Method and device for acquiring parameter sample of heat exchanger and storage medium

Abstract

The invention provides a heat exchanger parameter sample obtaining method and device and a storage medium, at least one sample set is obtained according to N initial parameter samples, a target sample set which corresponds to the nth initial parameter sample and comprises other initial parameter samples except the nth initial parameter sample is obtained from the at least one sample set, and n belongs to [1, N]; obtaining a prediction error of predicting the outlet temperature of the heat exchanger by using the target sample set, determining a sample screening model for screening parameter samples of which the prediction uncertainty meets a preset requirement according to the initial parameter samples and the prediction error of the target sample set corresponding to the initial parameter samples, and determining the outlet temperature of the heat exchanger on the basis of the sample screening model. Obtaining a target parameter sample from the at least one candidate parameter sample; the target parameter samples serve as at least part of samples used for training the outlet temperature prediction model. The target parameter sample is adaptively acquired according to the initial parameter sample, so that the problem that the coverage area of the parameter sample is not comprehensive when the parameter sample is fixedly set can be avoided.

Description

Technical Field

[0001] The present invention relates to the field of vehicle technology, and in particular to a method, device and storage medium for acquiring heat exchanger parameter samples. Background Art

[0002] Heat exchangers are used to exchange heat between key components such as motors, engines, and air conditioners. Their performance impacts a vehicle's energy efficiency and reliability. When designing and optimizing heat exchangers using computational fluid dynamics (CFD) methods, it's essential to predict the heat exchanger's outlet temperature.

[0003] In related art, heat exchanger parameters at fixed sampling points are usually obtained, and then a heat exchanger parameter database is established based on the heat exchanger parameters at the fixed sampling points, and a heat exchanger outlet temperature prediction model is trained based on the heat exchanger parameter database.

[0004] However, the data in the parameter database obtained by this fixed sampling method may not cover a comprehensive sampling area. The accuracy and applicability of the heat exchanger parameter database constructed based on this method are poor. The prediction results of the outlet temperature prediction model obtained by training the outlet temperature prediction model based on this parameter database are poor in accuracy. Summary of the Invention

[0005] In view of the above problems, embodiments of the present invention provide a method, device, and storage medium for obtaining heat exchanger parameter samples that overcome the above problems or at least partially solve the above problems.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] In a first aspect, an embodiment of the present application discloses a method for obtaining a heat exchanger parameter sample, comprising:

[0008] Obtaining N initial parameter samples, and obtaining at least one sample set based on the N initial parameter samples; the initial parameter samples include performance parameters of the heat exchanger, and N is an integer greater than or equal to 2;

[0009] For the nth initial parameter sample, obtaining a target sample set corresponding to the nth initial parameter sample from at least one sample set; the target sample set includes other initial parameter samples except the nth initial parameter sample; n∈[1,N];

[0010] Obtaining a prediction error of heat exchanger outlet temperature prediction using the target sample set;

[0011] Determining a sample screening model based on the initial parameter sample and the prediction error of the target sample set corresponding to the initial parameter sample, wherein the sample screening model is used to screen parameter samples whose prediction uncertainty meets preset requirements;

[0012] At least one candidate parameter sample is obtained, and a model is screened based on the sample, and a target parameter sample is obtained from the at least one candidate parameter sample; the target parameter sample is used as at least part of the sample for training the outlet temperature prediction model.

[0013] In a second aspect, an embodiment of the present application discloses a device for acquiring parameter samples of a heat exchanger, comprising: a first acquisition module for acquiring N initial parameter samples and obtaining at least one sample set based on the N initial parameter samples; the initial parameter samples include performance parameters of the heat exchanger, and N is an integer greater than or equal to 2; a second acquisition module for acquiring, for the nth initial parameter sample, a target sample set corresponding to the nth initial parameter sample from at least one sample set; the target sample set includes other initial parameter samples other than the nth initial parameter sample; n∈[1, N]; a third acquisition module for acquiring a prediction error for predicting the heat exchanger outlet temperature using the target sample set; a fourth acquisition module for determining a sample screening model based on the prediction errors between the initial parameter sample and the target sample set corresponding to the initial parameter sample, the sample screening model being used to screen parameter samples whose prediction uncertainty meets preset requirements;

[0014] The fifth acquisition module is used to obtain at least one candidate parameter sample and, based on the sample screening model, obtain a target parameter sample from at least one candidate parameter sample; the target parameter sample serves as at least a part of the sample for training the outlet temperature prediction model.

[0015] In a third aspect, an embodiment of the present application discloses an electronic device comprising a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the programs or instructions are executed by the processor, the steps of the method described in the first aspect are implemented.

[0016] In a fourth aspect, an embodiment of the present application discloses a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the method described in the first aspect are implemented.

[0017] In this embodiment, N initial parameter samples are obtained, and at least one sample set is obtained based on the N initial parameter samples; the initial parameter samples include performance parameters of the heat exchanger, and for the nth initial parameter sample, a target sample set corresponding to the nth initial parameter sample is obtained from at least one sample set, and the target sample set includes other initial parameter samples except the nth initial parameter sample, and then a prediction error of using the target sample set to predict the heat exchanger outlet temperature is obtained, and a sample screening model is determined based on the prediction error between the initial parameter sample and the target sample set corresponding to the initial parameter sample, and based on the sample screening model, the target parameter sample is obtained from at least one candidate parameter sample. In this embodiment, the target sample set corresponding to the nth initial parameter sample does not include the nth initial parameter sample, and its prediction error is the leave-one-out prediction error of the nth initial parameter sample. The leave-one-out prediction error reflects the prediction error when the nth initial parameter sample is excluded, and can characterize the generalization ability of the outlet temperature prediction model used to predict the outlet temperature on a single initial parameter sample. A sample screening model is determined based on the prediction error of the nth initial parameter sample and the target sample set corresponding to the nth initial parameter sample. The sample screening model can accurately identify the area where the prediction uncertainty meets the requirements. Based on the sample screening model, target parameter samples where the prediction uncertainty meets the requirements can be obtained from the candidate parameter samples. In other words, the target parameter samples obtained in this embodiment can cover the area where the prediction uncertainty meets the requirements, and can solve the problem of the method of fixedly setting sampling points in the related art, that the sampling points do not fully cover the area. For example, it can solve the problem in the related art that the sampling points may not cover the area where the prediction uncertainty meets the requirements. In addition, the target parameter sample of this embodiment is a parameter sample adaptively determined from the candidate parameter samples based on a sample screening model constructed according to the initial parameter sample and the prediction error. The target parameter sample is not a fixed parameter sample, but a parameter sample adaptively determined based on the initial parameter sample. Based on this embodiment, the problem of obtaining sampling points through a fixed sampling method in related technologies, which may lead to redundant data in the parameter sample, is solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flowchart of a method for obtaining a heat exchanger parameter sample provided by an embodiment of the present invention;

[0019] Figure 2 is a flowchart of another method for obtaining heat exchanger parameter samples provided by an embodiment of the present invention;

[0020] Figure 3 This is a flowchart of another method for obtaining heat exchanger parameter samples provided by an embodiment of the present invention;

[0021] Figure 4This is a flowchart of another method for obtaining heat exchanger parameter samples provided by an embodiment of the present invention;

[0022] Figure 5 This is a block diagram of a device for obtaining heat exchanger parameter samples provided by an embodiment of the present invention;

[0023] Figure 6 is a block diagram of an electronic device provided in an embodiment of the present application;

[0024] Figure 7 This is a block diagram of another electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0026] refer to Figure 1 , which shows a method for obtaining a heat exchanger parameter sample provided by an embodiment of the present application, the method comprising:

[0027] Step 101: Obtain N initial parameter samples, and obtain at least one sample set based on the N initial parameter samples.

[0028] Wherein, N is an integer greater than or equal to 2; the initial parameter sample includes performance parameters of the heat exchanger.

[0029] For example, the initial parameter sample may include at least one of the following performance parameters of the heat exchanger: fluid inlet temperature, flow rate, fin thickness, and material thermal conductivity.

[0030] The sample set includes at least some of the initial parameter samples. For example, initial parameter samples can be randomly sampled from the initial parameter samples, and the sample set can be constructed based on the sampled initial parameter samples. At least one sample set can be obtained by repeatedly sampling initial parameter samples and constructing sample sets.

[0031] Step 102: For the nth initial parameter sample, obtain a target sample set corresponding to the nth initial parameter sample from at least one sample set.

[0032] The target sample set includes other initial parameter samples except the nth initial parameter sample.

[0033] Where n∈[1, N], this step allows for a traversal of the initial parameter samples. For each traversed initial parameter sample, a target sample set corresponding to the traversed initial parameter sample is obtained from at least one sample set, and the prediction error for heat exchanger outlet temperature prediction using the target sample set is obtained. The target sample set corresponding to the traversed initial parameter sample includes all other initial parameter samples other than the traversed initial parameter sample.

[0034] For example, when n=1, the target sample set corresponding to the nth initial parameter sample obtained from at least one sample set includes all initial parameter samples except the first initial parameter sample. When n=2, the target sample set corresponding to the nth initial parameter sample obtained from at least one sample set includes all initial parameter samples except the second initial parameter sample. Similarly, a target sample set corresponding to each initial parameter sample can be obtained.

[0035] For example, obtaining at least one sample set based on N initial parameter samples includes: obtaining at least one initial sample set based on the N initial parameter samples; if for each n-th initial parameter sample, there exists an initial sample set that does not include the n-th initial parameter sample, determining the at least one initial sample set as the at least one sample set; if for any initial parameter sample z, there does not exist an initial sample set that does not include the initial parameter sample z, regenerating at least one sample set based on initial parameters other than the initial parameter sample z, and adding the regenerated sample set to the previously generated initial sample set to obtain the at least one sample set in this step. In this way, it is ensured that for each traversed initial parameter sample, a corresponding target sample set can be obtained.

[0036] For example, the initial parameter sample z is the first initial parameter sample. If there is no initial sample set that does not include the first initial parameter sample in at least one initial sample set, then at least one sample set is regenerated from the second to Nth initial parameter sample sets, and the regenerated sample set is added to the previously generated initial sample set to obtain at least one sample set in this step.

[0037] Step 103: Obtain the prediction error of the heat exchanger outlet temperature prediction using the target sample set.

[0038] For example, based on the target sample set, a reduced-order model for predicting the heat exchanger outlet temperature is constructed, and the prediction error of the reduced-order model is obtained. The prediction error of the reduced-order model is determined as the prediction error of the heat exchanger outlet temperature prediction using the target sample set.

[0039] In this embodiment, the target sample set is the sample set that does not include the nth initial parameter sample. Accordingly, the prediction error for heat exchanger outlet temperature prediction using the target sample set is the prediction error when outlet temperature prediction is performed using the target sample set that does not include the nth initial parameter sample. This prediction error is the leave-one-out prediction error for the nth initial parameter sample. This step allows for leave-one-out cross-validation to be performed on the initial parameter sample to obtain the prediction error corresponding to the initial parameter sample.

[0040] Step 104: Determine a sample screening model based on the initial parameter sample and the prediction error of the target sample set corresponding to the initial parameter sample.

[0041] The sample screening model is used to select parameter samples whose prediction uncertainty meets preset requirements. For example, the sample screening model is a Gaussian Process (GP) model, which can obtain the expected improvement value corresponding to the candidate parameter samples and determine the candidate parameter samples corresponding to the preset expected improvement value as the parameter samples that meet the preset requirements.

[0042] In this example, the sample screening model is a Gaussian process model; the initial parameter sample is used as the input parameter of the Gaussian process model, and the prediction error of the target sample set corresponding to the initial parameter sample is used as the output parameter of the Gaussian process model to train and obtain the Gaussian process model in this step.

[0043] For example, the sample screening model is a neural network model trained by initial parameter samples and prediction errors of target sample sets corresponding to the initial parameter samples.

[0044] For example, a covariance matrix of the Gaussian model is generated based on the initial parameter sample and the preset kernel function. The posterior mean function and variance function of the Gaussian model are generated based on the covariance matrix. The Gaussian process model with the covariance matrix, posterior mean function, and variance function is determined as the Gaussian process model in this step.

[0045] Step 105: Obtain at least one candidate parameter sample, and obtain a target parameter sample from the at least one candidate parameter sample based on the sample screening model.

[0046] The target parameter samples serve as at least part of the samples used to train the outlet temperature prediction model.

[0047] Wherein, the candidate parameter sample is different from the initial parameter sample. For example, a parameter sample different from the initial parameter sample can be collected and determined as the candidate parameter sample; or a parameter sample generation strategy can be used to generate a parameter sample different from the initial parameter sample and determine it as the candidate parameter sample, for example, by adjusting the initial parameter sample according to a preset step size to obtain a parameter sample different from the initial parameter sample.

[0048] For example, the candidate parameter samples are substituted into the posterior mean function and variance function in the Gaussian process model respectively to obtain the posterior mean and variance values ​​of the candidate parameter samples. According to the posterior mean and variance values ​​and the expected improvement value calculation formula of the Gaussian process model, the expected improvement value of the candidate parameter samples is obtained, and then the candidate parameter samples corresponding to the preset expected improvement value are determined as the target parameter samples.

[0049] For example, the target parameter samples are used together with the initial parameter samples to train the outlet temperature prediction model, or the outlet temperature prediction model can be trained separately.

[0050] For example, after obtaining the target parameter samples, the method may further include: constructing a sample library based on the target parameter samples and the initial parameter samples; or, adding the target parameter samples to the sample library consisting of the initial parameter samples to update the sample library. The data in the sample library is used to train an outlet temperature prediction model; the outlet temperature prediction model may be a reduced-order model.

[0051] Furthermore, after obtaining N initial parameter samples in step 101, a sample library including at least one initial parameter sample can be constructed. In this step, the target parameter sample is taken as a new parameter sample and added to the sample library including the initial parameter sample, thereby constructing a sample library based on the target parameter sample and the initial parameter sample.

[0052] As a core component of an automotive thermal management system, the heat exchanger is used to exchange heat between key components such as the motor, engine, and air conditioner to ensure that these components operate within an appropriate temperature range. The performance of the heat exchanger directly affects the energy efficiency and reliability of the vehicle. For example, the heat exchanger can be designed and optimized using CFD methods. During the CFD-based heat exchanger design and optimization process, the thermodynamic performance of the heat exchanger can be simulated using Finite Element Analysis (FEA) to obtain thermodynamic performance parameters such as the heat exchanger's outlet temperature. Based on the FEA method, the thermodynamic performance of the heat exchanger can be accurately simulated.

[0053] However, traditional finite element analysis (FEA) often involves numerically solving high-dimensional nonlinear partial differential equations. This process requires a large amount of data to be processed, especially in multi-condition design optimization applications or real-time control applications. This method requires processing even larger amounts of data, resulting in low processing efficiency and making it difficult to meet practical needs. Model order reduction (MOR) technology can address these issues with traditional FEA. Specifically, MOR improves computational efficiency by reducing the complexity of the computational model, making it possible for large-scale computing and real-time applications.

[0054] For example, a snapshot database can be constructed and the parameters in the snapshot database can be used to train the reduced-order model. In data-driven model reduction methods, efficiently constructing the snapshot database is key to improving the prediction accuracy and computational efficiency of the reduced-order model. In other words, the snapshot database is the foundation of reduced-order model analysis. Its core lies in capturing the key dynamic characteristics of the original high-dimensional system through sampling. Based on these key dynamic characteristics, reliable data is then obtained to support the construction of the low-dimensional reduced-order model.

[0055] In related technologies, parameter samples from the database used to construct low-dimensional reduced-order models are typically obtained through a one-time sampling method. However, this method requires fixing the distribution of sampling points at the initial stage of the calculation. The resulting sampling points cannot effectively represent the diversity and nonlinear characteristics of complex systems, resulting in significant limitations in this related art approach. Furthermore, this fixed sampling method may generate redundant parameter samples, which wastes computing resources. Furthermore, this fixed sampling method may leave critical areas uncovered, which reduces the versatility and predictive accuracy of the reduced-order model.

[0056] In this embodiment, N initial parameter samples are obtained, and at least one sample set is obtained based on the N initial parameter samples; the initial parameter samples include performance parameters of the heat exchanger, and for the nth initial parameter sample, a target sample set corresponding to the nth initial parameter sample is obtained from at least one sample set, and the target sample set includes other initial parameter samples except the nth initial parameter sample, and then a prediction error of using the target sample set to predict the heat exchanger outlet temperature is obtained, and a sample screening model is determined based on the prediction error between the initial parameter sample and the target sample set corresponding to the initial parameter sample, and based on the sample screening model, the target parameter sample is obtained from at least one candidate parameter sample. In this embodiment, the target sample set corresponding to the nth initial parameter sample does not include the nth initial parameter sample, and its prediction error is the leave-one-out prediction error of the nth initial parameter sample. The leave-one-out prediction error can reflect the prediction error when the nth initial parameter sample is excluded, and can characterize the generalization ability of the outlet temperature prediction model used to predict the outlet temperature on a single initial parameter sample. A sample screening model is determined based on the prediction error of the nth initial parameter sample and the target sample set corresponding to the nth initial parameter sample. The sample screening model can accurately identify the area where the prediction uncertainty meets the requirements. Based on the sample screening model, a target parameter sample where the prediction uncertainty meets the requirements can be obtained from the candidate parameter samples. In other words, the target parameter sample obtained in this embodiment can cover the area where the prediction uncertainty meets the requirements, and can solve the problem of the method of fixedly setting sampling points in the related art, that the sampling points do not fully cover the area. For example, it can solve the problem in the related art that the sampling points may not cover the area where the prediction uncertainty meets the requirements. In addition, the target parameter sample of this embodiment is a parameter sample adaptively determined from the candidate parameter samples based on a sample screening model constructed according to the initial parameter sample and the prediction error. The target parameter sample is not a fixed parameter sample, but a parameter sample adaptively determined based on the initial parameter sample. Based on this embodiment, the problem of obtaining sampling points through a fixed sampling method in related technologies, which may lead to redundant data in the parameter sample, is solved.

[0057] Referring to 3, the method may include the following steps:

[0058] Step 201: Obtain N initial parameter samples.

[0059] The initial parameter samples include performance parameters of the heat exchanger.

[0060] Step 202: Repeat the first operation to obtain at least one sample set.

[0061] The first operation includes: repeatedly performing the operation of randomly extracting initial parameter samples from N initial parameter samples until the total number of extracted initial parameter samples reaches a preset number, and constructing a sample set based on the preset number of extracted initial parameter samples.

[0062] For example, the preset number may be equal to the number of initial parameter samples in the original data set, and the number of initial parameter samples in the original data set may be equal to the number of at least one initial parameter sample obtained in step 201 .

[0063] Furthermore, a random sampling method with replacement is used to obtain randomly selected initial parameter samples, and then a set of a preset number of randomly selected initial parameter samples with replacement is determined as a sample set. Based on this method, after being selected, the initial parameter samples are not removed from the original data set including the initial parameter samples. In subsequent sampling, the initial parameter samples that have been selected may still be selected again.

[0064] In this embodiment, in the first operation, initial parameter samples are obtained through random sampling. A sample set is constructed based on the extracted initial parameter samples, and the first operation is then repeated to obtain at least one sample set. Constructing a sample set through random sampling and repeatedly performing the first operation of constructing a sample set to obtain at least one sample set ensures that the initial parameter samples in the sample set are uniformly distributed in the high-dimensional space, thus avoiding clustering.

[0065] Step 203 : For the nth initial parameter sample, obtain a target sample set corresponding to the nth initial parameter sample from at least one sample set.

[0066] The target sample set includes other initial parameter samples except the nth initial parameter sample; n∈[1,N].

[0067] For example, each initial parameter sample in the target sample set has a corresponding outlet temperature simulation value. Specifically, the initial parameter samples can be input into simulation software (e.g., CFD software) for simulating outlet temperature distribution to obtain the outlet temperature simulation value corresponding to each initial parameter sample in the target sample set.

[0068] Step 204 : Obtain a reduced-order model and a prediction error of the reduced-order model according to the target sample set and the outlet temperature simulation value corresponding to each initial parameter sample in the target sample set.

[0069] Among them, the reduced-order model is used to predict the heat exchanger outlet temperature.

[0070] For example, when there are multiple target sample sets, reduced-order models corresponding to the respective target sample sets may be established in parallel.

[0071] For example, a reduced-order model can be constructed using the bagging method. Specifically, for each target sample set, a low-dimensional reduced-order model is constructed by combining the Proper Orthogonal Decomposition (POD) method and the Galerkin projection.

[0072] Specifically, the outlet temperature simulation value of the initial parameter sample is extracted. The outlet temperature simulation value is not a single value, but can be the outlet temperature distribution value in the grid area divided by the heat exchanger. According to the outlet temperature simulation value of the heat exchanger of each initial parameter sample in the target sample set, the temperature field snapshot matrix S is constructed; where S∈R P×l ; P is the number of grid nodes, l is the number of initial parameter samples in the target sample set.

[0073] Furthermore, the temperature field snapshot matrix S is subjected to singular value decomposition (SVD), and the first Q (for example, 10) main modes are selected, and the POD basis matrix Φ is constructed based on these Q main modes.

[0074] Combined with the constructed POD basis matrix Φ, the high-dimensional equation is reduced to a 10-dimensional ordinary differential equation system through Galerkin projection, and the model coefficient matrix of the reduced-order model is obtained by solving it. The reduced-order model with this model coefficient matrix is ​​the reduced-order model in this step.

[0075] For example, there are multiple target sample sets corresponding to the nth initial parameter sample, and correspondingly, there are also multiple reduced-order models constructed. The prediction errors of multiple reduced-order models can be averaged to achieve comprehensive processing of the prediction results of multiple reduced-order models, and then the average prediction error obtained by the comprehensive processing is determined as the prediction error of the initial parameter sample corresponding to the target parameter sample.

[0076] For example, M sub-datasets are generated from N initial parameter samples by random sampling with replacement, where the number of initial parameter samples in each sample set is N.

[0077] For example, the sampling process can be implemented through the bootstrap method, for example, through the Bootstrap class of the scikit-learn library. By random sampling with replacement, it is possible to ensure that each subset covers the diversity of the original data.

[0078] Step 205 : Determine the prediction error of the reduced-order model as the prediction error of the heat exchanger outlet temperature prediction using the target sample set.

[0079] In this embodiment, a target sample set and the outlet temperature simulation value of each initial parameter sample in the target sample set are used to obtain a reduced-order model. The reduced-order model has the characteristics of high computational efficiency and low model complexity. Based on the reduced-order model, the heat exchanger outlet temperature can be predicted quickly and accurately, and the initial prediction error corresponding to the target sample set can be obtained quickly and accurately.

[0080] In the case that there are multiple target sample sets corresponding to the traversed initial parameter samples, after step 205, the following steps are further included:

[0081] Step 206 , averaging the prediction errors of multiple target sample sets corresponding to the nth initial parameter sample to obtain a prediction error average value.

[0082] According to the aforementioned embodiment, the prediction error of each target sample set is obtained respectively, and the prediction error is the prediction error of the reduced-order model obtained based on the target sample set.

[0083] Step 207: Determine the average value of the prediction errors as the prediction errors corresponding to the nth initial parameter sample.

[0084] The target sample set corresponding to the nth initial parameter sample is a sample set that does not contain the nth initial parameter sample. Then the prediction error of the target sample set corresponding to the nth initial parameter sample is the leave-one-out prediction error of the initial parameter sample.

[0085] The prediction error corresponding to the nth initial parameter sample can be obtained by averaging the prediction errors of each target sample set corresponding to the nth initial parameter sample. For example, the prediction error e(μ i ) can be obtained by the following method:

[0086]

[0087] in, is the predicted value, T i Indicates the actual value, L 2 Represents the norm.

[0088] Step 208: Determine a sample screening model based on the at least one initial parameter sample and the prediction error corresponding to each initial parameter sample in the at least one initial parameter sample.

[0089] For example, the sample screening model is a Gaussian process model, and a probability model is established. The objective function of the probability model obeys the Gaussian process, and the parameter value of the objective function is a random variable, and its distribution is determined by the variance and covariance of the Gaussian process model.

[0090] Furthermore, the objective function is a proxy model constructed using the Gaussian process model. The initial parameter sample is brought into the Gaussian process model, and its corresponding prediction error is determined as the current observation data. The proxy model of the Gaussian process model is trained to obtain a Gaussian process model including the trained objective function.

[0091] In this embodiment, the prediction errors of the target sample set corresponding to the traversed parameter samples are averaged to obtain the prediction error average value, and the prediction error average value is determined as the prediction error corresponding to the initial parameter sample; by averaging the initial prediction error, the prediction error corresponding to the initial parameter sample is obtained, and subsequent processing is performed based on the prediction error, which can reduce the variance of the prediction error and improve the robustness. Based on the initial parameter sample and the prediction error corresponding to the initial parameter sample, a Gaussian process model with high processing result accuracy can be obtained.

[0092] Step 209: Obtain at least one candidate parameter sample, and screen the model based on the sample to obtain an expected improvement value corresponding to each candidate parameter sample.

[0093] The expected improvement value is used to characterize the prediction uncertainty corresponding to the candidate parameter sample.

[0094] For example, the sample screening model is a Gaussian process model.

[0095] Step 210: Determine the candidate parameter sample corresponding to the preset expected improvement value as the target parameter sample.

[0096] For example, the preset expected improvement value may be the maximum expected improvement value among the expected improvement values ​​corresponding to the plurality of candidate parameter samples. In this case, the target parameter sample determined according to the preset expected improvement value is the parameter sample with the maximum prediction uncertainty.

[0097] The expected improvement value can reflect the prediction uncertainty of the model. In this embodiment, the candidate parameter samples corresponding to the preset expected improvement value are determined as target parameter samples. The target parameter samples are parameter samples that meet the prediction uncertainty corresponding to the preset expected improvement value. Based on this embodiment, target parameter samples that meet the prediction uncertainty requirements can be adaptively obtained.

[0098] For example, the outlet temperature prediction model trained with target parameter samples is a reduced-order model. Based on the method of obtaining target parameter samples in this embodiment, the sampling efficiency can be maximized, redundant data can be reduced, and parameter samples with fewer sampling points can be used to obtain a reduced-order model that meets the prediction uncertainty requirements.

[0099] For example, after step 201, the method may further include: obtaining at least one sample set based on N initial parameter samples; the initial parameter samples include performance parameters of the heat exchanger, and N is an integer greater than or equal to 2; for the nth initial parameter sample, obtaining a target sample set corresponding to the nth initial parameter sample from at least one sample set; the target sample set includes other initial parameter samples except the nth initial parameter sample; n∈[1, N]; obtaining a prediction error for predicting the outlet temperature of the heat exchanger using the target sample set; determining a sample screening model based on the prediction errors of the initial parameter samples and the target sample set corresponding to the initial parameter samples, the sample screening model being used to screen parameter samples whose prediction uncertainties meet preset requirements; obtaining at least one candidate parameter sample, and obtaining a target parameter sample from at least one candidate parameter sample based on the sample screening model; looping through the operations of obtaining at least one sample set based on the N initial parameter samples, and obtaining a target parameter sample from at least one candidate parameter sample based on the sample screening model, until a preset loop termination condition is met, and obtaining target parameter samples corresponding to each loop operation.

[0100] Furthermore, the operations of obtaining at least one sample set based on N initial parameter samples and obtaining a target parameter sample from at least one candidate parameter sample based on the sample screening model may specifically include the method of steps 202 to 210. By repeatedly executing steps 202 to 210, target parameter samples corresponding to each iteration can be obtained.

[0101] Furthermore, the preset loop termination condition may be: the number of loops reaches a number threshold, or the number of target parameter samples obtained reaches a preset target number threshold.

[0102] Among them, one loop operation can obtain at least one target parameter sample, and at least one target parameter sample can be obtained through at least one loop operation of this embodiment.

[0103] In this embodiment, the operation of obtaining at least one sample set based on at least one initial parameter sample, and obtaining a target parameter sample from at least one candidate parameter sample based on a Gaussian process model is executed in a loop until a preset loop termination condition is met, and at least one target parameter sample adaptively determined based on the initial parameter sample can be obtained.

[0104] Step 211: Acquire the outlet temperature simulation value corresponding to the target parameter sample.

[0105] For example, the target parameter sample is input into the outlet temperature simulation software to obtain the outlet temperature simulation value corresponding to the target parameter sample.

[0106] Step 212 : constructing a sample library based on the target parameter samples, the outlet temperature simulation values ​​corresponding to the target parameter samples, the initial parameter samples, and the outlet temperature simulation values ​​corresponding to the initial parameter samples.

[0107] Among them, the target parameter samples in the sample library, the outlet temperature simulation values ​​corresponding to the target parameter samples, the initial parameter samples, and the outlet temperature simulation values ​​corresponding to the initial parameter samples are used to train the outlet temperature prediction model.

[0108] For example, a sample library can be reconstructed using the target parameter samples, the outlet temperature simulation values ​​corresponding to the target parameter samples, the initial parameter samples, and the outlet temperature simulation values ​​corresponding to the initial parameter samples; or the target parameter samples and the outlet temperature simulation values ​​corresponding to the target parameter samples can be added to the sample library consisting of the initial parameter samples and the outlet temperature simulation values ​​corresponding to the initial parameter samples to update the sample library and obtain an updated sample library.

[0109] Furthermore, after obtaining the updated sample library, a reduced-order model may be constructed based on the updated sample library, and the reduced-order model may be used as an outlet temperature prediction model for predicting the outlet temperature of the heat exchanger.

[0110] Furthermore, after obtaining the updated sample library, the parameter samples in the updated samples can be determined as initial parameter samples, and then return to step 201, and the steps of obtaining the target parameter samples in the above embodiment are cyclically executed, and the steps of constructing the sample library are performed based on the target parameter samples, the outlet temperature simulation value corresponding to the target parameter samples, the initial parameter samples, and the outlet temperature simulation value corresponding to the initial parameter samples, so as to cyclically update the sample library so that the area covered by the parameter samples in the sample library becomes more and more comprehensive.

[0111] In this embodiment, the outlet temperature simulation value corresponding to the target parameter sample is obtained, and a sample library is constructed based on the target parameter sample, the outlet temperature simulation value corresponding to the target parameter sample, and the outlet temperature simulation value corresponding to the initial parameter sample. The sample library includes the initial parameter sample, the outlet temperature simulation value corresponding to the initial parameter sample, and the target parameter sample adaptively obtained based on the initial parameter sample, and the outlet temperature simulation value of the target parameter sample. The parameter samples in the sample library are not fixed, but are dynamically determined by the target parameter sample obtained by adaptation. The parameter samples can cover the area where the preset uncertainty meets the requirements, avoiding the problem of constructing a database based on fixed sampling points in the related art, which may result in redundant data in the database and incomplete coverage area. The outlet temperature prediction model is trained based on the sample library, and the prediction result of the outlet temperature prediction model obtained is highly accurate.

[0112] Reference Figure 3The method for obtaining the heat exchanger parameter sample of this embodiment may include the following steps:

[0113] Step S1: Obtain initial sample points.

[0114] The initial sample is the initial parameter sample in the aforementioned embodiment.

[0115] Step S2: Calculate the full-order solution of the initial sample points according to the preset numerical model.

[0116] Specifically, according to a preset outlet temperature simulation model, an outlet temperature simulation value corresponding to the initial sample is obtained.

[0117] Step S3: Based on the reduced-order model constructed using the bagging method, the prediction accuracy of the initial sample points is evaluated.

[0118] In this step, the reduced-order model constructed based on the Bagging method is used to obtain the leave-one-out prediction error corresponding to the initial sample point, and the leave-one-out prediction error is used to represent the prediction result of the initial sample point.

[0119] Step S4, determine whether the prediction accuracy meets the requirements, or whether the number of initial sample points that meet the requirements reaches a preset threshold; if so, end the process, otherwise enter step S5.

[0120] For example, if the prediction accuracy of the target sample set corresponding to each initial parameter sample reaches a preset prediction accuracy threshold, it is determined that the prediction accuracy meets the requirement; otherwise, it is determined that the prediction accuracy does not meet the requirement.

[0121] Step S5: Evaluate the prediction uncertainty of the candidate sample points based on the Bayesian optimization method.

[0122] In this embodiment, based on the Bayesian optimization method, the expected improvement value of the candidate sample point is obtained, and the prediction uncertainty of the candidate sample point is determined according to the expected improvement value.

[0123] In step S6, the sample point with the largest prediction uncertainty is determined as a new sample point, so as to update the sample point, and then return to step S2.

[0124] For example, the sample point with the largest expected improvement value is the sample point with the largest prediction uncertainty.

[0125] For example, refer to Figure 4 , the method may include the following steps:

[0126] Step F1, determining the parameter space, target tolerance and sampling upper limit of the heat exchanger.

[0127] The parameter space includes the parameter types and parameter value ranges of the parameter samples, and the sampling upper limit includes the total number of initial parameter samples and target parameter samples.

[0128] In step F2, a Latin hypercube sampling method is used to generate multiple initial sample points in the parameter space. The simulation data of the heat exchanger temperature field distribution corresponding to the initial sample points is obtained through a high-precision finite element method. An initial snapshot database is constructed based on the initial sample points and the corresponding simulation data of the heat exchanger temperature field distribution.

[0129] The initial sample point is the initial parameter sample in the aforementioned embodiment.

[0130] For example, N initial sample points μ can be generated by the Latin Hypercube Sampling (LHS) method i (i=1,2,…N), and obtain the initial sample point μ through high-precision numerical finite element simulation software. i The simulation data of the temperature field distribution of the heat exchanger under the action of T i (i=1, 2, ...N), and initializes a snapshot database based on the initial sample points and the simulated data of the heat exchanger temperature field distribution. In other words, the initialized snapshot database includes N initial sample points and the simulated data of the heat exchanger temperature field distribution corresponding to each initial sample point. The simulated data of the heat exchanger temperature field distribution is the simulated data of the heat exchanger outlet temperature distribution. Numerical finite element simulation can be performed using commercial CFD software.

[0131] In step F3, a reduced-order model is established for the initial sample points using the Bagging algorithm, and a leave-one-out cross-validation method is adopted to calculate the leave-one-out prediction error of each current initial sample point in combination with the reduced-order model.

[0132] The prediction error is retained, which is the prediction error in the above embodiment.

[0133] For example, the leave-one-out cross-validation method (LOO) is used to obtain the leave-one-out prediction error e(μ i ), for example, leave-one-out prediction error e(μ i ) can be expressed as:

[0134]

[0135] in, is the predicted data obtained using the reduced-order model, which is equivalent to the predicted result obtained by the reduced-order model; T i is the observed data, which is equivalent to the simulated value of the heat exchanger outlet temperature obtained by simulation software.

[0136] For example, step F3 may include the following sub-steps:

[0137] Sub-step F3.1, repeatedly extract subsets from the training data set.

[0138] The core of the bagging method is to repeatedly extract subsets from the training dataset through random sampling with replacement, where each sampling may result in a different subset of data. The extracted subset is the initial parameter sample extracted in the above embodiment. Specifically, random sampling is performed using the bootstrap sampling technique.

[0139] In the process of building the reduced-order model, the Bootstrap sampling method for random sampling is implemented through random sampling and replacement method, and the samples obtained are resampled samples.

[0140] In the processing process based on the Bagging method, the Bootstrap sampling method can be used to create multiple different training data subsets to train multiple models based on different training data subsets. The training data subset is at least one sample set in the aforementioned embodiment.

[0141] During this process, the number of samples can be the same as the size of the original dataset. This process is a random sampling process with replacement. In other words, once a sample is selected, it is not removed from the original dataset; the parameter sample may still be selected again in subsequent sampling. This process is a repetitive sampling process, repeating the above random sampling operation until the new data subset reaches a predetermined size, for example, the new data subset is the same size as the original dataset.

[0142] This embodiment includes multiple sampling processes, and generates multiple data subsets by repeatedly executing the Bootstrap sampling process multiple times. The data subsets are used to train multiple reduced-order models in the Bagging method.

[0143] In sub-step F3.2, a reduced-order model is trained independently using each data subset.

[0144] The reduced-order models are trained on different subsets of the data. Therefore, the bias and variance of each reduced-order model will be different, and the prediction accuracy of the model will also be different.

[0145] Sub-step F3.3: performing model aggregation processing on multiple reduced-order models.

[0146] For example, in a bagging ensemble, multiple reduced-order models are built in parallel, and their results are combined and output. Specifically, the prediction accuracy of each reduced-order model is obtained, and then the average prediction accuracy of all reduced-order models is calculated. This average prediction accuracy is determined as the final prediction accuracy, which is the model synthesis result of the reduced-order models. This allows for comprehensive processing of the output results of multiple models.

[0147] In step F4, a Bayesian optimization method is used to find new sample points based on the leave-one-out validation error predicted by the reduced-order model.

[0148] The newly added sample points are the target parameter samples in the aforementioned embodiment.

[0149] For example, step F4 may include the following sub-steps:

[0150] Sub-step F4.1, initialize the probability model.

[0151] Specifically, a probability model is established. For example, it is assumed that the objective function obeys a Gaussian process (GP), where any value of the objective function is a random variable whose distribution is determined by the mean, variance or covariance of the GP.

[0152] Furthermore, the objective function is a proxy model constructed using a Gaussian process model. For example, a small number of sample points can be randomly sampled, and these small number of sample points are the initial parameter samples in the aforementioned embodiment; and an initial GP model is constructed based on the small number of sample points.

[0153] Based on the current observation data, a Gaussian process model proxy model is trained. The GP prediction mean and variance are then obtained. Based on these, the expected improvement (EI) of the candidate sample points is calculated; the candidate sample points are the candidate parameter samples in the aforementioned embodiment. By maximizing the EI, new sample points are sampled and their true functions are evaluated. Finally, the new sample points are added to the observation set. The new sample points are the target parameter samples in the aforementioned embodiment, and the observation set is the initial database containing the initial sample points.

[0154] Sub-step F4.2, select the acquisition function.

[0155] In Bayesian optimization, the acquisition function guides the sampling process, specifically balancing exploration and exploitation. Exploration aims to explore a wider range of locations away from known data points, while exploitation seeks for better solutions near known data points.

[0156] For example, the acquisition function may be Expected Improvement (EI), Upper Confidence Bound (UCB), etc. In this embodiment, EI is preferably used as the acquisition function for Bayesian optimization. In other words, the EI value is preferably used to characterize the prediction uncertainty of the sampling point.

[0157] Sub-step F4.3, calculating the scores of all unevaluated sampling points according to the acquisition function, and determining the sampling point with the highest score as the new sampling point.

[0158] Among them, the unevaluated sampling points are the candidate parameter samples in the aforementioned embodiment; the scores of the unevaluated points are the expected improvement values ​​corresponding to the candidate sampling points in the aforementioned embodiment, and the newly added sampling points are the target parameter samples in the aforementioned embodiment.

[0159] In this example, the point with the highest score is selected for objective function evaluation. The score can be used to characterize the uncertainty of the prediction result. In this example, based on the current Gaussian model's prediction of the objective function and the uncertainty evaluation, a new sampling point corresponding to the highest score is obtained. This new sampling point represents the point with the maximum prediction uncertainty for the reduced-order model's prediction of the heat exchanger temperature field.

[0160] Substep F4.4, Evaluate and Update the Model, evaluates the objective function at the selected new point, obtains a new observation, and then uses this new observation to update the probability model.

[0161] For example, what is updated in this step is usually the mean and variance of the GP.

[0162] Sub-step F4.5, repeat sub-steps F4.2 to F4.4 until the computational budget limit is met and the relevant parameter point with the largest heat exchange rate prediction error is used as the latest sampling point of the reduced-order model snapshot matrix.

[0163] In this step, the heat exchange rate prediction error can be characterized by the EI value.

[0164] Step F5: Calculate the simulation data of the heat exchanger temperature field distribution corresponding to the newly added sample points by using the finite element analysis method, and update the snapshot database based on the newly added sample points and the simulation data of the heat exchanger temperature field distribution corresponding to the newly added sample points.

[0165] Step F6, determine whether the leave-one-out prediction error is less than the target tolerance, or the number of samples reaches the sampling upper limit; if so, end; otherwise return to step F3.

[0166] In this step, repeat F3 to F5 until the leave-one-out prediction error e(μ i) is less than the target tolerance, or the number of samples reaches the upper sampling limit, sampling is terminated.

[0167] The samples in this step include initial sample points and newly added sample points.

[0168] For example, sampling is prioritized in high-error or high-sensitivity areas to significantly reduce redundant sampling points while more accurately capturing the nonlinear characteristics of the heat exchanger system, thereby constructing a more compact and efficient snapshot database. The snapshot database obtained based on this embodiment not only reduces the computational cost of model order reduction but also improves the predictive performance of the reduced-order model in complex scenarios, providing important technical support for the rapid analysis and optimization of complex engineering problems in automotive heat exchangers under complex operating conditions.

[0169] This embodiment effectively utilizes existing sample data to track parameter sample points with the greatest prediction uncertainty. This maximizes sampling efficiency and significantly reduces the generation of redundant data, meeting the stringent prediction accuracy requirements of the reduced-order model with a minimum number of sample points. The method of this embodiment exhibits high versatility and robustness, making it widely applicable to the rapid modeling and analysis of complex, nonlinear heat exchanger systems.

[0170] The method of this embodiment is further described below. The method of this embodiment may include the following steps:

[0171] Step N1: determine the parameter space and perform initial sampling.

[0172] For example, according to the design requirements of an automobile heat exchanger, key parameters are selected as input variables, including fluid inlet temperature, flow rate, fin thickness, and material thermal conductivity.

[0173] For example, the parameter space dimension is 4-dimensional, and each parameter is independent and evenly distributed.

[0174] Step N2: Generate initial sample points.

[0175] For example, the initial sample points can be generated by Python Design of Experiments (pyDOE) library.

[0176] Specifically, the initial sampling number k can be set and the initial sample points can be generated by the LHS method to ensure that the sample points are evenly distributed in the high-dimensional space and avoid clustering. After the sample points are generated, they are input into the commercial CFD software for steady-state thermal flow field simulation to obtain the temperature field data T corresponding to each sample point. i , construct an initial snapshot database based on the sampling points and temperature field data.,The temperature field data includes the heat exchanger outlet temperature.

[0177] In step N3, a reduced-order model is constructed based on the bagging method for the initial sampling points, and the prediction accuracy corresponding to the initial sampling points is obtained.

[0178] First, from the initial k samples, M subsets are generated by random sampling with replacement, where each subset is of size k. For example, the sampling process is implemented using a bootstrap method, for example, the Bootstrap class in the scikit-learn library, to ensure that each subset covers the diversity of the original data.

[0179] Secondly, for each sub-dataset, a low-dimensional model is constructed using POD combined with Galerkin projection.

[0180] Specifically, extract the temperature field snapshot matrix S∈R in the sub-dataset N×k , where N is the number of grid nodes. SVD is performed on S, and the top 10 main modes (with cumulative energy > 99%) are selected to construct the POD basis matrix Φ. Projection is used to reduce the high-dimensional equation to a 10-dimensional system of ordinary differential equations, which are then solved to obtain the reduced-order model coefficient matrix.

[0181] The leave-one-out cross-validation method is used to evaluate the prediction error of each sub-model. i , calculate the predicted temperature field when it is not included in the sub-dataset And the normalized error is calculated according to the following formula:

[0182]

[0183] The final model output is the mean of the prediction results of all sub-models to reduce variance and improve robustness.

[0184] Step N4: perform Bayesian optimization sampling based on the prediction accuracy.

[0185] Specifically, a Gaussian process model (GP) proxy model is established through the gp_minimize function of the scikit-optimize library.

[0186] In this step, the input data is the 4-dimensional parameter space defined in step S1, and the output data is the prediction error e(μ) of the reduced-order model.

[0187] The Matérn 5 / 2 kernel is selected as the covariance function to capture nonlinear characteristics; EI is used as the acquisition function to balance exploration and utilization.

[0188] Furthermore, the EI value of the unsampled area is calculated, and the point μ with the largest EI is selected. newAs a newly added sample point, this point corresponds to the area where the reduced-order model prediction uncertainty is the largest.

[0189] Step N5, μ new Input CFD software for simulation to obtain the real temperature field data T new and add it to the snapshot database.

[0190] Step N6, returning to step N3, until the number of target sampling points reaches a preset number threshold, or the prediction error of the target sampling point meets the preset error requirement.

[0191] Specifically, in this step, steps S3-S5 are repeated, and one sample point is added in each iteration until one of the following conditions is met: the LOO error e(μ i ) is less than the preset prediction error, which can be 0.01; the total number of target samples reaches the preset number threshold N max =100.

[0192] In one application, the snapshot database obtained based on this embodiment contains 100 samples, covering the key areas of the parameter space, and the proportion of redundant data is less than 5%.

[0193] refer to Figure 5 , an embodiment of the present application provides a device for acquiring parameter samples of a heat exchanger, the device 30 including: a first acquisition module 301, used to acquire N initial parameter samples, and obtain at least one sample set based on the N initial parameter samples; the initial parameter samples include performance parameters of the heat exchanger, and N is an integer greater than or equal to 2; a second acquisition module 302, used to acquire, for the nth initial parameter sample, a target sample set corresponding to the nth initial parameter sample from at least one sample set; the target sample set includes other initial parameter samples except the nth initial parameter sample; n∈[1, N]; a third acquisition module 303, used to acquire a prediction error for predicting the outlet temperature of the heat exchanger using the target sample set; a fourth acquisition module 304, used to determine a sample screening model based on the prediction errors between the initial parameter sample and the target sample set corresponding to the initial parameter sample, the sample screening model being used to screen parameter samples whose prediction uncertainty meets preset requirements; a fifth acquisition module 305, used to acquire at least one candidate parameter sample, and based on the sample screening model, acquire a target parameter sample from the at least one candidate parameter sample; the target parameter sample serves as at least part of the sample for training the outlet temperature prediction model.

[0194] Optionally, the device 30 also includes: a sixth acquisition module, which is used to cyclically execute the operation of obtaining at least one sample set based on N initial parameter samples, and obtaining a target parameter sample from at least one candidate parameter sample based on a sample screening model, until a preset loop termination condition is met, and the target parameter sample corresponding to each loop operation is obtained.

[0195] Optionally, the third acquisition module 303 includes: a first acquisition submodule, used to obtain the reduced-order model and the prediction error of the reduced-order model based on the target sample set and the outlet temperature simulation value corresponding to each initial parameter sample in the target sample set; the reduced-order model is used to predict the outlet temperature of the heat exchanger; and a first determination submodule, used to determine the prediction error of the reduced-order model as the prediction error for predicting the outlet temperature of the heat exchanger using the target sample set.

[0196] Optionally, the device 30 also includes: a seventh acquisition module, which is used to obtain the outlet temperature simulation value corresponding to the target parameter sample after obtaining the target parameter sample from at least one candidate parameter sample based on the sample screening model; an eighth acquisition module, which is used to construct a sample library based on the target parameter sample, the outlet temperature simulation value corresponding to the target parameter sample, the initial parameter sample, and the outlet temperature simulation value corresponding to the initial parameter sample; wherein, the target parameter sample, the outlet temperature simulation value corresponding to the target parameter sample, the initial parameter sample, and the outlet temperature simulation value corresponding to the initial parameter sample in the sample library are used to train the outlet temperature prediction model.

[0197] Optionally, when there are multiple target sample sets corresponding to the nth initial parameter sample, the fourth acquisition module includes: a second acquisition submodule, used to average the prediction errors of multiple target sample sets corresponding to the nth initial parameter sample to obtain a prediction error average value; a second determination submodule, used to determine the prediction error average value as the prediction error corresponding to the nth initial parameter sample; and a third determination submodule, used to determine a sample screening model based on at least one initial parameter sample and the prediction errors corresponding to each initial parameter sample in the at least one initial parameter sample.

[0198] Optionally, the fifth acquisition module 305 includes: a third acquisition sub-module, used to obtain the expected improvement value corresponding to each candidate parameter sample based on the sample screening model; the expected improvement value is used to characterize the prediction uncertainty corresponding to the candidate parameter sample; and a fourth determination sub-module, used to determine the candidate parameter sample corresponding to the preset expected improvement value as the target parameter sample.

[0199] Optionally, the first acquisition module 301 includes: a repeated execution sub-module, used to repeatedly execute the first operation to obtain at least one sample set; wherein the first operation includes: repeatedly executing the operation of randomly extracting initial parameter samples from N initial parameter samples until the total number of extracted initial parameter samples reaches a preset number; and constructing a sample set based on the preset number of extracted initial parameter samples.

[0200] In this embodiment, in this embodiment, the target sample set corresponding to the nth initial parameter sample does not include the nth initial parameter sample, and its prediction error is the leave-one-out prediction error of the nth initial parameter sample. The leave-one-out prediction error reflects the prediction error when the nth initial parameter sample is excluded, and can characterize the generalization ability of the outlet temperature prediction model used to predict the outlet temperature on a single initial parameter sample. According to the nth initial parameter sample and the prediction error of the target sample set corresponding to the nth initial parameter sample, a sample screening model is determined. The sample screening model can accurately identify the area where the prediction uncertainty meets the requirements. Based on the sample screening model, a target parameter sample where the prediction uncertainty meets the requirements can be obtained from the candidate parameter samples. In other words, the target parameter sample obtained in this embodiment can cover the area where the prediction uncertainty meets the requirements, which can solve the problem of the method of fixedly setting sampling points in the related art, that the sampling points do not fully cover the area. For example, it can solve the problem in the related art that the sampling points may not cover the area where the prediction uncertainty meets the requirements. In addition, the target parameter sample of this embodiment is a parameter sample adaptively determined from the candidate parameter samples based on a sample screening model constructed according to the initial parameter sample and the prediction error. The target parameter sample is not a fixed parameter sample, but a parameter sample adaptively determined based on the initial parameter sample. Based on this embodiment, the problem of obtaining sampling points through a fixed sampling method in related technologies, which may lead to redundant data in the parameter sample, is solved.

[0201] Figure 6 4 is a block diagram of an electronic device 400 according to an exemplary embodiment. For example, the electronic device 400 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0202] Reference Figure 6 , electronic device 400 may include one or more of the following components: a processing component 402 , a memory 404 , a power component 406 , a multimedia component 408 , an audio component 410 , an input / output (I / O) interface 412 , a sensor component 414 , and a communication component 416 .

[0203] The processing component 402 generally controls the overall operation of the electronic device 400, such as operations associated with display, phone calls, data communications, camera operation, and recording operations. The processing component 402 may include one or more processors 420 to execute instructions to perform all or part of the steps of the above-described method. In addition, the processing component 402 may include one or more modules to facilitate interaction between the processing component 402 and other components. For example, the processing component 402 may include a multimedia module to facilitate interaction between the multimedia component 408 and the processing component 402.

[0204] The memory 404 is used to store various types of data to support operations on the electronic device 400. Examples of such data include instructions for any application or method operating on the electronic device 400, contact data, phone book data, messages, pictures, multimedia, etc. The memory 404 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0205] The power supply assembly 404 provides power to the various components of the electronic device 400. The power supply assembly 406 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 400.

[0206] The multimedia component 408 includes a screen that provides an output interface between the electronic device 400 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensor can not only sense the boundaries of touch or slide actions, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 408 includes a front camera and / or a rear camera. When the electronic device 400 is in an operating mode, such as a shooting mode or a multimedia mode, the front camera and / or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have focal length and optical zoom capabilities.

[0207] The audio component 410 is used to output and / or input audio signals. For example, the audio component 410 includes a microphone (MIC) that is used to receive external audio signals when the electronic device 400 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal can be further stored in the memory 404 or transmitted via the communication component 416. In some embodiments, the audio component 410 also includes a speaker for outputting audio signals.

[0208] I / O interface 412 provides an interface between processing component 402 and peripheral interface modules, such as a keyboard, click wheel, buttons, etc. These buttons may include but are not limited to: a home button, volume buttons, a start button, and a lock button.

[0209] The sensor assembly 414 includes one or more sensors for providing various aspects of status assessment for the electronic device 400. For example, the sensor assembly 414 can detect the open / closed state of the electronic device 400, the relative positioning of components, such as the display and keypad of the electronic device 400. The sensor assembly 414 can also detect changes in the position of the electronic device 400 or a component of the electronic device 400, the presence or absence of user contact with the electronic device 400, the orientation or acceleration / deceleration of the electronic device 400, and temperature changes of the electronic device 400. The sensor assembly 414 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 414 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 414 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0210] The communication component 416 is used to facilitate wired or wireless communication between the electronic device 400 and other devices. The electronic device 400 can access a wireless network based on a communication standard, such as WiFi, an operator network (such as 2G, 3G, 4G or 5G), or a combination thereof. In an exemplary embodiment, the communication component 416 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 416 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0211] In an exemplary embodiment, the electronic device 400 can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to implement a method for obtaining heat exchanger parameter samples provided in an embodiment of the present application.

[0212] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 404 including instructions, which can be executed by a processor 420 of an electronic device 400 to perform the above method. For example, the non-transitory storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0213] Figure 7 1 is a block diagram of an electronic device 500 according to an exemplary embodiment. For example, the electronic device 500 may be provided as a server. Figure 7 Electronic device 500 includes a processing component 522, which further includes one or more processors, and memory resources represented by memory 532 for storing instructions executable by processing component 522, such as applications. The applications stored in memory 532 may include one or more modules, each corresponding to a set of instructions. Furthermore, processing component 522 is configured to execute instructions to perform a method for obtaining heat exchanger parameter samples provided in an embodiment of the present application.

[0214] The electronic device 500 may further include a power supply component 526 configured to perform power management of the electronic device 500, a wired or wireless network interface 550 configured to connect the electronic device 500 to a network, and an input / output (I / O) interface 558. The electronic device 500 may operate based on an operating system stored in the memory 532, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, or the like.

[0215] An embodiment of the present application further provides a computer program product, including a computer program, and a method for obtaining heat exchanger parameter samples implemented when the computer program is executed by a processor.

[0216] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the application disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of the present application and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0217] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A method for obtaining a heat exchanger parameter sample, characterized in that: include: Obtaining N initial parameter samples, and obtaining at least one sample set based on the N initial parameter samples; The initial parameter sample includes performance parameters of the heat exchanger, where N is an integer greater than or equal to 2; For the nth initial parameter sample, obtaining a target sample set corresponding to the nth initial parameter sample from at least one sample set; the target sample set includes other initial parameter samples except the nth initial parameter sample; n∈[1,N]; Obtaining a prediction error of heat exchanger outlet temperature prediction using the target sample set; Determining a sample screening model based on the initial parameter sample and the prediction error of the target sample set corresponding to the initial parameter sample, wherein the sample screening model is used to screen parameter samples whose prediction uncertainty meets preset requirements; At least one candidate parameter sample is obtained, and a model is screened based on the sample, and a target parameter sample is obtained from the at least one candidate parameter sample; the target parameter sample is used as at least part of the sample for training the outlet temperature prediction model.

2. The method according to claim 1, characterized in that The method further comprises: The operation of obtaining at least one sample set based on the N initial parameter samples, and obtaining a target parameter sample from at least one candidate parameter sample based on the sample screening model is executed cyclically until a preset loop termination condition is met, and the target parameter sample corresponding to each loop operation is obtained.

3. The method according to claim 1, characterized in that Each of the initial parameter samples in the target sample set has a corresponding outlet temperature simulation value, and obtaining a prediction error for heat exchanger outlet temperature prediction using the target sample set includes: Obtaining a reduced-order model and a prediction error of the reduced-order model based on the target sample set and the outlet temperature simulation value corresponding to each initial parameter sample in the target sample set; the reduced-order model is used to predict the outlet temperature of the heat exchanger; The prediction error of the reduced-order model is determined as the prediction error of heat exchanger outlet temperature prediction using the target sample set.

4. The method according to claim 3, characterized in that After obtaining a target parameter sample from at least one of the candidate parameter samples based on the sample screening model, the method further includes: Obtaining an outlet temperature simulation value corresponding to the target parameter sample; constructing a sample library according to the target parameter sample, the outlet temperature simulation value corresponding to the target parameter sample, the initial parameter sample, and the outlet temperature simulation value corresponding to the initial parameter sample; The target parameter samples in the sample library, the outlet temperature simulation values ​​corresponding to the target parameter samples, the initial parameter samples, and the outlet temperature simulation values ​​corresponding to the initial parameter samples are used to train the outlet temperature prediction model.

5. The method according to claim 1, wherein In the case where there are multiple target sample sets corresponding to the nth initial parameter sample, determining the sample screening model according to the prediction errors between the initial parameter sample and the target sample set corresponding to the initial parameter sample includes: The prediction errors of multiple target sample sets corresponding to the nth initial parameter sample are averaged to obtain the average prediction error; Determining the prediction error average as the prediction error corresponding to the nth initial parameter sample; A sample screening model is determined based on at least one of the initial parameter samples and the prediction error corresponding to each of the at least one initial parameter sample.

6. The method according to claim 1, characterized in that The obtaining of a target parameter sample from at least one candidate parameter sample based on the sample screening model includes: Based on the sample screening model, an expected improvement value corresponding to each candidate parameter sample is obtained; the expected improvement value is used to characterize the prediction uncertainty corresponding to the candidate parameter sample; The candidate parameter sample corresponding to the preset expected improvement value is determined as the target parameter sample.

7. The method according to claim 1, characterized in that The step of obtaining at least one sample set based on the N initial parameter samples includes: Repeat the first operation to obtain at least one sample set; The first operation includes: repeatedly performing the operation of randomly extracting initial parameter samples from N initial parameter samples until the total number of extracted initial parameter samples reaches a preset number, and constructing a sample set based on the preset number of extracted initial parameter samples.

8. A device for obtaining heat exchanger parameter samples, characterized in that: include: A first acquisition module is configured to acquire N initial parameter samples and obtain at least one sample set based on the N initial parameter samples; The initial parameter sample includes performance parameters of the heat exchanger, where N is an integer greater than or equal to 2; A second acquisition module is configured to acquire, for the nth initial parameter sample, a target sample set corresponding to the nth initial parameter sample from at least one sample set; the target sample set includes other initial parameter samples except the nth initial parameter sample; n∈[1,N]; a third acquisition module, configured to acquire a prediction error of the heat exchanger outlet temperature prediction using the target sample set; a fourth acquisition module, configured to determine a sample screening model based on the initial parameter sample and the prediction error of the target sample set corresponding to the initial parameter sample, wherein the sample screening model is configured to screen parameter samples whose prediction uncertainty meets a preset requirement; a fifth acquisition module, configured to acquire at least one candidate parameter sample, and acquire a target parameter sample from the at least one candidate parameter sample based on the sample screening model; The target parameter samples serve as at least part of the samples for training the outlet temperature prediction model.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A readable storage medium, characterized in that: The readable storage medium stores a program or instruction, and when the program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images