A Machine Learning-Based Performance Prediction Method for Micro Thermoelectric Cooling Devices

Abstract

This invention relates to a machine learning-based method for predicting the performance of micro thermoelectric cooling devices. This method constructs a performance prediction model by integrating a multi-dimensional set of coupled parameters (including geometric parameters, material parameters, and operating condition parameters) of the micro thermoelectric cooling device. First, historical data is collected and standardized, and then divided into training and testing sets. Machine learning algorithms are then used for model training and parameter optimization, ultimately achieving intelligent prediction of cooling performance. This invention focuses on the design process of micro thermoelectric cooling devices, solving the problem that traditional prediction methods struggle to predict the performance of micro thermoelectric cooling devices, requiring multiple experimental measurements and sample fabrication. This significantly improves design efficiency and reduces the cost of the verification process.

Description

Technical Field

[0001] This invention relates to the field of micro thermoelectric cooling device technology, specifically a method for predicting the performance of micro thermoelectric cooling devices based on machine learning. Background Technology

[0002] Micro thermoelectric cooling devices (Micro TECs) are widely used in precision temperature control fields such as optical communication lasers and lidar due to their small size, fast response, and pollution-free characteristics.

[0003] The parameters of micro thermoelectric cooling devices are complex, and their performance is affected by the coupling of multiple parameters such as geometric parameters, material parameters, and operating conditions. The nonlinear coupling mechanism between parameters further increases the difficulty of predicting the performance of micro thermoelectric cooling devices.

[0004] Currently, the main method for predicting the performance of micro thermoelectric cooling devices is numerical analysis. Numerical simulation requires precise solution of the thermoelectric coupling equation, resulting in high computational complexity and long design iteration cycles. If traditional experimental testing methods are still used, it is necessary to fabricate samples and conduct multiple experiments for measurement. Due to the numerous design parameters of micro thermoelectric cooling devices, it is difficult to comprehensively consider them during the design process, leading to performance failures to meet application requirements. Therefore, the design and manufacturing process of micro thermoelectric cooling devices heavily relies on engineers' design knowledge and requires multiple experimental iterations and optimizations to ensure performance meets application requirements, which is not only costly but also inefficient. Summary of the Invention

[0005] To address the problems existing in the prior art, the purpose of this invention is to provide a machine learning-based method for predicting the performance of micro thermoelectric cooling devices, which can improve the efficiency of performance verification in the early stages of micro thermoelectric cooling device design, shorten the design iteration cycle, and reduce the cost caused by repeated sample processing and experimental testing.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A machine learning-based method for predicting the performance of micro thermoelectric cooling devices includes the following steps: Obtain a multidimensional coupling parameter set for a micro thermoelectric refrigeration device, which includes input and output features; Based on a multidimensional coupling parameter set, a dataset containing the correspondence between input features and output features is constructed; The dataset is preprocessed, and the preprocessed dataset is divided into training set and test set; The selected machine learning model is trained using the training set to obtain a performance prediction model; The input characteristics of the micro thermoelectric refrigeration device to be predicted are input into the performance prediction model to obtain the prediction results of its performance parameters.

[0007] Furthermore, the input features in the multidimensional coupling parameter set include geometric parameters, material parameters, and operating condition parameters, while the output features are performance parameters.

[0008] Furthermore, the geometric parameters include the thermoelectric arm height h, cross-sectional area A, and number of thermoelectric arm pairs N; the material parameters include the Seebeck coefficient α, electrical conductivity σ, and thermal conductivity κ; and the operating parameters include the input current I and the hot-end temperature T. h Performance parameters include cooling capacity Q.

[0009] Furthermore, constructing the dataset includes: establishing a structured database, collecting and integrating historical data of input and output features to form a labeled dataset. ; in, , including h, A, N, α, σ, κ, I, T h Eight dimensions; This represents the corresponding output feature.

[0010] Furthermore, preprocessing includes: unifying all physical quantities in the dataset to the International System of Units (SI) and normalizing the data after unification.

[0011] Furthermore, the formula for normalization is: ; in, It is a factor The normalized value, and They are respectively After normalization, the maximum and minimum values ​​of the data are mapped to the range [0,1].

[0012] Furthermore, the ratio of training set to test set data volume is 7:3.

[0013] Furthermore, the training steps include: Select a machine learning model, define the parameter network of the machine learning model, perform cross-validation hyperparameter optimization through grid search to determine the optimal parameters, and train the performance prediction model using the optimal parameters and the training set.

[0014] Furthermore, the parameter network includes penalty coefficients, kernel function, kernel function coefficients, and slack variables.

[0015] Furthermore, the machine learning model is a support vector regression model.

[0016] In summary, the present invention has the following advantages: This invention addresses the challenge of numerous parameters in the design of micro thermoelectric cooling devices by employing a machine learning model to establish a mapping relationship between the multi-parameter coupling and performance parameters of micro thermoelectric cooling devices. This enables the prediction of the performance of micro thermoelectric cooling devices, replacing the repeated experimental process in traditional micro thermoelectric cooling device design. This significantly improves design efficiency and reduces the cost of the verification process. Attached Figure Description

[0017] Figure 1 This is a schematic diagram illustrating the steps of a machine learning-based method for predicting the performance of a micro thermoelectric refrigeration device in this embodiment. Figure 2 This embodiment shows the predicted performance of a micro thermoelectric refrigeration device in the training set, based on a machine learning-based performance prediction method for micro thermoelectric refrigeration devices. Figure 3 This embodiment describes the predicted performance of a micro thermoelectric refrigeration device in a test set, based on a machine learning-based performance prediction method for micro thermoelectric refrigeration devices. Detailed Implementation

[0018] The present invention will now be described in further detail.

[0019] like Figure 1 The diagram illustrates a specific implementation of a machine learning-based performance prediction method for micro thermoelectric refrigeration devices according to the present invention. This embodiment of the machine learning-based performance prediction method for micro thermoelectric refrigeration devices includes the following steps: S1. Obtain the multidimensional coupling parameter set of the micro thermoelectric refrigeration device. The multidimensional coupling parameter set includes input features and output features. S2. Based on the multidimensional coupling parameter set, construct a dataset containing the correspondence between input features and output features; S3. Preprocess the dataset and divide the preprocessed dataset into training and test sets; S4. Train the selected machine learning model using the training set to obtain the performance prediction model; S5. Input the input characteristics of the micro thermoelectric refrigeration device to be predicted into the performance prediction model to obtain the prediction results of its performance parameters.

[0020] Specifically, the performance prediction method for micro thermoelectric refrigeration devices based on machine learning in this embodiment collected 27 sample data of micro thermoelectric refrigeration devices.

[0021] This embodiment presents a machine learning-based method for predicting the performance of a micro thermoelectric cooling device. In step S1, the multidimensional coupling parameter set includes geometric parameters, material parameters, operating condition parameters, and performance parameters. The geometric parameters, material parameters, and operating condition parameters are input features, and the performance parameters are output features. The geometric parameters include thermoelectric arm height h, cross-sectional area A, and the number of thermoelectric arm logarithms N; the material parameters include Seebeck coefficient α, electrical conductivity σ, and thermal conductivity κ; the operating condition parameters include input current I and hot-end temperature T. h Performance parameters include cooling capacity Q.

[0022] This embodiment presents a machine learning-based method for predicting the performance of micro thermoelectric cooling devices. Step S2, dataset construction, includes establishing a structured database and collecting and integrating historical data of the input and output features to form a labeled dataset. ; in, , including h, A, N, α, σ, κ, I, T h Eight dimensions; This represents the corresponding output feature.

[0023] This embodiment presents a machine learning-based method for predicting the performance of micro thermoelectric cooling devices. Step S3, data preprocessing and partitioning, includes: (1) Unify the units of physical quantities to the International System of Units (SI); (2) The data is processed using a normalization method, namely: ; in, It is a factor The normalized value, and They are respectively After normalization, the maximum and minimum values ​​of the data are mapped to the range [0,1].

[0024] (3) The ratio of the amount of data in the training set to the amount of data in the test set is 7:3. After the split, the number of samples in the training set is 20 and the number of samples in the test set is 7.

[0025] This embodiment presents a machine learning-based method for predicting the performance of micro thermoelectric cooling devices. Step S4, model building and training, includes the following steps: (1) Create a model; (2) Define the parameter network: The parameter network includes penalty coefficients, kernel function, kernel function coefficients, and slack variables; (3) Optimize hyperparameters through cross-validation using a grid search method; (4) Determine the optimal parameters and use the optimal parameters and training set to train the performance prediction model.

[0026] In this embodiment, a machine learning-based method for predicting the performance of a micro thermoelectric refrigeration device is described. In step S4, the model construction and training adopts the following machine learning model: Support Vector Regression (SVR).

[0027] This embodiment provides a method for predicting the performance of a micro thermoelectric refrigeration device based on machine learning. Step S5 includes predicting the performance of the micro thermoelectric refrigeration device by using a trained machine learning SVR model.

[0028] The machine learning-based performance prediction method for micro thermoelectric cooling devices described in this embodiment has achieved good performance in practical applications. In the training set, such as... Figure 2 As shown, the model's true and predicted values ​​on the training set (20 samples) are highly consistent, with a root mean square error (RMSE) of only 0.046. The absolute deviation for over 85% of the sample points is below 0.05, and the maximum deviation is controlled within the range of 0.15. On the test set, as... Figure 3 As shown, in the independent test set (7 samples), the average deviation (RMSE) between the predicted results and the true values ​​was 0.328, of which 71.4% of the samples (5 / 7) had a deviation of less than 0.3, proving that the model has industrial-grade reliability in unseen data.

[0029] This invention systematically integrates three major categories of multidimensional coupled parameters—geometric parameters, material parameters, and operating parameters—of micro thermoelectric cooling devices. It then utilizes machine learning algorithms to establish a nonlinear mapping model from input features to output performance (cooling capacity), achieving holistic modeling and learning of the complex coupling relationships between multiple parameters. This invention organically combines the systematic extraction and structured construction of multidimensional parameters with the nonlinear fitting and generalization prediction capabilities of machine learning models. This allows performance evaluation, which previously required engineers' experience, multiple experimental iterations, and complex numerical simulations, to be completed in a very short time using a trained model. This collaborative end-to-end approach—data acquisition, feature engineering, model training, and intelligent prediction—not only significantly improves design efficiency and reduces trial-and-error costs but also solves the long-standing technical challenges in the field of micro thermoelectric cooling devices, such as numerous parameters, complex couplings, and difficulty in rapid prediction, demonstrating significant synergistic effects.

[0030] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method for predicting the performance of micro thermoelectric cooling devices based on machine learning, characterized in that, Includes the following steps: Obtain a multidimensional coupling parameter set for a micro thermoelectric refrigeration device, which includes input and output features; Based on a multidimensional coupling parameter set, a dataset containing the correspondence between input features and output features is constructed; The dataset is preprocessed, and the preprocessed dataset is divided into training set and test set; The selected machine learning model is trained using the training set to obtain a performance prediction model; The input characteristics of the micro thermoelectric refrigeration device to be predicted are input into the performance prediction model to obtain the prediction results of its performance parameters.

2. The method for predicting the performance of a micro thermoelectric refrigeration device according to claim 1, characterized in that: The input features of the multidimensional coupled parameter set include geometric parameters, material parameters, and operating condition parameters, while the output features are performance parameters.

3. The method for predicting the performance of a micro thermoelectric refrigeration device according to claim 2, characterized in that: The geometric parameters include a thermoelectric arm height h, a cross-sectional area A, and a thermoelectric arm number N; the material parameters include a Seebeck coefficient α, an electrical conductivity σ, and a thermal conductivity κ; the working condition parameters include an input current I and a hot end temperature T h ; and the performance parameter includes a refrigerating capacity Q.

4. The method for predicting the performance of a micro thermoelectric refrigeration device according to claim 3, characterized in that: Building the dataset includes: establishing a structured database, collecting and integrating historical data of input and output features to form a labeled dataset. ; in, , including h, A, N, α, σ, κ, I, T h Eight dimensions; This represents the corresponding output feature.

5. The method for predicting the performance of a micro thermoelectric refrigeration device according to claim 1, characterized in that: Preprocessing includes: All physical quantities in the dataset are standardized to the International System of Units (SI), and the data after standardization is then normalized.

6. The method for predicting the performance of a micro thermoelectric refrigeration device according to claim 5, characterized in that: The formula for normalization is: ; in, It is a factor The normalized value, and They are respectively After normalization, the maximum and minimum values ​​of the data are mapped to the range [0,1].

7. The method for predicting the performance of a micro thermoelectric refrigeration device according to claim 1, characterized in that: The ratio of training set to test set data volume is 7:

3.

8. The method for predicting the performance of a micro thermoelectric refrigeration device according to claim 1, characterized in that: The training steps include: Select a machine learning model, define the parameter network of the machine learning model, perform cross-validation hyperparameter optimization through grid search to determine the optimal parameters, and train the performance prediction model using the optimal parameters and the training set.

9. The method for predicting the performance of a micro thermoelectric refrigeration device according to claim 8, characterized in that: The parameter network includes penalty coefficients, kernel function, kernel function coefficients, and slack variables.

10. The method for predicting the performance of a micro thermoelectric refrigeration device according to claim 1, characterized in that: The machine learning model is a support vector regression model.

Images