Phase change material assisted soil air heat exchanger heat transfer prediction method and system

Abstract

The present application relates to the technical field of phase change heat storage material heat transfer prediction, and discloses a soil-air heat exchanger heat transfer prediction method and system based on phase change material assistance, in particular to the soil-air heat exchanger heat transfer prediction method and system based on phase change material assistance.The present application combines a physical model with a data-driven method, generates high-quality synthetic data by adopting the way of constructing a calculation domain, and then trains a gated recurrent neural network by combining historical data, so as to significantly reduce the calculation time and professional threshold required for soil-air heat exchanger performance prediction, thereby providing an efficient prediction mode, which can quickly and accurately predict the heat transfer performance of the soil-air heat exchanger, and thus provide a reliable basis for system optimization design and real-time operation control.

Description

Technical Field

[0001] This invention relates to the field of heat transfer prediction technology for phase change thermal storage materials, and in particular to a method and system for predicting heat transfer in a soil-air heat exchanger based on phase change materials. Background Technology

[0002] Earth-Air Heat Exchangers (EAHEs), as energy-saving devices utilizing shallow geothermal energy, heat the air inside buried air ducts in winter and cool it in summer without consuming energy, thus attracting increasing attention. In recent years, with the development and application of renewable energy, Phase Change Materials (PCMs), due to their high energy density and isothermal properties, have been explored for application in EAHE systems to improve their thermal stability. However, without sufficient design calculations, PCMs and EAHE systems can suffer from low heat transfer efficiency and high costs in actual operation. Currently, performance prediction is widely achieved through manual calculations, which are inefficient and prone to errors.

[0003] Therefore, there is an urgent need for a method that uses advanced algorithms to intelligently and automatically predict the thermal performance of soil-air heat exchangers. Summary of the Invention

[0004] To overcome the problems of low efficiency and lack of reliability in the calculation of thermal performance in existing soil-air heat exchangers using phase change materials, this invention provides a method and system for predicting heat transfer in soil-air heat exchangers based on phase change materials.

[0005] In a first aspect, the present invention provides a method for predicting heat transfer in a soil-air heat exchanger based on phase change materials, comprising:

[0006] A computational domain is constructed along the axial direction of the pipeline containing the phase change material and the soil layer in which the pipeline is buried, and governing equations are constructed based on the physical parameters of each computational domain; the computational domain includes an air region, a phase change material region, and a soil region;

[0007] The predicted outlet temperature corresponding to the historical inlet temperature of the pipeline is calculated using the control equation, and a composite dataset is constructed based on the historical outlet temperature, the predicted outlet temperature, and the design parameters of the pipeline.

[0008] A gated recurrent neural network is trained based on the composite dataset to obtain a trained prediction model.

[0009] The predicted model is used to predict the heat transfer of soil-air heat exchangers containing phase change materials in the target area.

[0010] According to a specific implementation, in the above prediction method, the computational domain is constructed using the finite difference method, and the air region, phase change material region, and soil region are divided by a grid of preset length and preset width.

[0011] According to a specific implementation, in the above prediction method, the governing equation specifically includes:

[0012] Based on the radius, temperature, wind speed, density, specific heat capacity, and thermal conductivity of the air region, and combined with the radius of the phase change material region, an energy conservation equation is constructed, resulting in the first governing equation.

[0013] Based on the phase change material radius, specific heat capacity, density and temperature of the phase change material region, an energy conservation equation is constructed to obtain the second governing equation.

[0014] Based on the soil radius, specific heat capacity, density, and temperature of the soil region, an energy conservation equation is constructed, resulting in the third governing equation.

[0015] According to a specific implementation method, the above prediction method calculates the predicted outlet temperature corresponding to the historical inlet temperature using the control equation, specifically including:

[0016] The temperature of each computational domain is initialized based on the preset temperature boundary;

[0017] The historical inlet temperature at the first moment is input into the initialized computational domain, and the temperature of each computational domain is calculated by simultaneously solving the first, second, and third control equations.

[0018] The temperature of the last grid layer in the air region is used as the predicted outlet temperature for the corresponding time moment, and the calculation is repeated cyclically based on the historical inlet temperature for the next time moment until all time moments have been calculated.

[0019] According to one specific implementation, the prediction method described above, in which training a gated recurrent neural network based on the composite dataset specifically includes:

[0020] According to the time series, historical outlet temperature data is selected based on the first preset time window, and historical inlet temperature data and design parameters are selected based on the second preset time window as input, and predicted outlet temperature value is selected as output based on the third preset time window to train the gated recurrent neural network.

[0021] The Adam algorithm is used as the optimization algorithm during the training process to adjust the hyperparameters of the gated recurrent neural network;

[0022] Wherein, the first preset time window and the second preset time window have the same length, and the beginning of the second preset time window is located at the end of the first preset time window; the beginning of the third preset time window is the same as the beginning of the second preset time window, and the length of the third preset time window is less than that of the second preset time window.

[0023] According to one specific implementation, in the above prediction method, the gated recurrent neural network includes an update gate and a reset gate, and the calculation formula is as follows:

[0024] ,

[0025] ,

[0026] ,

[0027] ,

[0028] in, To update the gate output, for function, Update the weights of the gates to the hidden state. This is the hidden state from the previous moment. To update the weights of the inputs in the gate, Enter the current time. To update the door offset, To reset the gate output, To reset the weights of the gate to a hidden state, To reset the weights of the inputs in the gate, To reset the door offset, The current state is hidden. It is the hyperbolic tangent function. The weights input to the hidden state at the current time. This represents the weight of the previous hidden state within the current hidden state. This is the bias for the hidden state. Output for the current moment.

[0029] According to one specific implementation, in the above prediction method, the design parameters include pipe radius, pipe length, phase change material radius, and inlet wind speed.

[0030] According to one specific implementation, the prediction method further includes:

[0031] Based on the heat transfer prediction of soil-air heat exchangers containing different phase change materials in the target area, the computational domain is reconstructed, the corresponding predicted outlet temperature is recalculated, and the composite dataset is reconstructed.

[0032] The prediction model is adjusted by using transfer learning combined with a reconstructed composite dataset to obtain a prediction model for the corresponding phase change material.

[0033] According to a specific implementation method, the heat transfer prediction using a prediction model in the above prediction method specifically includes:

[0034] The pipeline design parameters of the target area and the inlet air temperature of the soil-air heat exchanger are input into the prediction model to obtain the outlet air temperature.

[0035] The heat transfer efficiency of the corresponding phase change material is calculated based on the inlet air temperature and the outlet air temperature.

[0036] Secondly, the present invention provides a soil-air heat exchanger heat transfer prediction system based on phase change material assisted, the system including a memory and a processor;

[0037] The memory is used to store computer programs; the processor is used to call and execute the computer programs so that the system executes the soil-air heat exchanger heat transfer prediction method based on phase change materials as described above.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0039] This invention combines a physical model with a data-driven approach, generating high-quality synthetic data by constructing a computational domain, and then training a gated recurrent neural network using historical data. This significantly reduces the computation time and expertise required for predicting the performance of soil-air heat exchangers, providing an efficient prediction method that can quickly and accurately predict the heat transfer performance of soil-air heat exchangers, thus providing a reliable basis for system optimization design and real-time operation control. Attached Figure Description

[0040] Figure 1 A schematic flowchart of a soil-air heat exchanger heat transfer prediction method based on phase change material assisted by an embodiment of the present invention is provided.

[0041] Figure 2 This is a schematic diagram illustrating the application scenario of the soil-air heat exchanger provided in an embodiment of the present invention;

[0042] Figure 3 This is a schematic diagram of the structure of the computational domain provided in an embodiment of the present invention;

[0043] Figure 4 This is a schematic diagram illustrating the comparison results of the computational domains provided in an embodiment of the present invention;

[0044] Figure 5 This is a schematic diagram illustrating the construction of a composite dataset provided in an embodiment of the present invention;

[0045] Figure 6 This is a schematic diagram of the structure of a gated recurrent neural network provided in an embodiment of the present invention;

[0046] Figure 7 This is a schematic diagram of the results of the trained prediction model provided in an embodiment of the present invention. Detailed Implementation

[0047] The present invention will now be described in further detail with reference to specific embodiments. However, this should not be construed as limiting the scope of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0048] Unless otherwise specified, the terms "upper," "lower," "left," "right," "center," "inner," and "outer," etc., used in the description of specific embodiments of the present invention to indicate orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings, or the orientation or positional relationship in which the product / equipment / device is usually placed during use. These terms are merely for the purpose of facilitating the description of the present invention or simplifying the description in specific embodiments, and for enabling those skilled in the art to quickly understand the solution, and do not indicate or imply that a particular device / component / element must have a specific orientation, or be constructed and operated in a specific positional relationship. Therefore, they should not be construed as limitations on the present invention.

[0049] Furthermore, the use of terms such as "horizontal," "vertical," "suspended," "parallel," and "coaxial" does not imply that the corresponding device / component / element must be absolutely horizontal, vertical, suspended, parallel, or coaxial. Slight tilt or deviation is permissible, as long as it does not affect the normal function of the relevant component. For example, "horizontal" simply means that its direction is more horizontal relative to "vertical," not that the structure must be perfectly horizontal; a slight tilt is acceptable. "Coaxial" means that two components are arranged as coaxially as possible, allowing them to move coaxially or approximately coaxially when their relative positions change. Alternatively, it can be simplified to mean that the corresponding device / component / element, when arranged in "horizontal," "vertical," "suspended," "parallel," or "coaxial" directions, can have an error / deviation of ±10% relative to the corresponding direction, more preferably within ±8%, more preferably within ±6%, more preferably within ±5%, and more preferably within ±4%. For example, the deviation in the "coaxial" direction is controlled within 0.2-1mm, preferably within 0.2-0.5mm. As long as the corresponding device / component / element is within the error / deviation range, it can still achieve its function in the solution of the present invention.

[0050] Furthermore, the use of terms such as "first," "second," and "third" in terminology is merely for distinguishing descriptions of identical or similar components and should not be interpreted as emphasizing or implying the relative importance of a particular component.

[0051] Furthermore, in the description of the technical solution of this invention, unless otherwise explicitly specified / limited / restricted, the terms "set up," "install," "connect," "link," "provided with," "laid out," and "arranged" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to connection methods commonly used in the art, such as welding, riveting, bolting, and threaded connections. Such connections can be mechanical, electrical, or communication connections; they can be direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components.

[0052] Existing soil-air heat exchanger (EAHE) designs rely on numerical simulation methods, such as the finite element method (FEM) or the finite volume method (FLC). These methods require detailed physical models and solving complex sets of partial differential equations. This process is computationally expensive and demands specialized knowledge of heat transfer and fluid dynamics from the user. When phase change materials (PCMs) are added to the system, the complexity and computational burden of traditional numerical methods increase further due to the nonlinear latent heat absorption and release involved in the phase change process. Furthermore, traditional design methods struggle to quickly assess system performance under different operating conditions or material configurations, exhibiting poor flexibility and adaptability. In this context, some design approaches are shifting towards utilizing advanced intelligent algorithms to optimize and predict the thermal performance of EAHE systems. Among these, recurrent neural networks (RNNs), as machine learning tools capable of processing sequential data, have shown great potential in predicting the dynamic behavior of complex systems.

[0053] Although extensive research has been conducted on heat transfer models, performance optimization, and performance prediction of air-to-soil (EAHE) heat exchangers, performance prediction for PAHE remains insufficient, lacking research on using neural network models for PAHE system performance prediction. Furthermore, current research on prediction using neural networks primarily focuses on short-term forecasts, with limited research on long-term predictions and issues with computational efficiency. Recurrent Neural Networks (RNNs), on the other hand, are widely used in time series data prediction, demonstrating excellent performance, particularly in long-term time series data prediction.

[0054] Based on this, embodiments of the present invention provide a data-driven method for predicting the performance of soil-air heat exchangers containing phase change materials. This method first establishes a physical model through numerical simulation, generating simulated data covering multiple operating conditions, and then integrates historical measured data to construct a training dataset. Subsequently, a gated recurrent neural network is used to learn from the data, ultimately obtaining a model capable of accurately and quickly predicting the heat transfer performance of the soil-air heat exchanger. Embodiments of the present invention primarily address the problems of high computational cost and difficulty in quickly evaluating different materials and operating conditions using traditional pure numerical simulation methods, thereby significantly improving the efficiency and flexibility of performance prediction and providing a practical tool for the optimized design and operational control of heat exchangers.

[0055] The embodiments of the present invention will be further described and illustrated in detail below with reference to specific implementation methods.

[0056] Please refer to Figure 1 The diagram illustrates a flow chart of a soil-air heat exchanger heat transfer prediction method based on phase change material assisted by an embodiment of the present invention. The prediction method includes:

[0057] Step 1: Construct a computational domain along the axial direction of the pipeline containing the phase change material and the soil layer in which the pipeline is buried, and construct the governing equations based on the physical parameters of each computational domain.

[0058] It is understood that the aforementioned pipes containing phase change materials are mainly horizontally buried underground pipes, with vertical pipes usually installed at their inlet and outlet positions for inputting air and outputting air to the connected air conditioning unit, respectively. These two parts are not considered in the embodiments of the present invention.

[0059] Furthermore, the phase change material is mainly cylindrical and located in the central part of the pipe.

[0060] Please refer to the details. Figure 2 The diagram illustrates a scenario of using the soil-air heat exchanger provided in an embodiment of the present invention.

[0061] It should be noted that in the process of constructing the computational domain, the embodiments of the present invention adopt the following well-verified basic assumptions: (1) Air is regarded as an incompressible fluid, the influence of changes in moisture content in the pipe on heat transfer is ignored, and only sensible heat exchange in the pipe is considered; (2) The air in the ventilation duct is regarded as uniformly mixed and there is no thermal stratification phenomenon; (3) It is assumed that the pipe is located in a constant temperature layer and the change of soil temperature along the vertical direction is ignored; (4) Soil and phase change material are simplified as isotropic media and the interfacial contact thermal resistance is ignored; (5) The influence of metal pipe wall thickness and air gravity field is not considered.

[0062] In one possible implementation, the computational domain is constructed using the finite difference method, and the air region, phase change material region, and soil region are divided by a grid of preset length and width. Specifically, the air region uses an axially uniform grid with a grid size of ( , The phase change material region and the soil region were divided into multiple layers of mesh in the radial direction. , ), and aligned with the axial grid. The phase change material radius is... The pipe radius is The radius of the air region is .

[0063] Please refer to Figure 3 This illustrates a schematic diagram of the computational domain structure provided in an embodiment of the present invention.

[0064] For the phase change process of PCM, this embodiment of the invention uses the equivalent heat capacity method to dynamically process the specific heat capacity and thermal conductivity, dividing them into three stages: solid state, phase change transition state, and liquid state. The specific formulas of the equivalent heat capacity method are shown in formulas (1) and (2).

[0065] (1)

[0066] (2)

[0067] in, This is the equivalent specific heat capacity of the phase change material. ; The specific heat capacity of a solid phase change material. ; The specific heat capacity of the liquid phase change material. ; The latent heat of phase change in phase change materials. It is a function of liquid phase fraction. The current temperature. The solidus temperature of the phase change material. The liquidus temperature of the phase change material;

[0068] The equivalent thermal conductivity of the phase change material is... ; The thermal conductivity of solid phase change materials, ; The thermal conductivity of liquid phase change materials, .

[0069] Furthermore, the governing equations specifically include:

[0070] Based on the radius, temperature, wind speed, density, specific heat capacity, and thermal conductivity of the air region, and combined with the radius of the phase change material region, an energy conservation equation is constructed, yielding the first governing equation, which is:

[0071] (3)

[0072] in,

[0073] Pipe radius, unit ;

[0074] The radius of the phase change material is in meters (m).

[0075] air density, unit ;

[0076] The specific heat capacity of air, in units ;

[0077] The thermal conductivity of air, in units of .

[0078] Air temperature, unit ;

[0079] The unit of time is seconds (s).

[0080] For inlet wind speed, unit ;

[0081] This is the axial coordinate (along the flow direction), in meters;

[0082] It is the source of heat and mass transfer through air convection.

[0083] Based on the phase change material radius, specific heat capacity, density, and temperature of the phase change material region, an energy conservation equation is constructed, resulting in the second governing equation, which is:

[0084] (4)

[0085] in, The specific heat capacity of a phase change material, in units of ; The density of the phase change material is expressed in units of... ; The temperature of the phase change material, in units ; The unit of time is seconds (s). Radial coordinates, in meters (m). The thermal conductivity of the phase change material; Let be the radius of the phase change material control volume. For axial coordinates, in meters (m).

[0086] Based on the soil radius, specific heat capacity, density, and temperature of the soil region, an energy conservation equation is constructed, resulting in the third governing equation, which is:

[0087] (5)

[0088] in, Specific heat capacity of soil, unit ; Soil density, unit ; Soil temperature, unit ; The radius of the soil control volume. The thermal conductivity of the soil. The unit of time is seconds (s). Radial coordinates, in meters (m). For axial coordinates, in meters (m).

[0089] Step 2: Calculate the predicted outlet temperature corresponding to the historical inlet temperature of the pipeline using the control equation, and construct a composite dataset based on the historical outlet temperature, the predicted outlet temperature, and the design parameters of the pipeline.

[0090] It should be noted that, in this embodiment of the invention, the unsteady-state term of the first governing equation for the air region adopts a first-order backward difference scheme, while the convection and diffusion terms adopt a second-order upwind scheme. At the inlet, since the convection and diffusion terms cannot be expressed using a second-order backward difference scheme, a first-order backward difference scheme is used at that location. The unsteady-state term of the second governing equation for the phase change material region adopts a first-order backward difference scheme, while the convection and diffusion terms adopt a second-order upwind scheme. The unsteady-state term of the third governing equation for the soil region adopts a first-order backward difference scheme, while the convection and diffusion terms adopt a second-order upwind scheme.

[0091] Furthermore, the predicted outlet temperature corresponding to the historical inlet temperature is calculated using the control equation, specifically including:

[0092] The temperature of each computational domain is initialized based on the preset temperature boundary;

[0093] The historical inlet temperature at the first moment is input into the initialized computational domain, and the temperature of each computational domain is calculated by simultaneously solving the first, second, and third control equations.

[0094] The temperature of the last grid layer in the air region is used as the predicted outlet temperature for the corresponding time moment, and the calculation is repeated cyclically based on the historical inlet temperature for the next time moment until all time moments have been calculated.

[0095] Specifically, both axial and radial boundaries of the computational domain can be set as adiabatic boundaries. A constant temperature boundary is set at the outer boundary, where the distance from the buried pipe wall is determined by the soil thermal diffusivity and thermal disturbance period. This boundary temperature is defined as the local annual average soil temperature, used as the soil temperature. Based on the governing equations of each part, the problem can be solved by creating Matlab code and using the TDMA algorithm. The specific calculation process is as follows: First, the temperature of all grids is initialized according to the initialization parameters. Then, the inlet temperature of the first time step is substituted into the governing equations of the discretized air region, phase change material region, and soil region for solving. The calculation is performed on all grids in the computational domain from top to bottom to obtain the temperature of all grids in the air region, phase change material region, and soil region at the current time. The temperature of the last layer of grids in the air region is output as the outlet temperature at the current time. Then, the calculation is repeated based on the inlet temperature of the next time step until all time steps have been calculated.

[0096] The experimental data from "Zhou Tiecheng. Thermal Performance Study of Soil-Air Heat Exchanger Assisted by Phase Change Thermal Storage" on soil-air heat exchangers assisted by phase change materials were used to verify the accuracy of the computational domain in this embodiment of the invention. The calculation results of the computational domain were compared with the experimental data, and the results are as follows: Figure 4 As shown, the calculation results of the computational domain fit the experimental data well. The average absolute error of the outlet temperature is 0.46°C, the maximum absolute error is 1.1°C, and the maximum absolute relative error is 1.79%. The above errors meet the accuracy requirements, indicating that the computational domain provided by the embodiments of the present invention can reflect the actual experimental situation well and can be used as the basis for the embodiments of the present invention.

[0097] Furthermore, in one possible implementation, based on the aforementioned verified computational domain, a large number of numerical simulations can be performed by changing input parameters such as the temperature and flow rate of the inlet air, the initial soil temperature, and the properties of the phase change material, generating synthetic data covering various possible operating conditions. Simultaneously, historical operational data from actual experiments or demonstration projects, particularly inlet and outlet air temperature records, are collected. The simulated data and measured data are combined to form a composite dataset containing input conditions (inlet temperature, flow rate, material parameters, etc.) and output targets (outlet temperature, etc.). No limitations are imposed in this embodiment of the invention.

[0098] Step 3: Train a gated recurrent neural network based on the composite dataset to obtain a trained prediction model.

[0099] Specifically, training a gated recurrent neural network based on the composite dataset includes:

[0100] According to the time series, historical outlet temperature data is selected based on the first preset time window, and historical inlet temperature data and design parameters are selected based on the second preset time window as input, and predicted outlet temperature value is selected as output based on the third preset time window to train the gated recurrent neural network.

[0101] The Adam algorithm is used as the optimization algorithm during the training process to adjust the hyperparameters of the gated recurrent neural network;

[0102] Wherein, the first preset time window and the second preset time window have the same length, and the beginning of the second preset time window is located at the end of the first preset time window; the beginning of the third preset time window is the same as the beginning of the second preset time window, and the length of the third preset time window is less than that of the second preset time window.

[0103] It is understandable that the gated recurrent neural network was chosen in this embodiment of the invention because it excels at processing time-series data and can capture the dynamic dependence of temperature changes in this embodiment. Since soil air exchangers typically contain thermal disturbances, the input-output configuration needs to be designed so that the gated recurrent neural network, after training, can learn the thermal inertia and delayed response characterizing the system's heat capacity and thermal resistance. Therefore, based on the different preset time windows mentioned above, the physical interpretability of the training process can be greatly improved, and the gated recurrent neural network can be guided to learn the thermodynamic essence of the system more efficiently.

[0104] For example, taking the time series by hour, let's illustrate with time t. We can use the time series from t-360 to t as the first preset time window, then select the corresponding historical outlet temperature data. Further, we can use the time series from t to t+360 as the second preset time window, then select the corresponding historical inlet temperature data and the pipeline design parameters (usually, the pipeline design parameters do not change; they are only used as training parameters for the gated recurrent neural network). Correspondingly, we can use the time series from t to t+180 as the third preset time window, and select the corresponding predicted outlet temperature value. Based on the composite dataset constructed above, such as... Figure 5 As shown.

[0105] In one possible implementation, embodiments of the present invention provide a reference to a composite dataset, taking into account the actual pipe diameter and phase change material radius in the computational domain. Set to 0.05-0.2m, air zone radius. Set to 0.05-0.3m, pipe length Set to 10-45m, inlet wind speed The velocity was set to 0.5-2.5 m / s, and the inlet temperature was taken as the hourly temperature of a typical meteorological year in Chongqing from May to September. The specific dataset construction is shown in Tables 1 and 2. To ensure the model's generality, the composite dataset was randomly divided into two parts, with 80% used for model training and 20% used for model performance validation.

[0106] Table 1. Illustration of the composite dataset used for training

[0107]

[0108] Table 2. Schematic diagram of the composite dataset used for testing.

[0109]

[0110] Furthermore, the gated recurrent neural network used in this embodiment of the invention is a simplified version of the Long Short-Term Memory (LSTM) neural network model, but retains the long-term memory capability of the LSTM model. The main change is that the input gate, forget gate, and output gate in the LSTM neural network cells are replaced with update gates and reset gates. The gated recurrent neural network controls the information flow by introducing a gating mechanism, thereby better capturing long-term dependencies. Its structure is as follows... Figure 6 As shown. The formulas for updating and resetting doors are:

[0111] ,

[0112] ,

[0113] ,

[0114] ,

[0115] in, To update the gate output, for function, Update the weights of the gates to the hidden state. This is the hidden state from the previous moment. To update the weights of the inputs in the gate, Enter the current time. To update the door offset, To reset the gate output, To reset the weights of the gate to a hidden state, To reset the weights of the inputs in the gate, To reset the door offset, The current state is hidden. It is the hyperbolic tangent function. The weights input to the hidden state at the current time. This represents the weight of the previous hidden state within the current hidden state. This is the bias for the hidden state. Output for the current moment.

[0116] Furthermore, a composite dataset was used to train and tune the gated recurrent neural network. The specific hyperparameter search range is shown in Table 3. The Adam algorithm was used as the optimization algorithm during neural network training. The feature data was normalized before training and denormalized after model output to obtain the final result.

[0117] Table 3. Schematic diagram of hyperparameter settings

[0118] Learning rate Number of iterations Batch quantity number of floors Number of neurons 0.0005 100 4 4 16 0.001 500 8 5 32 0.005 1000 16 6 64 0.01 1500 32 128

[0119] To better evaluate the model's predictive accuracy, the root mean square error (RMSE), mean absolute percentage error (MAPE), and coefficient of determination (R²) are used. 2 The model is evaluated using four performance metrics: mean absolute error (MAE), mean absolute error (MAE), and mean absolute error (MAE). The specific calculation formulas include:

[0120]

[0121]

[0122]

[0123]

[0124] in, It is the actual value. It is a predicted value. It is the average of the true values. It represents the number of samples.

[0125] After hyperparameter search, the results of the prediction model are as follows: Figure 7 As shown in the figure. The results indicate that the trained prediction model has good predictive performance, with an RMSE of 0.1561. The value is 0.9969, MAPE is 0.0039, and MAE is 0.1030.

[0126] In one possible implementation, to improve the generality of the prediction model, this embodiment of the invention also provides a transfer learning method to transfer a prediction model for one phase change temperature system to soil-air heat exchanger systems with other phase change temperatures. It is understood that transfer learning is a machine learning method that aims to improve the learning performance of a target task by utilizing knowledge learned from a source task, especially when the target task has limited data or high annotation costs. It effectively improves the model's generalization ability and learning efficiency by optimizing feature representations, sharing model parameters, or learning relationships between samples, and has achieved significant results in fields such as image recognition and natural language processing. Transfer learning is mainly defined as follows: given a source domain… and learning tasks Target domain and learning tasks Transfer learning aims to utilize and Transferring latent knowledge to improve new learning tasks Prediction function The performance, of which and .generally, The size is much larger than Size.

[0127] The transfer learning strategy used in this invention involves fine-tuning the overall parameters with a small learning rate. Using a lower learning rate than in the pre-training step is crucial. This is done to avoid model overfitting and divergence caused by unnecessary over-updating of model parameters. The learning rate used in this invention is 10. -5 Other hyperparameters should be set according to the hyperparameters of the original best prediction model.

[0128] Specifically, the above prediction methods also include:

[0129] Based on the heat transfer prediction of soil-air heat exchangers containing different phase change materials in the target area, the computational domain is reconstructed, the corresponding predicted outlet temperature is recalculated, and the composite dataset is reconstructed.

[0130] The prediction model is adjusted by using transfer learning combined with a reconstructed composite dataset to obtain a prediction model for the corresponding phase change material.

[0131] The above-mentioned PCM-soil-air heat exchanger, which can change the parameters of the phase change material, can also use the optimal neural network for prediction without having to train from scratch. The model of the already trained optimal neural network is used to fine-tune the overall parameters with a small learning rate.

[0132] For example, this embodiment of the invention uses paraffin as an example for illustration. Paraffin is used as a new phase change material for PCM-soil-air heat exchangers, and a dataset with fewer parameter combinations is used to train the trained prediction model. The total number of samples in the target domain is less than that in the source domain, with 1024 samples. The parameter combinations for constructing the dataset are shown in Table 4. The parameters of each part of the source domain and the target domain are shown in Table 5.

[0133] Table 4. Schematic diagram of experimental setup for the target domain dataset.

[0134]

[0135] Table 5. Parameter Comparison of Source and Target Domains

[0136]

[0137] Understandably, for amorphous phase change materials such as paraffin, the enthalpy method is needed to better reflect the phase change process. The core of the enthalpy method is to convert the temperature of the phase change material into its enthalpy at the beginning of each iteration, and then convert the calculated enthalpy back into temperature after each time step. The formula for calculating the enthalpy of a phase change material is:

[0138] ,

[0139] In the formula, The enthalpy value of non-amorphous phase change materials. The reference enthalpy value for non-amorphous phase change materials is expressed in J / kg. Reference temperature, K; Specific heat at constant pressure, J / (kg) K); The latent heat of phase change for amorphous phase change materials is expressed in J / kg. This is the current temperature.

[0140] Meanwhile, assuming that the heat release during the phase change process is linear, and that the temperature of the phase change material and the liquid fraction of the phase change material exhibit a linear relationship, the liquid fraction can be defined according to the following formula:

[0141] ,

[0142] In the formula,

[0143] T The current temperature;

[0144] The solidus temperature of the phase change material, in K;

[0145] The liquidus temperature of the phase change material is expressed in Kelvin (K).

[0146] Therefore, the latent heat of phase change in amorphous phase change materials can be expressed by the following formula:

[0147] ,

[0148] In the formula, is the latent heat capacity of the phase change material, in J / kg.

[0149] The training process, hyperparameter tuning, and model evaluation are the same as described above. With a sample size of 256, the RMSE of the prediction results using a transfer learning strategy that fine-tunes the overall parameters with a small learning rate is 0.1725. The RMSE was 0.9928, MAPE was 0.0053, and MAE was 0.1369. Compared to the original model without transfer learning and the model trained on target domain data, the RMSE decreased by 82.71% and increased by 5.44%, respectively. It increased by 30.94% and decreased by 0.07%, MAPE decreased by 83.17% and increased by 6.00%, and MAE decreased by 83.18% and increased by 7.37%.

[0150] Step 4: Use the prediction model to predict the heat transfer of the soil-air heat exchanger containing phase change material in the target area.

[0151] Furthermore, using predictive models for heat transfer prediction specifically includes:

[0152] The pipeline design parameters of the target area and the inlet air temperature of the soil-air heat exchanger are input into the prediction model to obtain the outlet air temperature.

[0153] The heat transfer efficiency of the corresponding phase change material is calculated based on the inlet air temperature and the outlet air temperature.

[0154] It is understood that the embodiments of the present invention can quickly and accurately predict the outlet air temperature of a soil-air heat exchanger under different inlet air conditions and different phase change material parameters. Furthermore, the method of calculating the heat transfer efficiency of a soil-air heat exchanger using inlet and outlet air temperatures provided in the embodiments of the present invention is merely a reference implementation method for heat transfer prediction.

[0155] In one or more embodiments, the heat exchange rate of the soil-air heat exchanger can be calculated by measuring the initial, undisturbed temperature of the soil at the far end, combined with the inlet and outlet air temperatures. This quantifies the actual heat exchange and the maximum possible heat exchange to predict the heat transfer of the soil-air heat exchanger in the target area. Furthermore, other heat transfer performance values ​​of the soil-air heat exchanger, such as the temperature fluctuation decay rate, can also be calculated, which are not limited in the embodiments of this invention.

[0156] Furthermore, based on the heat transfer performance values ​​of the soil-air heat exchanger obtained above, the structural design of the soil-air heat exchanger can be optimized, such as the radius and length of the pipes, the radius and properties of the phase change material, etc., thereby providing design guidance for the application of phase change materials.

[0157] Based on the above technical solution, this invention combines a physical model with a data-driven method, generates high-quality synthetic data by constructing a computational domain, and then trains a gated recurrent neural network using historical data. This significantly reduces the computation time and professional threshold required for predicting the performance of soil-air heat exchangers, thereby providing an efficient prediction method that can quickly and accurately predict the heat transfer performance of soil-air heat exchangers, thus providing a reliable basis for system optimization design and real-time operation control.

[0158] On the other hand, embodiments of the present invention also provide a soil-air heat exchanger heat transfer prediction system based on phase change material, the system including a memory and a processor;

[0159] The memory is used to store computer programs; the processor is used to call and execute the computer programs so that the system executes the soil-air heat exchanger heat transfer prediction method based on phase change materials as described above.

[0160] In this embodiment of the invention, the processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0161] The various methods, steps, and logic diagrams disclosed in the embodiments of this invention can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The processor reads information from the storage medium and, in conjunction with its hardware, completes the steps of the above methods.

[0162] The storage medium can be memory, such as volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.

[0163] Among them, non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory.

[0164] Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), sync link dynamic random access memory (SLDRAM), and direct memory bus RAM (DRRAM).

[0165] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0166] It should be understood that the system disclosed in the embodiments of the present invention can be implemented in other ways. For example, the division of units is merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the communication connection between units may be through some interface, server, or indirect coupling or communication connection, and may be electrical or other forms.

[0167] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one processing unit. The integrated unit described above can be implemented in hardware or as a software functional unit.

[0168] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0169] Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner for ease of understanding.

[0170] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting heat transfer in a soil-air heat exchanger based on phase change materials, characterized in that, include: A computational domain is constructed along the axial direction of the pipeline containing the phase change material and the soil layer in which the pipeline is buried, and governing equations are constructed based on the physical parameters of each computational domain; the computational domain includes an air region, a phase change material region, and a soil region; The predicted outlet temperature corresponding to the historical inlet temperature of the pipeline is calculated using the control equation, and a composite dataset is constructed based on the historical outlet temperature, the predicted outlet temperature, and the design parameters of the pipeline. A gated recurrent neural network is trained based on the composite dataset to obtain a trained prediction model. The predicted model was used to predict the heat transfer of a soil-air heat exchanger containing phase change material in the target area. The computational domain is constructed using the finite difference method, and the air region, phase change material region, and soil region are divided by a grid of preset length and preset width. The governing equations specifically include: Based on the radius, temperature, wind speed, density, specific heat capacity, and thermal conductivity of the air region, and combined with the radius of the phase change material region, an energy conservation equation is constructed, yielding the first governing equation, which is: in, Pipe radius, unit ; The radius of the phase change material is in meters (m). air density, unit ; The specific heat capacity of air, in units ; The thermal conductivity of air, in units of ; Air temperature, unit ; The unit of time is seconds (s). For inlet wind speed, unit ; This is the axial coordinate, i.e., along the flow direction, in meters (m). It serves as a source for heat and mass transfer via air convection. Based on the phase change material radius, specific heat capacity, density, and temperature of the phase change material region, an energy conservation equation is constructed, resulting in the second governing equation, which is: in, The specific heat capacity of a phase change material is expressed in units of... ; The density of the phase change material is expressed in units of... ; The temperature of the phase change material, in units ; The unit of time is seconds (s). Radial coordinates, in meters (m). The thermal conductivity of the phase change material; Let be the radius of the phase change material control volume. Here are the axial coordinates, in meters (m). Based on the soil radius, specific heat capacity, density, and temperature of the soil region, an energy conservation equation is constructed, resulting in the third governing equation, which is: in, Specific heat capacity of soil, unit ; Soil density, unit ; Soil temperature, unit ; The radius of the soil control volume. The thermal conductivity of the soil. The unit of time is seconds (s). Radial coordinates, in meters (m). Here are the axial coordinates, in meters (m). The predicted outlet temperature corresponding to the historical inlet temperature is calculated using the control equation, specifically including: The temperature of each computational domain is initialized based on the preset temperature boundary; The historical inlet temperature at the first moment is input into the initialized computational domain, and the temperature of each computational domain is calculated by simultaneously solving the first, second, and third control equations. The temperature of the last grid layer in the air region is used as the predicted outlet temperature for the corresponding time moment, and the calculation is repeated cyclically based on the historical inlet temperature for the next time moment until all time moments have been calculated.

2. The method for predicting heat transfer in a soil-air heat exchanger based on phase change materials according to claim 1, characterized in that, Training a gated recurrent neural network based on the composite dataset specifically includes: According to the time series, historical outlet temperature data is selected based on the first preset time window, and historical inlet temperature data and design parameters are selected based on the second preset time window as input, and predicted outlet temperature value is selected as output based on the third preset time window to train the gated recurrent neural network. The Adam algorithm is used as the optimization algorithm during the training process to adjust the hyperparameters of the gated recurrent neural network; Wherein, the first preset time window and the second preset time window have the same length, and the beginning of the second preset time window is located at the end of the first preset time window; the beginning of the third preset time window is the same as the beginning of the second preset time window, and the length of the third preset time window is less than that of the second preset time window.

3. The method for predicting heat transfer in a soil-air heat exchanger based on phase change materials according to claim 2, characterized in that, The gated recurrent neural network includes an update gate and a reset gate, and the calculation formula is as follows: ， ， ， ， in, To update the gate output, for function, Update the weights of the gates to the hidden state. This is the hidden state from the previous moment. To update the weights of the inputs in the gate, Enter the current time. To update the door offset, To reset the gate output, To reset the weights of the gate to a hidden state, To reset the weights of the inputs in the gate, To reset the door offset, The current state is hidden. It is the hyperbolic tangent function. The weights input to the hidden state at the current time. This represents the weight of the previous hidden state within the current hidden state. This is the bias for the hidden state. Output for the current moment.

4. The method for predicting heat transfer in a soil-air heat exchanger based on phase change materials according to claim 2, characterized in that, The design parameters include pipe radius, pipe length, phase change material radius, and inlet wind speed.

5. The method for predicting heat transfer in a soil-air heat exchanger based on phase change materials according to claim 4, characterized in that, The method further includes: Based on the heat transfer prediction of soil-air heat exchangers containing different phase change materials in the target area, the computational domain is reconstructed, the corresponding predicted outlet temperature is recalculated, and the composite dataset is reconstructed. The prediction model is adjusted by using transfer learning combined with a reconstructed composite dataset to obtain a prediction model for the corresponding phase change material.

6. The method for predicting heat transfer in a soil-air heat exchanger based on phase change materials according to claim 5, characterized in that, Using predictive models for heat transfer prediction specifically includes: The pipeline design parameters of the target area and the inlet air temperature of the soil-air heat exchanger are input into the prediction model to obtain the outlet air temperature. The heat transfer efficiency of the corresponding phase change material is calculated based on the inlet air temperature and the outlet air temperature.

7. A soil-air heat exchanger heat transfer prediction system based on phase change material, characterized in that, The system includes a memory and a processor; The memory is used to store computer programs; the processor is used to call and execute the computer programs so that the system performs the soil-air heat exchanger heat transfer prediction method based on phase change material assisted by any one of claims 1 to 6.

Images