High-efficiency heat dissipation management method and system for gallium nitride driver

Abstract

The present application relates to the technical field of gallium nitride drivers, and discloses an efficient heat dissipation management method and system for gallium nitride drivers, the method comprising: performing three-dimensional modeling and meshing on the gallium nitride driver to construct an initial heat dissipation model; performing feature extraction and transformation analysis on temperature field distribution data in the initial heat dissipation model to generate a temperature sensor layout scheme; collecting multi-operating condition temperature data according to the temperature sensor layout scheme, and performing local weighted regression analysis and thermal characteristic parameter extraction on the multi-operating condition temperature data to obtain characteristic parameters of the heat dissipation system; performing difference calculation on the characteristic parameters of the heat dissipation system and the initial heat dissipation model, and iteratively optimizing the heat dissipation parameters through a polynomial temperature regression method to obtain a target heat dissipation model; performing heat dissipation performance calculation and scheme evaluation based on the target heat dissipation model, and outputting a heat dissipation system optimization scheme, thereby realizing accurate updating of the heat dissipation model and improving the model prediction accuracy.

Description

Technical Field

[0001] The present application relates to the technical field of gallium nitride drivers, and in particular to a method and system for efficient heat dissipation management of gallium nitride drivers. Background Art

[0002] Under high-power operating conditions, the heating problem of the driver has become increasingly prominent, greatly affecting the reliability and service life of the device. Traditional heat dissipation management methods mainly rely on fixed heat dissipation models and empirical parameters, which are difficult to adapt to the dynamic heat dissipation requirements of GaN drivers under different working conditions.

[0003] At present, although the heat dissipation parameter identification technology of GaN drivers can determine the heat dissipation parameters through temperature sensor measurement data and heat conduction models, it faces problems such as measurement data uncertainty and nonlinear correlation in practical applications. These problems seriously affect the stability and accuracy of the heat dissipation parameter identification results, and it is impossible to provide an accurate heat dissipation management strategy for the driver. Summary of the invention

[0004] The present application provides a highly efficient heat dissipation management method and system for a gallium nitride driver, thereby achieving accurate updating of a heat dissipation model and improving model prediction accuracy.

[0005] In a first aspect, the present application provides a highly efficient heat dissipation management method for a gallium nitride driver, the highly efficient heat dissipation management method for a gallium nitride driver comprising: Perform 3D modeling and meshing of GaN drivers to build an initial heat dissipation model; Performing feature extraction and transformation analysis on the temperature field distribution data in the initial heat dissipation model to obtain temperature field monitoring parameters, and generating a temperature sensor layout plan based on the temperature field monitoring parameters; Collecting temperature data of multiple working conditions according to the temperature sensor arrangement scheme, and performing local weighted regression analysis and thermal characteristic parameter extraction on the temperature data of multiple working conditions to obtain characteristic parameters of the heat dissipation system; Calculating the difference between the characteristic parameters of the heat dissipation system and the initial heat dissipation model, and iteratively optimizing the heat dissipation parameters by a polynomial temperature regression method to obtain a target heat dissipation model; Based on the target heat dissipation model, heat dissipation performance calculation and solution evaluation are performed, and a heat dissipation system optimization solution is output.

[0006] A second aspect of the present application provides a high-efficiency heat dissipation management device for a gallium nitride driver, the high-efficiency heat dissipation management device for a gallium nitride driver comprising: Modeling module, used to perform 3D modeling and meshing of GaN drivers and build an initial heat dissipation model; An analysis module, used for performing feature extraction and transformation analysis on the temperature field distribution data in the initial heat dissipation model to obtain temperature field monitoring parameters, and generating a temperature sensor arrangement scheme based on the temperature field monitoring parameters; A collection module, used for collecting temperature data of multiple working conditions according to the temperature sensor arrangement scheme, and performing local weighted regression analysis and thermal characteristic parameter extraction on the temperature data of multiple working conditions to obtain characteristic parameters of the heat dissipation system; An optimization module, used for calculating the difference between the characteristic parameters of the heat dissipation system and the initial heat dissipation model, and iteratively optimizing the heat dissipation parameters by a polynomial temperature regression method to obtain a target heat dissipation model; The output module is used to calculate the heat dissipation performance and evaluate the solution based on the target heat dissipation model, and output the heat dissipation system optimization solution.

[0007] A third aspect of the present application provides an electronic device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the electronic device executes the above-mentioned efficient heat dissipation management method of the gallium nitride driver.

[0008] A fourth aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned efficient heat dissipation management method for a gallium nitride driver.

[0009] Compared with the prior art, the present application has the following beneficial effects: by combining the gallium nitride driver with the virtual heat dissipation unit to construct the initial heat dissipation model, and adopting a multi-level grid division strategy, the boundary prediction accuracy of the heat dissipation model is improved, making the model closer to the actual heat dissipation situation. The temperature field feature extraction method based on Copula function and MLDWT wavelet transform is combined with the IPCA-OPCE optimization algorithm to optimize the sensor layout, which solves the nonlinear correlation problem of temperature field monitoring data and ensures the reasonable distribution of temperature monitoring points. The local weighted LR-SVR algorithm is used to analyze the temperature data of multiple working conditions, and the thermal characteristic parameters are extracted by constructing a local regression model, which overcomes the limitations of traditional parameter identification methods in the data processing process. The polynomial temperature regression method is introduced to iteratively optimize the heat dissipation parameters. Through parameter sensitivity analysis and objective function optimization, the accurate update of the heat dissipation model is realized, and the model prediction accuracy is improved. The Kriging proxy model is used for parallel calculation of heat dissipation performance and scheme evaluation. The model parameters are continuously updated through the online learning mechanism, and the dynamic adjustment of the heat dissipation optimization scheme is realized, ensuring the optimization effect of the heat dissipation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0011] The structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantive technical significance. Any structural modification, change in proportion or adjustment of size, without affecting the effects and purposes that can be achieved by the present invention, should still fall within the scope of the technical contents disclosed by the present invention.

[0012] Figure 1 It is a flow chart of a highly efficient heat dissipation management method for a gallium nitride driver provided by an embodiment of the present invention; Figure 2 is a schematic block diagram of the structure of a high-efficiency heat dissipation management device for a gallium nitride driver provided by an embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0013] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0014] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0015] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0016] It should be further understood that the term "and / or" used in the specification and appended claims of this application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations. Figure 1 , an embodiment of the efficient heat dissipation management method of the gallium nitride driver in the embodiment of the present application includes: Step 100: Perform three-dimensional modeling and meshing on the gallium nitride driver to construct an initial heat dissipation model; It is understandable that the execution subject of the present application may be a high-efficiency heat dissipation management device for a gallium nitride driver, or a terminal or a server, which is not limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0017] Specifically, the device structure of the gallium nitride driver is three-dimensionally scanned and modeled to obtain its geometric data, and the mesh density of the geometric data is calculated to obtain the mesh parameter distribution, so as to ensure that in the subsequent simulation, accurate heat dissipation analysis can be performed according to the specific morphology and thermal conductivity characteristics of the device. Through the meshing process, the geometric structure of the gallium nitride driver is discretized into several small units. Based on the mesh parameter distribution, the gallium nitride driver is divided by a tetrahedral meshing method to form a basic mesh unit. Tetrahedral meshing can effectively adapt to complex geometric shapes and has high flexibility in calculation. Corresponding thermophysical parameters are assigned to each basic mesh unit to obtain a physical property mesh containing thermal conductivity. For the boundary area of ​​the physical property mesh, a virtual heat dissipation unit is configured to simulate the heat dissipation process at the boundary part of the physical property mesh. By configuring the virtual heat dissipation unit in the boundary area, a more accurate heat conduction analysis result is obtained by taking into account the heat exchange in the actual heat dissipation environment. The boundary mesh structure and the physical property mesh are topologically related to ensure that the heat transfer between the heat dissipation unit and the physical property mesh is coherent and accurate, and a hybrid mesh system is obtained. A heat source distribution model is constructed based on the hybrid mesh system. The heat source distribution model describes the location and distribution of heat sources inside the GaN driver, and provides heat source space data for heat dissipation analysis. The heat source distribution model is used to analyze how the heat generated inside the driver is distributed in different areas when it is working, and the power density of the heat source space data is calculated to obtain an accurate power distribution model to describe the heat distribution generated by the driver under different working conditions. The heat source nodes in the heat source model are mapped to the grid nodes in the hybrid grid system to obtain the corresponding relationship between the nodes. Through node mapping, the interaction between the heat source and the grid is clarified, and the thermal resistance matrix is ​​constructed. The thermal resistance matrix is ​​used to represent the path and difficulty of heat transfer from the heat source to the grid unit. The initial thermal resistance structure reveals the flow and distribution characteristics of heat between each unit by calculating the thermal resistance relationship between the heat source and the grid node. When analyzing the initial thermal resistance structure, the boundary heat exchange condition is set. The boundary heat exchange condition mainly considers the heat exchange process with the environment, including heat loss, radiation and other factors. By setting the boundary heat exchange conditions, the heat exchange behavior between the GaN driver and the external environment is simulated in the thermal analysis to obtain the boundary heat transfer model, which describes the heat transfer process between the driver surface and the external environment. After coupling the boundary heat transfer model with the power distribution model, a complete thermal control equation is obtained. The thermal control equation is the core mathematical model that describes the internal heat transfer process of the driver. It is used to calculate the temperature distribution under different working conditions. The temperature variation coefficient is configured for the thermal control equation to reflect the influence of temperature on the thermal conductivity of the material. The thermal conduction parameter model is combined with the initial thermal resistance structure to obtain the initial heat dissipation model, which includes multiple aspects such as temperature field distribution, heat source distribution and heat dissipation path.

[0018] Step 200: extract features and perform transformation analysis on the temperature field distribution data in the initial heat dissipation model to obtain temperature field monitoring parameters, and generate a temperature sensor layout plan based on the temperature field monitoring parameters; Specifically, the temperature field distribution data in the initial heat dissipation model is input into the Copula function for correlation analysis. The Copula function is a statistical method used to capture the correlation between different temperature variables, thereby constructing a temperature correlation matrix and revealing the interdependence of temperature changes between different regions. The temperature correlation matrix is ​​subjected to characteristic decomposition operation to obtain the main characteristic vector, and the most significant pattern and change trend in the temperature field distribution are identified. The temperature field distribution data is subjected to MLDWT (multi-layer discrete wavelet transform) multi-layer wavelet decomposition operation. By decomposing the temperature field distribution signal at multiple levels, the wavelet spectrum coefficients of different frequency components are obtained. By further processing the wavelet spectrum coefficients, the energy distribution is calculated to obtain the energy characteristic matrix, which reflects the energy distribution of different frequency components in the temperature field and reveals which areas have higher thermal energy density. The main characteristic vector of the temperature field and the energy characteristic matrix are feature fused to obtain the temperature field monitoring parameters, which help determine the law of temperature change and hot spots. Based on the temperature field monitoring parameters, an orthogonal polynomial function is constructed, which is used to describe the change characteristics of the temperature field, and a chaotic coefficient group is generated through the orthogonal polynomial function. The introduction of chaotic system is to simulate the complex dynamic behavior of temperature field. The chaotic coefficient group can capture the complexity and nonlinear characteristics in temperature distribution. The IPCA-OPCE optimization calculation is performed on the chaotic coefficient group to find the optimal solution in the multidimensional data space to obtain the temperature target monitoring point, which represents the temperature area that needs special attention in the heat dissipation system, usually those areas with large temperature changes or close to the thermal limit. For these monitoring points, spatial redundancy analysis is performed to ensure the reliability and accuracy of the monitoring system. In the redundancy analysis process, by calculating the redundancy, the redundant information between multiple monitoring points is identified, and unnecessary or duplicated monitoring points are removed to obtain an efficient monitoring point set with minimum redundancy. Sensor selection and matching is performed based on the monitoring point set. According to the selected monitoring point set, the temperature sensor is selected according to multiple factors such as temperature range, accuracy requirements, response time, etc., to ensure that the selected sensor can accurately capture the temperature changes of the target monitoring point and has sufficient sensitivity and stability. After the sensor is selected, the spatial layout calculation is performed. The position of the sensor is reasonably allocated in the design space of the GaN driver to ensure that each monitoring point can be effectively covered by the sensor and to avoid interference or repeated measurements between sensors. When calculating the spatial layout, factors such as the heat dissipation structure inside the driver, the location of the heat source, and the boundary conditions are taken into account to ensure that the layout of the temperature sensor is efficient and reliable in practical applications. A temperature sensor layout scheme based on the temperature field characteristics is obtained.

[0019] The chaotic coefficient group is compressed by principal component dimension, and the principal component coefficient matrix that best represents the data variation is extracted, thereby reducing the computational complexity and removing redundant information. The principal component coefficient matrix is ​​subjected to orthogonal transformation. The data is converted from the original space to an orthogonal space, so that each principal component is independent of each other, reducing the correlation, and obtaining an orthogonal feature matrix. The eigenvalues ​​of the orthogonal feature matrix are calculated to obtain an eigenvalue sequence, which reveals the importance of each principal component in the temperature field. The size of the eigenvalue directly reflects the contribution of each principal component to the temperature field. Based on the eigenvalue sequence, a covariance structure is constructed to obtain the temperature spatial distribution. The covariance structure reflects the mutual relationship and degree of variation between different regions in the temperature field. Based on the temperature spatial distribution, an orthogonal expansion calculation is performed to obtain a polynomial coefficient matrix. Through orthogonal expansion, the spatial distribution of the temperature field is represented at a higher level, which improves the accuracy of heat dissipation management. The polynomial coefficient matrix describes the changes in the temperature field in different dimensions. On this basis, a sensitivity analysis is performed on the polynomial coefficient matrix to obtain the temperature sensitivity distribution, which reveals the degree of influence of temperature changes in different regions on the overall heat dissipation performance. Based on the temperature sensitivity distribution, the temperature gradient field is calculated to obtain the gradient feature matrix. The gradient field reflects the rate and direction of temperature change, and can help identify areas with faster temperature changes, which are usually hot spots for heat dissipation. The temperature target monitoring points are obtained by searching for extreme values ​​of the gradient feature matrix. Extreme value search helps to find places where temperature changes drastically, which are the areas that need to be monitored the most, ensuring that overtemperature can be detected in time, so that effective heat dissipation measures can be taken. The temperature target monitoring points are mapped to the actual spatial coordinates to obtain a point distribution map, showing the distribution of each monitoring point in the physical space. By calculating the distance between these points, the point spacing data is obtained, which helps to evaluate whether the spacing between sensors is reasonable, ensure that each monitoring point can be effectively covered, and avoid redundant or overlapping monitoring areas. By calculating the point spacing, it is determined whether the position of the sensor needs to be adjusted to optimize the monitoring accuracy. Based on the point spacing data, redundant monitoring points are screened and deleted from the temperature target monitoring points to obtain the initial monitoring sequence. The initial monitoring sequence is analyzed for coverage to evaluate whether the sensor layout scheme can fully cover the hot spots in the entire temperature field to ensure that no key areas are missed. Through coverage analysis, the coverage index is obtained to evaluate the effectiveness of the monitoring system. According to the coverage index, the initial monitoring sequence is optimized and supplemented to ensure that the layout of monitoring points can cover all important areas while meeting the accuracy requirements and avoid redundant sensor layout as much as possible. The monitoring point set is obtained.

[0020] Step 300: collecting temperature data of multiple working conditions according to the temperature sensor arrangement scheme, and performing local weighted regression analysis and thermal characteristic parameter extraction on the temperature data of multiple working conditions to obtain characteristic parameters of the heat dissipation system; It should be noted that, according to the temperature sensor arrangement scheme, the temperature data sequence of the gallium nitride driver under different power conditions is collected. The temperature data sequence is filtered for noise, and irrelevant interference signals are removed to ensure the accuracy and reliability of the temperature data. After the noise is filtered, the time domain characteristics of the temperature are extracted by further processing the data to reveal the basic laws of temperature changes, including information such as temperature rise, fall, and fluctuation. The temperature time domain characteristics are input into the local weighted kernel function for processing in the high-dimensional data space. The local weighted kernel function is a nonlinear mapping method that can map the temperature time domain characteristics to a higher-dimensional space, in which the nonlinear relationship between the data is displayed. After mapping to the high-dimensional space, in order to improve the computational efficiency and accuracy of the model, the high-dimensional data space is divided into local regions, and the data is divided into multiple small subspaces. The data points in each subspace have stronger similarity physically or mathematically, thereby simplifying the subsequent regression analysis process. The local weight value is calculated for the data points in each subspace to reflect the importance of each data point in the local area. According to the local weight value, the weight parameter matrix is ​​obtained to describe the weight distribution of the data points in the local area. Based on the weight parameter matrix, the LR-SVR (local regression support vector regression) algorithm model is constructed. LR-SVR is a method that combines local regression and support vector machine. It can perform regression analysis in local areas, capture the nonlinear characteristics of data, and thus improve regression accuracy. Through the LR-SVR algorithm model, a local regression function is obtained, which can predict the thermal response of the cooling system based on the input temperature characteristic data. The local regression function is used to extract thermal characteristic data. The thermal characteristic data includes the thermal resistance and thermal capacitance data of the cooling system. The thermal resistance data group describes the resistance of the cooling path, and the thermal capacitance data group reflects the thermal storage capacity of the cooling system. By extracting these two sets of data, the relationship between temperature change and the cooling system is understood, providing a specific basis for system optimization. After combining the characteristics of the thermal resistance data group and the thermal capacitance data group, a comprehensive thermal parameter matrix is ​​obtained. The parameter correlation is calculated based on the thermal parameter matrix. By calculating the correlation between parameters such as thermal resistance and thermal capacitance, the mutual influence between different thermal characteristics is found. For example, higher thermal resistance in some areas will affect the effectiveness of thermal capacitance. By calculating the correlation coefficient, the potential bottlenecks and optimization space of the cooling system are revealed. The calculated phase relationship array is feature reconstructed and the correlation coefficient is mapped into a more operable form to improve its expressiveness and interpretability, thus obtaining the characteristic parameters of the cooling system.

[0021] Step 400, calculating the difference between the characteristic parameters of the heat dissipation system and the initial heat dissipation model, and iteratively optimizing the heat dissipation parameters by a polynomial temperature regression method to obtain a target heat dissipation model; Specifically, by constructing a parameter difference matrix between the characteristic parameters of the heat dissipation system and the initial heat dissipation model, the differences between the key parameters such as temperature field, heat flux density and thermal resistance in the actual heat dissipation system and the preliminary model are clarified, the deviation between the system prediction and the actual operation is revealed, and an error data group is formed to reflect the deviation of each parameter in the model. Based on the error data group, a temperature polynomial regression function is established. The regression function generates a set of basis function parameters by fitting the error data, thereby correcting the prediction results of the initial model to make it closer to the temperature performance of the actual heat dissipation system. The basis function parameters are optimized and calculated to obtain a regression coefficient group. The regression coefficient is refined by an iterative optimization method to reduce the error between the model prediction and the actual data. By optimizing the regression coefficient, the model's fit is effectively improved, making the model's prediction more accurate under different working conditions. The regression coefficient group and the error data group are weighted to form a weighted combination to obtain a model correction function. The model correction function can dynamically adjust the values ​​of each parameter in the initial heat dissipation model to make it more in line with the changes in the temperature field in actual applications and provide a more accurate heat dissipation management solution. Temperature prediction and error statistics are performed based on the correction function. Through temperature prediction, the temperature distribution under different conditions is calculated, and the difference between the current model and the actual cooling system is evaluated through error statistics. Through these predictions and statistical data, the temperature deviation data is obtained to reflect the difference between the model prediction and the actual temperature field. The temperature deviation data is gradient calculated to obtain the parameter sensitivity matrix. By analyzing the change trend of the temperature deviation, the parameters in the cooling system that are sensitive to temperature changes are found, and it is revealed which small changes in parameters will have a greater impact on the temperature prediction. Based on the parameter sensitivity matrix, an optimization function is constructed to obtain the objective function. The optimization function is a mathematical expression that combines the sensitivity of each parameter with the temperature prediction deviation to seek the optimal parameter correction. When solving the objective function, the optimization algorithm adjusts the model parameters according to the sensitivity and error data of each parameter until the temperature prediction error is minimized. The solution to the objective function obtained is the correction of the cooling parameter, indicating how the parameters in the initial model should be adjusted to make the prediction result more accurate. Based on the cooling parameter correction value, the model parameters of the initial cooling model are updated to obtain the target cooling model.

[0022] Step 500: Calculate heat dissipation performance and evaluate solutions based on the target heat dissipation model, and output a heat dissipation system optimization solution.

[0023] Specifically, the performance parameters of the target heat dissipation model are used to construct a feature vector space. The feature vector space is a mathematical representation of the performance of the heat dissipation system, in which each feature vector represents the performance index of the heat dissipation model under specific working conditions, such as heat flux density, temperature distribution, thermal resistance and other key factors. By converting these performance parameters into vectors, data analysis and pattern recognition are effectively performed. The obtained feature vector group contains the heat dissipation characteristics under all possible working conditions. By calculating the distance matrix of the feature vector group, the sample spacing data is obtained. The similarity between different samples is evaluated, and the spacing data reflects the degree of difference between different heat dissipation working conditions. Based on the sample spacing data, the cluster center points are initialized to obtain the initial cluster nodes. Cluster analysis is an unsupervised learning method that classifies similar samples into one category, thereby simplifying the complexity of the problem. Samples are classified on the initial cluster nodes to obtain preliminary classification results. According to the similarity between each sample and each cluster node, the sample is assigned to the closest cluster node. By analyzing these classification results, it is identified which cluster nodes represent similar heat dissipation modes and which nodes form the wrong category due to improper initial point selection. The coordinates of the preliminary clustering nodes are updated to obtain the optimized clustering nodes. The optimized clustering nodes will reassign the sample categories to ensure that each sample is assigned to the most appropriate category to obtain the final clustering data. The final clustering data is input into the Kriging proxy model for processing. The Kriging proxy model is a tool for regression analysis that can build a statistical model based on existing data and predict the performance of the cooling system under other working conditions by interpolation. By taking the final clustering data as input, the Kriging model can generate a cooling performance sample set that reflects the distribution of cooling performance under different working conditions. In order to improve the computational efficiency, the cooling performance sample set is parallelized to obtain the performance calculation results corresponding to each sample. Parallel computing can significantly shorten the time required for large-scale data calculation and improve the efficiency of the entire calculation process. The Kriging proxy model is trained based on the performance calculation results. The parameters of the Kriging model are adjusted by inputting a known cooling performance sample set so that it can more accurately predict the cooling performance under unknown working conditions. By continuously optimizing the model parameters, the Kriging proxy model can more accurately predict the performance of the cooling system under different design schemes. The trained Kriging model is used as a performance prediction model to provide a basis for evaluating the cooling system solution. Different cooling solutions are evaluated using the trained performance prediction model. The advantages and disadvantages of different design solutions in terms of cooling efficiency and thermal stability are compared to select the optimal cooling system design. By evaluating different solutions, the cooling configuration with the best performance under various working conditions is found to ensure the stability and efficiency of the GaN driver under different working conditions. Through calculation and analysis, the cooling system optimization solution is output.

[0024] In the embodiment of the present application, the initial heat dissipation model is constructed by combining the gallium nitride driver with the virtual heat dissipation unit, and a multi-level grid division strategy is adopted to improve the boundary prediction accuracy of the heat dissipation model, making the model closer to the actual heat dissipation situation. The temperature field feature extraction method based on Copula function and MLDWT wavelet transform is combined with the IPCA-OPCE optimization algorithm to optimize the sensor layout, which solves the nonlinear correlation problem of temperature field monitoring data and ensures the reasonable distribution of temperature monitoring points. The local weighted LR-SVR algorithm is used to analyze the temperature data of multiple working conditions, and the thermal characteristic parameters are extracted by constructing a local regression model, which overcomes the limitations of traditional parameter identification methods in the data processing process. The polynomial temperature regression method is introduced to iteratively optimize the heat dissipation parameters. Through parameter sensitivity analysis and objective function optimization, the accurate update of the heat dissipation model is realized, and the model prediction accuracy is improved. The Kriging proxy model is used for parallel calculation of heat dissipation performance and scheme evaluation. The model parameters are continuously updated through the online learning mechanism, and the dynamic adjustment of the heat dissipation optimization scheme is realized, ensuring the optimization effect of the heat dissipation system.

[0025] In a specific embodiment, the process of executing step 100 may specifically include the following steps: Perform three-dimensional scanning and modeling on the device structure of the GaN driver to obtain structural geometric data, and perform grid density calculation on the structural geometric data to obtain grid parameter distribution; Based on the grid parameter distribution, the GaN driver is tetrahedron-meshed to obtain basic grid units, and thermophysical parameters are assigned to the basic grid units to obtain a physical property grid including thermal conductivity. A virtual heat dissipation unit is configured on the boundary area of ​​the physical property grid to obtain a boundary grid structure, and a topological relationship is constructed between the boundary grid structure and the physical property grid to obtain a hybrid grid system; A heat source distribution model is constructed based on a hybrid grid system to obtain heat source spatial data, and power density is calculated on the heat source spatial data to obtain a power distribution model; Map the positions of the heat source nodes of the power distribution model to the grid nodes of the hybrid grid system to obtain the node correspondence, and construct a thermal resistance matrix based on the node correspondence to obtain the initial thermal resistance structure; The boundary heat exchange conditions are set for the initial thermal resistance structure to obtain the boundary heat transfer model, and the boundary heat transfer model is coupled with the power distribution model to obtain the heat conduction control equation; The temperature variation coefficient is configured for the heat conduction control equation to obtain a heat conduction parameter model, and the heat conduction parameter model is combined with the initial thermal resistance structure to obtain an initial heat dissipation model.

[0026] Specifically, the device structure of the gallium nitride driver is modeled by three-dimensional scanning. The geometric data of the driver device is obtained by using three-dimensional scanning technology, and the details of the driver's appearance and internal structure are captured by laser scanning or computer tomography. With these data, a highly accurate three-dimensional model is constructed to represent the spatial geometry of the gallium nitride driver. The grid density of the structural geometric data is calculated to determine the size and distribution of the grid units during the simulation process. Through reasonable distribution, the simulation calculation can maintain high accuracy without excessive calculation. The grid parameter distribution is set according to the geometric complexity of the device and the hot areas that need to be considered. The heat source area and the heat sink structure require finer grids, while other areas are divided by coarser grids. Tetrahedral meshing is performed based on the grid parameter distribution. Tetrahedral meshing is a three-dimensional meshing method that divides the entire driver model into multiple small tetrahedral units. Each small unit describes the physical field in three-dimensional space according to its shape, size and connection relationship between adjacent units. The boundaries of each tetrahedral unit are connected by common vertices or boundaries to form a grid composed of many tetrahedrons. For each basic grid unit, the corresponding thermophysical parameters such as thermal conductivity, specific heat capacity and density are assigned to it. Thermal conductivity represents the thermal conductivity of the material, while specific heat capacity is the ability of the material to absorb heat. By assigning these thermophysical parameters to each grid unit, a physical property grid containing thermal conductivity is formed. Virtual heat dissipation units are configured for the boundary areas of the physical property grid. The heat dissipation process of the driver occurs at its edge or in contact with the outside world, especially around the heat source area. The heat dissipation treatment of the boundary area requires special attention. The virtual heat dissipation units in these areas are used to simulate the heat exchange with the external environment and are processed using thermal conduction boundary conditions. After configuring the virtual heat dissipation units, the boundary grid structure and the physical property grid are topologically constructed to form a hybrid grid system. The hybrid grid system combines the physical property grid with the boundary heat dissipation unit, and ensures the interconnection and consistency of physical property transmission of each grid unit through topological relationships. In this way, the thermal characteristics of the GaN driver in different areas are described, providing a global grid system for heat dissipation simulation. Based on the hybrid grid system, a heat source distribution model is constructed to simulate the spatial distribution of the heat source inside the GaN driver. The distribution of the heat source is related to the current density and power input. By analyzing the spatial data of the heat source, the power density of the heat source is calculated, that is, the amount of heat generated per unit volume. The power density is expressed as ,in are spatial coordinates, is the power density. This model is implemented by considering the working conditions of each component of the GaN driver and combining it with the actual situation of current or power input. By calculating the heat source distribution, the power distribution model is obtained to reflect the distribution of heat sources in various parts of the driver. The heat source nodes of the power distribution model are mapped to the grid nodes of the hybrid grid system to obtain the node correspondence. The heat source data of the power distribution is combined with the grid nodes to clarify the heat transfer relationship between each heat source and the corresponding grid unit in the subsequent heat conduction calculation. The thermal resistance matrix is ​​constructed based on the node correspondence. The thermal resistance matrix is ​​an important component of the heat dissipation model, which describes the transfer resistance of heat flow between different substances. It means that the calculation formula is: ; in, is the path length of heat flow transfer, is the thermal conductivity of the material, is the cross-sectional area of ​​heat flow transfer. The initial thermal resistance structure is obtained by calculating the thermal resistance between each node. The boundary heat exchange conditions are set for the initial thermal resistance structure to establish a boundary heat transfer model. The boundary conditions include parameters such as the temperature and heat flux density of the heat exchange, which are used to simulate the heat exchange with the environment. The boundary conditions include constant temperature boundaries and constant heat flow boundaries. The boundary heat transfer model is coupled with the power distribution model to obtain the heat conduction control equation. The heat conduction control equation describes the temperature distribution of each point inside the driver under the conditions of known heat source distribution and thermal resistance. The heat conduction control equation is: ; in, is the thermal conductivity of the material, is the temperature, is the power density. The equation is solved according to different heat source distributions and physical parameters to obtain the distribution of the temperature field. The heat conduction parameter model is refined by configuring the temperature variation coefficient for the heat conduction control equation. The temperature variation coefficient represents the rate at which the temperature changes over time and is used in the dynamic heat conduction model. After combining the heat conduction parameter model with the initial thermal resistance structure, the initial heat dissipation model is obtained.

[0027] In a specific embodiment, the process of executing step 200 may specifically include the following steps: The temperature field distribution data of the initial heat dissipation model is input into the Copula function for correlation analysis to obtain a temperature correlation matrix, and the temperature correlation matrix is ​​subjected to eigendecomposition operation to obtain a main eigenvector; Perform MLDWT multi-layer wavelet decomposition operation on the temperature field distribution data to obtain the wavelet spectrum coefficients, and then calculate the energy distribution based on the wavelet spectrum coefficients to obtain the energy characteristic matrix; The main feature vector and the energy feature matrix are fused to obtain the temperature field monitoring parameters, and an orthogonal polynomial function is constructed based on the temperature field monitoring parameters to obtain a chaotic coefficient group; Perform IPCA-OPCE optimization calculation on the chaotic coefficient group to obtain the temperature target monitoring points, and perform spatial redundancy analysis on the temperature target monitoring points to obtain the monitoring point set; Sensor selection and matching is performed based on the monitoring point set to obtain a sensor parameter group, and the sensor parameter group and the monitoring point set are spatially arranged to obtain a temperature sensor arrangement plan.

[0028] Specifically, a correlation analysis is performed on the temperature field distribution data of the initial heat dissipation model to identify the heat conduction relationship between different areas. The temperature field data of the initial heat dissipation model is input into the Copula function for correlation analysis. The Copula function is a mathematical tool that can describe the correlation between different variables and is suitable for the joint distribution of multidimensional data. Through analysis, the temperature correlation matrix in the temperature field data is obtained, which reflects the correlation between each point in the temperature field. Temperature correlation matrix is a symmetric matrix where each element Indicates location and location The correlation between them is calculated as: ; in, Indicates temperature point and The covariance between and Temperature points and The standard deviation of . After correlation analysis through the Copula function, the temperature correlation matrix is ​​obtained, which can help understand the law of temperature distribution. MLDWT (Multi-layer Discrete Wavelet Transform) operation is performed on the temperature field distribution data. Wavelet transform is a signal processing method that can analyze different frequency components. Through multi-layer wavelet decomposition, the temperature field data is decomposed into components of different scales and frequencies. MLDWT is an extension of the traditional wavelet transform. It can decompose signals at different scale levels, thereby revealing the changing characteristics of the temperature field at different frequencies. When applying MLDWT to decompose the temperature field data, a series of wavelet spectrum coefficients are obtained. ,in represents the number of wavelet layers, Represents the frequency components at this level. These wavelet spectrum coefficients reflect the energy distribution of the temperature field at various scales. Based on these wavelet spectrum coefficients, the energy distribution is calculated to obtain the energy characteristic matrix , describing how the energy of the temperature field is distributed at different frequencies and scales. The main feature vector is fused with the energy feature matrix to obtain the temperature field monitoring parameters , this parameter set can fully describe the spatial characteristics of the temperature field and the heat conduction mode. Based on the temperature field monitoring parameters, an orthogonal polynomial function is constructed. , this function can transform the monitoring parameters into a more concise and regular expression form through orthogonal transformation to obtain a chaotic coefficient group. For example, the orthogonal polynomial is expressed as: ; in, are the coefficients of the polynomial, It is an orthogonal basis function constructed based on the temperature field monitoring parameters. In this way, the complex data of the temperature field can be converted into a more standardized mathematical expression. The chaotic coefficient group is optimized by IPCA-OPCE to obtain the temperature target monitoring point. The IPCA-OPCE (Iterative Principal Component Analysis-Optimized Principal Component Feature Extraction) method is used for calculation to find the best monitoring point distribution through iterative optimization. Through this method, the temperature target monitoring point set is calculated. , that is, those points in the temperature field that are most sensitive to temperature changes. These monitoring points are located in heat source areas or places where heat flow changes greatly. Spatial redundancy analysis is performed on the target monitoring points to determine whether there are repeated or redundant monitoring points. By calculating the distance between monitoring points, monitoring points with repeated information are identified. For example, if the distance between two monitoring points is less than a certain threshold, it is considered that their temperature monitoring functions overlap, so one of them is deleted. Through redundancy analysis, a set of optimized monitoring points is obtained. . Based on the set of monitoring points, sensor selection and matching are performed. The selection of sensors is determined based on parameters such as the location of the monitoring point, temperature range, and accuracy requirements. For example, if the monitoring point is located in a high temperature area, it is necessary to select a sensor that can withstand higher temperatures; if the monitoring point location is relatively narrow, it is necessary to select a miniaturized sensor. The sensor parameter group S includes information such as the sensor model, accuracy, and range. These parameters need to be matched with the requirements of the monitoring point. Based on the set of monitoring points and the sensor parameter group, spatial layout calculations are performed to determine the location of each sensor so that it can cover all important monitoring points. The layout calculation involves optimization algorithms, such as genetic algorithms or particle swarm optimization, through which an optimal sensor layout plan is obtained. The final output of the temperature sensor layout plan The location and related parameters of each sensor are shown.

[0029] In a specific embodiment, the execution step performs IPCA-OPCE optimization calculation on the chaotic coefficient group to obtain the temperature target monitoring point, and performs spatial redundancy analysis on the temperature target monitoring point to obtain the monitoring point set, which may specifically include the following steps: The chaotic coefficient group is compressed into principal component dimensions to obtain a principal component coefficient matrix, and the principal component coefficient matrix is ​​orthogonally transformed to obtain an orthogonal characteristic matrix; The eigenvalues ​​of the orthogonal eigenmatrix are calculated to obtain the eigenvalue sequence, and the covariance structure is constructed based on the eigenvalue sequence to obtain the temperature spatial distribution; Based on the temperature spatial distribution, an orthogonal expansion calculation is performed to obtain a polynomial coefficient matrix, and a sensitivity analysis is performed on the polynomial coefficient matrix to obtain a temperature sensitivity distribution; The temperature gradient field is calculated based on the temperature sensitivity distribution to obtain the gradient characteristic matrix, and the extreme value search of the gradient characteristic matrix is ​​performed to obtain the temperature target monitoring point; The temperature target monitoring points are mapped to spatial coordinates to obtain a point distribution map, and the distance of the point distribution map is calculated to obtain point spacing data; Based on the point spacing data, redundant monitoring points are screened and deleted from the temperature target monitoring points to obtain an initial monitoring sequence; The initial monitoring sequence is subjected to coverage analysis to obtain a coverage index, and the initial monitoring sequence is optimized and supplemented based on the coverage index to obtain a set of monitoring points.

[0030] Specifically, the chaotic coefficient group is compressed by the principal component dimension to better capture the main features in the temperature field and remove redundant data. The chaotic coefficient group is obtained by an orthogonal polynomial function, which contains complex temperature distribution information. In order to reduce the dimension of the data and highlight the main features, the principal component analysis method is used to compress the chaotic coefficient group. The high-dimensional data is projected into a lower-dimensional space through linear transformation to retain as much variance information as possible. In this process, the chaotic coefficient matrix is ​​constructed. , where each element Represents the correlation of chaotic coefficients under different variables. By performing eigenvalue decomposition on the matrix, the principal component coefficient matrix is ​​obtained: , where each column is a principal component eigenvector of the data. Each column of the matrix represents a principal component direction of the data, and the eigenvalue represents the importance of each principal component. Based on the principal component coefficient matrix , perform an orthogonal transformation, and project the original data into a new coordinate system so that each dimension is independent of each other. , is obtained through the orthogonalization process, satisfying relationship, among which is the unit matrix. After orthogonal transformation, the orthogonal characteristic matrix is ​​obtained Contains the data distribution after orthogonal transformation, in which there is no correlation between the features, making the data analysis more concise. Perform eigenvalue calculation to determine the importance of each feature to the temperature field distribution. Through eigenvalue decomposition, obtain the eigenvalue sequence , indicating the importance of different features in the temperature field. According to the size of the eigenvalue, the main heat source area and heat flow distribution are determined. The covariance matrix is ​​constructed based on these eigenvalues , which reflects the variance relationship between each point in the temperature field, and can help the system reveal the characteristics of the temperature spatial distribution. The calculation of the covariance structure is achieved through the following formula: ; in, Indicates temperature point and The covariance between them. The analysis of the covariance structure can reveal the correlation between different regions in the temperature field. Based on the spatial distribution of the temperature field, the orthogonal expansion calculation is performed to transform the complex temperature field distribution into a more concise expression. By mapping the temperature spatial distribution into the polynomial space, the polynomial coefficient matrix is ​​obtained. , where each column represents the expansion coefficient of the temperature field under different orders of polynomials. In order to ensure the accurate expression of the temperature field, a sensitivity analysis is performed on the polynomial coefficient matrix to evaluate the contribution of each coefficient to the change of the temperature field. The sensitivity is obtained by calculating the partial derivative of the coefficient matrix, for example: ; in, Represents temperature field For polynomial coefficients Through sensitivity analysis, the temperature sensitivity distribution matrix is ​​obtained , revealing the degree of response of each temperature point to different factors. Based on the temperature sensitivity distribution, the temperature gradient field is calculated. Temperature gradient field It refers to the rate and direction of temperature change, which is calculated by the spatial derivative of the temperature field data. The gradient feature matrix G represents the gradient information of each point in the temperature field, and its calculation formula is: ; By performing extreme value search on the gradient feature matrix, we can identify the areas with the most dramatic temperature changes. These areas are the key points for heat dissipation and temperature management. The purpose of extreme value search is to find the maximum and minimum values ​​in the temperature gradient field. These points represent the parts with the most dramatic temperature changes, i.e., the temperature target monitoring points. After mapping the temperature target monitoring points to spatial coordinates, we can obtain the point distribution map. , which shows the location of the selected monitoring points in space. In order to optimize the arrangement of monitoring points, the distance between these points is calculated to obtain the point spacing data , which represents the distance between two monitoring points. The distance calculation formula is: ; in, and Monitoring points and The three-dimensional spatial coordinates of the monitoring points. By calculating the point spacing data, we can identify which monitoring points are too close to each other and thus redundant. Based on the point spacing data, we can filter out redundant monitoring points. Redundant monitoring points are located in close proximity and can be identified by setting a distance threshold. , delete the monitoring points that are too close to obtain the initial monitoring sequence The initial monitoring sequence is analyzed for coverage to ensure that the selected monitoring points can cover the key areas of the entire temperature field. The coverage analysis is performed by calculating the size of the area covered by each monitoring point, and finally the coverage index is obtained. , which measures the coverage of monitoring points in space. Based on the coverage index, the initial monitoring sequence is optimized and supplemented to obtain the final set of monitoring points. .

[0031] In a specific embodiment, the process of executing step 300 may specifically include the following steps: Based on the temperature sensor layout scheme, the temperature data sequence of the GaN driver under different power conditions is collected, and the temperature data sequence is subjected to noise filtering and feature extraction to obtain the temperature time domain features; The temperature time domain features are input into the local weighted kernel function to obtain a high-dimensional data space, and the high-dimensional data space is divided into local regions to obtain a feature subspace; Calculate the local weight value for the data points in the feature subspace to obtain the weight parameter matrix, and build the LR-SVR algorithm model based on the weight parameter matrix to obtain the local regression function; Extract thermal characteristic data through local regression function to obtain thermal resistance data group and thermal capacitance data group, and perform feature combination on the thermal resistance data group and the thermal capacitance data group to obtain a thermal parameter matrix; The parameter correlation is calculated based on the thermal parameter matrix to obtain a phase relationship array, and the phase relationship array is feature reconstructed to obtain characteristic parameters of the heat dissipation system.

[0032] Specifically, based on the temperature sensor arrangement scheme, the temperature data of the GaN driver is collected under different power conditions. Noise filtering is performed on these data. Noise filtering methods include low-pass filter, Gaussian filter and wavelet transform, among which low-pass filter can effectively eliminate high-frequency noise. Input filtering algorithm to obtain a smooth temperature signal . The temperature data sequence is feature extracted to obtain the temperature time domain features. The temperature time domain features include the mean, variance, kurtosis, skewness and instantaneous changes in the time domain. Perform time domain feature extraction to obtain a set of time domain feature vectors that describe the temperature variation pattern. For example, the mean of the temperature data is calculated using the following formula: ; in, is the number of data points, It is The temperature value at each moment. Similarly, the variance of the temperature data is calculated using the following formula: ; The temperature time domain features are input into the local weighted kernel function for analysis. The local weighted kernel function is a technology used for high-dimensional data space construction and feature analysis. It uses a weighted function to locally model the data and retain the local features of the data. Measuring the similarity between sample points, using Gaussian kernel function: ; in, is the sample point and The Euclidean distance between is the bandwidth parameter of the kernel function. By inputting the temperature time domain features into the local weighted kernel function, a high-dimensional data space is obtained. The high-dimensional data space is divided into local regions to obtain the feature subspace. The division of the feature subspace is usually based on clustering algorithms or adaptive segmentation techniques. The divided feature subspace It can reflect the data distribution characteristics of different regions. For the data points in the feature subspace, their local weight values ​​are calculated. These weight values ​​reflect the importance of each data point in the local area. The calculation of the local weight value is realized by the weighted kernel function, and the local weight is calculated by the following formula: ; in, is a data point in the subspace, is the center point of the subspace, is the weighted kernel function. The weight parameter matrix obtained by calculation , which helps the system determine the contribution of each data point in the regression model. Based on the local weight parameter matrix , construct a local regression model, and use the local regression support vector regression (LR-SVR) algorithm. LR-SVR is a modeling method that combines local weighting and support vector regression, which can handle nonlinear problems. In the LR-SVR model, the input data points are nonlinearly mapped through the support vector machine, and the regression function is trained by minimizing the loss function. The optimization goal of the regression model is to minimize the following objective function: ; in, is the weight vector of the support vector machine, is the bias term, is the regularization parameter, is a relaxation variable used to process the error of data points. By training the LR-SVR model, a local regression function is obtained to predict and extract thermal characteristic data, such as thermal resistance data and thermal capacitance data. The thermal characteristic data extracted from the local regression function contains the thermal conductivity characteristics of the driver under different power conditions. These data include thermal resistance and heat capacity , these two parameters describe the material's resistance to heat flow and the material's ability to store heat. The calculation formulas for thermal resistance and thermal capacity are: ; in, is the temperature difference, is the heat flow; the heat capacity is given by: ; in, is the time interval. By combining the thermal resistance and thermal capacitance data, a matrix containing multiple thermal parameters is obtained. , describes the thermal characteristics of GaN drivers. Based on the thermal parameter matrix Calculate parameter correlation. Calculate correlation coefficient matrix , and its calculation formula is: ; in, and are the first and Thermal parameters, and is the standard deviation of the corresponding thermal parameter, is their covariance. By reconstructing the correlation coefficient matrix, the thermal parameters most relevant to the heat dissipation performance are screened out to obtain the final characteristic parameters of the heat dissipation system.

[0033] In a specific embodiment, the process of executing step 400 may specifically include the following steps: Constructing a parameter difference matrix between characteristic parameters of the heat dissipation system and the initial heat dissipation model to obtain an error data set, and establishing a temperature polynomial regression function based on the error data set to obtain basis function parameters; The basis function parameters are optimized and calculated to obtain a regression coefficient group, and the regression coefficient group is weightedly combined with the error data group to obtain a model correction function; Based on the model correction function, temperature prediction and error statistics are performed to obtain temperature deviation data, and the gradient of the temperature deviation data is calculated to obtain the parameter sensitivity matrix; An optimization function is constructed based on the parameter sensitivity matrix to obtain the objective function formula, and the objective function formula is solved to obtain the parameter correction amount to obtain the heat dissipation parameter correction value; The model parameters of the initial heat dissipation model are updated based on the heat dissipation parameter correction values ​​to obtain a target heat dissipation model.

[0034] Specifically, the characteristic parameters of the cooling system are compared with the parameters of the initial cooling model to construct a difference matrix. The difference matrix reflects the deviation between the characteristics of the actual operation of the cooling system and the initial model assumptions, which can help the system identify the deficiencies of the model. Suppose the characteristic parameters of the cooling system are , the parameters of the initial heat dissipation model are , then the difference matrix between them is calculated by the following formula: ; in, is the error data set, which contains the difference between each parameter in the actual cooling system and in the initial model. , the error is modeled by establishing a temperature polynomial regression function. The polynomial regression function can capture the relationship between temperature and various heat dissipation parameters. Suppose the predicted value of temperature is , the error data set is , construct a polynomial regression function: ; in, are the basis function parameters of the regression function, representing the impact of each error term on temperature change. These basis function parameters are solved by minimizing the error between the temperature prediction value and the actual measurement value. Minimize the objective function Using the least squares method: ; in, It is The actual temperature of the data point, is the predicted temperature calculated by the regression function. Get the basis function parameters After that, these basis function parameters are optimized. By adjusting these parameters, the error between the predicted temperature and the actual temperature is minimized. The optimization calculation is implemented by gradient descent or other optimization algorithms. The parameter values ​​are updated by calculating the gradient of the objective function: ; Update the parameters by iteratively , until the error is minimized. The update rule is: ; in, is the learning rate, which determines the step size of each update. and error data set The weighted combination is performed to obtain the model correction function. The correction function is used to correct the initial heat dissipation model so that it can better fit the behavior of the actual heat dissipation system. The formula for the weighted combination is expressed as: ; in, is a weighted coefficient that reflects the contribution of different error terms to the correction of the heat dissipation model. The correction function is used to adjust the parameters of the heat dissipation model according to the errors in the actual system, so that it can more accurately reflect the heat dissipation performance under actual working conditions. Based on the model correction function, temperature prediction is performed and errors are statistically analyzed. By applying the corrected model to different power conditions, new temperature prediction values ​​are obtained. For each predicted value , and the actual measured value The difference between , the calculation formula is: ; By comparing the temperature deviation data Perform gradient calculations to obtain a parameter sensitivity matrix that reflects the impact of heat dissipation parameter changes on temperature deviations and helps the system identify parameters that have a greater impact on temperature changes. The calculation formula is: ; in, is the temperature deviation Relative to heat dissipation parameters The gradient of When the temperature changes, the temperature deviation changes. Finally, an optimization function is constructed to optimize the parameters of the heat dissipation model. The goal of the optimization function is to minimize the temperature deviation of the heat dissipation system while taking into account the impact of changes in various parameters on the heat dissipation performance. The objective function is expressed as: ; in, is the regularization coefficient, which controls the influence of the heat dissipation parameter sensitivity on the optimization target. By solving the objective function, the parameter correction value is obtained , used to update the parameters in the initial heat dissipation model. Based on the heat dissipation parameter correction value , update the parameters of the initial heat dissipation model , get the target heat dissipation model The update formula is: ; Through the above calculations and optimizations, the target heat dissipation model is obtained, which can effectively predict the heat dissipation performance of the GaN driver under different working conditions.

[0035] In a specific embodiment, the process of executing step 500 may specifically include the following steps: The performance parameters of the target heat dissipation model are used to construct a feature vector space to obtain a feature vector group, and a distance matrix is ​​calculated for the feature vector group to obtain sample spacing data; Initialize the cluster center point based on the sample spacing data to obtain the initial cluster node, and classify the samples of the initial cluster node to obtain the initial classification result; Update the cluster center coordinates of the initial classification results to obtain optimized cluster nodes, and redistribute sample categories based on the optimized cluster nodes to obtain the final cluster data; The final clustering data is input into the Kriging proxy model to obtain a heat dissipation performance sample set, and the heat dissipation performance sample set is parallel calculated to obtain a performance calculation result; The Kriging proxy model is trained based on the performance calculation results to obtain a performance prediction model, which is then used to evaluate the solution and output a cooling system optimization solution.

[0036] Specifically, the performance parameter feature vector space of the target heat dissipation model is constructed. Assume that the performance parameters of the target heat dissipation model include a set of numerical values , each Represents a heat dissipation parameter, such as thermal resistance, thermal capacitance, power density, etc. Organize these performance parameters into a high-dimensional vector These vectors are used as feature vectors to represent different operating states of the cooling system. When constructing the feature vector space, the cooling parameter vectors under different working conditions are combined into a data matrix ,in is the sample size, is the dimension of each sample. By calculating the distance between these feature vectors, we get the distance matrix, which reflects the differences in cooling systems under different working conditions. The calculation formula uses the Euclidean distance metric: ; in, Represents the feature vector and The distance between and They are and The first feature vector By calculating the distances between all samples, we get the complete distance matrix . Initialize cluster centers based on the distance matrix. Clustering algorithms, such as K-means clustering, initialize cluster nodes by selecting samples with a longer distance as initial cluster centers. Suppose you choose Cluster Centers , and assign all samples to the cluster center closest to them. For each sample , by calculating its distance to all cluster centers, select the nearest cluster center To classify: ; At this point, the initial classification results are obtained, and the samples are divided into Classes, each class contains a set of similar heat dissipation parameter vectors. Update the cluster center coordinates for the initial classification results to obtain optimized cluster nodes. Calculate the mean vector of each class based on the current classification results, that is, the center of each class: ; in, It belongs to The sample set of the class, is the number of samples in this class, It is The optimized cluster center of the class. Reassign the sample categories based on the new cluster center, that is, recalculate the distance between each sample and each cluster center and assign it to the closest cluster center until the cluster center no longer changes. Input the final cluster data into the Kriging proxy model for heat dissipation performance evaluation. The Kriging model is an interpolation method based on spatial data. It predicts the performance of unknown areas by considering the spatial correlation between sample data. Input the heat dissipation performance sample set into the Kriging proxy model to predict the performance of the heat dissipation system under different power conditions. Assume is the final clustering dataset, and the Kriging model performs spatial interpolation using the following formula: ; in, is the predicted thermal performance, is the mean function, It is a test point and the covariance vector between known sample points, is the covariance matrix between known sample points, is the heat dissipation performance value of the known sample. By training the Kriging proxy model, the predicted performance data under different heat dissipation conditions are obtained. Parallel calculation is performed on the heat dissipation performance sample set to improve the calculation efficiency, especially when the sample set is large. The parallel calculation method divides the sample set into multiple subsets, performs Kriging predictions on each subset, and then merges the results. By performing parallel calculation on the heat dissipation performance sample set, the performance calculation results of the heat dissipation system under different conditions are obtained. Based on the performance calculation results, the Kriging proxy model is trained to obtain the performance prediction model of the heat dissipation system. The model predicts the performance of the heat dissipation system according to different working condition parameters and provides support for the optimal design of the heat dissipation system. The training process of the performance prediction model includes selecting a suitable covariance function, calculating the correlation between sample points, and optimizing the model parameters by minimizing the prediction error. The trained performance prediction model is used to evaluate the heat dissipation system scheme. By evaluating different design schemes, the optimal heat dissipation design scheme is found. For example, by changing the position, size or material of the radiator and inputting them into the performance prediction model, the corresponding heat dissipation performance prediction value is obtained, and then the pros and cons of the scheme are evaluated. The heat dissipation system optimization scheme is output to guide the design and improvement of the actual heat dissipation system.

[0037] The above describes the efficient heat dissipation management method of the gallium nitride driver in the embodiment of the present application. The following describes the efficient heat dissipation management device 10 of the gallium nitride driver in the embodiment of the present application. Figure 2 In the embodiment of the present application, an efficient heat dissipation management device 10 of a gallium nitride driver includes: A modeling module 11 is used to perform three-dimensional modeling and meshing of the gallium nitride driver and construct an initial heat dissipation model; The analysis module 12 is used to perform feature extraction and transformation analysis on the temperature field distribution data in the initial heat dissipation model to obtain temperature field monitoring parameters, and generate a temperature sensor layout plan based on the temperature field monitoring parameters; The acquisition module 13 is used to collect temperature data of multiple working conditions according to the temperature sensor arrangement scheme, and perform local weighted regression analysis and thermal characteristic parameter extraction on the temperature data of multiple working conditions to obtain characteristic parameters of the heat dissipation system; The optimization module 14 is used to calculate the difference between the characteristic parameters of the heat dissipation system and the initial heat dissipation model, and iteratively optimize the heat dissipation parameters through a polynomial temperature regression method to obtain a target heat dissipation model; The output module 15 is used to calculate the heat dissipation performance and evaluate the solution based on the target heat dissipation model, and output the heat dissipation system optimization solution.

[0038] Through the cooperation of the above components, the initial heat dissipation model is constructed by combining the GaN driver with the virtual heat dissipation unit, and a multi-level grid division strategy is adopted to improve the boundary prediction accuracy of the heat dissipation model, making the model closer to the actual heat dissipation situation. The temperature field feature extraction method based on Copula function and MLDWT wavelet transform is combined with the IPCA-OPCE optimization algorithm to optimize the sensor layout, which solves the nonlinear correlation problem of temperature field monitoring data and ensures the reasonable distribution of temperature monitoring points. The local weighted LR-SVR algorithm is used to analyze the temperature data of multiple working conditions, and the thermal characteristic parameters are extracted by constructing a local regression model, which overcomes the limitations of traditional parameter identification methods in the data processing process. The polynomial temperature regression method is introduced to iteratively optimize the heat dissipation parameters. Through parameter sensitivity analysis and objective function optimization, the accurate update of the heat dissipation model is achieved, and the model prediction accuracy is improved. The Kriging proxy model is used for parallel calculation of heat dissipation performance and scheme evaluation. The model parameters are continuously updated through the online learning mechanism, and the dynamic adjustment of the heat dissipation optimization scheme is achieved, ensuring the optimization effect of the heat dissipation system.

[0039] See also Figure 3 , Figure 3 This is a schematic block diagram of the structure of an electronic device 300 provided in an embodiment of the present application. The electronic device 300 includes a processor 301 and a memory 302. The processor 301 and the memory 302 are connected via a device bus 303, wherein the memory 302 may include a non-volatile storage medium and an internal memory.

[0040] The non-volatile storage medium can store a computer program. The computer program includes program instructions, and when the program instructions are executed by the processor 301, the processor 301 can execute any of the above-mentioned efficient heat dissipation management methods for the gallium nitride driver.

[0041] The processor 301 is used to provide computing and control capabilities to support the operation of the entire electronic device 300 .

[0042] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor 301, the processor 301 can execute any of the above-mentioned efficient heat dissipation management methods for the gallium nitride driver.

[0043] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a partial structure related to the present application scheme, and does not constitute a limitation on the electronic device 300 involved in the present application scheme. The specific electronic device 300 may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0044] It should be understood that the processor 301 may be a central processing unit (CPU), and the processor 301 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0045] It should be noted that those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the electronic device 300 described above can refer to the corresponding process of the efficient heat dissipation management method of the aforementioned gallium nitride driver, and will not be repeated here.

[0046] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by one or more processors, the one or more processors implement the efficient heat dissipation management method of the gallium nitride driver provided in the embodiment of the present application.

[0047] The computer-readable storage medium may be an internal storage unit of the electronic device 300 in the aforementioned embodiment, such as a hard disk or memory of the electronic device 300. The computer-readable storage medium may also be an external storage device of the electronic device 300, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped with the electronic device 300.

[0048] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0049] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable an electronic device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0050] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An efficient heat dissipation management method for a gallium nitride driver, characterized in that: The method comprises: Perform 3D modeling and meshing of GaN drivers to build an initial heat dissipation model; Performing feature extraction and transformation analysis on the temperature field distribution data in the initial heat dissipation model to obtain temperature field monitoring parameters, and generating a temperature sensor arrangement plan based on the temperature field monitoring parameters; Collecting temperature data of multiple working conditions according to the temperature sensor arrangement scheme, and performing local weighted regression analysis and thermal characteristic parameter extraction on the temperature data of multiple working conditions to obtain characteristic parameters of the heat dissipation system; Calculating the difference between the characteristic parameters of the heat dissipation system and the initial heat dissipation model, and iteratively optimizing the heat dissipation parameters by a polynomial temperature regression method to obtain a target heat dissipation model; Based on the target heat dissipation model, heat dissipation performance calculation and solution evaluation are performed, and a heat dissipation system optimization solution is output.

2. The efficient heat dissipation management method of the gallium nitride driver according to claim 1, characterized in that: The three-dimensional modeling and meshing of the gallium nitride driver are performed to construct an initial heat dissipation model, including: Performing three-dimensional scanning modeling on the device structure of the gallium nitride driver to obtain structural geometric data, and performing grid density calculation on the structural geometric data to obtain grid parameter distribution; Based on the grid parameter distribution, the gallium nitride driver is tetrahedron-meshed to obtain basic grid units, and thermophysical parameters are assigned to the basic grid units to obtain a physical property grid including thermal conductivity; Performing virtual heat dissipation unit configuration on the boundary area of ​​the physical property grid to obtain a boundary grid structure, and constructing a topological relationship between the boundary grid structure and the physical property grid to obtain a hybrid grid system; Building a heat source distribution model based on the hybrid grid system to obtain heat source space data, and performing power density calculation on the heat source space data to obtain a power distribution model; Position mapping is performed on the heat source nodes of the power distribution model and the grid nodes of the hybrid grid system to obtain a node correspondence relationship, and a thermal resistance matrix is ​​constructed based on the node correspondence relationship to obtain an initial thermal resistance structure; Setting boundary heat exchange conditions for the initial thermal resistance structure to obtain a boundary heat transfer model, and coupling analysis of the boundary heat transfer model with the power distribution model to obtain a heat conduction control equation; The temperature variation coefficient is configured for the heat conduction control equation to obtain a heat conduction parameter model, and the heat conduction parameter model is combined with the initial thermal resistance structure to obtain an initial heat dissipation model.

3. The efficient heat dissipation management method of a gallium nitride driver according to claim 2, characterized in that: The feature extraction and transformation analysis of the temperature field distribution data in the initial heat dissipation model are performed to obtain temperature field monitoring parameters, and a temperature sensor arrangement scheme is generated based on the temperature field monitoring parameters, including: Inputting the temperature field distribution data of the initial heat dissipation model into the Copula function for correlation analysis to obtain a temperature correlation matrix, and performing eigendecomposition operation on the temperature correlation matrix to obtain a main eigenvector; Performing MLDWT multi-layer wavelet decomposition operation on the temperature field distribution data to obtain wavelet spectrum coefficients, and performing energy distribution calculation based on the wavelet spectrum coefficients to obtain an energy characteristic matrix; The main feature vector and the energy feature matrix are subjected to feature fusion to obtain temperature field monitoring parameters, and an orthogonal polynomial function is constructed based on the temperature field monitoring parameters to obtain a chaotic coefficient group; Performing IPCA-OPCE optimization calculation on the chaotic coefficient group to obtain temperature target monitoring points, and performing spatial redundancy analysis on the temperature target monitoring points to obtain a monitoring point set; Sensor selection and matching is performed based on the monitoring point set to obtain a sensor parameter group, and spatial arrangement calculation is performed on the sensor parameter group and the monitoring point set to obtain a temperature sensor arrangement plan.

4. The efficient heat dissipation management method of a gallium nitride driver according to claim 3, characterized in that: The IPCA-OPCE optimization calculation is performed on the chaotic coefficient group to obtain the temperature target monitoring point, and the spatial redundancy analysis is performed on the temperature target monitoring point to obtain the monitoring point set, including: The chaotic coefficient group is subjected to principal component dimension compression to obtain a principal component coefficient matrix, and the principal component coefficient matrix is ​​subjected to orthogonal transformation processing to obtain an orthogonal characteristic matrix; Calculating eigenvalues ​​of the orthogonal eigenmatrix to obtain an eigenvalue sequence, and constructing a covariance structure based on the eigenvalue sequence to obtain a temperature spatial distribution; Performing orthogonal expansion calculation based on the temperature spatial distribution to obtain a polynomial coefficient matrix, and performing sensitivity analysis on the polynomial coefficient matrix to obtain a temperature sensitivity distribution; Calculating the temperature gradient field based on the temperature sensitivity distribution to obtain a gradient characteristic matrix, and performing extreme value search on the gradient characteristic matrix to obtain a temperature target monitoring point; Mapping the temperature target monitoring points to spatial coordinates to obtain a point distribution map, and performing distance calculation on the point distribution map to obtain point spacing data; Filtering redundant monitoring points based on the point spacing data, and deleting the redundant monitoring points from the temperature target monitoring points to obtain an initial monitoring sequence; A coverage analysis is performed on the initial monitoring sequence to obtain a coverage index, and the initial monitoring sequence is optimized and supplemented based on the coverage index to obtain a monitoring point set.

5. The efficient heat dissipation management method of a gallium nitride driver according to claim 4, characterized in that: The method of collecting temperature data of multiple working conditions according to the temperature sensor arrangement scheme, and performing local weighted regression analysis and thermal characteristic parameter extraction on the temperature data of multiple working conditions to obtain characteristic parameters of the heat dissipation system includes: Based on the temperature sensor arrangement scheme, a temperature data sequence of the gallium nitride driver under different power conditions is collected, and noise filtering and feature extraction are performed on the temperature data sequence to obtain a temperature time domain feature; Inputting the temperature time domain feature into a local weighted kernel function to obtain a high-dimensional data space, and performing local area division on the high-dimensional data space to obtain a feature subspace; Calculating local weight values ​​for the data points in the feature subspace to obtain a weight parameter matrix, and constructing an LR-SVR algorithm model based on the weight parameter matrix to obtain a local regression function; Extract thermal characteristic data through the local regression function to obtain a thermal resistance data group and a thermal capacitance data group, and perform feature combination on the thermal resistance data group and the thermal capacitance data group to obtain a thermal parameter matrix; The parameter correlation is calculated based on the thermal parameter matrix to obtain a correlation array, and the characteristics of the correlation array are reconstructed to obtain characteristic parameters of the heat dissipation system.

6. The efficient heat dissipation management method of a gallium nitride driver according to claim 5, characterized in that: The difference calculation between the characteristic parameters of the heat dissipation system and the initial heat dissipation model is performed, and the heat dissipation parameters are iteratively optimized by a polynomial temperature regression method to obtain a target heat dissipation model, including: Constructing a parameter difference matrix between the characteristic parameters of the heat dissipation system and the initial heat dissipation model to obtain an error data group, and establishing a temperature polynomial regression function based on the error data group to obtain basis function parameters; Optimizing the basis function parameters to obtain a regression coefficient group, and weightedly combining the regression coefficient group with the error data group to obtain a model correction function; Performing temperature prediction and error statistics based on the model correction function to obtain temperature deviation data, and performing gradient calculation on the temperature deviation data to obtain a parameter sensitivity matrix; An optimization function is constructed based on the parameter sensitivity matrix to obtain an objective function formula, and the objective function formula is solved to obtain a parameter correction amount to obtain a heat dissipation parameter correction value; The model parameters of the initial heat dissipation model are updated based on the heat dissipation parameter correction value to obtain a target heat dissipation model.

7. The efficient heat dissipation management method of a gallium nitride driver according to claim 6, characterized in that: The heat dissipation performance calculation and solution evaluation based on the target heat dissipation model and output of the heat dissipation system optimization solution include: Constructing a feature vector space with the performance parameters of the target heat dissipation model to obtain a feature vector group, and calculating a distance matrix for the feature vector group to obtain sample spacing data; Initializing cluster center points based on the sample spacing data to obtain initial cluster nodes, and classifying samples on the initial cluster nodes to obtain initial classification results; Updating the cluster center coordinates of the initial classification results to obtain optimized cluster nodes, and reallocating sample categories based on the optimized cluster nodes to obtain final cluster data; Inputting the final clustering data into a Kriging proxy model to obtain a heat dissipation performance sample set, and performing parallel calculation on the heat dissipation performance sample set to obtain a performance calculation result; The Kriging proxy model is trained based on the performance calculation results to obtain a performance prediction model, and the performance prediction model is used to evaluate the solution and output a cooling system optimization solution.

8. An efficient heat dissipation management device for a gallium nitride driver, characterized in that: The method for efficiently managing heat dissipation of a gallium nitride driver according to any one of claims 1 to 7 is used, wherein the device for efficiently managing heat dissipation of the gallium nitride driver comprises: Modeling module, used to perform 3D modeling and meshing of GaN drivers and build an initial heat dissipation model; An analysis module, used for performing feature extraction and transformation analysis on the temperature field distribution data in the initial heat dissipation model to obtain temperature field monitoring parameters, and generating a temperature sensor arrangement scheme based on the temperature field monitoring parameters; A collection module, used for collecting temperature data of multiple working conditions according to the temperature sensor arrangement scheme, and performing local weighted regression analysis and thermal characteristic parameter extraction on the temperature data of multiple working conditions to obtain characteristic parameters of the heat dissipation system; An optimization module, used for calculating the difference between the characteristic parameters of the heat dissipation system and the initial heat dissipation model, and iteratively optimizing the heat dissipation parameters by a polynomial temperature regression method to obtain a target heat dissipation model; The output module is used to calculate the heat dissipation performance and evaluate the solution based on the target heat dissipation model, and output the heat dissipation system optimization solution.

9. An electronic device, characterized in that: The electronic device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the electronic device to execute the efficient heat dissipation management method for the gallium nitride driver according to any one of claims 1 to 7.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the efficient heat dissipation management method of the gallium nitride driver according to any one of claims 1 to 7 is implemented.

Images