Air conditioning system energy efficiency prediction method and system

Abstract

The invention provides an air conditioning system energy efficiency prediction method and system. The method relates to the technical field of energy efficiency prediction of the air conditioning system and comprises the steps that a physical topology network of the air conditioning system is constructed, component relation modeling is carried out on the physical topology network to form a dynamic relation matrix, and the component relation modeling adopts a hybrid modeling mechanism. According to the energy efficiency prediction method and system for the air conditioning system, dynamic modeling is carried out on the relation of components such as a compressor, an evaporator, a condenser and a throttling device of the air conditioning system by introducing a hybrid modeling mechanism combining a graph neural network and a physical rule, and a dynamic relation matrix is formed. And in combination with self-adaptive data screening and dynamic coupling solution of the multi-dimensional operating parameters, accurate description of the operating state of the system and real-time prediction of energy efficiency change are realized, and the accuracy of energy efficiency prediction and the response speed of the system are improved.

Description

Technical Field

[0001] This invention relates to the field of air conditioning system energy efficiency prediction technology, specifically to an air conditioning system energy efficiency prediction method and system. Background Technology

[0002] Currently, air conditioning systems are widely used in residential, commercial, and industrial sectors. With increasing energy consumption, the energy efficiency of air conditioning systems has become a major concern. Traditional air conditioning systems typically rely on set temperature and humidity levels for regulation. Modern air conditioning systems employ multiple sensors to monitor environmental parameters such as temperature, humidity, and pressure in real time. By collecting data, the system can dynamically adjust its operating mode according to changes to achieve energy savings. Most systems rely on preset operating strategies for control, adjusting internal components such as compressors, fans, and heat exchangers based on environmental changes to optimize energy efficiency.

[0003] However, existing technologies still have significant shortcomings. Current energy efficiency prediction for air conditioning systems is mainly based on fixed models or empirical rules. It relies on simple environmental data such as temperature and humidity as input, lacking a comprehensive consideration of the interrelationships between various components of the air conditioning system. Existing technologies fail to effectively integrate the interrelationships of internal components and the diversity of external environmental conditions. The operating state of an air conditioning system can vary significantly under different environments, and existing static models cannot accurately reflect these changes. The accuracy of energy efficiency prediction is limited, and it cannot provide precise optimization solutions under complex and variable environmental conditions. Existing technologies cannot meet the real-time adjustment and precise control requirements of intelligent air conditioning systems, resulting in low energy efficiency performance in actual operation. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and system for predicting the energy efficiency of air conditioning systems. The technical problem this invention aims to solve is: how to address the issues of low energy efficiency prediction accuracy, static models, and lack of real-time adjustment in air conditioning systems under complex operating conditions by integrating graph neural networks and physical rules into an energy efficiency prediction method process.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for predicting the energy efficiency of an air conditioning system, comprising: S1. Construct the physical topology network of the air conditioning system, and perform component relationship modeling on the physical topology network to form a dynamic relationship matrix. The component relationship modeling adopts a hybrid modeling mechanism. S2. The multi-dimensional operating parameters of the air conditioning system are analyzed and processed to form a dynamic and effective parameter set. The analysis and processing adopts an adaptive data filtering mechanism. S3. Input the dynamic effective parameter set into the dynamic relation matrix for solution processing to form a dynamic coupling factor. The solution processing includes a steady-state relation analysis algorithm and a dynamic parameter identification mechanism. S4. The dynamic coupling factor and the dynamic effective parameter set are fused and inferred to form an energy efficiency prediction value. The fusion and inference process adopts a multi-source information fusion energy efficiency prediction model. S5. Perform multi-objective optimization decision processing on the energy efficiency prediction value to form a real-time control strategy. The multi-objective optimization decision processing includes a rolling optimization algorithm and a strategy fine-tuning mechanism.

[0006] Preferably, the physical topology network includes a compressor, an evaporator, a condenser, and a throttling device. The hybrid modeling mechanism includes a graph neural network and physical rules. The graph neural network includes node feature learning and edge message passing. The physical rules include the law of conservation of energy and the law of conservation of mass. The node feature learning dynamically models the physical topology network to form energy transfer paths. The edge message passing updates the node energy information of the energy transfer paths to form optimized energy exchange paths. The physical rules serve as constraints during the dynamic modeling and node energy information updating processes, adjusting edge weights and node energy states in real time. The synergistic effect of the graph neural network and physical rules forms the dynamic relationship matrix.

[0007] Preferably, the multidimensional operating parameters include internal operating state parameters and external environment parameters. The adaptive data filtering mechanism performs signal quality assessment on the multidimensional operating parameters based on real-time signal quality assessment, and forms a quality score based on the signal quality assessment. The signal quality assessment scores the multidimensional operating parameters through signal detection to form the quality score. The quality score is then subjected to data smoothing and repair to form the dynamic effective parameter set. The data smoothing and repair adopts a weighted average method.

[0008] Preferably, the model formula of the steady-state relation analytical algorithm is: .

[0009] in, The input heat, measured in J / s, represents the heat input in the air conditioning system. Fluid flow rate, unit: This represents the mass of fluid flowing through the system per unit time. Specific heat capacity, unit: This represents the heat absorbed by a unit mass of fluid when the temperature changes by 1 K. The fluid input temperature, expressed in Kelvin (K), represents the temperature at which the fluid enters the air conditioning system. The fluid output temperature, expressed in Kelvin (K), represents the temperature of the fluid when it leaves the air conditioning system. It is a system load adaptability function that adjusts the heat exchange efficiency based on the external ambient temperature, system internal pressure, and humidity, and is dimensionless.

[0010] Preferably, the steady-state relationship analysis algorithm adopts a steady-state model based on the physical rules. The steady-state model performs heat transfer calculations on the dynamic effective parameter set to obtain the input heat. The heat transfer calculation uses temperature difference, fluid flow rate, and specific heat capacity as key parameters.

[0011] Preferably, the dynamic parameter identification mechanism performs real-time error correction on the input heat to form a heat input estimate. The real-time error correction compares the input heat with the steady-state model by absolute error, and forms the heat input estimate through the absolute error comparison. The heat input estimate is then optimized by nonlinear parameters to form the dynamic coupling factor. The nonlinear parameter optimization performs nonlinear regression on the heat input estimate, and the key parameters are adjusted based on the nonlinear regression to form the dynamic coupling factor.

[0012] Preferably, the multi-source information fusion energy efficiency prediction model includes a dynamic adjustment mechanism and a model correction mechanism, wherein the dynamic adjustment mechanism includes time series analysis and dynamic weight adjustment.

[0013] Preferably, the fusion inference process includes the following steps: S41. Perform the time series analysis on the dynamic coupling factor and the dynamic effective parameter set, and form a predicted trend value based on the time series analysis; S42. The predicted trend value is dynamically weighted to form a weighted predicted trend value. The dynamic weighting adjustment is based on the dynamic coupling factor and the dynamic effective parameter set to adjust the weight coefficients of the predicted trend value, and the weighted predicted trend value is formed by the weight coefficient adjustment. S43. The weighted prediction trend value is corrected to form the energy efficiency prediction value. The correction process adopts the model correction mechanism. The model correction mechanism compares the weighted prediction trend value with the multidimensional operating parameters by relative error, and forms the energy efficiency prediction value by the relative error comparison.

[0014] Preferably, the rolling optimization algorithm performs long-term trend adjustment on the energy efficiency prediction value based on historical data, forms a long-term energy efficiency trend term through the long-term trend adjustment, extracts trend components from the historical data to form historical trend features, performs trend fitting on the energy efficiency prediction value based on the historical trend features to form a long-term energy efficiency trend term, and performs time series residual correction on the long-term energy efficiency trend term to form the optimized energy efficiency prediction value.

[0015] Preferably, the strategy fine-tuning mechanism includes real-time error adjustment and load adaptability optimization. The real-time error adjustment adjusts the optimized energy efficiency prediction value in real time to form a short-term optimized prediction value. The real-time error adjustment calculates the error between the optimized energy efficiency prediction value and the dynamic effective parameter set, and forms an error result through the error calculation. Based on the error result, the optimized energy efficiency prediction value is calibrated, and a short-term optimized prediction value is formed through the calibration. The efficiency is corrected based on the dynamic effective parameter set, and the real-time control strategy is formed through the efficiency correction. The efficiency correction adopts a load adaptability function.

[0016] An air conditioning system energy efficiency prediction system includes the following working modules: A topology construction module is provided, which models the component relationships of the physical topology network of the air conditioning system. The topology construction module forms a dynamic relationship matrix through the component relationship modeling, which includes graph neural networks and physical rules. The parameter analysis module analyzes and processes the multi-dimensional operating parameters of the air conditioning system. The parameter analysis module forms a dynamic and effective parameter set through the analysis and processing, and the analysis and processing adopts an adaptive data filtering mechanism. A dynamic relation solving module inputs the dynamic effective parameter set into the dynamic relation matrix for solving. The dynamic relation solving module forms a dynamic coupling factor through the solving process. The solving process includes a steady-state relation analysis algorithm and a dynamic parameter identification mechanism. The fusion inference module performs fusion inference processing on the dynamic coupling factor and the dynamic effective parameter set, and the fusion inference module generates an energy efficiency prediction value through the fusion inference processing. The fusion inference processing includes a dynamic adjustment mechanism and a model correction mechanism. An optimization decision module performs multi-objective optimization decision processing on the energy efficiency prediction value. The optimization decision module forms a real-time control strategy through the multi-objective optimization decision processing, which includes a rolling optimization algorithm and a strategy fine-tuning mechanism.

[0017] This invention provides a method and system for predicting the energy efficiency of an air conditioning system. It has the following beneficial effects: This air conditioning system energy efficiency prediction method and system introduces a hybrid modeling mechanism combining graph neural networks and physical rules to dynamically model the relationships between components such as the compressor, evaporator, condenser, and throttling device of the air conditioning system, forming a dynamic relationship matrix. By combining adaptive data filtering and dynamic coupling solution of multi-dimensional operating parameters, it achieves accurate description of the system's operating state and real-time prediction of energy efficiency changes, improving the accuracy of energy efficiency prediction and the system's response speed.

[0018] A collaborative mechanism combining a multi-source information fusion energy efficiency prediction model and a rolling optimization algorithm enables dynamic adjustment of predicted values ​​and fine-tuning of strategies. This collaborative mechanism comprehensively considers historical trends, environmental changes, and real-time load characteristics to continuously optimize the control strategy of the air conditioning system, improving energy efficiency and operational stability, and achieving a combined effect of energy saving, consumption reduction, and intelligent control. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the system structure of the present invention; Figure 2 This is an overall flowchart of the method of the present invention; Figure 3 This is a schematic diagram illustrating the component relationship modeling of the present invention; Figure 4 This is a flowchart of the parameter analysis and processing of the present invention; Figure 5 This is a flowchart illustrating the integration of reasoning and optimization decision-making in this invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1 like Figure 1-5As shown, this embodiment of the invention provides a method for predicting the energy efficiency of an air conditioning system, including: S1. Constructing a physical topology network of the air conditioning system, and performing component relationship modeling on the physical topology network to form a dynamic relationship matrix. The component relationship modeling adopts a hybrid modeling mechanism. The physical topology network includes a compressor, evaporator, condenser, and throttling device. The hybrid modeling mechanism includes a graph neural network and physical rules. The graph neural network includes node feature learning and edge message passing. The physical rules include the law of conservation of energy and the law of conservation of mass. Node feature learning performs dynamic modeling on the physical topology network to form energy transfer paths. Edge message passing updates the node energy information of the energy transfer paths to form optimized energy exchange paths. The physical rules serve as constraints during the dynamic modeling and node energy information updating process, adjusting the edge weights and node energy states in real time. The synergistic effect of the graph neural network and physical rules forms a dynamic relationship matrix.

[0022] S2. The multi-dimensional operating parameters of the air conditioning system are analyzed and processed to form a dynamic effective parameter set. The analysis and processing adopts an adaptive data filtering mechanism. The multi-dimensional operating parameters include internal operating status parameters and external environmental parameters. The adaptive data filtering mechanism performs signal quality assessment on the multi-dimensional operating parameters based on real-time signal quality assessment, and generates a quality score based on the signal quality assessment. The signal quality assessment scores the multi-dimensional operating parameters through signal detection to form a quality score. The quality score is then smoothed and repaired to form a dynamic effective parameter set. The data smoothing and repair adopts a weighted average method.

[0023] S3. The dynamic effective parameter set is input into the dynamic relation matrix for solution processing to form the dynamic coupling factor. The solution processing includes a steady-state relation analytical algorithm and a dynamic parameter identification mechanism. The model formula of the steady-state relation analytical algorithm is: .

[0024] in, The input heat, measured in J / s, represents the heat input in the air conditioning system. Fluid flow rate, unit: This represents the mass of fluid flowing through the system per unit time. Specific heat capacity, unit: This represents the heat absorbed by a unit mass of fluid when the temperature changes by 1 K. The fluid input temperature, expressed in Kelvin (K), represents the temperature at which the fluid enters the air conditioning system. The fluid output temperature, in Kelvin (K), represents the temperature of the fluid when it leaves the air conditioning system. It is a system load adaptability function that adjusts the heat exchange efficiency based on the external ambient temperature, system internal pressure, and humidity, and is dimensionless.

[0025] The steady-state relation analysis algorithm adopts a steady-state model based on physical rules. The steady-state model calculates the heat transfer of the dynamic effective parameter set to obtain the input heat. The heat transfer calculation uses temperature difference, fluid flow rate and specific heat capacity as key parameters.

[0026] The dynamic parameter identification mechanism performs real-time error correction on the input heat to form a heat input estimate. The real-time error correction compares the input heat with the steady-state model by absolute error, and forms a heat input estimate by absolute error comparison. The heat input estimate is then optimized by nonlinear parameters to form a dynamic coupling factor. The nonlinear parameter optimization performs nonlinear regression on the heat input estimate, and the key parameters are adjusted based on the nonlinear regression to form the dynamic coupling factor.

[0027] S4. The dynamic coupling factor and the dynamic effective parameter set are fused and inferred to form the energy efficiency prediction value. The fusion and inference process adopts a multi-source information fusion energy efficiency prediction model. The multi-source information fusion energy efficiency prediction model includes a dynamic adjustment mechanism and a model correction mechanism. The dynamic adjustment mechanism includes time series analysis and dynamic weight adjustment. The fusion and inference process includes the following steps: S41. Perform time series analysis on the dynamic coupling factor and the dynamic effective parameter set, and generate predicted trend values ​​based on the time series analysis.

[0028] S42. Dynamically adjust the predicted trend value to form a weighted predicted trend value. The dynamic weight adjustment is based on the dynamic coupling factor and the dynamic effective parameter set to adjust the weight coefficient of the predicted trend value, and the weighted predicted trend value is formed by adjusting the weight coefficient.

[0029] S43. The weighted forecast trend value is corrected to form the energy efficiency forecast value. The correction process adopts a model correction mechanism, which compares the relative error between the weighted forecast trend value and the multi-dimensional operating parameters to form the energy efficiency forecast value.

[0030] S5. A multi-objective optimization decision-making process is performed on the energy efficiency forecast values ​​to form a real-time control strategy. This process includes a rolling optimization algorithm and a strategy fine-tuning mechanism. The rolling optimization algorithm performs long-term trend adjustment on the energy efficiency forecast values ​​based on historical data, forming a long-term energy efficiency trend term. This adjustment extracts trend components from historical data to form historical trend features. Based on these features, the energy efficiency forecast values ​​are trend-fitted to form a long-term energy efficiency trend term. Time-series residual correction is then applied to the long-term energy efficiency trend term to form an optimized energy efficiency forecast value. The strategy fine-tuning mechanism includes real-time error adjustment and load adaptability optimization. Real-time error adjustment adjusts the optimized energy efficiency forecast values ​​to form short-term optimized forecast values. It also calculates the error between the optimized energy efficiency forecast values ​​and the dynamic effective parameter set, generating an error result. Based on this error result, the optimized energy efficiency forecast values ​​are calibrated to form short-term optimized forecast values. Efficiency correction is then applied to the short-term optimized forecast values ​​based on the dynamic effective parameter set, forming a real-time control strategy. The efficiency correction uses a load adaptability function.

[0031] An air conditioning system energy efficiency prediction system includes the following working modules: The topology building module models the component relationships in the physical topology network of the air conditioning system. The topology building module forms a dynamic relationship matrix through component relationship modeling, which includes graph neural networks and physical rules.

[0032] The parameter analysis module analyzes and processes the multi-dimensional operating parameters of the air conditioning system. The parameter analysis module forms a dynamic and effective parameter set through analysis and processing, and the analysis and processing adopts an adaptive data filtering mechanism.

[0033] The dynamic relation solving module inputs the dynamic effective parameter set into the dynamic relation matrix for solving. The dynamic relation solving module forms a dynamic coupling factor through the solving process, which includes a steady-state relation analysis algorithm and a dynamic parameter identification mechanism.

[0034] The fusion inference module performs fusion inference processing on the dynamic coupling factors and the dynamic effective parameter set. The fusion inference module generates energy efficiency prediction values ​​through fusion inference processing, which includes dynamic adjustment mechanisms and model correction mechanisms.

[0035] The optimization decision module performs multi-objective optimization decision processing on the energy efficiency prediction values. The optimization decision module forms a real-time control strategy through multi-objective optimization decision processing, which includes a rolling optimization algorithm and a strategy fine-tuning mechanism.

[0036] This invention improves the physical consistency and computational accuracy of energy efficiency prediction models by combining graph neural networks with physical rules. An adaptive data filtering mechanism enhances the stability and robustness of operating parameter processing. Real-time correction of system heat transfer characteristics is achieved through the coordinated application of dynamic parameter identification and steady-state analysis. Multi-source information fusion and time series analysis improve the timeliness and comprehensiveness of energy efficiency prediction.

[0037] By analyzing historical time-series data, patterns and trends in system operation are identified. Specifically, time-series analysis is used to study the temporal variation characteristics of dynamic coupling factors and dynamic effective parameter sets, extracting relevant trend information. The time-series analysis process includes data collection and preprocessing, trend and seasonality analysis, model building and training, trend prediction generation, and result analysis and application. Historical data is collected and cleaned, and smoothed to ensure data quality. Long-term trends and seasonal fluctuations in the data are identified through visualization analysis. Appropriate prediction models are selected and trained based on historical data to build models capable of predicting the future. The predicted trend values ​​generated by the model provide a basis for energy efficiency forecasting and data support for subsequent system adjustment and optimization decisions, promoting dynamic optimization of energy efficiency and continuous improvement of system performance.

[0038] By introducing rolling optimization and strategy fine-tuning mechanisms, a self-learning and adaptive closed-loop control system is constructed. The overall approach has good versatility and scalability, and can be applied to various types of air conditioning systems and energy efficiency management scenarios.

[0039] Example 2 This embodiment is based on an energy efficiency prediction method and system for air conditioning systems. By constructing a physical topology network of the air conditioning system and modeling the component relationships, it provides an accurate dynamic relationship matrix foundation for energy efficiency prediction. The specific implementation method is as follows: This embodiment focuses on a commercial split-type air conditioning system with a rated cooling capacity of 12kW. The system mainly consists of a variable frequency compressor, a finned condenser, a throttling device, and a plate evaporator, using R32 as the refrigerant. The piping is connected with copper pipes, with a total length of approximately 15.8m. Pressure, temperature, and flow sensors are installed at each node to collect real-time operating data.

[0040] The system's operating data under rated cooling conditions of 27℃ indoors and 35℃ outdoors are as follows: The compressor suction pressure is 0.81 MPa, and the discharge pressure is 2.62 MPa. The condenser inlet air temperature is 34.5℃, and the outlet air temperature is 39.1℃. The evaporator inlet air temperature is 26.7℃, and the outlet air temperature is 13.5℃. The refrigerant mass flow rate is 0.021 kg / s. The total system input power is 3.42 kW.

[0041] 1. Physical Topology Network Construction Based on the actual piping connections of the air conditioning system, the compressor, condenser, throttling device, and evaporator are denoted as nodes N1, N2, N3, and N4, respectively. The refrigerant flow direction is used as the edge direction. An initial physical topology network is constructed based on the existence of fluid connectivity, resulting in the following topology matrix:

[0042] in, This indicates that there is a refrigerant flow path from the i-th component to the j-th component. This indicates that there is no direct connection between the two. The topology matrix fully reflects the closed-loop flow structure of the compressor-condenser-throttling device-evaporator-compressor.

[0043] 2. Feature settings for nodes and edges Based on the above topology, a feature vector containing actual measurement data is introduced for each node. Taking the rated operating condition as an example, the following features are set for each node: Compressor: suction pressure 0.80MPa, discharge pressure 2.60MPa, input power 3.40kW, discharge temperature 82.5℃.

[0044] Condenser: Inlet refrigerant temperature 81.0℃, outlet refrigerant temperature 45.3℃, condenser inlet air temperature 34.2℃, outlet air temperature 38.9℃.

[0045] Throttling device: Inlet pressure 2.55MPa, outlet pressure 0.82MPa, refrigerant temperature before throttling 44.8℃.

[0046] Evaporator: Inlet refrigerant temperature 6.5℃, outlet refrigerant temperature 12.3℃, evaporator inlet air temperature 26.8℃, outlet air temperature 13.6℃.

[0047] Node features are input into the graph neural network in vector form during implementation. Each node feature has an 8-dimensional dimension, including physical quantities such as pressure, temperature, flow rate, and local power.

[0048] The edge characteristics are set based on mass flow rate and pressure drop: The mass flow rate of the N1-N2 side is 0.020 kg / s, and the pressure drop is 0.05 MPa. The mass flow rate of the N4-N1 side is maintained at 0.020 kg / s, and the pressure drop is approximately 0.03 MPa, ensuring that the entire loop meets the mass conservation constraint.

[0049] 3. Hybrid modeling mechanism and dynamic relation matrix formation After completing the physical topology network and setting the node and edge features, the above network is input into a hybrid modeling mechanism for component relationship modeling. The hybrid modeling mechanism consists of two parts: graph neural networks and physical rules. Graph Neural Network Part: The input layer uses 8-dimensional node features, and the hidden layer has 2 layers with 64 neurons in each layer. The activation function is a modified linear function, and the learning rate is set to 0.001.

[0050] The energy and mass relationships between adjacent nodes are updated through edge message passing, optimizing the energy exchange path. After each message pass, the energy state of a node is dynamically updated, reflecting a more reasonable heat exchange process and gradually approaching the actual energy transfer path.

[0051] Physical rules and constraints: In the process of modeling component relationships, the output of the graph neural network is modified using the laws of conservation of energy and mass as constraints. Law of conservation of mass: Ensure that the deviation between the inlet mass flow rate and the outlet mass flow rate at each node does not exceed 1%.

[0052] Law of Conservation of Energy: The heat exchange quantity of each heat exchange component is calculated using the following formula:

[0053] The heat transfer capacity of the heat exchange components is calculated and verified. On the evaporator side, when the specific heat capacity is taken... At that time, the relative deviation between the heat exchange calculated based on the air-side temperature difference and flow rate and the cooling capacity converted from the electrical input power is controlled within 2.5%.

[0054] By adjusting edge weights and node energy states in real time using physical rules, the accuracy and stability of the system during dynamic modeling are ensured. The error in node energy states gradually decreases after each training iteration.

[0055] After about 400 iterations, the node energy balance error gradually decreased from the initial 6.1% to 1.9%, and the mass balance error stabilized at around 0.8%. At this point, the component relationship modeling was considered to have converged, and the resulting dynamic relationship matrix reflected the coupling strength between the components under the operating conditions.

[0056] In the converged dynamic relation matrix, the edge weight coefficients of condenser node N2 and evaporator node N4 increased from the initial 0.72 to 0.87, and the edge weight coefficients of compressor node N1 and condenser node N2 increased from 0.78 to 0.90. Numerical results show that, after introducing physical conservation constraints, the model more accurately characterizes the actual energy transfer relationship of the refrigerant between the main heat exchange components, providing a reliable structural basis for subsequent energy efficiency predictions based on the dynamic relation matrix.

[0057] This embodiment constructs the physical topology network of an air conditioning system and models the relationships between its components. A hybrid modeling mechanism combining graph neural networks and physical rules is employed to capture the energy and mass transfer relationships between the major components. Model training results show that the node energy balance error is reduced to 1.9%, and the mass balance error is controlled within 0.8%. The obtained dynamic relationship matrix accurately reflects the coupling characteristics between the components, providing a reliable data foundation for subsequent energy efficiency prediction and optimization control.

[0058] Example 3 This embodiment is based on the method and system for predicting the energy efficiency of air conditioning systems. By substituting the measured operating condition data into the steady-state relation analytical algorithm model formula and comparing it with the measured heat input, the accuracy of the steady-state relation analytical algorithm model formula in calculating the heat input of the air conditioning system is verified.

[0059] 1. Experimental conditions and data sources The tested equipment is an air conditioning system prototype with a rated cooling capacity of 3.5kW and refrigerant R410A. The main components include a compressor, evaporator, condenser, and throttling device.

[0060] Environment and operating conditions: The laboratory environment temperature was 32℃, the relative humidity was about 55%, the operating mode was stable cooling, the continuous operation was 3600s, and the sampling cycle was 1s.

[0061] Sensors and data acquisition systems: Pressure sensors are used to collect compressor suction and discharge pressures; K-type thermocouples are used to collect evaporator and condenser inlet and outlet temperatures as well as ambient temperatures; mass flow meters are used to collect refrigerant mass flow rates; power meters are used to collect compressor power; and hygrometers and speed sensors are used to collect ambient humidity and fan speed.

[0062] All signals are processed by an adaptive data filtering mechanism to complete signal quality assessment and data smoothing repair, forming a dynamic and effective parameter set.

[0063] 2. Dynamic effective parameter set Statistical analysis of 3600 sets of sampled data yielded the following average values ​​for key parameters: Refrigerant mass flow rate: The refrigerant mass flow rate was obtained by taking the arithmetic mean of the data over a period of 3600 seconds, which was measured by a mass flow meter. The result was 0.018 kg / s.

[0064] Condenser inlet temperature: The average value of the temperature measured by the thermocouple at the condenser inlet is 41.3℃.

[0065] Condenser outlet temperature: The average value of the temperature measured by the thermocouple at the condenser outlet is 35.8℃.

[0066] Compressor power: The compressor power was measured by an electric power meter and the average value was taken to obtain a compressor power of 980W.

[0067] Ambient temperature and humidity: measured by environmental sensors and used to calculate the load adaptability function, the ambient temperature was found to be 32.5℃ and the humidity to be 55.6%.

[0068] Convert the temperature into the temperature difference form required for thermodynamic calculations: Since this embodiment treats the system input heat as an absolute temperature difference, the actual calculation will use:

[0069] To facilitate comparison with previous tests, in this embodiment, a typical operating temperature difference of 6.5K was taken as the calculation sample based on multiple tests. The calculation sample was derived from the weighted average results of multiple stable operating conditions and was marked accordingly in the test records.

[0070] 3. Steady-state relation analysis algorithm A steady-state model based on physical rules is adopted. This model calculates the input heat using a dynamic set of effective parameters, with temperature difference, fluid flow rate, and specific heat capacity as key parameters. The model formula for the steady-state relation analytical algorithm is as follows: .

[0071] in, The input heat, measured in J / s, represents the heat input in the air conditioning system. Fluid flow rate, unit: This represents the mass of fluid flowing through the system per unit time. Specific heat capacity, unit: This represents the heat absorbed by a unit mass of fluid when the temperature changes by 1 K. The fluid input temperature, expressed in Kelvin (K), represents the temperature at which the fluid enters the air conditioning system. The fluid output temperature, in Kelvin (K), represents the temperature of the fluid when it leaves the air conditioning system. It is a system load adaptability function that adjusts the heat exchange efficiency based on the external ambient temperature, system internal pressure, and humidity, and is dimensionless.

[0072] Specific heat capacity values ​​and unit conversions: According to the property data sheet of refrigerant R410A, its specific heat capacity is approximately 1.54 within the temperature and pressure range specified in this embodiment. Converted to J / (kg·K), it is 1.54×10 3 =1540J / (kg·K).

[0073] Values ​​of the load adaptability function: Based on the external ambient temperature of 32.5℃, ambient humidity of 55.6%, and the current system load rate of approximately 0.92, the following results were obtained using the pre-calibrated load adaptability curve: , is a dimensionless correction coefficient.

[0074] Substitute the known data into the calculation:

[0075] 4. Dynamic parameter identification mechanism The dynamic parameter identification mechanism is used to correct the input heat in real time. Its main function is to dynamically correct the estimated heat input based on the error between the actually measured dynamic parameters and the steady-state model predictions during the operation of the air conditioning system. The specific process is as follows: Real-time data acquisition and comparison: The system acquires real-time data through sensors and compares it with the predicted values ​​of the steady-state model.

[0076] Error identification and correction: The real-time error value is calculated, which is the difference between the actual input heat and the value calculated by the steady-state model. The error is fed back to the dynamic parameter identification mechanism to correct the system's heat input.

[0077] Dynamic optimization: The dynamic parameter identification mechanism optimizes the error value and adjusts the input heat estimate based on the error to ensure accurate control and high efficiency of the system under various operating conditions.

[0078] Continuous feedback: The system continuously adjusts errors and optimizes input heat estimates to achieve real-time dynamic optimization, ultimately improving the energy efficiency prediction and control accuracy of the air conditioning system.

[0079] The dynamic coupling factor is a key factor obtained through nonlinear regression analysis of dynamic parameters. It is used to regulate and optimize the relationships between multiple variables in an air conditioning system. Its specific functions are as follows: Coupling adjustment: The dynamic coupling factor can adjust the nonlinear relationship between key parameters such as heat input, fluid flow rate, and temperature difference to ensure the stable operation of the air conditioning system under various dynamic conditions.

[0080] Nonlinear regression: Through nonlinear regression analysis, the dynamic coupling factor optimizes the estimation of input heat and eliminates errors caused by dynamic changes in the system.

[0081] Dynamic response: The dynamic coupling factor adjusts the system's response according to changes in external factors such as ambient temperature and humidity, maintaining efficient energy efficiency prediction and precise control.

[0082] In this embodiment, during air conditioning system operation, the dynamic parameter identification mechanism corrects errors based on the difference between real-time sensor data and the steady-state model, and optimizes the input heat estimate through nonlinear regression, ultimately forming a dynamic coupling factor. This dynamic coupling factor adjusts key parameters such as fluid flow rate and temperature to ensure the system adapts to changes in the external environment and maintains optimal operating conditions.

[0083] 5. Comparison with measured values ​​and error calculation The compressor power was measured by an electric power meter, and after energy conversion and system loss calculation, an experimental reference value for the input heat under the same operating conditions was obtained. .

[0084] Error calculation: Since 1J / s = 1W, therefore .

[0085]

[0086]

[0087] In summary, the calculation results show that when using the steady-state relational analytical algorithm model formula to solve for the input heat, the input heat obtained based on parameters such as refrigerant mass flow rate, inlet and outlet temperatures, and load correction function is approximately 171.2W. The relative error between this and the measured heat input of 176.3W obtained by measuring and converting through a power meter is approximately 2.9%.

[0088] The above results show that the steady-state relational analytical algorithm model formula accurately represents the energy transfer relationship of the air conditioning system under steady-state conditions, with good calculation accuracy and engineering applicability. Combined with the dynamic parameter identification mechanism and dynamic optimization of the dynamic coupling factor, it improves the system's accuracy and response capability for input heat estimation, providing support for the system's real-time energy efficiency prediction and dynamic adjustment.

[0089] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for predicting the energy efficiency of an air conditioning system, characterized in that, include: S1. Construct the physical topology network of the air conditioning system, and perform component relationship modeling on the physical topology network to form a dynamic relationship matrix. The component relationship modeling adopts a hybrid modeling mechanism. S2. The multi-dimensional operating parameters of the air conditioning system are analyzed and processed to form a dynamic and effective parameter set. The analysis and processing adopts an adaptive data filtering mechanism. S3. Input the dynamic effective parameter set into the dynamic relation matrix for solution processing to form a dynamic coupling factor. The solution processing includes a steady-state relation analysis algorithm and a dynamic parameter identification mechanism. S4. The dynamic coupling factor and the dynamic effective parameter set are fused and inferred to form an energy efficiency prediction value. The fusion and inference process adopts a multi-source information fusion energy efficiency prediction model. S5. Perform multi-objective optimization decision processing on the energy efficiency prediction value to form a real-time control strategy. The multi-objective optimization decision processing includes a rolling optimization algorithm and a strategy fine-tuning mechanism.

2. The method for predicting the energy efficiency of an air conditioning system according to claim 1, characterized in that: The physical topology network includes a compressor, evaporator, condenser, and throttling device. The hybrid modeling mechanism includes a graph neural network and physical rules. The graph neural network includes node feature learning and edge message passing. The node feature learning dynamically models the physical topology network to form energy transfer paths. The edge message passing updates the node energy information of the energy transfer paths to form optimized energy exchange paths. The physical rules serve as constraints during the dynamic modeling and node energy information updating process, adjusting edge weights and node energy states in real time. The synergistic effect of the graph neural network and the physical rules forms the dynamic relationship matrix.

3. The method for predicting the energy efficiency of an air conditioning system according to claim 1, characterized in that: The multidimensional operating parameters include internal operating state parameters and external environment parameters. The adaptive data filtering mechanism performs signal quality assessment on the multidimensional operating parameters and forms a quality score based on the signal quality assessment. The signal quality assessment scores the multidimensional operating parameters by signal detection to form the quality score. The quality score is then subjected to data smoothing and repair to form the dynamic effective parameter set. The data smoothing and repair adopts a weighted average method.

4. The method for predicting the energy efficiency of an air conditioning system according to claim 2, characterized in that: The steady-state relation analysis algorithm adopts a steady-state model based on the physical rules. The steady-state model performs heat transfer calculations on the dynamic effective parameter set to obtain the input heat. The heat transfer calculation uses temperature difference, fluid flow rate, and specific heat capacity as key parameters.

5. The method for predicting the energy efficiency of an air conditioning system according to claim 4, characterized in that: The dynamic parameter identification mechanism performs real-time error correction on the input heat to form a heat input estimate. The real-time error correction compares the input heat with the steady-state model by absolute error, and forms the heat input estimate by absolute error comparison. The heat input estimate is then optimized by nonlinear parameters to form the dynamic coupling factor. The nonlinear parameter optimization performs nonlinear regression on the heat input estimate, and the key parameters are adjusted based on the nonlinear regression to form the dynamic coupling factor.

6. The method for predicting the energy efficiency of an air conditioning system according to claim 1, characterized in that: The multi-source information fusion energy efficiency prediction model includes a dynamic adjustment mechanism and a model correction mechanism. The dynamic adjustment mechanism includes time series analysis and dynamic weight adjustment.

7. The method for predicting the energy efficiency of an air conditioning system according to claim 6, characterized in that: The fusion reasoning process includes the following steps: S41. Perform the time series analysis on the dynamic coupling factor and the dynamic effective parameter set, and form a predicted trend value based on the time series analysis; S42. The predicted trend value is dynamically weighted to form a weighted predicted trend value. The dynamic weighting adjustment is based on the dynamic coupling factor and the dynamic effective parameter set to adjust the weight coefficients of the predicted trend value, and the weighted predicted trend value is formed by the weight coefficient adjustment. S43. The weighted prediction trend value is corrected to form the energy efficiency prediction value. The correction process adopts the model correction mechanism. The model correction mechanism compares the weighted prediction trend value with the multidimensional operating parameters by relative error, and forms the energy efficiency prediction value by the relative error comparison.

8. The method for predicting the energy efficiency of an air conditioning system according to claim 1, characterized in that: The rolling optimization algorithm performs long-term trend adjustment on the energy efficiency prediction value based on historical data, and forms a long-term energy efficiency trend term through the long-term trend adjustment. The long-term trend adjustment extracts trend components from the historical data to form historical trend features. Based on the historical trend features, the energy efficiency prediction value is trend-fitted to form a long-term energy efficiency trend term. The long-term energy efficiency trend term is then corrected for time series residuals to form the optimized energy efficiency prediction value.

9. The method for predicting the energy efficiency of an air conditioning system according to claim 8, characterized in that: The strategy fine-tuning mechanism performs real-time error adjustment on the optimized energy efficiency prediction value to form a short-term optimized prediction value. The real-time error adjustment calculates the error between the optimized energy efficiency prediction value and the dynamic effective parameter set, and generates an error result based on the error result. The optimized energy efficiency prediction value is calibrated based on the error result, and a short-term optimized prediction value is generated through the calibration. The efficiency is then corrected based on the dynamic effective parameter set, and the real-time control strategy is formed through the efficiency correction. The efficiency correction adopts a load adaptability function.

10. An air conditioning system energy efficiency prediction system, implemented according to any one of claims 1-9, characterized in that, Includes the following working modules: A topology construction module is provided, which models the component relationships of the physical topology network of the air conditioning system. The topology construction module forms a dynamic relationship matrix through the component relationship modeling, which includes graph neural networks and physical rules. The parameter analysis module analyzes and processes the multi-dimensional operating parameters of the air conditioning system. The parameter analysis module forms a dynamic and effective parameter set through the analysis and processing, and the analysis and processing adopts an adaptive data filtering mechanism. A dynamic relation solving module inputs the dynamic effective parameter set into the dynamic relation matrix for solving. The dynamic relation solving module forms a dynamic coupling factor through the solving process. The solving process includes a steady-state relation analysis algorithm and a dynamic parameter identification mechanism. The fusion inference module performs fusion inference processing on the dynamic coupling factor and the dynamic effective parameter set, and the fusion inference module generates an energy efficiency prediction value through the fusion inference processing. The fusion inference processing includes a dynamic adjustment mechanism and a model correction mechanism. An optimization decision module performs multi-objective optimization decision processing on the energy efficiency prediction value. The optimization decision module forms a real-time control strategy through the multi-objective optimization decision processing, which includes a rolling optimization algorithm and a strategy fine-tuning mechanism.

Images