A surface temperature field reconstruction method for an air-cooled heat exchanger and a related device

Abstract

The application provides a surface temperature field reconstruction method of an air-cooled heat exchanger and related devices, and belongs to the technical field of power station air-cooling system state monitoring. The application extracts sparse measurement point data from the true value of the surface temperature field under various working conditions, which is recorded as fourth data; direction field constraint conditions are constructed based on the second data and the third data; the direction field constraint conditions are added to the original graph attention network to obtain an improved graph attention network; the improved graph attention network is trained by using the first data and the fourth data to obtain a trained improved graph attention network; the surface sensor temperature value of the air-cooled heat exchanger is collected, the surface sensor temperature value of the air-cooled heat exchanger is input into the trained improved graph attention network for reconstruction to obtain a reconstructed temperature field. The application solves the problem that the specific air flow direction and circulating water flow direction structure prior of the air-cooled heat exchanger are not utilized, resulting in a large deviation between the reconstruction result and the true physical distribution.

Description

Technical Field

[0001] This invention belongs to the field of power plant air-cooled system condition monitoring technology, specifically relating to a method and related device for reconstructing the surface temperature field of an air-cooled heat exchanger. Background Technology

[0002] The surface temperature distribution of air-cooled heat exchangers is crucial for monitoring the thermal state of radiators and diagnosing localized heat recirculation and fin fouling. In practical engineering, due to space and cost constraints, only 3 to 5 temperature sensors are typically installed in each heat dissipation sector. Traditional interpolation methods (Kriging and inverse distance weighting) cannot capture the non-uniform temperature distribution caused by heat recirculation and localized fouling.

[0003] In recent years, deep learning methods have been attempted for physical field reconstruction, but the following problems exist: (1) General neural networks (convolutional neural network CNN and multi-layer perceptron MLP) do not utilize the unique air flow direction and circulating water flow direction structure prior of air-cooled heat exchangers, resulting in a large deviation between the reconstruction results and the real physical distribution.

[0004] (2) Assuming that the location of the measuring point is fixed and known, it cannot adapt to random layouts installed on different sites.

[0005] (3) There is a lack of a self-verification mechanism for whether the reconstructed temperature field satisfies the continuity of heat flow. Summary of the Invention

[0006] The purpose of this invention is to provide a method and related apparatus for reconstructing the surface temperature field of an air-cooled heat exchanger, which solves the problem in the prior art that does not utilize the unique air flow direction and circulating water flow direction structure of the air-cooled heat exchanger, resulting in a large deviation between the reconstruction result and the actual physical distribution.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for reconstructing the surface temperature field of an air-cooled heat exchanger, comprising the following steps: The true values ​​of the surface temperature field under several operating conditions, as well as the air flow direction field on the heat exchanger surface and the water flow direction field inside the heat exchanger tubes, are recorded as the first data, the second data and the third data, respectively. Sparse measurement point data are extracted from the true values ​​of the surface temperature field under various working conditions and recorded as the fourth data. Based on the second and third data, construct the orientation field constraints; By adding the aforementioned orientation field constraint to the original graph attention network, an improved graph attention network is obtained. Using the first and fourth data, the improved graph attention network is trained to obtain the trained improved graph attention network; The temperature values ​​of the air-cooled heat exchanger surface sensor are collected and input into the trained and improved graph attention network for reconstruction to obtain the reconstructed temperature field.

[0008] A further improvement of this invention is that obtaining the true values ​​of the surface temperature field under several working conditions specifically involves: The true values ​​of the surface temperature field under several working conditions were obtained through computational fluid dynamics simulation or high-density infrared thermography experiments.

[0009] A further improvement of the present invention is that the acquisition of the air flow direction field on the surface of the heat exchanger and the water flow direction field inside the heat exchanger tubes specifically involves: The airflow direction field on the heat exchanger surface and the water flow direction field inside the heat exchanger tubes are obtained through geometric modeling.

[0010] A further improvement of the present invention is that, The loss function used when training the improved graph attention network is the comprehensive loss function, and the expression of the comprehensive loss function is as follows:

[0011] in, For the comprehensive loss function, This is the mean square error of the conventional temperature field reconstruction. These are the weighting coefficients for the heat flow continuity loss function. This is the heat flow continuity loss function; The expression for the heat flow continuity loss function is:

[0012] in, Let the heat flow continuity loss function be , For set Total number of internal grid nodes For divergence operators, For local heat flux density vector, h The local convective heat transfer coefficient is... The ambient air reference temperature To improve the graph attention network, the grid points predicted are obtained. i , j Temperature at )

[0013] A further improvement of this invention lies in that the temperature values ​​of the air-cooled heat exchanger surface sensor are collected and input into a trained, improved graph attention network for reconstruction to obtain the reconstructed temperature field, specifically as follows: The sparse measurement point data extraction module is used to collect the temperature values ​​of the surface sensors of the air-cooled heat exchanger. The temperature values ​​of the surface sensors of the air-cooled heat exchanger are input into the trained and improved graph attention network, and reconstructed through a single forward propagation to obtain the reconstructed temperature field.

[0014] A further improvement of the present invention is that, before the temperature value of the air-cooled heat exchanger surface sensor is collected and input into the trained improved graph attention network for reconstruction, the optimal installation position of the air-cooled heat exchanger surface sensor is determined by using the trained improved graph attention network.

[0015] A further improvement of the present invention is that, after obtaining the reconstructed temperature field, the maximum temperature, temperature gradient distribution, and area of ​​the high-temperature region are determined using the reconstructed temperature field.

[0016] In a second aspect, the present invention provides a surface temperature field reconstruction system for an air-cooled heat exchanger, comprising: The data acquisition module is used to acquire the true values ​​of the surface temperature field under several operating conditions, as well as the air flow direction field on the heat exchanger surface and the water flow direction field inside the heat exchanger tubes, which are respectively denoted as the first data, the second data and the third data. Sparse measurement point data are extracted from the true values ​​of the surface temperature field under various working conditions and recorded as the fourth data. The constraint construction module is used to construct orientation field constraints based on the second and third data. The network improvement module is used to add the orientation field constraint to the original graph attention network to obtain an improved graph attention network. The network training module is used to train the improved graph attention network using the first and fourth data to obtain the trained improved graph attention network. The reconstruction module is used to collect the temperature values ​​of the surface sensors of the air-cooled heat exchanger, and input the temperature values ​​of the surface sensors of the air-cooled heat exchanger into the trained and improved graph attention network for reconstruction to obtain the reconstructed temperature field.

[0017] Thirdly, the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the air-cooled heat exchanger surface temperature field reconstruction method described above.

[0018] Fourthly, the present invention provides a storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method for reconstructing the surface temperature field of an air-cooled heat exchanger.

[0019] Compared with the prior art, the present invention has the following beneficial effects: The proposed method for reconstructing the surface temperature field of an air-cooled heat exchanger involves two aspects. First, it acquires the true values ​​of the surface temperature field under several operating conditions, as well as the air flow direction field on the heat exchanger surface and the water flow direction field inside the heat exchanger tubes, denoted as the first data, the second data, and the third data, respectively. This demonstrates that the invention simultaneously acquires two types of flow direction physical fields (air flow direction field and water flow direction field), introducing real heat transfer priors into the graph attention network. This eliminates the reliance on simply fitting temperature values, supplementing heat transfer law information from the data source level and providing a physical basis for improving the accuracy of the air-cooled heat exchanger surface temperature field reconstruction. Second, it adds directional field constraints to the original graph attention network, resulting in an improved graph attention network. This operation not only allows heat transfer information to propagate along the main heat flow direction but also filters out invalid feature interactions that lack spatial heat transfer correlation, thereby improving the accuracy of the air-cooled heat exchanger surface temperature field reconstruction.

[0020] Furthermore, this invention discloses a comprehensive loss function expression used when training the improved graph attention network. As can be seen from the comprehensive loss function expression, this invention considers the heat flow continuity loss, and this design can ensure the physical consistency of the reconstructed temperature field.

[0021] Furthermore, this invention discloses a method for acquiring temperature values ​​from a surface sensor of an air-cooled heat exchanger. Before inputting these temperature values ​​into a trained, improved graph attention network for reconstruction, the optimal installation position of the surface sensor is determined using the trained, improved graph attention network. This operation can improve the accuracy of the reconstructed temperature field of the air-cooled heat exchanger surface. Attached Figure Description

[0022] Figure 1 This is a flowchart of the air-cooled heat exchanger surface temperature field reconstruction method of the present invention; Figure 2 This is a schematic diagram of the surface temperature field reconstruction system of the air-cooled heat exchanger of the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device of the present invention. Detailed Implementation

[0023] To further understand the content of this invention, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments are merely illustrative and not limiting of the invention.

[0024] Example 1: The flowchart of the air-cooled heat exchanger surface temperature field reconstruction method of the present invention is as follows: Figure 1 As shown, the method for reconstructing the surface temperature field of an air-cooled heat exchanger according to the present invention includes the following steps: S1. Obtain the true values ​​of the surface temperature field under several operating conditions, as well as the air flow direction field on the heat exchanger surface and the water flow direction field inside the heat exchanger tubes, and record them as the first data, the second data and the third data respectively; S2. Extract sparse measurement point data from the true values ​​of the surface temperature field under various working conditions, and record it as the fourth data. S3. Based on the second and third data, construct the orientation field constraints; S4. Add the aforementioned orientation field constraint to the original graph attention network to obtain an improved graph attention network; S5. Using the first and fourth data, train the improved graph attention network to obtain the trained improved graph attention network; S6. Collect the temperature values ​​of the surface sensors of the air-cooled heat exchanger, and input the temperature values ​​of the surface sensors of the air-cooled heat exchanger into the trained and improved graph attention network for reconstruction to obtain the reconstructed temperature field.

[0025] Example 2: A schematic diagram of the surface temperature field reconstruction system of the air-cooled heat exchanger of the present invention is shown below. Figure 2 As shown, the air-cooled heat exchanger surface temperature field reconstruction system of the present invention includes: The data acquisition module is used to acquire the true values ​​of the surface temperature field under several operating conditions, as well as the air flow direction field on the heat exchanger surface and the water flow direction field inside the heat exchanger tubes, which are respectively denoted as the first data, the second data and the third data. The sparse measurement point data extraction module is used to extract sparse measurement point data from the true value of the surface temperature field under various working conditions, and is referred to as the fourth data. The constraint construction module is used to construct orientation field constraints based on the second and third data. The network improvement module is used to add the orientation field constraint to the original graph attention network to obtain an improved graph attention network. The network training module is used to train the improved graph attention network using the first and fourth data to obtain the trained improved graph attention network. The reconstruction module is used to collect the temperature values ​​of the surface sensors of the air-cooled heat exchanger, and input the temperature values ​​of the surface sensors of the air-cooled heat exchanger into the trained and improved graph attention network for reconstruction to obtain the reconstructed temperature field.

[0026] Example 3: The method for reconstructing the surface temperature field of an air-cooled heat exchanger according to the present invention includes the following steps: S1. Obtain the true values ​​of the surface temperature field under several operating conditions, as well as the air flow direction field on the heat exchanger surface and the water flow direction field inside the heat exchanger tubes, and record them as the first data, the second data and the third data respectively.

[0027] This step involves obtaining the true values ​​of the surface temperature field under several operating conditions, specifically: The true values ​​of the surface temperature field under several working conditions were obtained through computational fluid dynamics simulation or high-density infrared thermography experiments.

[0028] This step involves obtaining the airflow direction field on the heat exchanger surface and the water flow direction field inside the heat exchanger tubes, specifically: The airflow direction field on the heat exchanger surface and the water flow direction field inside the heat exchanger tubes are obtained through geometric modeling.

[0029] S2. Extract sparse measurement point data from the true values ​​of the surface temperature field under various working conditions, and record it as the fourth data.

[0030] S3. Based on the second and third data, construct the orientation field constraints.

[0031] S4. Add the aforementioned orientation field constraint to the original graph attention network to obtain the improved graph attention network.

[0032] S5. Using the first and fourth data, train the improved graph attention network to obtain the trained improved graph attention network.

[0033] In this step, the loss function used when training the improved graph attention network is the comprehensive loss function, which is expressed as follows:

[0034] in, For the comprehensive loss function, This is the mean square error of the conventional temperature field reconstruction. These are the weighting coefficients for the heat flow continuity loss function. This is the heat flow continuity loss function.

[0035] The expression for the heat flux continuity loss function is:

[0036] in, Let the heat flow continuity loss function be , For set Total number of internal grid nodes For divergence operators, For local heat flux density vector, h The local convective heat transfer coefficient is... The ambient air reference temperature To improve the graph attention network, the grid points predicted are obtained ( i , j Temperature at )

[0037] S6. Collect the temperature values ​​of the surface sensors of the air-cooled heat exchanger, and input the temperature values ​​of the surface sensors of the air-cooled heat exchanger into the trained and improved graph attention network for reconstruction to obtain the reconstructed temperature field.

[0038] In this step, the temperature values ​​of the air-cooled heat exchanger surface sensor are collected and input into the trained and improved graph attention network for reconstruction, resulting in the reconstructed temperature field. Specifically: Temperature values ​​from sensors on the surface of an air-cooled heat exchanger are collected and input into a trained, improved graph attention network. The reconstructed temperature field is obtained through a single forward propagation.

[0039] In this step, the temperature values ​​of the air-cooled heat exchanger surface sensor are collected. Before the air-cooled heat exchanger surface sensor temperature values ​​are input into the trained improved graph attention network for reconstruction, the optimal installation position of the air-cooled heat exchanger surface sensor is determined using the trained improved graph attention network.

[0040] After obtaining the reconstructed temperature field in this step, the maximum temperature, temperature gradient distribution, and area of ​​the high-temperature region are determined using the reconstructed temperature field.

[0041] Example 4: The method of the present invention will be described in detail below. The method for reconstructing the surface temperature field of an air-cooled heat exchanger of the present invention includes the following steps: Step 1: Construct a dataset containing prior knowledge of heat flow direction A typical air-cooled heat dissipation unit (such as a complete sector) was selected as the research object. The true values ​​(first data) of the surface temperature field under different operating conditions were obtained through CFD (Computational Fluid Dynamics) simulation or high-density infrared thermography experiments. (H and W are the grid height and width).

[0042] Simultaneously acquire the airflow direction field on the heat exchanger surface (second data). (Unit vector, along the direction of airflow) and the direction of water flow inside the heat exchanger tubes. (Along the longitudinal direction of the finned tube, it is usually orthogonal to the air direction or at a certain angle). The directional field (air flow direction field and water flow direction field inside the tube, third data) can be directly obtained by geometric modeling, does not change with the operating conditions, and serves as a fixed physical prior.

[0043] Under each operating condition, K measurement points (K=3~8) are randomly selected in the simulation, and the temperature values ​​at the measurement points are recorded. and coordinates of the measuring points (Fourth data) A total of N training samples were generated (N≥5000).

[0044] Step 2: Design a graph attention network based on orientation field constraints The core improvement logic in this step is as follows: Based on the second data (air flow direction field) and the third data (water flow direction field inside the pipe), direction field constraints are constructed, and direction field constraints are added to the original graph attention network to obtain the improved graph attention network.

[0045] The structure of the improved graph attention network is explained below: Graph construction module: The entire surface temperature field is considered as a graph G=(V,E), where node V corresponds to each grid point. i , j Edge E is used to connect adjacent grid points. The orientation field vector at each node is pre-stored. and .

[0046] Measurement point encoder: Using the fourth data (temperature values ​​and coordinates of K sparse measurement points), the initial features of the nodes are mapped through MLP. For nodes without measurement points, the initial features are zero.

[0047] Directional Graph Attention Layer: Designing a direction-aware attention mechanism. Based on traditional graph attention, a direction field constraint is added using second and third data. node i For adjacent nodes j The attention coefficient depends not only on feature similarity, but also on two points (nodes). i For adjacent nodes j The angle between the line connecting the two points and the direction of airflow. ,node i and adjacent nodes j The formula for calculating the original attention score between them is:

[0048] in, For nodes i and adjacent nodes j The original attention scores between them For linear units with leakage correction, This is the transposed attention projection weight vector. For nodes i The node latent feature vector after uniform linear transformation Adjacent nodesj The node latent feature vector after uniform linear transformation Let be the directional similarity function. For nodes i and adjacent nodes j The spatial geometric angle between them.

[0049] Directional similarity function The design forces information to propagate preferentially along the airflow and waterflow directions, which conforms to the actual laws of heat diffusion.

[0050] Multi-layer propagation and dimensionality-upgrading output module: Stacks multiple directional graph attention layers (3-5 layers) to achieve temperature information propagation from sparse measurement points to all nodes in the graph. Finally, a fully connected mapping is used to obtain the temperature prediction value of each node. .

[0051] Step 3: Design a comprehensive loss function for self-verification of heat flow continuity In addition to the standard mean square error of temperature field reconstruction This introduces a loss of heat flow continuity.

[0052] Calculate the local heat flux density vector using the reconstructed temperature field. , where λ is the equivalent thermal conductivity of the fin (known).

[0053] At sparse measurement points, there are also measured temperatures. The actual heat flux density at that point can be estimated. (Measured via finite difference or sensor heat flux accessory; ignore this part if none is available).

[0054] In regions without measurement points, the divergence of the heat flow field must conform to thermal equilibrium: for finned units, the net heat flow divergence should be equal to the convective heat transfer with the air.

[0055] The expression for the comprehensive loss function is:

[0056] in, For the comprehensive loss function, This is the mean square error of the conventional temperature field reconstruction. These are the weighting coefficients for the heat flow continuity loss function. This is the heat flow continuity loss function.

[0057] The expression for the heat flux continuity loss function is:

[0058] in, Let the heat flow continuity loss function be , For set Total number of internal grid nodes For divergence operators, For local heat flux density vector, h The local convective heat transfer coefficient (can be pre-calculated based on empirical correlations or CFD). The ambient air reference temperature To improve the graph attention network, the grid points predicted are obtained ( i , j Temperature at )

[0059] Step 4: Measurement Point Layout Optimization Submodule To improve reconstruction accuracy, this invention also provides a method for active placement of measurement points based on information entropy, specifically: During the offline phase, a pre-trained, improved graph attention network is used to calculate the contribution of each grid point to the reconstruction uncertainty (Monte Carlo dropout estimation), recommending sensor placement in areas of highest uncertainty. This process can be combined with field installation to output a list of suggested measurement point locations.

[0060] Step 5: Online Reconfiguration and Monitoring Application In actual operation, the temperature values ​​and coordinates of K sensors installed on the surface of the heat exchanger are read, input into a trained graph attention network, and the full surface temperature field is output in a single forward propagation (reconstructing the temperature field in milliseconds).

[0061] The maximum temperature, temperature gradient distribution, and area of ​​the high-temperature zone are calculated using the reconstructed temperature field, which can be used to determine heat recirculation, dirt, or louver abnormalities.

[0062] Example 5: The following describes the method of this invention using a heat dissipation sector of an indirect air-cooled system of a 660MW unit as an example. This sector contains 20×8 finned tube bundle units, with air flowing from bottom to top and water flowing horizontally inside the tubes. On-site, only five PT100 temperature sensors (at the four corners and the center) were evenly spaced within this sector. The following sections describe the data set construction, network configuration, and training results.

[0063] A. Dataset Construction The simulation used ANSYS Fluent to simulate 4300 operating conditions (inlet water temperature, ambient temperature, wind speed, and louver opening changes). The surface temperature field of each condition was output as a 60×40 grid, and the airflow direction field (uniformly vertical upwards) and the water flow direction field inside the pipe (horizontally from left to right) were extracted. Five measuring points were randomly selected in the simulation (the locations correspond to the actual sensor locations). A total of 4000 training samples and 300 test samples were generated.

[0064] B. Network Configuration The graph has 2400 nodes, with edges connecting the top, bottom, left, and right neighbors. Longitudinal edges are added along the air direction (up / down) and the horizontal direction (water flow) to accelerate information propagation (allowing single-step jumps across multiple nodes). The angle weights in the directional attention coefficient are... This maximizes the weights propagating in both directions. The network has 4 layers and 64 hidden feature dimensions. The Adam optimizer is used with a learning rate of 1e-3.

[0065] C. Training Results The method of this invention achieves a total temperature RMSE (Root Mean Square Error) of 0.65℃ on the test set, with an average deviation of 0.08m at the location of the highest temperature point. Two existing methods were selected for comparative experiments: The Kriging interpolation RMSE is 2.34℃, and the average deviation of the highest temperature point location is 0.52m.

[0066] The RMSE of the undirected prior GAT (Graph Attention Network, which represents the same graph structure but whose attention coefficients do not contain a direction field) is 1.21℃.

[0067] Meanwhile, the heat flow continuity loss function of the present invention reduces the divergence error of the reconstructed temperature field by 60%.

[0068] To verify the effectiveness of the air-cooled heat exchanger surface temperature field reconstruction method of the present invention, online measurements were conducted in this embodiment. The specific process is as follows: The trained and improved graph attention network was deployed to a field PLC and industrial computer to read temperature values ​​from five sensors in real time (sampling period of 1 second). The network was then used to reconstruct the temperature field and output a full-sector temperature cloud map. On a certain day under high-temperature conditions, the network automatically identified a localized high-temperature area (8°C higher than the average temperature) in the southwest corner of the sector and issued an alarm indicating that the louvers in that area might be stuck. Inspection confirmed a malfunction in the louver's drive mechanism; after repair, the sector temperature returned to normal.

[0069] Compared with the prior art, the present invention has the following advantages: A. Significantly reduce monitoring bias With 5 sparse measurement points, the root mean square error (RMSE) of the full-field reconstruction is reduced by 70% compared to Kriging and by 45% compared to graph neural networks without direction prior.

[0070] B. Strong physical consistency The directional graph attention mechanism (direction perception attention mechanism) forces information to propagate along the direction of air and water flow, reconstructs the field to conform to the physical laws of thermal diffusion, and avoids unreasonable lateral jaggedness.

[0071] C. Measurement point adaptive Graph attention networks can accept any number and location of measurement points as input, without retraining, and adapt to different sensor layouts in the field.

[0072] D. Self-verification of heat flow continuity No additional sensors are required; the reconstructed field alone can be used to verify whether the heat exchanger is in thermal equilibrium, providing a confidence index.

[0073] E. Guidelines for Measurement Point Layout Active learning methods can reduce reconstruction errors caused by improper measurement point locations, further improving engineering practicality.

[0074] Example 6: Please see Figure 3 As shown, the present invention also provides an electronic device 100 for a method of reconstructing the surface temperature field of an air-cooled heat exchanger; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.

[0075] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the air-cooled heat exchanger surface temperature field reconstruction method described in Embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device 100 (such as audio data), etc. In addition, the memory 101 may include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other non-volatile solid-state storage device.

[0076] The at least one processor 102 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor. The processor 102 is the control center of the electronic device 100, connecting various parts of the electronic device 100 via various interfaces and lines.

[0077] The memory 101 in the electronic device 100 stores multiple instructions to implement a method for reconstructing the surface temperature field of an air-cooled heat exchanger, and the processor 102 can execute the multiple instructions to achieve the following: The true values ​​of the surface temperature field under several operating conditions, as well as the air flow direction field on the heat exchanger surface and the water flow direction field inside the heat exchanger tubes, are recorded as the first data, the second data and the third data, respectively. Sparse measurement point data are extracted from the true values ​​of the surface temperature field under various working conditions and recorded as the fourth data. Based on the second and third data, construct the orientation field constraints; By adding the aforementioned orientation field constraint to the original graph attention network, an improved graph attention network is obtained. Using the first and fourth data, the improved graph attention network is trained to obtain the trained improved graph attention network; The temperature values ​​of the air-cooled heat exchanger surface sensor are collected and input into the trained and improved graph attention network for reconstruction to obtain the reconstructed temperature field.

[0078] Example 7: If the modules / units integrated in the electronic device 100 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0079] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0080] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0081] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0082] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for reconstructing the surface temperature field of an air-cooled heat exchanger, characterized in that, Includes the following steps: The true values ​​of the surface temperature field under several operating conditions, as well as the air flow direction field on the heat exchanger surface and the water flow direction field inside the heat exchanger tubes, are recorded as the first data, the second data and the third data, respectively. Sparse measurement point data are extracted from the true values ​​of the surface temperature field under various working conditions and recorded as the fourth data. Based on the second and third data, construct the orientation field constraints; By adding the aforementioned orientation field constraint to the original graph attention network, an improved graph attention network is obtained. Using the first and fourth data, the improved graph attention network is trained to obtain the trained improved graph attention network; The temperature values ​​of the air-cooled heat exchanger surface sensor are collected and input into the trained and improved graph attention network for reconstruction to obtain the reconstructed temperature field.

2. The method for reconstructing the surface temperature field of an air-cooled heat exchanger according to claim 1, characterized in that, The specific steps for obtaining the true values ​​of the surface temperature field under several working conditions are as follows: The true values ​​of the surface temperature field under several working conditions were obtained through computational fluid dynamics simulation or high-density infrared thermography experiments.

3. The method for reconstructing the surface temperature field of an air-cooled heat exchanger according to claim 1, characterized in that, The acquisition of the airflow direction field on the heat exchanger surface and the water flow direction field inside the heat exchanger tubes specifically involves: The airflow direction field on the heat exchanger surface and the water flow direction field inside the heat exchanger tubes are obtained through geometric modeling.

4. The method for reconstructing the surface temperature field of an air-cooled heat exchanger according to claim 1, characterized in that, The loss function used when training the improved graph attention network is the comprehensive loss function, and the expression of the comprehensive loss function is as follows: in, For the comprehensive loss function, This is the standard mean square error for temperature field reconstruction. These are the weighting coefficients for the heat flow continuity loss function. This is the heat flow continuity loss function; The expression for the heat flow continuity loss function is: in, Let the heat flow continuity loss function be , For set Total number of internal grid nodes For divergence operators, For local heat flux density vector, h The local convective heat transfer coefficient is... The ambient air reference temperature To improve the graph attention network, the grid points predicted are obtained ( i , j Temperature at ) 5. The method for reconstructing the surface temperature field of an air-cooled heat exchanger according to claim 1, characterized in that, The temperature values ​​of the air-cooled heat exchanger surface sensor are collected and input into a trained, improved graph attention network for reconstruction to obtain the reconstructed temperature field. Specifically: Temperature values ​​from sensors on the surface of an air-cooled heat exchanger are collected and input into a trained, improved graph attention network. The reconstructed temperature field is obtained through a single forward propagation.

6. The method for reconstructing the surface temperature field of an air-cooled heat exchanger according to claim 1, characterized in that, Before the temperature values ​​of the air-cooled heat exchanger surface sensor are input into the trained improved graph attention network for reconstruction, the optimal installation position of the air-cooled heat exchanger surface sensor is determined using the trained improved graph attention network.

7. The method for reconstructing the surface temperature field of an air-cooled heat exchanger according to claim 1, characterized in that, After obtaining the reconstructed temperature field, the maximum temperature, temperature gradient distribution, and area of ​​the high-temperature region are determined using the reconstructed temperature field.

8. A surface temperature field reconstruction system for an air-cooled heat exchanger, characterized in that, include: The data acquisition module is used to acquire the true values ​​of the surface temperature field under several operating conditions, as well as the air flow direction field on the heat exchanger surface and the water flow direction field inside the heat exchanger tubes, which are respectively denoted as the first data, the second data and the third data. The sparse measurement point data extraction module is used to extract sparse measurement point data from the true value of the surface temperature field under various working conditions, and is referred to as the fourth data. The constraint construction module is used to construct orientation field constraints based on the second and third data. The network improvement module is used to add the orientation field constraint to the original graph attention network to obtain an improved graph attention network. The network training module is used to train the improved graph attention network using the first and fourth data to obtain the trained improved graph attention network. The reconstruction module is used to collect the temperature values ​​of the surface sensors of the air-cooled heat exchanger, and input the temperature values ​​of the surface sensors of the air-cooled heat exchanger into the trained and improved graph attention network for reconstruction to obtain the reconstructed temperature field.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the air-cooled heat exchanger surface temperature field reconstruction method according to any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the air-cooled heat exchanger surface temperature field reconstruction method according to any one of claims 1 to 7.

Images