Supercritical carbon dioxide heat exchange prediction method and device, electronic equipment, computer readable storage medium and program product
By preprocessing and using deep learning techniques on the supercritical carbon dioxide heat transfer prediction model, the accuracy problem of predicting supercritical carbon dioxide heat transfer performance in existing technologies has been solved, achieving high-precision prediction under complex operating conditions and improving the reliability of equipment design and safe operation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUANENG NUCLEAR ENERGY TECH RES INST CO LTD
- Filing Date
- 2026-02-04
- Publication Date
- 2026-05-12
Smart Images
Figure CN122024906A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of supercritical carbon dioxide heat exchange performance prediction technology in the thermal engineering field, and more specifically, to a supercritical carbon dioxide heat exchange prediction method, device, electronic device, computer-readable storage medium, and program product. Background Technology
[0002] Supercritical carbon dioxide (S-CO2) exhibits broad application prospects in high-efficiency energy conversion fields such as advanced nuclear energy systems, solar thermal power generation, and low-temperature waste heat recovery due to its relatively low critical parameters (critical temperature 31.0℃, critical pressure 7.38MPa) and excellent thermophysical properties. In its supercritical state, S-CO2 combines the high diffusivity of a gas with the high density of a liquid, enabling compact system design and high thermal efficiency operation. However, accurately predicting its heat transfer performance faces significant technical challenges.
[0003] Currently, the prediction of supercritical carbon dioxide heat transfer performance mainly relies on empirical or semi-empirical relationships based on experimental data fitting. Multiple dedicated correlations need to be selected for complex operating conditions such as different flow directions and heat exchanger shapes. Moreover, the selection process is highly dependent on human experience, lacks uniformity and adaptability, and is difficult to accurately characterize nonlinear heat transfer behavior under drastic changes in physical properties and the coupling effect of multiple factors, resulting in low prediction accuracy and limited applicability. Summary of the Invention
[0004] In view of this, the purpose of the present invention is to provide a method, apparatus, electronic device, computer-readable storage medium and program product for predicting supercritical carbon dioxide heat transfer, which can improve the prediction accuracy of heat transfer behavior under complex operating conditions.
[0005] To achieve the above objectives, the technical solutions adopted in the embodiments of the present invention are as follows: In a first aspect, the present invention provides a method for predicting supercritical carbon dioxide heat transfer, the method comprising: The received operating condition information to be predicted is preprocessed to obtain operating condition processing information; the operating condition information to be predicted includes thermophysical parameters, flow parameters, heat exchanger morphology parameters and flow direction parameters. The thermophysical parameters characterize the physical properties of the supercritical carbon dioxide, the flow parameters characterize the motion state of the supercritical carbon dioxide, the heat exchanger morphology parameters characterize the heat exchange environment, and the flow direction parameters characterize the relative relationship between the flow direction of the supercritical carbon dioxide and the direction of gravity. The heat transfer coefficient corresponding to the operating condition information is obtained by using a pre-trained heat transfer prediction model based on the operating condition information.
[0006] In an optional implementation, the heat exchanger morphology parameters include the heat exchanger type; the step of using a pre-trained heat transfer prediction model to predict the heat transfer coefficient corresponding to the operating condition information based on the operating condition processing information, includes: Based on the heat exchanger type in the operating condition information to be predicted, a matching heat exchanger prediction model is determined from multiple pre-trained heat exchanger prediction models; the heat exchanger type corresponds one-to-one with the heat exchanger prediction model. The operating condition processing information is input into a matching heat transfer prediction model for prediction to obtain the heat transfer coefficient corresponding to the operating condition information to be predicted.
[0007] In an optional implementation, the heat transfer prediction model includes an input layer, multiple hidden layers, and an output layer, wherein the multiple hidden layers adopt a fully connected structure; the step of inputting the operating condition processing information into a matching heat transfer prediction model for prediction to obtain the heat transfer coefficient corresponding to the operating condition information to be predicted includes: The operating condition processing information is input into the input layer of the heat transfer prediction model; The working condition processing information is mapped into working condition features using multiple neurons in the input layer; The heat transfer characteristics of the operating conditions are captured layer by layer by using a fully connected structure with multiple hidden layers and a nonlinear activation function, and the heat transfer feature vector is output by the last hidden layer. The output layer is used to convert the heat transfer feature vector into the heat transfer coefficient corresponding to the operating condition information to be predicted.
[0008] In an optional implementation, the heat transfer prediction model is trained in the following manner: The original operating condition data is cleaned and normalized to obtain a standard training dataset; Build multiple initial deep learning models with different numbers of hidden layers; The initial deep learning models are trained using the standard training dataset to obtain optimized deep learning models. The heat transfer prediction model is determined from multiple optimized deep learning models based on a preset complexity threshold and a preset error threshold.
[0009] In an optional implementation, the method further includes: If the operating data for the target heat exchanger type is less than the training data threshold, the target heat exchanger prediction model is obtained from the heat exchanger prediction model corresponding to a heat exchanger type other than the target heat exchanger type. The operating condition data of the target heat exchanger type is cleaned and normalized to obtain the target operating condition training data. The target heat exchange prediction model is fine-tuned by transfer learning based on the training data of the target operating condition to obtain the heat exchange prediction model corresponding to the target heat exchanger type.
[0010] In an optional implementation, the preprocessing of the received operating condition information to be predicted to obtain operating condition processing information includes: The thermophysical parameters, flow parameters, heat exchanger morphology parameters, and flow direction parameters are normalized to obtain the operating condition processing information.
[0011] In a second aspect, the present invention provides a supercritical carbon dioxide heat exchange prediction device, the device comprising: The processing module is used to preprocess the received operating condition information to be predicted to obtain operating condition processing information. The operating condition information to be predicted includes thermophysical parameters, flow parameters, heat exchanger morphology parameters and flow direction parameters. The thermophysical parameters characterize the physical properties of the supercritical carbon dioxide, the flow parameters characterize the motion state of the supercritical carbon dioxide, the heat exchanger morphology parameters characterize the heat exchange environment, and the flow direction parameters characterize the relative relationship between the flow direction of the supercritical carbon dioxide and the direction of gravity. The prediction module is used to make predictions based on the operating condition processing information using a pre-trained heat transfer prediction model, and obtain the heat transfer coefficient corresponding to the operating condition information to be predicted.
[0012] Thirdly, the present invention provides an electronic device including a processor and a memory, wherein the memory stores a computer program executable by the processor, and the processor can execute the computer program to implement the supercritical carbon dioxide heat exchange prediction method described in any of the foregoing embodiments.
[0013] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the supercritical carbon dioxide heat exchange prediction method as described in any of the foregoing embodiments.
[0014] Fifthly, the present invention provides a program product that, when executed by a processor, implements the supercritical carbon dioxide heat transfer prediction method as described in any of the foregoing embodiments.
[0015] Compared to existing technologies, the supercritical carbon dioxide heat transfer prediction method, apparatus, electronic device, computer-readable storage medium, and program product provided in this invention preprocess the operating condition information to be predicted, including thermophysical parameters characterizing the physical properties of supercritical carbon dioxide, flow parameters characterizing its motion state, heat exchanger morphology parameters characterizing the heat transfer environment, and flow direction parameters characterizing the relative relationship between the flow direction and the gravity direction. Then, a pre-trained heat transfer prediction model is used to predict the heat transfer coefficient based on the operating condition information, comprehensively reflecting the coupled influence between changes in physical properties, flow state, structural features, and spatial arrangement. The heat transfer prediction model, by learning the nonlinear mapping relationships in a large amount of operating condition data, effectively overcomes the limitations of traditional empirical formulas that rely on manual judgment and single applicable conditions, improving the prediction accuracy of heat transfer behavior under complex operating conditions and enhancing the modeling capability under the synergistic effect of multiple factors. This provides more reliable technical support for the design optimization and safe operation of thermal equipment.
[0016] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description
[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This diagram illustrates a flowchart of an industrial product sales forecasting method provided by an embodiment of the present invention.
[0019] Figure 2 A network structure diagram of the heat transfer prediction model provided in an embodiment of the present invention is shown.
[0020] Figure 3 This diagram illustrates another flowchart of the industrial product sales forecasting method provided in an embodiment of the present invention.
[0021] Figure 4 This diagram illustrates another flowchart of the industrial product sales forecasting method provided in an embodiment of the present invention.
[0022] Figure 5 A block diagram of a supercritical carbon dioxide heat exchange prediction device provided in an embodiment of the present invention is shown.
[0023] Figure 6 A block diagram of an electronic device provided in an embodiment of the present invention is shown. Detailed Implementation
[0024] 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0025] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.
[0026] It should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0027] The inventors discovered through research that predicting the thermal properties of supercritical carbon dioxide faces the following technical challenges: 1. Dramatic Changes in Physical Properties: In the near-critical region (near the critical point), the thermophysical properties of S-CO2, such as density, viscosity, and specific heat, undergo drastic nonlinear changes with temperature. For example, even a small change in pressure or temperature near the critical point can cause a sharp change in density. This change leads to drastic fluctuations in heat transfer performance, easily resulting in heat transfer deterioration, i.e., a sharp drop in the heat transfer coefficient, which severely reduces heat exchange efficiency and threatens equipment safety.
[0028] 2. Influence of Complex Flow States: The overall flow direction has a significant impact on S-CO2 heat transfer. In vertical risers, the buoyancy effect is in the same direction as the flow acceleration effect, which may exacerbate heat transfer degradation; while in vertical downcomers, the buoyancy effect is in the opposite direction to the flow acceleration effect, which enhances heat transfer. Horizontal and vertical pipes differ in their heat transfer characteristics; vertical pipes exhibit higher peak wall temperatures under heat transfer degradation modes.
[0029] 3. Diversity of heat exchanger shapes: Different heat exchanger shapes significantly affect the heat transfer performance of S-CO2. For example, spiral tube heat exchangers, due to the secondary flow effect, can enhance fluid turbulence and strengthen heat transfer; while straight tube heat exchangers have relatively weak heat transfer performance. This diversity of shapes increases the difficulty of predicting the heat transfer coefficient.
[0030] Based on this, embodiments of the present invention provide a method, apparatus, electronic device, computer-readable storage medium, and program product for predicting supercritical carbon dioxide heat transfer. This method preprocesses the operating condition information to be predicted, including thermophysical parameters characterizing the physical properties of supercritical carbon dioxide, flow parameters characterizing its motion state, heat exchanger morphology parameters characterizing the heat transfer environment, and flow direction parameters characterizing the relative relationship between the flow direction and the gravity direction, to obtain operating condition processing information. A pre-trained heat transfer prediction model then predicts the heat transfer coefficient based on this operating condition processing information, comprehensively reflecting the coupled effects between changes in physical properties, flow state, structural features, and spatial arrangement. The heat transfer prediction model, by learning nonlinear mapping relationships from a large amount of operating condition data, effectively overcomes the limitations of traditional empirical formulas that rely on manual judgment and single applicable conditions. This improves the prediction accuracy of heat transfer behavior under complex operating conditions and enhances the modeling capability under the synergistic effect of multiple factors, providing more reliable technical support for the design optimization and safe operation of thermal equipment.
[0031] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0032] Please refer to Figure 1 , Figure 1 A schematic flowchart of an industrial product sales forecasting method provided by an embodiment of the present invention is shown. The method includes the following steps: Step S300: Preprocess the received operating condition information to be predicted to obtain operating condition processing information; the operating condition information to be predicted includes thermophysical parameters, flow parameters, heat exchanger morphology parameters and flow direction parameters. Thermophysical parameters characterize the physical properties of supercritical carbon dioxide, flow parameters characterize the motion state of supercritical carbon dioxide, heat exchanger morphology parameters characterize the heat exchange environment, and flow direction parameters characterize the relative relationship between the flow direction of supercritical carbon dioxide and the direction of gravity.
[0033] In this embodiment of the invention, the operating condition information to be predicted is user-input data, including but not limited to thermophysical parameters, flow parameters, heat exchanger morphology parameters, and flow direction parameters. The received multidimensional parameters are normalized to obtain the operating condition processing information.
[0034] Thermophysical parameters are used to characterize the physical properties of supercritical carbon dioxide under specific conditions, including but not limited to temperature, pressure, density, specific heat capacity, and thermal conductivity. Flow parameters are used to describe the motion state of the fluid (supercritical carbon dioxide), including but not limited to mass flow rate, flow velocity, and Reynolds number.
[0035] Heat exchanger morphology parameters characterize the geometry and structure of the heat exchange environment, including but not limited to heat exchanger type, pipe diameter, pipe length, and helical tube pitch. Flow direction parameters define the relative relationship between the supercritical carbon dioxide flow direction and the direction of gravity, including but not limited to flow direction and angle of inclination. Examples include upward flow, downward flow, or horizontal flow.
[0036] Step S310: The heat transfer coefficient corresponding to the operating condition information is obtained by using the pre-trained heat transfer prediction model based on the operating condition processing information.
[0037] In this embodiment of the invention, after obtaining the operating condition processing information, the operating condition processing information is input into a pre-trained heat transfer prediction model. The coupling relationship between various parameters is abstracted and mapped at multiple levels through the internal neural network structure of the heat transfer prediction model, and then the predicted value of the heat transfer coefficient under the corresponding operating condition is output.
[0038] It should be noted that the heat transfer prediction model is a deep learning model that has been trained on a large amount of historical experimental or simulation data, and has the ability to capture nonlinear heat transfer laws. The training process ensures that the heat transfer prediction model can learn the heat transfer behavior patterns of supercritical carbon dioxide under different pressures, temperatures, flow rates, and geometric configurations, and has the ability to identify heat transfer deterioration or enhancement phenomena in regions of drastic changes in physical properties.
[0039] In summary, the supercritical carbon dioxide heat transfer prediction method provided in this invention preprocesses the operating condition information to be predicted, including thermophysical parameters characterizing the physical properties of supercritical carbon dioxide, flow parameters characterizing its motion state, heat exchanger morphology parameters characterizing the heat transfer environment, and flow direction parameters characterizing the relative relationship between the flow direction and the gravity direction. Based on this pre-trained heat transfer prediction model, the method predicts the heat transfer coefficient, comprehensively reflecting the coupled effects between changes in physical properties, flow state, structural features, and spatial arrangement. By learning the nonlinear mapping relationships in a large amount of operating condition data, the heat transfer prediction model effectively overcomes the limitations of traditional empirical formulas that rely on manual judgment and single applicable conditions. This improves the prediction accuracy of heat transfer behavior under complex operating conditions, enhances the modeling capability under the synergistic effect of multiple factors, and provides more reliable technical support for the design optimization and safe operation of thermal equipment.
[0040] Optionally, the heat exchanger morphology parameters include the heat exchanger type. With a heat transfer prediction model pre-trained for each heat exchanger type, the following is a possible implementation of how to generate the heat transfer coefficient based on the heat exchanger type. Figure 1 The sub-steps of step S310 may include: Step S311: Determine the matching heat exchange prediction model from multiple pre-trained heat exchange prediction models based on the heat exchanger type in the operating condition information to be predicted; the heat exchanger type corresponds one-to-one with the heat exchange prediction model.
[0041] In the embodiments of this invention, the flow and heat transfer behaviors of supercritical carbon dioxide in different types of heat exchangers vary significantly. This is mainly due to the fundamental differences in the geometry, flow channel layout, and wall contact methods of various heat exchangers. For example, plate heat exchangers, shell-and-tube heat exchangers, and spiral tube heat exchangers exhibit highly non-uniform turbulence intensity, boundary layer development, and local heat transfer performance due to differences in their internal flow channel dimensions, degree of curvature, and surface area density.
[0042] In this context, a single general-purpose model cannot adequately account for the heat transfer characteristics of all heat exchanger types. Therefore, it is necessary to construct differentiated prediction mechanisms for different heat exchanger types. Using heat exchanger type as a classification criterion allows for the selective application of prediction models. This can be understood as incorporating heat exchanger type into the input parameter system, enabling the system to identify applicable models, thereby improving the relevance and accuracy of subsequent prediction processes.
[0043] It should be understood that multiple heat transfer prediction models correspond to different heat exchanger types and have been independently trained on historical data for their respective equipment types, enabling them to capture the heat transfer patterns under specific structures. Upon receiving new operating condition information, the heat exchanger type is extracted from this information, and using this as an index, a perfectly matching heat transfer prediction model is retrieved from a pre-defined model library. This one-to-one correspondence means that each heat exchanger type is bound to a dedicated prediction model, avoiding bias caused by cross-type misuse.
[0044] Step S312: Input the operating condition processing information into the matching heat transfer prediction model for prediction to obtain the heat transfer coefficient corresponding to the operating condition information to be predicted.
[0045] In this embodiment of the invention, the operating condition processing information is input into the matching heat transfer prediction model. The matching heat transfer prediction model performs internal calculations on the received data based on the nonlinear heat transfer mapping relationship learned during its training process for a specific heat exchanger type, and finally predicts the heat transfer coefficient that matches the current operating condition.
[0046] In this process, since the heat transfer prediction model itself has been optimized for the operating characteristics of a specific type of heat exchanger, its output results can more accurately reflect the actual heat transfer state, especially showing stronger adaptability in areas with drastic changes in physical properties or under complex flow conditions.
[0047] As can be seen, the embodiments of the present invention correspond one-to-one with the heat exchanger type in the heat exchanger morphology parameters and pre-trained heat exchange prediction models. Based on the heat exchanger type in the operating condition information to be predicted, a matching heat exchange prediction model is determined from multiple models. Then, the operating condition processing information is input into the matching model for prediction. This enables the selection of a dedicated model for the structural characteristics of different heat exchanger types, improves the pertinence and accuracy of heat transfer coefficient prediction, and enhances the adaptability of the prediction method to diverse heat exchanger types.
[0048] Optionally, the heat transfer prediction model includes an input layer, multiple hidden layers, and an output layer, with the hidden layers employing a fully connected structure. Regarding how to generate the heat transfer coefficient using the heat transfer prediction model, a possible implementation is provided below. The sub-steps of step S312 may include: Step S312-1: Input the operating condition processing information into the input layer of the heat exchange prediction model.
[0049] Step S312-2: Use multiple neurons in the input layer to map the working condition processing information into working condition features.
[0050] In this embodiment of the invention, the network structure of the heat transfer prediction model is as follows: Figure 2 As shown, it includes an input layer, multiple hidden layers, and an output layer. The multiple hidden layers adopt a fully connected structure. The input layer, hidden layers, and output layer all have multiple neurons. The neurons in each layer establish weighted connections with all neurons in the next layer, thereby ensuring that the input information can be fully propagated and combined in the network.
[0051] Operating condition processing information is fed into the input layer in the form of numerical vectors. This information includes preprocessed normalized values of thermophysical parameters, flow parameters, heat exchanger morphology parameters, and flow direction parameters. The number of neurons in the input layer matches the dimension of the input vector, with each neuron corresponding to an operating condition variable. Dimension mapping and channel allocation are performed using the input layer to obtain the operating condition features.
[0052] Step S312-3: The heat transfer law is captured layer by layer by using the fully connected structure of multiple hidden layers and nonlinear activation functions to capture the operating condition characteristics, and the heat transfer feature vector is output by the last hidden layer.
[0053] In this embodiment of the invention, each hidden layer, after receiving the features output from the previous layer, performs a linear transformation through a weight matrix and introduces a nonlinear activation function (such as a Rectified Linear Unit (ReLU) or Sigmoid) to break the limitations of linear expression, thereby enabling the model to fit complex nonlinear functions. As the signal is transmitted layer by layer in the network, low-order features are gradually integrated into high-order abstract features, for example, evolving from a single-parameter response to pattern recognition of typical heat transfer phenomena. Finally, the heat transfer feature vector output by the last hidden layer condenses the core information closely related to the target heat transfer coefficient, reflecting the model's learning results on the heat transfer mechanism under a specific heat exchanger type.
[0054] It should be noted that compared to functions such as Sigmoid, the ReLU function is computationally simple and converges quickly. Therefore, the ReLU function can be preferentially used as the activation function for hidden layers, as its piecewise linearity can effectively alleviate the gradient vanishing problem and enhance the nonlinear expressive power of the network.
[0055] Step S312-4: Use the output layer to convert the heat transfer feature vector into the heat transfer coefficient corresponding to the operating condition information to be predicted.
[0056] In this embodiment of the invention, the output layer is configured with output neurons. The output layer receives the heat transfer feature vector from the last hidden layer. Through learnable weights and parameters, the high-order abstract heat transfer feature vector is restored into specific physical performance indicators. The output can directly correspond to the heat transfer coefficient used in actual engineering.
[0057] As can be seen, the embodiments of the present invention input the operating condition processing information into the input layer of the heat transfer prediction model, and use the input layer neurons to map it into operating condition features. Then, through multiple hidden layers with fully connected structures combined with nonlinear activation functions, the features are extracted and transformed layer by layer. Finally, the output layer transforms the heat transfer feature vector into the heat transfer coefficient. This can effectively capture the nonlinear heat transfer law of supercritical carbon dioxide under complex operating conditions and improve the accuracy and adaptability of heat transfer coefficient prediction.
[0058] Alternatively, regarding how to train the heat transfer prediction model, the following is one possible implementation method. Please refer to... Figure 3 The training process for the heat transfer prediction model includes the following steps: Step S100: Clean and normalize the original working condition data to obtain a standard training dataset.
[0059] In this embodiment of the invention, before training the heat transfer prediction model, technicians obtain the original monitorable parameters of the supercritical carbon dioxide heat transfer process under different operating conditions from the existing experimental database, including pressure, temperature, mass flow rate, heat flux density, etc. Based on physical mechanism analysis and correlation statistical analysis, the degree of correlation between each parameter and the measured heat transfer coefficient is analyzed. Through comprehensive judgment, key parameters (such as thermophysical parameters, flow parameters, heat exchanger morphology parameters and flow direction parameters) that have a significant impact on heat transfer behavior are screened out. The observed values corresponding to the key parameters and their matching measured heat transfer coefficients are extracted from the experimental data to form the original operating condition data for model training.
[0060] Next, the original operating data is cleaned. This cleaning process identifies and removes outliers, missing values, or data points that significantly deviate from physical laws. The remaining valid data after cleaning is then uniformly normalized, scaling each parameter to a similar order of magnitude to eliminate training instability caused by excessive differences in the dimensions of the dependent variable. This processed dataset constitutes the standard training dataset, serving as the unified input basis for subsequent model training and ensuring data consistency and reliability throughout the training process.
[0061] Step S110: Build multiple initial deep learning models with different numbers of hidden layers.
[0062] In this embodiment of the invention, based on the TensorFlow 2.10.0 framework and its provided Keras high-level API, multiple initial deep learning models with different numbers of hidden layers are built. The initial deep learning models adopt a feedforward neural network architecture, and the structure of their input and output layers is determined according to the dimension of the information processed under the working conditions and the prediction target. The number of hidden layers in the middle is set differently as a variable design parameter.
[0063] Leveraging Keras's ability to support rapid modeling, initial deep learning models with varying hidden layers were constructed. For example, some initial deep learning models had fewer hidden layers (e.g., 2-3 layers), suitable for capturing simpler nonlinear relationships; others contained more hidden layers (e.g., 6-8 layers), aiming to enhance the expressive power of deep heat transfer mechanisms. Each type of model maintained a basically consistent fully connected structure in terms of network topology, but by adjusting the number of layers, a series of candidate models with different representation capacities were formed.
[0064] It should be understood that each initial deep learning model maintains a consistent input layer structure (corresponding to thermophysical parameters, flow parameters, heat exchanger shape and flow direction parameters) and output layer design (predicting heat transfer coefficient) to ensure the comparability of the control experiment.
[0065] Step S120: Train each initial deep learning model according to the standard training dataset to obtain the optimized deep learning model.
[0066] In this embodiment of the invention, the standard training dataset is divided into a training set, a validation set, and a test set to ensure that each data subset is representative of the working conditions and thus guarantees the model's generalization ability. For each initial deep learning model containing different numbers of hidden layers and node (i.e., neuron) distributions, the training set data is used for model training.
[0067] During training, the loss function can adopt the mean squared logarithmic error (MSLE) to effectively suppress the excessive influence of large numerical samples on the loss calculation and improve the model's ability to control the relative error of the heat transfer coefficient change trend. At the same time, the Adam optimizer is selected for parameter update. The advantages of the Adam optimizer's momentum method and RMSProp algorithm are utilized, along with its adaptive learning rate mechanism, combined with the estimation of the first and second moments of the gradient, to improve the convergence efficiency and stability of the model in complex nonlinear spaces.
[0068] During the iterative training of the model, based on the performance on the validation set (such as loss value, mean absolute error (MAE), root mean square error (RMSE), etc.), key hyperparameters such as learning rate and batch size are dynamically adjusted, and multiple rounds of hyperparameter tuning are conducted to prevent overfitting and accelerate convergence. After each round of tuning, the optimal performance of the model on the validation set is recorded, and several candidate models that perform well in terms of convergence, stability, and prediction accuracy are selected.
[0069] Finally, by comparing the error metrics (such as mean absolute error (MAE) and root mean square error (RMSE) between the prediction results of each model on the independent test set and the experimental measurements, the prediction accuracy of the model is verified, the performance evaluation of the trained model is completed, the structural configuration with the best generalization ability is determined, and thus a fully trained and optimized deep learning model is obtained.
[0070] Step S130: Determine the heat transfer prediction model from multiple optimized deep learning models based on a preset complexity threshold and a preset error threshold.
[0071] In this embodiment of the invention, heat transfer prediction models are selected by comprehensively considering both the model's prediction accuracy and structural complexity. A preset error threshold is used to limit the maximum allowable prediction deviation of the model on the validation set, ensuring that the selected models have sufficient accuracy. A preset complexity threshold is used to control the total number of parameters or hidden layers of the model, avoiding the selection of overly complex models that could lead to increased inference latency or deployment difficulties.
[0072] In this process, the model that meets the error requirements and has the lowest complexity, or the model that achieves the optimal trade-off between accuracy and complexity, is selected as the final heat transfer prediction model. This achieves synergistic optimization of model performance and practicality, ensuring that the selected model has both high predictive ability and is suitable for promotion and application in actual engineering scenarios.
[0073] It should be noted that when training the corresponding heat exchange prediction model for each heat exchanger type, the original operating condition data obtained should be the experimental data corresponding to each heat exchanger type, and steps S100-S130 should be executed for each heat exchanger type to obtain the heat exchange prediction model corresponding to each heat exchanger type.
[0074] As can be seen, the embodiments of the present invention build and train multiple initial deep learning models with different numbers of hidden layers based on a deep learning framework. By comparing the performance of each optimized deep learning model under preset complexity thresholds and preset error thresholds, the optimal model that balances computational efficiency and prediction accuracy is selected as the heat exchange prediction model. This can systematically balance model performance and resource consumption, improve the stability and applicability of the model in actual engineering scenarios, facilitate subsequent deployment and continuous optimization in a distributed environment, and provide reliable data support for the design and safe operation of thermal equipment.
[0075] Optionally, regarding how to train a heat transfer prediction model using transfer learning, the following is a possible implementation method. Please refer to... Figure 4 The method also includes the following steps: Step S200: If the operating data of the target heat exchanger type is less than the training data threshold, obtain the target heat exchanger prediction model from the heat exchanger prediction model corresponding to heat exchanger types other than the target heat exchanger type.
[0076] In this embodiment of the invention, when the operating data (new heat exchanger morphology, extreme flow conditions, etc.) of the target heat exchanger type is less than the training data threshold, it indicates that the type of heat exchanger has not accumulated enough experimental data in practical applications to support independent modeling.
[0077] In this case, obtaining the target heat exchange prediction model from the heat exchange prediction model corresponding to heat exchanger types other than the target heat exchanger type is essentially introducing a pre-trained model that has been trained on sufficient experimental data of other heat exchanger types (such as straight tube heat exchangers, conventional flow conditions, etc.), and using it as the initial model under the new operating conditions. This allows it to inherit the learning results of the general laws of supercritical carbon dioxide heat exchange performance, providing a high starting point for subsequent adaptation.
[0078] Step S210: Clean and normalize the operating condition data of the target heat exchanger type to obtain the target operating condition training data.
[0079] In this embodiment of the invention, the operating data of the target heat exchanger type is cleaned and normalized to remove abnormal samples, fill in missing fields, and eliminate the problems of dimensional differences and numerical range inconsistencies among parameters through standardization.
[0080] It should be understood that the generated target working condition training data has data distribution characteristics that match the model input space, which can effectively support the parameter fine-tuning process in the subsequent transfer learning stage and ensure that a limited amount of target domain data can still have a substantial optimization effect on the model.
[0081] Step S220: Based on the training data of the target operating condition, the target heat exchange prediction model is fine-tuned by transfer learning to obtain the heat exchange prediction model corresponding to the target heat exchanger type.
[0082] In this embodiment of the invention, the training data for the target operating condition is divided into a training set, a validation set, and a test set. The target heat transfer prediction model is fine-tuned based on the training set, applying the pre-trained model to novel or scarce operating scenarios such as spiral tube heat exchangers or extreme flow conditions. During fine-tuning, some hidden layer weights of the original model are frozen; for example, only the hidden layer near the output layer and the output layer parameters are updated. A small learning rate is used to control the parameter adjustment range, preventing the model from deviating from the known general heat transfer laws due to a small number of samples.
[0083] After training the model using the training set of the target operating condition, the fine-tuned model is validated using the validation set of the target operating condition to evaluate its performance. Then, the validated model is tested using the test set of the target operating condition to calculate prediction error metrics (such as MAE and RMSE) and verify the model's accuracy. Finally, the heat transfer prediction model corresponding to the target heat exchanger type is obtained.
[0084] As can be seen, the embodiments of the present invention utilize the learning results of existing models on the heat transfer laws of supercritical carbon dioxide. Through transfer learning technology, the existing models are fine-tuned under the condition of scarce target operating data. This retains the basic characterization ability of the heat transfer behavior while achieving targeted adaptation and optimization for new operating conditions. It effectively improves the modeling ability and prediction reliability of the model in scenarios with insufficient data, and enhances the adaptability and generalization performance of the heat transfer prediction model to new or rare heat exchanger types.
[0085] Alternatively, one possible implementation method for how to perform preprocessing is provided below. Figure 1 The sub-steps of step S300 may include: The thermophysical parameters, flow parameters, heat exchanger morphology parameters, and flow direction parameters are normalized to obtain operating condition processing information.
[0086] In this embodiment of the invention, the operating condition information to be predicted includes thermophysical parameters, flow parameters, heat exchanger morphology parameters, and flow direction parameters. These parameters have different physical dimensions and numerical ranges. Directly inputting them into the model may cause some parameters to dominate the feature learning process due to their large order of magnitude.
[0087] In this context, normalizing the thermophysical parameters, flow parameters, heat exchanger morphology parameters, and flow direction parameters maps each parameter to a similar numerical range, effectively eliminating dimensional differences and scale imbalances between different variables. The normalized data exhibits consistent distribution characteristics, and as input to the subsequent heat transfer prediction model, it helps improve the stability and convergence speed of model training, while also enhancing the ability to capture the changing trends of the heat transfer coefficient under multiple operating conditions.
[0088] Based on the same inventive concept, the basic principle and technical effects of the supercritical carbon dioxide heat exchange prediction device provided in this embodiment are the same as those in the above embodiments. For the sake of brevity, any parts not mentioned in this embodiment can be referred to the corresponding content in the above embodiments.
[0089] Please refer to Figure 5 , Figure 5 This is a block diagram of a supercritical carbon dioxide heat exchange prediction device 400 provided in an embodiment of the present invention. The supercritical carbon dioxide heat exchange prediction device 400 includes a processing module 410 and a prediction module 420.
[0090] The processing module 410 is used to preprocess the received operating condition information to be predicted to obtain operating condition processing information. The operating condition information to be predicted includes thermophysical parameters, flow parameters, heat exchanger morphology parameters and flow direction parameters. The thermophysical parameters characterize the physical properties of supercritical carbon dioxide, the flow parameters characterize the motion state of supercritical carbon dioxide, the heat exchanger morphology parameters characterize the heat exchange environment, and the flow direction parameters characterize the relative relationship between the flow direction of supercritical carbon dioxide and the direction of gravity. The prediction module 420 is used to make predictions based on the operating condition processing information using a pre-trained heat transfer prediction model to obtain the heat transfer coefficient corresponding to the operating condition information to be predicted.
[0091] In summary, the supercritical carbon dioxide heat transfer prediction device provided in this invention preprocesses the operating condition information to be predicted, including thermophysical parameters characterizing the physical properties of supercritical carbon dioxide, flow parameters characterizing its motion state, heat exchanger morphology parameters characterizing the heat transfer environment, and flow direction parameters characterizing the relative relationship between the flow direction and the gravity direction. Based on this pre-trained heat transfer prediction model, it predicts the heat transfer coefficient, comprehensively reflecting the coupled effects between changes in physical properties, flow state, structural features, and spatial arrangement. By learning the nonlinear mapping relationships in a large amount of operating condition data, the heat transfer prediction model effectively overcomes the limitations of traditional empirical formulas that rely on manual judgment and single applicable conditions. This improves the prediction accuracy of heat transfer behavior under complex operating conditions and enhances the modeling capability under the synergistic effect of multiple factors, providing more reliable technical support for the design optimization and safe operation of thermal equipment.
[0092] Optionally, the heat exchanger morphology parameters include the heat exchanger type. The prediction module 420 is specifically used to determine a matching heat exchanger prediction model from multiple pre-trained heat exchanger prediction models based on the heat exchanger type in the operating condition information to be predicted; the heat exchanger type corresponds one-to-one with the heat exchanger prediction model; the operating condition processing information is input into the matching heat exchanger prediction model for prediction to obtain the heat transfer coefficient corresponding to the operating condition information to be predicted.
[0093] Optionally, the heat transfer prediction model includes an input layer, multiple hidden layers, and an output layer, with the hidden layers employing a fully connected structure. The prediction module 420 is specifically used to input operating condition processing information into the input layer of the heat transfer prediction model; map the operating condition processing information into operating condition features using multiple neurons in the input layer; capture the heat transfer law layer by layer using the fully connected structure and nonlinear activation function of the multiple hidden layers, and output a heat transfer feature vector from the last hidden layer; and convert the heat transfer feature vector into the heat transfer coefficient corresponding to the operating condition information to be predicted using the output layer.
[0094] Optionally, the prediction module 420 is also used to clean and normalize the original operating data to obtain a standard training dataset; build multiple initial deep learning models with different numbers of hidden layers; train each initial deep learning model according to the standard training dataset to obtain an optimized deep learning model; and determine the heat transfer prediction model from multiple optimized deep learning models according to a preset complexity threshold and a preset error threshold.
[0095] Optionally, the prediction module 420 is further configured to: if the operating condition data of the target heat exchanger type is less than the training data threshold, obtain the target heat exchanger prediction model from the heat exchanger prediction model corresponding to a heat exchanger type other than the target heat exchanger type; clean and normalize the operating condition data of the target heat exchanger type to obtain the target operating condition training data; and perform transfer learning fine-tuning on the target heat exchanger prediction model based on the target operating condition training data to obtain the heat exchanger prediction model corresponding to the target heat exchanger type.
[0096] Optionally, the processing module 410 is specifically used to normalize the thermophysical parameters, flow parameters, heat exchanger morphology parameters, and flow direction parameters to obtain operating condition processing information.
[0097] Please refer to Figure 6 This is a block diagram illustrating an electronic device 500 provided in an embodiment of the present invention. The electronic device 500 includes, but is not limited to, a personal computer (PC), a personal digital assistant (PDA), a laptop computer, a tablet computer, and a server. The electronic device 500 includes a memory 510, a processor 520, and a communication module 530. The memory 510, processor 520, and communication module 530 are electrically connected directly or indirectly to each other to achieve data transmission or interaction. For example, these components can be electrically connected to each other through one or more communication buses or signal lines.
[0098] The memory 510 is used to store programs or data. The memory 510 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.
[0099] The processor 520 is used to read / write data or programs stored in the memory 510 and perform corresponding functions. For example, when a computer program stored in the memory 510 is executed by the processor 520, the supercritical carbon dioxide heat exchange prediction method disclosed in the above embodiments can be implemented.
[0100] The communication module 530 is used to establish a communication connection between the electronic device 500 and other communication terminals via a network, and to send and receive data via the network.
[0101] It should be understood that, Figure 6 The structure shown is only a schematic diagram of the electronic device 500. The electronic device 500 may also include components that are larger than those shown. Figure 6 The more or fewer components shown, or having the same Figure 6 The different configurations shown. Figure 6 The components shown can be implemented using hardware, software, or a combination thereof.
[0102] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor 520, implements the supercritical carbon dioxide heat exchange prediction method disclosed in the above embodiments.
[0103] This invention also provides a program product that, when executed by processor 520, implements the supercritical carbon dioxide heat transfer prediction method disclosed in the above embodiments.
[0104] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0105] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0106] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion 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 this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0107] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for predicting supercritical carbon dioxide heat transfer, characterized in that, The method includes: The received operating condition information to be predicted is preprocessed to obtain operating condition processing information; the operating condition information to be predicted includes thermophysical parameters, flow parameters, heat exchanger morphology parameters and flow direction parameters. The thermophysical parameters characterize the physical properties of the supercritical carbon dioxide, the flow parameters characterize the motion state of the supercritical carbon dioxide, the heat exchanger morphology parameters characterize the heat exchange environment, and the flow direction parameters characterize the relative relationship between the flow direction of the supercritical carbon dioxide and the direction of gravity. The heat transfer coefficient corresponding to the operating condition information is obtained by using a pre-trained heat transfer prediction model based on the operating condition information.
2. The method for predicting supercritical carbon dioxide heat transfer according to claim 1, characterized in that, The heat exchanger morphology parameters include the heat exchanger type; the step of using a pre-trained heat transfer prediction model to predict the heat transfer coefficient corresponding to the operating condition information based on the operating condition information includes: Based on the heat exchanger type in the operating condition information to be predicted, a matching heat exchanger prediction model is determined from multiple pre-trained heat exchanger prediction models; the heat exchanger type corresponds one-to-one with the heat exchanger prediction model. The operating condition processing information is input into a matching heat transfer prediction model for prediction to obtain the heat transfer coefficient corresponding to the operating condition information to be predicted.
3. The method for predicting supercritical carbon dioxide heat transfer according to claim 2, characterized in that, The heat transfer prediction model includes an input layer, multiple hidden layers, and an output layer, with the hidden layers employing a fully connected structure. The step of inputting the operating condition processing information into the matching heat transfer prediction model for prediction, to obtain the heat transfer coefficient corresponding to the operating condition information to be predicted, includes: The operating condition processing information is input into the input layer of the heat transfer prediction model; The working condition processing information is mapped into working condition features using multiple neurons in the input layer; The heat transfer characteristics of the operating conditions are captured layer by layer by using a fully connected structure with multiple hidden layers and a nonlinear activation function, and the heat transfer feature vector is output by the last hidden layer. The output layer is used to convert the heat transfer feature vector into the heat transfer coefficient corresponding to the operating condition information to be predicted.
4. The method for predicting supercritical carbon dioxide heat transfer according to any one of claims 1-3, characterized in that, The heat transfer prediction model was trained in the following way: The original operating condition data is cleaned and normalized to obtain a standard training dataset; Build multiple initial deep learning models with different numbers of hidden layers; The initial deep learning models are trained using the standard training dataset to obtain optimized deep learning models. The heat transfer prediction model is determined from multiple optimized deep learning models based on a preset complexity threshold and a preset error threshold.
5. The method for predicting supercritical carbon dioxide heat transfer according to claim 2, characterized in that, The method further includes: If the operating data for the target heat exchanger type is less than the training data threshold, the target heat exchanger prediction model is obtained from the heat exchanger prediction model corresponding to a heat exchanger type other than the target heat exchanger type. The operating condition data of the target heat exchanger type is cleaned and normalized to obtain the target operating condition training data. The target heat exchange prediction model is fine-tuned by transfer learning based on the training data of the target operating condition to obtain the heat exchange prediction model corresponding to the target heat exchanger type.
6. The method for predicting supercritical carbon dioxide heat transfer according to claim 1, characterized in that, The preprocessing of the received operating condition information to be predicted to obtain operating condition processing information includes: The thermophysical parameters, flow parameters, heat exchanger morphology parameters, and flow direction parameters are normalized to obtain the operating condition processing information.
7. A supercritical carbon dioxide heat exchange prediction device, characterized in that, The device includes: The processing module is used to preprocess the received operating condition information to be predicted to obtain operating condition processing information. The operating condition information to be predicted includes thermophysical parameters, flow parameters, heat exchanger morphology parameters and flow direction parameters. The thermophysical parameters characterize the physical properties of the supercritical carbon dioxide, the flow parameters characterize the motion state of the supercritical carbon dioxide, the heat exchanger morphology parameters characterize the heat exchange environment, and the flow direction parameters characterize the relative relationship between the flow direction of the supercritical carbon dioxide and the direction of gravity. The prediction module is used to make predictions based on the operating condition processing information using a pre-trained heat transfer prediction model, and obtain the heat transfer coefficient corresponding to the operating condition information to be predicted.
8. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a computer program that can be executed by the processor to implement the supercritical carbon dioxide heat exchange prediction method according to any one of claims 1-6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the supercritical carbon dioxide heat transfer prediction method as described in any one of claims 1-6.
10. A program product, characterized in that, When the program product is executed by the processor, it implements the supercritical carbon dioxide heat transfer prediction method as described in any one of claims 1-6.