioning system in its specific installation environment and current state to achieve optimal performance. Summary of the Invention

[0011] To address the shortcomings of existing technologies, the purpose of this invention is to provide a dynamic refrigerant charging process based on whole-system simulation operation. This process constructs a digital twin model of the air conditioning system to be charged, performs dynamic charging during simulation operation, and uses real-time feedback data to drive online optimization of the model, ultimately achieving personalized and high-precision refrigerant charging for each specific air conditioning system.

[0012] The deeper purpose of this invention is to transform the traditional "open-loop" refrigerant charging process into a "closed-loop interactive, self-optimizing" intelligent process. By constructing a real-time data link between the physical entity and the virtual model, the system can autonomously "explore" and "discover" its optimal energy-efficient operating state during the charging process, thereby achieving perfect self-adaptation to individual differences.

[0013] The above-mentioned objective of this invention is achieved through the following technical solutions: On one hand, this invention discloses a refrigerant dynamic charging process based on whole-machine simulated operation, applicable to air conditioning or heat pump systems, including the following steps: Step S1, Construction Phase: Connect the air conditioning system to be charged to the refrigerant charging machine and establish a digital twin simulation model corresponding to the physical entity of the air conditioning system to be charged; the digital twin simulation model is configured to receive the real-time operating parameters of the air conditioning system to be charged and dynamically correct the model parameters. Specifically, the digital twin simulation model is a gray-box or white-box model built based on first-principles equations (such as the conservation equations of mass, energy, and momentum). Its initial parameters are derived from the design values ​​or historical statistical data of the same type of air conditioning system. The core of this model lies in its online-correctable "twin" characteristic, that is, it can automatically adjust the key parameters inside the model (such as the heat transfer coefficient of the heat exchanger, the volumetric efficiency and isentropic efficiency of the compressor, and the pressure drop coefficient of the pipeline) based on the real-time feedback data of the physical entity, so that the output behavior of the model is highly consistent with the actual behavior of the physical entity.

[0014] Preferably, the digital twin simulation model adopts a modular architecture, including at least a compressor module, a condenser module, an evaporator module, a throttling mechanism module, and a connecting pipeline module. Each module has a built-in mathematical model based on physical mechanisms. For example, the compressor module includes characteristic curves or mapping tables describing its volumetric efficiency and isentropic efficiency; the heat exchanger module uses the distributed parameter method or the moving boundary method to establish its dynamic heat transfer model. This modular and mechanism-based modeling approach ensures that even with limited data, the model can make reasonable predictions based on fundamental thermodynamic principles, exhibiting strong generalization ability and interpretability.

[0015] Step S2, Simulation Operation Phase: Start the air conditioning system to be charged and run it in the preset simulation operating mode; collect the first set of operating parameters of the air conditioning system to be charged in the simulation operating mode in real time, and synchronously input the first set of operating parameters into the digital twin simulation model to initialize and calibrate the model; Preferably, the preset simulation operating mode is designed to stimulate the dynamic response of the system at different operating points to obtain sufficient information for model parameter identification. This mode includes not only the basic operating modes of cooling or heating, but also the active stimulation of adjustable components inside the system (such as electronic expansion valves and fans).

[0016] Furthermore, the "active excitation" can be achieved by applying a pseudo-random binary sequence (PRBS) signal to the frequency of the electronic expansion valve or compressor to stimulate the system's dynamic response across the entire frequency domain of interest. This allows for a more accurate identification of the system's dynamic characteristic parameters, such as time constant and gain. This enables the calibrated digital twin model to not only match the physical entity in steady state but also maintain a high degree of consistency during dynamic processes, laying a more solid foundation for subsequent dynamic loading optimization.

[0017] Step S3, Dynamic Charging and Optimization Stage: The refrigerant charging machine is controlled to dynamically charge the air conditioning system to be charged with refrigerant at a preset step size and rate; during the charging process, the second set of operating parameters of the air conditioning system to be charged is continuously collected in real time and fed back to the calibrated digital twin simulation model; based on the real-time feedback of the second set of operating parameters, the digital twin simulation model calculates and predicts the trend of system performance indicators with the refrigerant charging amount online, and dynamically generates the optimal charging target amount; This stage is the core of the invention. The calibrated model is equivalent to a "virtual copy" that can accurately reflect the characteristics of the physical system. While charging the real system with refrigerant, the model also receives the "virtual charging" signal simultaneously, and predicts in advance the amount of refrigerant corresponding to the peak performance of the system (such as energy efficiency ratio EER or coefficient of performance COP) based on the changes in boundary conditions fed back by the real system.

[0018] The key here lies in "prediction." The model doesn't simply replicate the measurements of the physical entity; instead, it utilizes its built-in physical mechanisms and identified individualized parameters to deduce the future performance trajectory of the system under current changing trends. For example, the model can analyze the rate of increase in subcooling (SC) with increasing charge (first derivative) and the rate of change of this rate (second derivative), and, combined with the decreasing trend of superheat, predict the "inflection point" region where the energy efficiency ratio (EER) will shift from increasing to decreasing. This predictive capability is the core of this invention's superiority over traditional feedback control.

[0019] Step S4, Convergence Determination Stage: Compare the real-time acquired second set of operating parameters with the system performance index predicted by the digital twin simulation model. When the deviation between the two is within the preset convergence threshold range, and the system performance index reaches or approaches the predicted extreme value region, the filling is determined to be complete. This step ensures the reliability of the charging termination point through a dual verification process combining virtual and real-world testing. On one hand, it verifies whether the response of the real system matches the trajectory predicted by the model. On the other hand, it confirms that the system has entered its optimal performance region through multiple independent physical criteria (such as the subcooling slope and the inflection point of the exhaust temperature).

[0020] This multi-criteria fusion convergence decision mechanism has produced unexpected synergistic effects. A single physical criterion (such as the supercooling slope) may lead to misjudgments under certain operating conditions due to noise interference or sensor accuracy limitations. A single model prediction may also contain errors under certain boundary conditions. This invention combines the two: using model prediction to guide the macroscopic direction and using multiple independent physical characteristic signals for microscopic confirmation, essentially providing a "double insurance" for the charge termination decision. Experiments have shown that the robustness of this fused criterion far exceeds that of any single criterion, reducing the misjudgment rate to near zero and ensuring that each charge accurately converges to the true energy-optimal region.

[0021] Step S5, Termination Stage: Control the refrigerant charging machine to stop charging and record the final charging amount.

[0022] As a further technical solution of the present invention: the preset simulation operating mode in step S2 includes: cooling mode or heating mode; and by adjusting the opening of the electronic expansion valve, the indoor fan speed and the outdoor fan speed in the air conditioning system to be charged, at least two different system internal throttling or air volume combination states are constructed to stimulate the dynamic response characteristics of the system.

[0023] Furthermore, the at least two different system internal throttling or airflow combination states are set such that the main state parameters of the system (such as evaporation temperature, condensation temperature, superheat, and subcooling) change significantly, with the change amplitude being more than 5 times the sensor measurement noise, to ensure that the data used for model parameter identification has a sufficient signal-to-noise ratio. For example, the electronic expansion valve can be adjusted from the minimum opening to the maximum opening in steps, or the indoor fan can be operated at high, medium, and low fan speeds respectively.

[0024] In a preferred embodiment, the simulated operating mode further includes adjusting the compressor operating frequency. For inverter air conditioners, by changing the compressor frequency, system characteristics under different load conditions can be simulated, thereby more comprehensively identifying the compressor efficiency curve and the system performance at different pressure ratios. This allows the calibrated digital twin model to have high fidelity across the entire operating range.

[0025] As a further technical solution of the present invention: the second set of operating parameters in step S3 includes at least: compressor suction pressure, compressor discharge pressure, compressor suction temperature, compressor discharge temperature, condenser middle temperature, condenser outlet temperature, evaporator outlet temperature, electronic expansion valve opening, compressor operating frequency, and compressor input power.

[0026] Furthermore, preferably, the second set of operating parameters also includes outdoor ambient temperature and indoor ambient temperature. These two parameters serve as important boundary condition inputs to the model, which can further improve the accuracy of model predictions and allow the model to compensate for environmental changes, making the optimization process unaffected by minor fluctuations in ambient temperature.

[0027] These parameters form the basis for constructing the system's thermodynamic state and calculating performance indicators. To ensure data accuracy and real-time performance, the sampling frequency of the pressure and temperature sensors should be no less than 1 Hz, and the measurement accuracy should be better than ±0.5% and ±0.2℃, respectively. Setting the temperature at the center of the condenser is to more accurately calculate the refrigerant subcooling, especially when using large-diameter heat exchangers, where the outlet temperature may not fully represent the temperature of the saturated liquid.

[0028] The temperature sensor in the middle of the condenser is preferably installed at 80% to 90% of the refrigerant flow path, where the refrigerant has typically just transitioned from a two-phase state to a single-phase liquid state. The measured temperature here is closest to the saturation temperature, allowing for the calculation of the subcooling closest to the true value. This subtle improvement in sensor placement significantly enhances the accuracy of subcooling calculation, thereby improving the overall precision of the optimization algorithm.

[0029] As a further technical solution of the present invention: the specific method for dynamically generating the optimal filling target amount using the digital twin simulation model in step S3 includes: Step S31: Based on the second set of operating parameters with real-time feedback, calculate the real-time subcooling (SC) and real-time superheat (SH); Wherein, subcooling SC = saturation temperature corresponding to condenser outlet pressure - condenser outlet temperature; superheat SH = compressor suction temperature - saturation temperature corresponding to evaporator outlet pressure.

[0030] Step S32: Correlate the real-time subcooling (SC) and real-time superheat (SH) with the charge change (Δm) and generate the real-time subcooling change curve SC=f(Δm) and the real-time superheat change curve SH=g(Δm) online. The online fitting can employ algorithms such as sliding window least squares or Kalman filtering to fit data points collected within a recent time window (e.g., the last 10 injection steps) to eliminate random noise interference and update the curve function and its derivative in real time.

[0031] The preferred online fitting algorithm is recursive least squares with a forgetting factor (FFRLS). The introduction of the forgetting factor λ (0.95 < λ < 1) allows the algorithm to automatically reduce the influence weight of historical data on the current fitting result, thereby enabling the model to quickly track real-time changes in system characteristics (e.g., subtle changes in the heat exchanger's heat exchange capacity during charging), ensuring that the fitted SC=f(Δm) and SH=g(Δm) curves always best reflect the "current" state of the system.

[0032] Step S33: The digital twin simulation model predicts the inflection point of the system's energy efficiency ratio (EER) or coefficient of performance (COP) based on the preset objective function and the derivative changes of SC=f(Δm) and SH=g(Δm), and determines the predicted refrigerant quantity corresponding to the inflection point as the optimal charging target quantity.

[0033] The objective function is typically defined as the ratio of system cooling capacity (or heating capacity) to compressor input power, i.e., EER or COP. The model has a pre-existing EER / COP estimation model built on thermodynamic principles, which takes real-time collected pressure, temperature, power, and current SC and SH values ​​as input. By analyzing the current form and trends of SC=f(Δm) and SH=g(Δm) (e.g., slower SC growth, slower SH decline), the model infers changes in system volumetric efficiency and isentropic efficiency, thus predicting the peak point and its corresponding charge amount before the actual EER / COP reaches its peak. This is significantly more advanced and predictive than traditional methods that can only "review" historical data or rely on fixed target values.

[0034] In a preferred embodiment, the digital twin model integrates a Physically Based Neural Network (PINN) as a fast predictor. This PINN is pre-trained on a large amount of offline simulation data, learning the intrinsic mapping relationship between the peak EER and the shapes of the SC and SH curves under different system characteristics. After model calibration in step S2, the calibrated key system parameters are used to fine-tune the weights of specific layers in the PINN. During dynamic filling, the PINN takes the real-time fitted SC=f(Δm) and SH=g(Δm) curve features (such as the current slope and curvature) as input and directly outputs the predicted optimal filling amount M_target. Simultaneously, a simplified model based on a rigorous physical mechanism runs in parallel to verify the PINN's predictions in real time. When the predictions of the two models are consistent (the deviation is less than a preset threshold, such as 5g), the system adopts the fast prediction result of the PINN, significantly improving the optimization speed; when discrepancies occur, the result of the physical model is used to ensure the reliability of the decision. This hybrid modeling strategy of "data-driven acceleration and physical model as a safety net" is a key innovation of this invention. Its synergistic effect is to obtain the high-speed computing power of neural networks while retaining the reliability and interpretability of physical models, thus achieving extreme optimization of both filling efficiency and accuracy.

[0035] As a further technical solution of the present invention: the convergence determination in step S4 further includes: monitoring the slope d(SC) / d(Δm) of the real-time subcooling change curve SC=f(Δm), and triggering the first charging termination signal the instant when the slope changes from a positive value to zero or from zero to a negative value.

[0036] The physical mechanism is as follows: when the refrigerant charge is insufficient, increasing the refrigerant will significantly improve the condenser's liquid seal effect and increase the subcooling, thus d(SC) / d(Δm) will be a large positive value. When the subcooling reaches near the optimal point, the condenser's heat exchange area has been fully utilized. Continuing to increase the refrigerant mainly leads to an increase in condensing pressure, while the subcooling increases extremely slowly or even stops, meaning the slope approaches zero. If excessive charging continues, the condensing pressure will rise sharply, potentially causing a slight decrease in subcooling, and the slope will become negative. Therefore, the instant the slope changes from positive to zero or from zero to negative is an important indicator for identifying the optimal charging point. To prevent misjudgment, a weighted average of the slopes from multiple consecutive points can be used for determination.

[0037] The physical significance of this criterion lies in its direct correlation with the condenser's "effective utilization rate." When d(SC) / d(Δm) approaches zero, it means that the added refrigerant can no longer improve the subcooling effect by forming a longer liquid seal, but instead begins to "encroach" on the condenser's effective heat exchange area, leading to an increase in condensing pressure. Therefore, this inflection point essentially indicates the critical state of condenser utilization rate transitioning from "gain" to "saturation," a precursor to an impending decline in energy efficiency.

[0038] As a further technical solution of the present invention: the convergence determination in step S4 further includes: monitoring the changing trend of the compressor exhaust temperature, and triggering a second charging termination signal when the compressor exhaust temperature shows a turning point of first decreasing and then increasing during the charging process.

[0039] The physical mechanism is as follows: In the initial stage of charging, the system has insufficient refrigerant, resulting in high suction superheat and thus a higher exhaust temperature. As the refrigerant is added, the suction superheat decreases, the compression ratio becomes more reasonable, and the exhaust temperature gradually decreases. When the charge exceeds the optimal point, the condensing pressure increases, the compression ratio increases, and the exhaust temperature rebounds. Therefore, the inflection point where the exhaust temperature changes from decreasing to increasing (i.e., the point where the first derivative changes from negative to positive) is another important indicator that the system has reached its thermodynamic optimum. This criterion is independent of the subcooling slope criterion, and their combined use can improve the robustness of the determination.

[0040] The unique value of the discharge temperature inflection point criterion lies in its comprehensive reflection of the compressor's "workload" and "suction state." Discharge temperature is the result of the combined effects of the compressor's suction state (pressure and temperature) and discharge pressure (i.e., condensing pressure). Its inflection point signifies that the system has found an optimal balance: at this point, the suction superheat has decreased to a reasonable level, while the discharge pressure has not yet excessively increased. The compressor exhibits the highest volumetric efficiency and lowest power consumption near this point. Therefore, this criterion directly points to the optimal region of the compressor's own operating efficiency, forming an excellent complement to the subcooling slope criterion, which reflects condenser utilization. The synergistic use of these two criteria—one focusing on the "condensing side" and the other on the "compression process"—jointly ensures optimal overall system energy efficiency.

[0041] As a further technical solution of the present invention: before starting the air conditioning system to be charged in step S2, the method further includes: performing a vacuuming operation on the air conditioning system to be charged and maintaining the vacuum level below a preset threshold for at least 15 minutes to check the airtightness of the system. The preset threshold is usually an absolute pressure below 50 Pa to ensure that the system is dry and free of impurities, and to prevent non-condensable gases from affecting the charging accuracy and system operating performance.

[0042] As a further technical solution of the present invention: the digital twin simulation model is built into the controller of the refrigerant charging machine, or built into a host computer that is communicatively connected to the air conditioning system to be charged and the refrigerant charging machine. When built into a host computer, the host computer can be an industrial tablet PC, PC, or server, which communicates in real time with the data acquisition unit and control unit via wired or wireless means (such as Wi-Fi, Bluetooth, industrial Ethernet) to realize centralized calculation of the model and command issuance.

[0043] Preferably, the host computer is an edge computing gateway deployed at the production site or maintenance service point. This edge gateway not only runs the digital twin simulation model but also interacts with the factory's Manufacturing Execution System (MES) or the cloud-based after-sales service system. It can bind the unique identifier (such as serial number) of each air conditioner to the final accurate filling quantity and upload it to the quality traceability database. Simultaneously, it can download the latest model library or algorithm updates from the cloud, enabling continuous evolution of its capabilities. This edge computing architecture ensures the real-time nature of the filling process (without relying on the cloud network) while achieving centralized data management and continuous model optimization. It perfectly embodies the concept of the Industrial Internet in the filling process, and its synergistic effect is the seamless integration of single-point intelligence and system intelligence.

[0044] On the other hand, the present invention also discloses a refrigerant dynamic charging system for implementing the above-mentioned refrigerant dynamic charging process based on whole-machine simulated operation, comprising: The air conditioning system to be charged includes a compressor, condenser, evaporator, electronic expansion valve, and corresponding temperature and pressure sensors; In an improved embodiment, high-precision digital temperature and pressure sensors are pre-installed at key locations of the air conditioning system to be charged, and connected to an external data acquisition unit via standardized communication interfaces (such as RS485 or CAN bus). This provides a hardware foundation for achieving fast and reliable data acquisition, avoiding problems such as poor contact or accuracy loss caused by temporarily adding sensors.

[0045] The refrigerant charging unit is connected to the refrigerant circulation loop of the air conditioning system to be charged via an openable and closable pipeline, and is used to accurately control and measure the amount of refrigerant charged. The refrigerant charging unit typically includes a refrigerant source, a charging pump, a high-precision mass flow meter (accuracy better than ±0.5%), and a charging valve composed of a solenoid valve or an electric valve.

[0046] The refrigerant charging unit is also equipped with a high-precision electronic expansion valve, which is used to precisely control the charging rate during the charging process, achieving a smooth transition from rapid coarse charging to fine-tuning. During the optimization phase, the electronic expansion valve can perform micro-step charging as low as 0.1 grams per charge according to the instructions of the digital twin model, greatly improving the resolution of the optimization process and the final charging accuracy.

[0047] The data acquisition and control unit is electrically connected to each sensor of the air conditioning system to be charged and the refrigerant charging unit. It is used to collect operating parameters in real time and issue control commands. This unit can be an industrial controller based on a PLC or embedded system. It is responsible for the acquisition, filtering and conversion of analog / digital signals, and executes control commands issued by the upper-level decision-making unit (such as a digital twin simulation unit), such as starting / stopping charging and adjusting the charging rate.

[0048] The data acquisition and control unit features high-speed synchronous sampling, ensuring that all sensor signals (pressure, temperature, power, flow) are acquired simultaneously with a timestamp error of less than 1 millisecond. This provides high-quality synchronous data for subsequent accurate online calculations and dynamic analysis of the digital twin model, avoiding phase errors and calculation distortions caused by data asynchrony.

[0049] The digital twin simulation unit interacts with the data acquisition and control unit, and has a built-in dynamic simulation model of the air conditioning system to be charged. It receives real-time data, performs model calibration, predicts the optimal charging volume, and feeds back decision information to the data acquisition and control unit. This unit is the "brain" of the entire system; its internal model includes thermodynamic component models, as well as parameter identification and optimization algorithms.

[0050] The digital twin simulation unit also includes a self-diagnostic module. This module continuously monitors the residuals between the model's predicted values ​​and the measured values. When the residuals increase abnormally over a certain period, the self-diagnostic module will issue an alarm, indicating potential problems such as sensor failure, system leakage, or model mismatch, and can preliminarily locate the fault source based on the residual characteristics. For example, if the evaporator outlet temperature sensor reading continuously deviates from the model's predicted value, the system may indicate "evaporator temperature sensor abnormality" or "abrupt change in evaporator heat transfer performance." This self-diagnostic capability significantly improves the system's reliability and provides data support for subsequent maintenance.

[0051] On the other hand, the present invention also discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a dynamic refrigerant charging process based on whole-machine simulation operation as described above.

[0052] Compared with the prior art, the present invention includes at least one of the following beneficial technical effects: 1. Achieving truly personalized and precise charging: This invention no longer relies on statistically significant standard charging amounts or fixed target values. Instead, through real-time interaction between a digital twin model and the physical entity during the charging process, it "tailor-makes" the optimal charging amount for each air conditioning system, perfectly adapting to all individual characteristics caused by manufacturing tolerances, component differences, pipe lengths, and installation environments. For example, for two air conditioners of the same model, due to slight differences in compressor efficiency or heat exchanger performance, the process of this invention may determine slightly different optimal charging amounts for them, enabling both to reach their respective optimal energy efficiency states—something traditional methods cannot achieve. The unexpected effect of this personalized and precise charging is that it greatly reduces the dispersion of product performance at the factory. On traditional production lines, even if all components are qualified, the final overall energy efficiency may fluctuate significantly due to the cumulative effect of manufacturing tolerances. After applying this invention, each product is adjusted to its own optimal point, resulting in a revolutionary improvement in the energy efficiency consistency of off-line products. This has immeasurable value for enhancing brand image and meeting high-standard energy efficiency certification requirements.

[0053] 2. Full Lifecycle Adaptability: Whether it's new machine production, installation and commissioning, or refrigerant replacement after many years of use, this process can dynamically optimize the refrigerant quantity for the current system state. Natural performance degradation (such as compressor wear and heat exchanger dust accumulation) is captured in real-time and reflected in the optimization process, ensuring the air conditioner always operates at its optimal state. For example, the heat exchanger efficiency of an air conditioner used for five years may decrease. During maintenance and recharging, this process recalibrates the model based on current operating parameters to find a new, optimal refrigerant quantity that matches the current system, rather than blindly charging according to the factory value on the nameplate. This synergistic effect of full lifecycle adaptability transforms the "one-time lifespan" of the air conditioning system into "multiple lifecycles." When system performance declines due to aging, the traditional approach is often to directly replace the system or accept the reality of reduced energy efficiency. However, this invention, by precisely matching the needs of an aging system, maximizes its remaining potential, delays the impact of performance degradation on the user experience, maximizes the utilization of equipment value, and reduces unnecessary resource waste and premature scrapping.

[0054] 3. Significantly Improved Filling Efficiency and Accuracy: Compared to traditional intermittent filling and detection methods, the dynamic continuous filling of this invention, combined with model prediction, greatly shortens the time required to find the optimal filling point. The traditional "fill-wait-measure-refill" cycle requires multiple system stabilizations, taking tens of minutes or even longer. This invention simultaneously completes data acquisition, model calibration, and trend prediction during the dynamic filling process, ensuring continuous and uninterrupted filling, typically completing optimization within minutes. By capturing multiple characteristic signals such as subcooling slope changes and exhaust temperature inflection points, filling accuracy can be controlled within ±5g, far exceeding the ±15g or even worse control levels of existing technologies. Even more surprisingly, because the optimization process is continuous and guided by model prediction, it avoids the risk of "overshooting" the optimal point due to discrete measurements, a risk inherent in traditional methods. Model prediction can indicate the location of the optimal point in advance; as the system approaches this point, the control unit automatically reduces the filling rate, making fine adjustments with extremely high resolution until the target is accurately hit. This combination of "predictive deceleration" and "precise hit" is something that pure feedback control methods cannot achieve. It enables the filling process to have both high speed and high precision, and the two are no longer mutually exclusive.

[0055] 4. Driving Intelligent Transformation of the Refrigerant Charging Process: This invention introduces advanced control concepts such as digital twins, dynamic optimization, and multi-source information fusion into the traditional refrigerant charging field, constructing an intelligent closed loop of "collection-modeling-decision-execution," providing key technical support for the intelligent upgrading of air conditioning manufacturing and the digital transformation of after-sales service. This not only improves the quality and energy efficiency of individual products but also lays the data foundation for achieving full lifecycle product quality traceability and performance optimization. The deeper significance of this transformation lies in its shift from a simple "physical filling" step to a crucial "data generation and value extraction" step. Each charging operation generates a high-value dataset containing the individual characteristics of the air conditioning system, the optimal charging amount, and its relationship with energy efficiency. This data can be used upwards to optimize product design, providing feedback to the R&D department to improve the matching of heat exchangers and compressors in the next generation of products; downwards, it can guide after-sales service, providing users with more accurate maintenance suggestions. Therefore, this invention is not only a process improvement but also a key link in building an enterprise data closed loop and realizing intelligent manufacturing, generating synergistic value far exceeding the charging process itself. Attached Figure Description

[0056] Figure 1 This is a process flow diagram of the present invention.

[0057] Figure 2 This is a system structure block diagram of the present invention.

[0058] Reference numerals: 100, Air conditioning system to be charged; 110, Compressor; 120, Four-way valve; 130, Outdoor heat exchanger; 140, Electronic expansion valve; 150, Indoor heat exchanger; 200, Refrigerant charging unit; 210, Refrigerant source; 220, Charging pump; 230, Mass flow meter; 240, Charging valve; 300, Data acquisition and control unit; 400, Digital twin simulation unit. Detailed Implementation

[0059] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0060] In the description of this application, it should be noted that the terms "upper," "lower," "inner," "outer," "top / bottom," etc., indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0061] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installed," "equipped with," "sleeved / connected," "connected," etc., should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0062] This invention discloses a refrigerant dynamic charging process based on whole-system simulation operation, applicable to air conditioning or heat pump systems, comprising the following steps: Step S1, Construction Phase: Connect the air conditioning system 100 to be charged to the refrigerant charging machine, and establish a digital twin simulation model corresponding to the physical entity of the air conditioning system 100 to be charged; the digital twin simulation model is configured to receive the real-time operating parameters of the air conditioning system 100 to be charged and dynamically correct the model parameters. In this step, the charging pipeline of the refrigerant charging unit 200 is first connected to the process valve port of the air conditioning system 100 to be charged via a quick-connect fitting or a shut-off valve, typically on the suction or discharge side of the compressor 110. Simultaneously, ensure that the data acquisition and control unit 300 has established a reliable electrical connection with all sensors (pressure, temperature) and actuators (compressor inverter, electronic expansion valve driver, fan driver) within the air conditioning system 100 to be charged. After connection, input the model information of the air conditioning system 100 to be charged into the human-machine interface of the host computer or controller. The system then retrieves the corresponding basic simulation model from its internal database. This basic model includes the topology and component design parameters of the air conditioning model (such as compressor displacement, heat exchanger pipe length and inner diameter, electronic expansion valve model, etc.), but the performance parameters (such as the heat transfer coefficient UA of the heat exchanger, the compressor volumetric efficiency η_v, and the isentropic efficiency η_is coefficients) are set to factory nominal values ​​or empirical values, which will be calibrated in subsequent steps.

[0063] To further improve the efficiency and accuracy of model calibration, step S1 also includes a "model pre-matching" sub-step. The data acquisition and control unit 300 briefly starts the compressor 110 to acquire a set of system static pressure data within a very short time (e.g., 5 seconds). This data is used to quickly filter multiple candidate model versions in the database, selecting the one that best matches the current system's static characteristics as the base model. This provides a better starting point for subsequent fine calibration, thereby reducing the number of iterations and time required for subsequent parameter identification.

[0064] Step S2, Simulation Operation Phase: Start the air conditioning system 100 to be charged and run it in the preset simulation operating mode; collect the first set of operating parameters of the air conditioning system 100 to be charged in the simulation operating mode in real time, and synchronously input the first set of operating parameters into the digital twin simulation model to initialize and calibrate the model; After confirming the system's airtightness, the data acquisition and control unit 300 issues a command to start the compressor 110 and fan, and sets the state of the four-way valve 120, causing the system to operate in a preset simulated operating mode, such as cooling mode. To fully stimulate the system, the controller will change the opening degree of the electronic expansion valve 140 and / or the fan speed according to a preset sequence. For example, first, the outdoor and indoor fans are kept at their rated speeds, and the electronic expansion valve 140 is set sequentially to 200 steps, 350 steps, and 500 steps, running stably for 3-5 minutes at each opening degree, recording the average operating parameters under each stable state. The first set of operating parameters of the air conditioning system 100 to be charged under the simulated operating mode is collected in real time, and the first set of operating parameters is synchronously input into the digital twin simulation model for initialization and calibration; after receiving multiple sets of steady-state data, the digital twin simulation unit 400 starts the parameter identification algorithm. For example, using data such as compressor suction pressure, discharge pressure, suction temperature, discharge temperature, and power collected at different operating degrees as targets, and employing optimization algorithms (such as genetic algorithms or particle swarm optimization) to adjust key parameters within the model (such as the UA values ​​of the condenser and evaporator, and the coefficients of the compressor efficiency curve), the sum of squared errors between the model's simulation outputs (such as pressure and temperature) and the measured data is minimized. When the average error under all operating conditions is lower than a preset threshold (e.g., pressure error <1%, temperature error <0.5℃), the model initialization calibration is complete. At this point, the digital twin model can accurately reflect the physical characteristics of "this" specific air conditioning system under its current state.

[0065] In a preferred embodiment, the initial calibration process focuses not only on steady-state error but also on dynamic response characteristics. The digital twin simulation unit 400 acquires dynamic response data (such as the intake pressure change curve) of the system at a higher sampling frequency (e.g., 10Hz) at the instant the electronic expansion valve opening switches. The model parameter identification algorithm simultaneously optimizes the model's performance at multiple steady-state points and during the dynamic transition from one steady state to another, ensuring that the calibrated model accurately simulates both the system's steady-state performance and its dynamic behavior. This precise calibration of dynamic characteristics provides a more reliable model foundation for trend-based predictions during subsequent dynamic charging.

[0066] Step S3, Dynamic Charging and Optimization Stage: The refrigerant charging machine is controlled to dynamically charge the air conditioning system 100 to be charged with refrigerant at a preset step size and rate; during the charging process, the second set of operating parameters of the air conditioning system 100 to be charged is continuously collected in real time and fed back to the calibrated digital twin simulation model; based on the real-time feedback of the second set of operating parameters, the digital twin simulation model calculates and predicts the trend of system performance indicators with the refrigerant charging amount online, and dynamically generates the optimal charging target amount; After model calibration, the system automatically enters the dynamic charging phase. The data acquisition and control unit 300 first stabilizes the air conditioning system under a baseline operating condition, for example, setting the electronic expansion valve 140 opening to 400 steps and the compressor 110 operating frequency to 60Hz. After 2 minutes of stable operation, the controller sends a command to the refrigerant charging unit 200 to open the charging valve 240 and the charging pump 220. The charging pump 220 charges the system with refrigerant at a constant rate (e.g., 2g / s), and the mass flow meter 230 feeds back the cumulative charging amount Δm and instantaneous flow rate to the controller in real time. The controller collects a second set of operating parameters every second and packages it for transmission to the digital twin simulation unit 400.

[0067] The digital twin simulation unit 400 runs a real-time optimization algorithm. Upon receiving a new set of data, it first calculates the current SC and SH, adding (Δm, SC) and (Δm, SH) as new data points to their respective dynamic data sequences. Then, using a recursive least squares method with a forgetting factor, it performs online fitting of SC=f(Δm) and SH=g(Δm), updating their first derivatives f'(Δm) and g'(Δm). Simultaneously, based on the currently acquired pressure, temperature, and power data, combined with the model's internal thermodynamic relationships, it estimates the system performance (EER_actual) at the current moment online. The core optimization logic of the model is based on a pre-stored mapping of the relationship between "EER and SC / SH". This mapping shows that, with other system conditions remaining constant, EER increases with increasing SC, but reaches a peak after a certain point due to changes in the compression ratio as SH decreases. Based on the current real-time estimated EER_actual, the trend of the fitted SC=f(Δm) curve (especially its second derivative information), and the trend of the SH=g(Δm) curve, the model predicts the trajectory of EER after a future injection of Δm. The model continuously updates its prediction of the EER peak and outputs the predicted Δm value corresponding to that peak (i.e., the optimal injection target amount M_target) to the data acquisition and control unit 300. For example, the model may display in real time: "The predicted optimal injection target amount is 35g more than the current injection amount."

[0068] During the optimization process, the digital twin simulation unit 400 also executes a "confidence assessment" subroutine. This subroutine calculates an "optimal filling target prediction confidence level" in real time based on the quality of the current fitted curve (e.g., R-squared value), the number of data points, and the historical consistency of the model's prediction results. When the confidence level is low, the system automatically reduces the filling rate or triggers a short system stabilization period to acquire more high-quality data points until the confidence level rises above a preset threshold. This dynamic adjustment mechanism ensures that the optimization process is both fast and robust, avoiding erroneous decisions when data quality is poor, and is another key design feature ensuring process reliability.

[0069] Step S4, Convergence Determination Stage: Compare the real-time acquired second set of operating parameters with the system performance index predicted by the digital twin simulation model. When the deviation between the two is within the preset convergence threshold range, and the system performance index reaches or approaches the predicted extreme value region, the filling is determined to be complete. The data acquisition and control unit 300 monitors multiple convergence conditions in real time and makes a comprehensive judgment using either AND logic or weighted voting logic. Specific criteria are as follows: Model prediction consistency criterion: Calculate the average deviation ΔEER between the EER_actual estimated based on real-time data and the EER_target predicted by the digital twin model over the past 5 seconds. When ΔEER is less than a preset threshold (e.g., 0.05), and the slope of change of EER_target is less than a minimum value (e.g., 0.001 / s) for three consecutive samples, it indicates that the system performance has stabilized near the peak value predicted by the model.

[0070] Supercooling slope criterion: Monitor the first derivative f'(Δm) of SC=f(Δm). When f'(Δm) changes from greater than 0.01 to less than 0.01, and finally becomes less than 0.005 in 5 consecutive sampling points, the first charging termination signal T1 is triggered.

[0071] Exhaust temperature inflection point criterion: Monitor the trend of exhaust temperature T1. Use second derivative or pattern recognition methods to determine the inflection point. When it is detected that after a period of decrease or stabilization, T1 begins to show values ​​at three consecutive sampling points that are all higher than the average of the previous three sampling points, it is determined that the exhaust temperature has passed the inflection point and has begun to rise again, triggering the second charging termination signal T2.

[0072] When the model prediction consistency criterion is met, and at least one of T1 and T2 is triggered, the system finally determines that the filling is complete.

[0073] In a highly preferred embodiment, when all three criteria are met, the system determines that "optimal charging is complete" and records the final charging amount at this point as the "golden charging amount." If both T1 and T2 are triggered but the model prediction consistency criterion is not fully met (e.g., ΔEER is slightly large), the system enters a brief "observation period," stopping charging while continuing to monitor system stability. If ΔEER converges to within the threshold during the observation period, charging is confirmed to be complete; if ΔEER diverges, interference is suspected, the system records the anomaly, and prompts for manual intervention. This multi-level, fault-tolerant convergence determination strategy exhibits strong robustness and adaptability in actual industrial production environments, effectively coping with complex situations such as power fluctuations and environmental airflow interference. This synergistic effect is unmatched by single criteria or simple combinations of criteria.

[0074] Step S5, Termination Stage: The refrigerant charging machine is stopped, and the final charge amount is recorded. The controller immediately sends a stop command to the refrigerant charging unit 200, shutting down the charging pump 220 and the charging valve 240. The system displays the final charge amount m_final (usually in grams) accumulated by the mass flow meter 230 on the interface and associates it with the unique identifier (such as the serial number) of the air conditioning system, storing it in the local database or uploading it to the cloud server for quality traceability and subsequent analysis.

[0075] After the termination phase, the system can optionally perform a "filling effect verification" step. After filling is complete and the system has been running stably for 1-2 minutes, the data acquisition and control unit 300 collects another set of operating parameters and inputs them into the digital twin model. The model calculates a "verified energy efficiency ratio" and compares it with the optimal energy efficiency ratio predicted during the optimization process. If the two are highly consistent (deviation less than 2%), the filling verification is successful; if the deviation is large, the system generates a "filling quality warning," prompting operators or subsequent quality inspection stages to pay close attention. This verification step provides a final closed-loop confirmation of filling quality, ensuring that products flowing into the next process or delivered to customers are of high quality.

[0076] In one embodiment of the present invention, the preset simulation operating mode in step S2 includes either a cooling mode or a heating mode; and by adjusting the opening degree of the electronic expansion valve 140, the indoor fan speed, and the outdoor fan speed within the air conditioning system 100 to be charged, at least two different combinations of throttling or airflow within the system are constructed to stimulate the dynamic response characteristics of the system. This helps to more comprehensively identify system model parameters. For example, by changing the fan speed, the heat transfer coefficient correlation of the heat exchanger under different air-side flow rates can be identified; by changing the opening degree of the electronic expansion valve, the flow characteristics of the throttling mechanism can be identified, thereby enabling the digital twin model to have high fidelity across the entire operating range.

[0077] In one embodiment of the present invention, the second set of operating parameters mentioned in step S3 includes at least: compressor 110 suction pressure, compressor 110 discharge pressure, compressor 110 suction temperature, compressor 110 discharge temperature, condenser mid-temperature, condenser outlet temperature, evaporator outlet temperature, electronic expansion valve 140 opening, compressor 110 operating frequency, and compressor 110 input power. These parameters are the basis for characterizing the system's operating status and calculating performance indicators. In particular, the setting of the "condenser mid-temperature" is for more accurate calculation of subcooling. In large air conditioning systems, the refrigerant typically passes through three regions in the condenser: overheating, two-phase, and subcooling. The outlet temperature sensor may be installed at the rear, and the measured temperature is the subcooled liquid temperature, which can be directly used to calculate the subcooling accurately. However, in some small systems, the condenser outlet may still be in the two-phase region, and using the outlet temperature to calculate the subcooling will produce errors. By setting a temperature measuring point in the middle of the condenser (usually located at the end of the two-phase region), the saturation temperature can be determined more reliably, thereby calculating a more accurate subcooling.

[0078] In one embodiment of the present invention, the specific method for dynamically generating the optimal filling target amount using the digital twin simulation model in step S3 includes: Step S31: Based on the second set of operating parameters with real-time feedback, calculate the real-time subcooling (SC) and real-time superheat (SH); Step S32: Correlate the real-time subcooling (SC) and real-time superheat (SH) with the charge change (Δm) and generate the real-time subcooling change curve SC=f(Δm) and the real-time superheat change curve SH=g(Δm) online. To ensure the stability and accuracy of the fitted curve, this embodiment employs recursive least squares (RLS) with a forgetting factor for online fitting. The forgetting factor λ is typically set between 0.9 and 1, for example, λ = 0.95. This means that the weight of earlier historical data gradually decreases, allowing the model to better track the real-time changes of the current system. For SC = f(Δm), a quadratic polynomial model is used for fitting: SC = a·Δm² + b·Δm + c. By iteratively updating the coefficients a, b, and c for new data points, its first derivative d(SC) / d(Δm) = 2a·Δm + b can be directly calculated.

[0079] To address potential sensor noise or transient disturbances, step S32 also includes a "data cleaning and filtering" sub-step. Before using new data points (Δm, SC) and (Δm, SH) for fitting, they are first passed through an outlier detector based on model predictions. This detector uses the current model's predicted values ​​for SC or SH and their prediction intervals to determine whether the new data point falls within a reasonable confidence interval (e.g., a 95% confidence interval). If a new data point is identified as an outlier (e.g., caused by transient electromagnetic interference), it is discarded and not included in the update of the fitting parameters. This mechanism greatly improves the robustness of the fitted curve to noise and ensures the smoothness of the optimization process.

[0080] Step S33: The digital twin simulation model predicts the inflection point of the system's energy efficiency ratio (EER) or coefficient of performance (COP) based on the preset objective function and the derivative changes of SC=f(Δm) and SH=g(Δm), and determines the predicted refrigerant quantity corresponding to the inflection point as the optimal charging target quantity.

[0081] The digital twin model contains a pre-stored physics-based EER estimator. This estimator utilizes real-time collected compressor input power W and estimates the cooling capacity Q_e using the model. The cooling capacity Q_e is estimated based on the heat exchange on the evaporator side, which can be dynamically calculated from the real-time superheat SH, suction pressure, and the evaporator UA value obtained during the model calibration phase. Therefore, real-time EER ≈ Q_e / W. The process of the model predicting the optimal charge target amount essentially extrapolates the change in EER by using the relationship between SC and SH and the charge amount. The model assumes that near the optimal charge point, the rate of increase in cooling capacity will slow down due to the decrease in evaporation temperature, while the rate of increase in compressor power consumption will accelerate due to the increase in condensing pressure, and the ratio of these two, EER, will peak. By analyzing the current values ​​and rates of change of SC=f(Δm) (reflecting the condenser-side state) and SH=g(Δm) (reflecting the evaporator-side state), the model can predict in advance the Δm value that will maximize the EER. For example, the model calculates d(EER) / d(Δm) = ∂EER / ∂SC * d(SC) / d(Δm) + ∂EER / ∂SH * d(SH) / d(Δm) using the built-in sensitivity analysis module. Setting d(EER) / d(Δm) = 0, the predicted optimal injection amount can be solved.

[0082] In one embodiment of the present invention, the convergence determination in step S4 further includes: monitoring the slope d(SC) / d(Δm) of the real-time subcooling change curve SC=f(Δm), and triggering a first charging termination signal the instant the slope changes from a positive value to zero or from zero to a negative value. When the rate at which subcooling increases with the charging amount slows to zero or begins to decrease, it usually means that the effective heat exchange area in the condenser has been fully utilized, and continued charging will lead to a rapid increase in condensing pressure, which is an important indicator of approaching the optimal charging point.

[0083] In one embodiment of the present invention, the convergence determination in step S4 further includes: monitoring the changing trend of the exhaust temperature of compressor 110, and triggering a second charging termination signal when the exhaust temperature of compressor 110 shows a turning point of first decreasing and then increasing during the charging process. The lowest point of exhaust temperature usually corresponds to the optimal state of the system's intake superheat and compression ratio, which is one of the regions with the highest system efficiency. To avoid false triggering caused by signal noise, this embodiment uses a moving average method to smooth the exhaust temperature data and calculates its first derivative. A valid turning point is confirmed only when the first derivative changes from a negative value to a positive value and the positive value continues to exceed a preset time threshold (e.g., 5 seconds).

[0084] In one embodiment of the present invention, before starting the air conditioning system 100 to be charged in step S2, the method further includes: performing a vacuuming operation on the air conditioning system 100 to be charged and maintaining the vacuum level below a preset threshold for at least 15 minutes to check the airtightness of the system. This is the basis for ensuring the purity of the charging process and the reliable operation of the system in the future. The preset threshold is typically an absolute pressure of 50 Pa or lower.

[0085] In one embodiment of the present invention, the digital twin simulation model is embedded in the controller of the refrigerant charging machine, or embedded in a host computer that communicates with the air conditioning system 100 to be charged and the refrigerant charging machine. This flexible deployment method facilitates the integration of production lines and the upgrading of on-site maintenance equipment. For production line applications, the digital twin simulation unit 400 can be deployed on the factory's central control server to manage multiple charging stations on multiple production lines simultaneously; for on-site maintenance, it can be integrated into the embedded controller of a portable charging machine to achieve integrated operation.

[0086] Example 1 This embodiment provides a dynamic refrigerant charging process based on whole-system simulation operation. This process is implemented on a variable frequency air conditioning system that includes a variable frequency compressor and an electronic expansion valve. The entire charging process is uniformly scheduled by a host computer (such as an industrial computer) that integrates data acquisition, control algorithms, and a digital twin model.

[0087] Reference Figure 2 The refrigerant dynamic charging system of the present invention includes: The air conditioning system 100 to be charged consists of a compressor 110 (with an intake and exhaust port), a four-way valve 120, an outdoor heat exchanger 130, an electronic expansion valve 140, an indoor heat exchanger 150, and connecting pipes. Temperature sensors T (such as T1 exhaust temperature, T2 condenser outlet temperature, and T3 evaporator outlet temperature) and pressure sensors P (such as P1 intake pressure and P2 exhaust pressure) are installed at key locations in the system.

[0088] The refrigerant charging unit 200 includes a refrigerant source 210, a charging pump 220, a mass flow meter 230, and a charging valve 240. The charging line is connected to the process port (such as the compressor suction side) of the air conditioning system 100 to be charged via a quick-connect fitting.

[0089] The data acquisition and control unit 300 uses a high-precision data acquisition card and is electrically connected to all sensors, the electronic expansion valve 140, the inverter of the compressor 110, and the pump and valve of the refrigerant charging unit 200.

[0090] The digital twin simulation unit 400, built into the host computer, runs dynamic simulation software for the air conditioning system based on a physical model (such as the moving boundary method or the finite volume method). This unit exchanges data with the data acquisition and control unit 300 in real time via a high-speed data bus.

[0091] Based on the above system, referring to Figure 1 The specific process steps of this invention are as follows: Step S1: Build Phase The operator connects the air conditioning system 100 to be charged and the refrigerant charging unit 200 via piping. In the digital twin simulation unit 400 of the host computer, the corresponding basic simulation model is called according to the model of the air conditioning system 100 to be charged (such as compressor displacement and heat exchanger specifications). This model includes the mass, energy, and momentum conservation equations for each component, but the key parameters (such as the heat transfer coefficient of the heat exchanger and the volumetric efficiency coefficient of the compressor) are initially empirical values.

[0092] Step S2: Simulation Run Phase (Model Initialization and Calibration) After the air conditioning system is evacuated to 100% capacity and the pressure is maintained within acceptable limits, the system is started.

[0093] The host computer issues commands through the data acquisition and control unit 300 to make the air conditioning system operate in a preset "cooling simulation condition". In order to fully stimulate the dynamic characteristics of the system, the host computer controls the electronic expansion valve 140 to run stably for a period of time (e.g., 3 minutes) at three preset different opening degrees (e.g., opening degree A: 200 steps, opening degree B: 350 steps, opening degree C: 500 steps).

[0094] At each opening degree, the first set of operating parameters is collected and recorded in real time: compressor suction pressure P1, discharge pressure P2, compressor discharge temperature T1, condenser outlet temperature T2, evaporator outlet temperature T3, compressor operating frequency f, and compressor input power W.

[0095] The data acquisition and control unit 300 sends these initial operating parameter sets to the digital twin simulation unit 400 in real time. Based on this data under multiple operating conditions, the simulation unit 400 uses parameter identification algorithms such as Kalman filtering or recursive least squares method to correct key parameters such as the heat transfer coefficient and pressure drop coefficient of its internal model online until the deviation between the pressure and temperature values ​​output by the simulation model and the actual acquired values ​​is less than 1%. At this point, a "digital twin" that can accurately reflect the current physical characteristics of "this" specific air conditioning system is completed.

[0096] Step S3: Dynamic Filling and Optimization Stage The host computer controls the air conditioning system to remain stable under preset baseline conditions (e.g., the electronic expansion valve opens 400 steps and the compressor runs at 60Hz).

[0097] Subsequently, the host computer sends a command to the refrigerant charging unit 200 to open the charging valve 240 and the charging pump 220, and to begin charging the system with refrigerant in constant small steps (e.g., 5 grams / cycle) and at a rate (e.g., 2 grams / second). The mass flow meter 230 feeds back the charging volume data Δm to the host computer in real time.

[0098] During the filling process, the data acquisition and control unit 300 continuously acquires the second set of operating parameters (same as the first set of operating parameters) at a frequency of 1Hz. This real-time data is continuously fed into the calibrated digital twin simulation unit 400.

[0099] The optimization algorithm within the digital twin simulation unit 400 has begun operating: First, based on the real-time pressure and temperature, calculate the real-time subcooling SC = T_saturated condensation temperature (corresponding to P2) - T2, and the real-time superheat SH = T3 - T_saturated evaporation temperature (corresponding to P1).

[0100] Then, SC and SH are mapped to the cumulative charge amount Δm, and real-time curves of SC=f(Δm) and SH=g(Δm) are dynamically constructed in memory, and their first derivatives f'(Δm) and g'(Δm) are calculated.

[0101] Meanwhile, the simulation model predicts the real-time trend of the system's Energy Efficiency Ratio (EER) based on real-time updated boundary conditions (such as a gradually increasing refrigerant quantity). The core logic of the model is: when the charge quantity changes from insufficient to sufficient, the subcooling degree (SC) will increase accordingly, and the EER will also increase rapidly; when the charge quantity enters the excess region, the increase in power consumption caused by the increase in condensing pressure will exceed the small gain in cooling capacity, the growth of EER will slow down, and eventually an inflection point will appear and begin to decline. The goal of the simulation model is to predict the location where the EER inflection point will appear based on the current state of SC=f(Δm), and dynamically set the refrigerant quantity corresponding to that location as the "optimal target charge quantity".

[0102] Step S4: Convergence Determination Phase The host computer uses multiple criteria to determine the convergence of the filling process: The main criterion is that the deviation between the predicted EER value output by the digital twin simulation unit 400 and the actual EER value calculated based on real-time power and cooling capacity (estimated by the model) is consistently less than 0.05, and the slope of the predicted EER change is close to zero.

[0103] Auxiliary criterion 1 (supercooling slope criterion): Monitor the value of f'(Δm). When it changes from greater than 0.01 (positive value) to less than 0.01 and approaches 0 (for example, the slope of three consecutive acquisitions is less than 0.005), it indicates that the growth of SC has become extremely slow and has reached the inflection point region. At this time, the first charging termination signal is triggered.

[0104] Auxiliary Criterion 2 (Exhaust Temperature Criterion): Monitor the trend of exhaust temperature T1. During the charging process, T1 is observed to initially show a slow decreasing trend. When the value of T1 at the current moment is higher than the average of the previous two moments, it is determined that the exhaust temperature has passed its lowest point and has begun to rebound. At this time, the second charging termination signal is triggered.

[0105] When the main criterion and at least one auxiliary criterion are satisfied simultaneously, the host computer finally determines that the filling is complete.

[0106] Step S5: Termination Phase The host computer immediately sends a command to the refrigerant charging unit 200 to shut down the charging pump 240 and the charging valve 240, stopping the charging process. The system records the final charging amount m_final accumulated by the mass flow meter 230, binds this data to the air conditioning system's serial number, and uploads it to the production management database or saves it as a maintenance record.

[0107] Example 2 This embodiment is basically the same as Embodiment 1, except that the optimization of step S2, "simulation operation phase", is different.

[0108] Considering the limited adjustment capabilities of some air conditioning systems (such as fixed-frequency air conditioners), this embodiment modifies heat exchange conditions by controlling the speeds of the indoor and outdoor fans, thereby stimulating the system's dynamic response. Specifically: In cooling mode, the outdoor fan and indoor fan are first set to high speed, and data is collected after stable operation; then, the outdoor fan is kept at high speed while the indoor fan is switched to low speed, and data is collected again; next, the indoor fan is kept at low speed while the outdoor fan is switched to low speed, and data is collected again. These three sets of data under different airflow rates can provide sufficient stimulation for the digital twin model to identify the heat exchanger's heat exchange characteristics under different airflow velocities, thus achieving model calibration. This approach expands the scope of application of the invention, and is particularly suitable for air conditioning systems equipped only with fixed-speed compressors and simple throttling elements.

[0109] In another variation, for constant-speed systems that cannot be actively adjusted, the heat exchange conditions can be altered during the simulation phase by artificially blocking the air inlet of the outdoor heat exchanger, thereby obtaining data under different operating conditions for model calibration. The flexibility of this invention's process allows it to adapt to various types of air conditioning systems.

[0110] Example 3 This embodiment is basically the same as Embodiment 1, except that the method for predicting the optimal filling amount using a digital twin model in step S3, "Dynamic Filling and Optimization Stage," is described in more detail.

[0111] In this embodiment, the digital twin simulation model integrates a fast predictor based on an artificial neural network (ANN). This ANN model is pre-trained on a large amount of offline simulation data, which covers the optimal charging point characteristics under different combinations of compressor efficiency, heat exchanger UA value, pipe length, and ambient temperature. After model calibration in step S2, the calibrated key parameters (such as UA value and efficiency coefficient) are input into the ANN model to fine-tune the weights of some output layers, making them more suitable for the current system.

[0112] During dynamic filling, the ANN model takes real-time collected pressure, temperature, SC, SH, and the current cumulative filling volume Δm as input and directly outputs a predicted optimal filling volume M_target. Simultaneously, a simplified physics-based model runs in parallel to validate the ANN's output. When the difference between the two models' predicted values ​​M_target is less than a preset threshold (e.g., 5g), the system uses the ANN's prediction as the final target; if the difference is too large, a safety mechanism is triggered, filling is paused, and the multi-step prediction results from the physics-based model are used as the decision-making basis, while abnormal data is recorded for subsequent analysis. This hybrid modeling approach of "data-driven + physics model" balances the rapid response capability of the ANN with the reliability and interpretability of the physics model, further improving the efficiency and robustness of the optimization process.

[0113] In practical tests, this hybrid model demonstrated a synergistic effect that surpassed that of a single model. When faced with novel systems whose distribution differed slightly from the training data, the physical model provided the ANN with reliable "physical common sense" constraints, preventing it from making predictions that violated thermodynamic principles. Conversely, when dealing with complex nonlinear relationships, the ANN compensated for the insufficient accuracy of the simplified physical model. Working together, the two improved the prediction accuracy of the optimal charge amount by more than 30% compared to using either model alone, while also increasing the prediction speed by 5 times.

[0114] Example 4 This embodiment provides a computer-readable storage medium, such as a USB flash drive, portable hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, on which a computer program is stored. When this program is executed by a processor, it can realize the dynamic refrigerant charging process based on whole-machine simulation operation as described in Embodiment 1, 2, or 3. For example, by inserting the storage medium into the card reader of the controller of the refrigerant charging machine, the controller can automatically execute the above steps S1 to S5 to complete the intelligent charging of the air conditioning system connected to it.

[0115] The program stored on this storage medium, in addition to the core filling process control logic, also includes an online-upgradeable "digital twin model library." This model library is periodically updated from a cloud server, containing the topology and basic model parameters of the latest air conditioning system models. This allows filling equipment using the process of this invention to always remain in sync with product updates, without frequent hardware replacements, and possesses excellent scalability and future adaptability.

[0116] In summary, this invention, by creatively introducing a digital twin model and a dynamic optimization algorithm, completely transforms the traditional "preset target, open-loop input" mode of refrigerant charging, achieving intelligent charging with "real-time interaction and closed-loop optimization." This process significantly improves the accuracy and adaptability of charging, and is of great significance for enhancing the performance of air conditioning products, saving energy and reducing emissions, and promoting intelligent manufacturing.

[0117] The existing technology follows a "measure accurately before stopping" approach, meaning that charging stops once a preset, fixed target value is reached for a certain parameter (pressure, subcooling). The present invention, however, adopts a "charge while observing, dynamically optimizing" approach, using a digital twin model to "understand" the individual characteristics of the current system in real time and "predict" its future optimal state. Applying digital twin technology to the real-time charging process and performing online optimization based on real-time data-driven models is not something that can be achieved through logical analysis and reasoning based on existing technologies, as those skilled in the art would find difficult to grasp. No existing technical literature provides any inspiration for combining digital twins, dynamic optimization, and refrigerant charging processes. On the contrary, existing technologies generally pursue more precise static target values ​​or more complex lookup methods, which run counter to the technical approach of this invention.

[0118] As described above, this invention achieves personalized and precise filling, full lifecycle adaptability, and significant improvements in filling efficiency and accuracy, while also driving the intelligent transformation of the filling process. These effects are unattainable by existing technologies. For example, traditional methods based on subcooling calibrate the target value for a "standard machine" under standard operating conditions, failing to adapt to individual differences. In contrast, this invention, through real-time calibration using a digital twin model, generates a unique, dynamically changing "optimal subcooling curve" for each machine, rather than a static "optimal subcooling point." This fundamentally overcomes the limitations of existing technologies, bringing about a qualitative leap.

[0119] Another unexpected global effect of this invention's process is the establishment of a continuously optimizing data loop on the production line. As thousands of air conditioners undergo charging using this process, the system's backend accumulates massive amounts of correlation data between "individual system characteristics" and "optimal charging amounts." Through big data analysis, subtle deviations in the performance parameters of certain batches or components (such as compressors and heat exchangers) from certain suppliers can be identified, along with their impact on the optimal charging amount. These insights can be fed back in real-time to upstream component suppliers and product design departments, enabling quality improvement and design optimization from the source. This transforms the invention from a simple process improvement tool into a core engine driving the entire enterprise's quality and technological innovation system, generating value far exceeding the charging process itself.

[0120] The implementation principle of this invention is as follows: This invention discloses a dynamic refrigerant charging process based on whole-system simulation operation, belonging to the field of air conditioning and heat pump technology. It aims to solve the problems of existing refrigerant charging methods that rely on static target values ​​and cannot adapt to individual system differences and operating condition changes. This process constructs a digital twin simulation model of the air conditioning system 100 to be charged. Under preset simulated operating conditions, the system is started and first operating parameters are collected to initialize and calibrate the model. Then, refrigerant is dynamically charged at preset step sizes, and real-time second operating parameters are continuously collected and fed back to the model. Based on real-time data, the model online fits the curves of subcooling, superheating, and other parameters changing with the charging amount, predicting the refrigerant amount at which the system's energy efficiency ratio reaches its optimal value as the best charging target. When the real-time collected parameters converge with the extreme value region predicted by the model, the charging is considered complete. This invention achieves "tailor-made" dynamic and precise charging for each specific air conditioning system, significantly improving charging accuracy and system operating efficiency.

[0121] The embodiments described herein are preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape, and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A refrigerant dynamic charging process based on whole-machine simulated operation, applied to air conditioning or heat pump systems, characterized in that, Includes the following steps: Step S1, Construction Stage: Connect the air conditioning system (100) to be charged to the refrigerant charging machine, and establish a digital twin simulation model corresponding to the physical entity of the air conditioning system (100) to be charged; the digital twin simulation model is configured to receive the real-time operating parameters of the air conditioning system (100) to be charged and dynamically correct the model parameters; Step S2, Simulation Operation Phase: Start the air conditioning system (100) to be charged and run it in the preset simulation operating mode; The first set of operating parameters of the air conditioning system (100) to be charged is collected in real time under the simulated operating condition mode, and the first set of operating parameters is synchronously input into the digital twin simulation model to initialize and calibrate the model; Step S3, Dynamic charging and optimization stage: Control the refrigerant charging machine to dynamically charge the air conditioning system (100) to be charged with refrigerant at a preset step size and rate; during the charging process, continuously collect the second set of operating parameters of the air conditioning system (100) to be charged in real time and feed them back to the calibrated digital twin simulation model. The digital twin simulation model is based on the real-time feedback of the second set of operating parameters, calculates and predicts the trend of system performance indicators with the refrigerant charge online, and dynamically generates the optimal charge target amount; Step S4, Convergence Determination Stage: Compare the real-time acquired second set of operating parameters with the system performance index predicted by the digital twin simulation model. When the deviation between the two is within the preset convergence threshold range, and the system performance index reaches or approaches the predicted extreme value region, the filling is determined to be complete. Step S5, Termination Stage: Control the refrigerant charging machine to stop charging and record the final charging amount.

2. The refrigerant dynamic charging process based on whole-machine simulation operation according to claim 1, characterized in that, The preset simulation operating mode in step S2 includes: cooling mode or heating mode; and by adjusting the opening of the electronic expansion valve (140), the indoor fan speed and the outdoor fan speed in the air conditioning system (100) to be charged, at least two different system internal throttling or air volume combination states are constructed to stimulate the dynamic response characteristics of the system.

3. The refrigerant dynamic charging process based on whole-machine simulation operation according to claim 1, characterized in that, The second set of operating parameters mentioned in step S3 includes at least: compressor (110) suction pressure, compressor (110) discharge pressure, compressor (110) suction temperature, compressor (110) discharge temperature, condenser middle temperature, condenser outlet temperature, evaporator outlet temperature, electronic expansion valve (140) opening degree, compressor (110) operating frequency and compressor (110) input power.

4. The refrigerant dynamic charging process based on whole-machine simulation operation according to claim 3, characterized in that, The specific method for dynamically generating the optimal injection target amount using the digital twin simulation model in step S3 includes: Step S31: Based on the second set of operating parameters with real-time feedback, calculate the real-time subcooling (SC) and real-time superheat (SH); Step S32: Correlate the real-time subcooling (SC) and real-time superheat (SH) with the charge change (Δm) and generate the real-time subcooling change curve SC=f(Δm) and the real-time superheat change curve SH=g(Δm) online. Step S33: The digital twin simulation model predicts the inflection point of the system's energy efficiency ratio (EER) or coefficient of performance (COP) based on the preset objective function and the derivative changes of SC=f(Δm) and SH=g(Δm), and determines the predicted refrigerant quantity corresponding to the inflection point as the optimal charging target quantity.

5. The refrigerant dynamic charging process based on whole-machine simulation operation according to claim 4, characterized in that, The convergence determination in step S4 further includes: monitoring the slope d(SC) / d(Δm) of the real-time subcooling change curve SC=f(Δm), and triggering the first charging termination signal the instant the slope changes from a positive value to zero or from zero to a negative value.

6. The refrigerant dynamic charging process based on whole-machine simulation operation according to claim 4, characterized in that, The convergence determination in step S4 also includes: monitoring the trend of the compressor (110) exhaust temperature change, and triggering a second charging termination signal when the compressor (110) exhaust temperature shows a turning point of first decreasing and then increasing during the charging process.

7. A refrigerant dynamic charging process based on whole-machine simulated operation according to any one of claims 1-6, characterized in that, Before starting the air conditioning system (100) to be charged in step S2, the method further includes: performing a vacuuming operation on the air conditioning system (100) to be charged and maintaining the vacuum level below a preset threshold for at least 15 minutes to check the airtightness of the system.

8. A refrigerant dynamic charging process based on whole-machine simulated operation according to any one of claims 1-6, characterized in that, The digital twin simulation model is built into the controller of the refrigerant charging machine, or into a host computer that is communicatively connected to the air conditioning system (100) to be charged and the refrigerant charging machine.

9. A refrigerant dynamic charging system for implementing the refrigerant dynamic charging process based on whole-machine simulated operation as described in any one of claims 1-8, characterized in that, include: An air conditioning system (100) to be charged includes a compressor (110), a condenser, an evaporator, an electronic expansion valve (140), and corresponding temperature and pressure sensors; The refrigerant charging unit (200) is connected to the refrigerant circulation loop of the air conditioning system (100) to be charged via an openable and closable pipe, and is used to accurately control and measure the amount of refrigerant charged. The data acquisition and control unit (300) is electrically connected to each sensor of the air conditioning system (100) to be charged and the refrigerant charging unit (200) for real-time acquisition of operating parameters and issuance of control commands; The digital twin simulation unit (400) interacts with the data acquisition and control unit (300) and has a built-in dynamic simulation model of the air conditioning system (100) to be filled. It is used to receive real-time data, perform model calibration, predict the optimal filling amount, and feed back decision information to the data acquisition and control unit (300).

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of a dynamic refrigerant charging process based on whole-machine simulation operation as described in any one of claims 1 to 8.

Images