High performance energy transfer control method for gallium nitride power modules

Abstract

The application relates to the technical field of intelligent power management, and particularly discloses an efficient energy transmission control method for a gallium nitride power module, which comprises the following steps: analyzing a temperature difference effective value exceeding a preset gradient threshold value through a classification algorithm to obtain a position coordinate of an overheating point; processing temperature data of the overheating point and adjacent collection points by using a discretized heat conduction equation to obtain a heat diffusion trend vector; evaluating a temperature rise risk score of the overheating point according to the heat diffusion trend vector, obtaining the overheating point according to the temperature rise risk score, and taking the overheating point and the adjacent collection points as a high-risk area; and re-distributing the power of all chips in the power module with the purpose of reducing the power of gallium nitride chips in the high-risk area. The efficient energy transmission control method for the gallium nitride power module solves the problem that current power modules are difficult to prevent device failure caused by local overheating according to the characteristics of gallium nitride chips.

Description

Technical Field

[0001] This invention relates to the field of intelligent power management technology, and more specifically, to a high-efficiency energy transfer control method for gallium nitride power modules. Background Technology

[0002] Gallium nitride (GaN) charger power modules, as core components of next-generation fast-charging devices, are becoming a key technology driving the rapid development of consumer electronics, power tools, and new energy vehicle charging due to their significant advantages of high frequency, high efficiency, and miniaturization. Their efficient energy transfer capabilities directly determine charging speed, device size, and long-term reliability. Current GaN charger power modules typically integrate multiple chips, including multiple high-voltage GaN power switches, rectifier control chips, buck converters, and MCUs, among others.

[0003] When gallium nitride (GaN) chips, such as high-voltage GaN power switches, operate at high frequencies and high currents, their on-resistance and switching losses increase significantly with rising local temperatures. This increased temperature further amplifies these losses, creating a vicious cycle. This strong thermo-electric coupling characteristic causes the effective temperature difference between different locations on the GaN chip surface to widen rapidly, resulting in extremely steep temperature gradients in hotspot areas. Traditional control methods that uniformly measure the temperature of the entire power module cannot capture these rapidly changing local thermal distribution characteristics, leading to delayed or misjudgments of truly dangerous hotspots. Due to the excessively large and varied temperature gradients at hotspots, the system struggles to accurately determine within milliseconds which specific area is becoming the source of thermal runaway, nor can it predict which adjacent areas will rapidly diffuse heat. This local thermal imbalance is further exacerbated by uneven current distribution in power modules with multiple GaN chips connected in parallel or operating in a multi-phase manner. This causes some GaN chips to experience prolonged overstress and premature aging, or even sudden failure, making it difficult for the power module to achieve efficient energy transfer. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, this invention provides a high-efficiency energy transfer control method for gallium nitride power modules, aiming to solve the problems mentioned above in the prior art.

[0005] The technical solution adopted by this invention to solve its technical problem is: a high-efficiency energy transfer control method for gallium nitride power modules, comprising the following steps:

[0006] S1: Collect temperature data from multiple sampling points on the surface of the power module to obtain the temperature distribution matrix on the surface of the power module;

[0007] S2: Calculate the effective value of the temperature difference between adjacent sampling points in the temperature distribution matrix, and combine all the effective values ​​of temperature difference to generate a temperature spatial distribution feature matrix;

[0008] S3: If there is a valid temperature difference value in the temperature spatial distribution feature matrix that exceeds the preset gradient threshold, then the valid temperature difference value exceeding the preset gradient threshold is analyzed by a classification algorithm to obtain the coordinates of the hot spot location;

[0009] S4: The temperature data of the hot spot and its adjacent collection points are processed by the discretized heat conduction equation to obtain the heat diffusion trend vector;

[0010] S5: Evaluate the temperature rise risk score of the hot spot based on the heat diffusion trend vector, filter out the hot spots corresponding to the temperature rise risk scores that exceed the preset risk threshold, and regard the hot spots and the adjacent collection points as high-risk areas.

[0011] S6: With the aim of reducing the power of gallium nitride chips in high-risk areas, the power of all chips in the power module is reallocated, and the updated power of each chip is used as the output configuration of the power module.

[0012] Preferably, in step S1, temperature data is collected by a sensor array with multiple acquisition points deployed on the surface of the gallium nitride power module to generate a temperature distribution matrix on the surface of the power module.

[0013] Optionally, in step S1, the step of generating the temperature distribution matrix on the surface of the power module includes:

[0014] For the temperature data, the sensor array is mapped into a two-dimensional matrix according to its physical arrangement on the surface of the gallium nitride power module to form a temperature distribution matrix;

[0015] The temperature data in the temperature distribution matrix is ​​represented by T1(i,j), where (i,j) represents the physical arrangement position of the sensor array on the surface of the gallium nitride power module, i represents the i-th row of the sensor array, j represents the j-th column of the sensor array, and i and j are positive integers.

[0016] Specifically, in step S2, temperature data between adjacent sampling points are obtained from the temperature distribution matrix, and the initial temperature difference of the current sampling point is determined by the difference calculation method of subtracting the temperature data of the previous sampling point from the temperature data of the current sampling point; the Euclidean norm of all the effective values ​​of the initial temperature difference of the current sampling point is calculated to obtain the effective value of the temperature difference.

[0017] For the effective value of the temperature difference, the sensor array is mapped into a two-dimensional matrix according to its physical arrangement on the surface of the gallium nitride power module to obtain the temperature spatial distribution feature matrix.

[0018] The effective value of the temperature difference in the temperature spatial distribution feature matrix is ​​represented by T2(i,j), and the physical arrangement position of the sensor array on the surface of the gallium nitride power module is represented by (i,j), where i represents the i-th row of the sensor array, j represents the j-th column of the sensor array, i is a positive integer, and j is a positive integer.

[0019] It is worth noting that in step S3, the effective value of the temperature difference in the temperature spatial distribution feature matrix is ​​compared with a preset gradient threshold. If the effective value of the temperature difference is greater than the preset gradient threshold, the sampling point corresponding to the effective value of the temperature difference is marked as an out-of-limit location point.

[0020] The effective temperature difference values ​​corresponding to the out-of-limit locations are extracted, and the effective temperature difference values ​​are processed using the K-means clustering algorithm. The number of clusters is set to two, and the cluster labels and cluster center values ​​are obtained by fitting the data.

[0021] Based on the cluster center value, the cluster with the highest value is marked as overheated;

[0022] Obtain the corresponding overheating location points from the overheating classification, map them to the physical arrangement position (i,j) of the sensor array on the power module surface, and determine the physical arrangement position (i,j) as the coordinates of the overheating hot spot.

[0023] Preferably, in step S4, temperature data of adjacent collection points are obtained based on the coordinates of the hot spot;

[0024] The coordinates of the hot spot, the temperature data of the hot spot, the coordinates of the sampling points adjacent to the hot spot, and the temperature data of the sampling points adjacent to the hot spot are combined to generate a complete temperature dataset containing the coordinates of the hot spot, the temperature data of the hot spot, the coordinates of the sampling points adjacent to the hot spot, and the temperature data of the sampling points adjacent to the hot spot.

[0025] For the complete temperature dataset, a temperature grid array was created using NumPy;

[0026] Fill the temperature grid array with initial temperature values ​​from the complete temperature dataset;

[0027] For the temperature grid array, iterative temperature update calculations are performed to obtain a new temperature grid array, wherein the temperature update calculation is as follows: ,in Indicates the new temperature value. The old temperature value is represented by dt, the time step by dt, and the thermal conductivity by k. c represents density, and c represents specific heat capacity. express Spatial second derivative along the x-direction, express The spatial second derivative along the y-direction is obtained by using the NumPy finite difference method for each temperature in the temperature grid array. Calculate the spatial second derivatives along the x-direction and along the y-direction;

[0028] The NumPy finite difference method is used to analyze each temperature in the temperature grid array. Calculate the spatial second derivatives along the x-direction and along the y-direction;

[0029] Temperature The spatial second derivative along the x-direction multiplied by the thermal conductivity is used as the diffusion component of that temperature along the x-direction. The temperature in the temperature grid array is then... The spatial second derivative along the y-direction multiplied by the thermal conductivity is taken as the diffusion component of that temperature along the y-direction.

[0030] The heat diffusion trend vector is obtained by fusing the diffusion direction components along the x-direction and the diffusion direction components along the y-direction using SciPy vector addition.

[0031] Optionally, in step S5, the magnitude of the heat diffusion trend vector is calculated as the temperature change amplitude, and the temperature rise risk score of each hot spot is calculated based on the temperature change amplitude, wherein the formula for the temperature rise risk score is S=ΔTw, where S represents the risk score, ΔT represents the temperature change amplitude, and w represents the preset weight.

[0032] For the risk score of temperature rise, a preset risk threshold is used for comparison. If the score exceeds the preset risk threshold, the corresponding hot spot is selected as a high-risk point.

[0033] High-risk points and their adjacent collection points are integrated into high-risk areas, and all high-risk areas are combined to generate a high-risk list.

[0034] Specifically, in step S6, the current power value of each gallium nitride chip in the high-risk area is extracted and averaged to obtain the high-risk average power; the high-risk average power is calculated by subtracting a preset power threshold to obtain the difference, and then multiplied by the number of gallium nitride chips in the high-risk area to obtain the remaining power share.

[0035] Each chip outside the high-risk area and each non-gallium nitride chip within the high-risk area are considered as non-high-risk chips. The current power value of each non-high-risk chip is obtained, and the sum of the current power values ​​of all non-high-risk chips is calculated as the non-high-risk power sum.

[0036] The first power allocation weight for each non-high-risk chip is determined by the ratio of its current power value to the total power of all non-high-risk chips. The power increase value for each non-high-risk chip is obtained by multiplying the remaining power share by the first power allocation weight. The power adjustment value for each non-high-risk chip is obtained by adding the current power value to the power increase value.

[0037] The sum of the current power values ​​of all gallium nitride chips in the high-risk area is calculated as the power sum of the first high-risk area; the ratio of the current power value of each gallium nitride chip in the high-risk area to the power sum of the first high-risk area is used as the second power allocation weight of the corresponding gallium nitride chip.

[0038] The difference between the current total output power of the power module and the sum of the power adjustment values ​​of all non-high-risk chips is used as the power sum of the second high-risk area. The power adjustment value of the corresponding gallium nitride chip in the high-risk area is obtained by multiplying the power sum of the second high-risk area by the second power allocation weight.

[0039] The power adjustment values ​​of chips in non-high-risk areas and the power adjustment values ​​of gallium nitride chips in high-risk areas are used as the output configuration of the power module.

[0040] The beneficial effects of this invention are as follows: In the high-efficiency energy transfer control method for the gallium nitride power module, a temperature distribution matrix is ​​formed by collecting multi-point temperature data on the surface of the power module, and then a temperature spatial distribution feature matrix composed of the effective values ​​of the temperature difference between adjacent collection points is constructed. When there is an effective value of temperature difference exceeding a preset gradient threshold, a classification algorithm is used to accurately locate the coordinates of the hot spot. Subsequently, the temperature data of the hot spot and its adjacent points are processed based on the discretized heat conduction equation to obtain a heat diffusion trend vector, and the temperature rise risk score of each hot spot is calculated accordingly, and high-risk areas exceeding the risk threshold are screened out. With the goal of significantly reducing the power of gallium nitride chips in high-risk areas, the power of all chips in the power module is redistributed to form an optimized power configuration output. This solution realizes a complete closed-loop logic from intelligent sensing of abnormal temperature gradients, accurate location of hot spots, quantitative assessment of heat diffusion trends to active power regulation in high-risk areas, effectively preventing device failure caused by local overheating and improving the operational reliability and lifespan of the power module. Attached Figure Description

[0041] Figure 1 This is a flowchart of a high-efficiency power transfer control method for gallium nitride power modules.

[0042] Figure 2 This is a flowchart of the steps in step S2.

[0043] Figure 3 This is a flowchart of the steps in step S3.

[0044] Figure 4 This is a flowchart of the steps in step S5. Detailed Implementation

[0045] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be noted that these descriptions are for the purpose of aiding understanding the present invention, but do not constitute a limitation thereof. Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0046] Combination Figures 1 to 4 The method for efficient power transfer control of a gallium nitride power module, as shown, includes the following steps:

[0047] S1: Collect temperature data from multiple sampling points on the surface of the power module to obtain the temperature distribution matrix on the surface of the power module;

[0048] S2: Calculate the effective value of the temperature difference between adjacent sampling points in the temperature distribution matrix, and combine all the effective values ​​of temperature difference to generate a temperature spatial distribution feature matrix;

[0049] S3: If there is a valid temperature difference value in the temperature spatial distribution feature matrix that exceeds the preset gradient threshold, then the valid temperature difference value exceeding the preset gradient threshold is analyzed by a classification algorithm to obtain the coordinates of the hot spot location;

[0050] S4: The temperature data of the hot spot and its adjacent collection points are processed by the discretized heat conduction equation to obtain the heat diffusion trend vector;

[0051] S5: Evaluate the temperature rise risk score of the hot spot based on the heat diffusion trend vector, filter out the hot spots corresponding to the temperature rise risk scores that exceed the preset risk threshold, and regard the hot spots and the adjacent collection points as high-risk areas.

[0052] S6: With the aim of reducing the power of gallium nitride chips in high-risk areas, the power of all chips in the power module is reallocated, and the updated power of each chip is used as the output configuration of the power module.

[0053] In the efficient energy transfer control method for the gallium nitride power module, a temperature distribution matrix is ​​formed by collecting multi-point temperature data on the surface of the power module. This matrix is ​​then used to construct a temperature spatial distribution feature matrix composed of the effective temperature differences between adjacent collection points. When an effective temperature difference exceeds a preset gradient threshold, a classification algorithm is used to accurately locate the coordinates of the hot spot. Subsequently, the temperature data of the hot spot and its adjacent points are processed based on the discretized heat conduction equation to obtain a heat diffusion trend vector. Based on this vector, a temperature rise risk score for each hot spot is calculated, and high-risk areas exceeding the risk threshold are selected. With the goal of significantly reducing the power of the gallium nitride chips in the high-risk areas, the power of all chips within the power module is redistributed to form an optimized power configuration output. This solution realizes a complete closed-loop logic from intelligent sensing of abnormal temperature gradients, accurate location of hot spots, quantitative assessment of heat diffusion trends, to proactive power regulation in high-risk areas. This effectively prevents device failure caused by localized overheating and improves the operational reliability and lifespan of the power module.

[0054] It is worth noting that in step S1, temperature data is collected by a sensor array with multiple acquisition points deployed on the surface of the gallium nitride power module to generate a temperature distribution matrix on the surface of the power module.

[0055] Preferably, in step S1, the step of generating the temperature distribution matrix on the surface of the power module includes:

[0056] For the temperature data, the sensor array is mapped into a two-dimensional matrix according to its physical arrangement on the surface of the gallium nitride power module to form a temperature distribution matrix;

[0057] The temperature data in the temperature distribution matrix is ​​represented by T1(i,j), where (i,j) represents the physical arrangement position of the sensor array on the surface of the gallium nitride power module, i represents the i-th row of the sensor array, and j represents the j-th column of the sensor array.

[0058] For example, by uniformly arranging a 4x4 sensor array on the surface of the power module, with each sensor corresponding to a sampling point, these sampling points can accurately capture the heat changes of the power module, thereby monitoring the temperature gradient in real time.

[0059] Specifically, when mapping these temperature data to a two-dimensional matrix, the matrix elements are corresponding to the physical arrangement of the sensor array on the surface of the gallium nitride power module. For example, the first row and first column of the sensor array corresponds to matrix T1(1,1), with a value of 85℃, and the first row and second column is T1(1,2)=92℃. This process is repeated to fill the entire matrix, forming a 4x4 temperature distribution matrix. This matrix not only visually represents the heat map but also facilitates subsequent analysis of thermal non-uniformity.

[0060] Optionally, in step S2, temperature data between adjacent sampling points are obtained from the temperature distribution matrix, and the initial temperature difference of the current sampling point is determined by the difference calculation method of subtracting the temperature data of the previous sampling point from the temperature data of the current sampling point; the Euclidean norm of all the effective values ​​of the initial temperature difference of the current sampling point is calculated to obtain the effective value of the temperature difference.

[0061] For the effective value of the temperature difference, the sensor array is mapped into a two-dimensional matrix according to its physical arrangement on the surface of the gallium nitride power module to obtain the temperature spatial distribution feature matrix.

[0062] The effective value of the temperature difference in the temperature spatial distribution feature matrix is ​​represented by T2(i,j), and the physical arrangement position of the sensor array on the surface of the gallium nitride power module is represented by (i,j), where i represents the i-th row of the sensor array, j represents the j-th column of the sensor array, i is a positive integer, and j is a positive integer.

[0063] In one possible implementation, when extracting temperature data of adjacent acquisition points from the temperature distribution matrix, it is based on the physical arrangement of the sensor array. For example, for an acquisition point T1(2,2), its adjacent previous acquisition points could be T1(2,1) and T1(1,2).

[0064] Specifically, the initial temperature difference is calculated by subtracting the temperature data from the previous acquisition point from the current acquisition point's temperature data. This method highlights changes in the thermal gradient. For example, for adjacent points within a row, subtracting T1(1,2) from T1(2,2) yields a difference of 3℃, and subtracting T1(2,2) from T1(2,1) yields a difference of 4℃. This reflects the heat propagation process from point T1(1,2) to point T1(2,2), and from point T1(2,1) to point T1(2,2). Without calculating these differences, potential faults caused by thermal inhomogeneity might be overlooked, such as the accumulation of thermal stress in the high-voltage gallium nitride power switches inside the power module. Then, the calculation... 5℃ is obtained, which is the effective value of the temperature difference used to comprehensively reflect the temperature difference between the two directions to T1(2,2).

[0065] The mapping process for the effective temperature difference value requires constructing a two-dimensional matrix according to the physical location of the sensor array. For example, the calculated effective temperature difference value is placed into T2(i,j), where (i,j) corresponds to the row and column positions of the array, i and j are positive integers, and both i and j are greater than 0. This matrix can comprehensively capture the spatial thermal distribution characteristics. In practice, this helps optimize the thermal management of the charging system and avoid efficiency degradation caused by overheating. The resulting T2 matrix after mapping is such as T2(1,1)=0 (as the starting point), T2(1,2)=5℃, T2(2,1)=3℃, etc. This matrix is ​​visualized as a thermal gradient map to help engineers obtain temperature distribution data. In particular, for T2(1,1) when i=1 and j=1, since the sampling point (1,1) is the starting point, there is actually no previous sampling point. In this embodiment, the previous sampling point of the starting point (1,1) is assigned a fixed value of 0. The handling of the previous acquisition point in special cases such as (1,1) is to ensure the integrity of the matrix and the consistency of the calculation. If the previous acquisition point is not assigned a fixed value of 0, it may cause the subsequent difference chain to break, affecting the overall feature analysis. In addition, for T2(i,j) with i=1 or j=1, there is only one initial temperature difference, so this initial temperature difference can be directly used as the effective value of the temperature difference of this acquisition point.

[0066] Specifically, in step S3, the effective value of the temperature difference in the temperature spatial distribution feature matrix is ​​compared with a preset gradient threshold. If the effective value of the temperature difference is greater than the preset gradient threshold, the sampling point corresponding to the effective value of the temperature difference is marked as an out-of-limit location point.

[0067] The effective temperature difference values ​​corresponding to the out-of-limit locations are extracted, and the effective temperature difference values ​​are processed using the K-means clustering algorithm. The number of clusters is set to two, and the cluster labels and cluster center values ​​are obtained by fitting the data.

[0068] Based on the cluster center value, the cluster with the highest value is marked as overheated;

[0069] Obtain the corresponding overheating location points from the overheating classification, map them to the physical arrangement position (i,j) of the sensor array on the power module surface, and determine the physical arrangement position (i,j) as the coordinates of the overheating hot spot.

[0070] In one possible implementation, the process of comparing the effective value of the temperature difference in the temperature spatial distribution feature matrix with a preset gradient threshold can begin by setting the threshold to, for example, 8°C. This threshold is based on empirical data from power modules operating under high loads. For instance, in gallium nitride power modules of electric vehicle charging piles, if an effective value of a temperature difference exceeds this threshold, it indicates an abnormal thermal gradient, possibly stemming from localized heat accumulation during high-current conversion. This comparison is achieved by traversing each element in the temperature spatial distribution feature matrix. For example, assuming matrix T2 contains element T2(2,3)=9°C, and the threshold is 8°C, the sampling point corresponding to this element is marked as an out-of-limit location. This marking can be represented using simple Boolean labels, such as setting the location to 1 to indicate an out-of-limit condition, thereby highlighting potential areas of thermal non-uniformity.

[0071] In one possible implementation, after extracting the effective temperature difference values ​​corresponding to the out-of-limit locations, the K-means clustering algorithm is used for processing. It's important to understand that K-means is an unsupervised learning method that iteratively optimizes data points to assign them to specified clusters. Here, two clusters are set to divide the effective temperature difference values ​​into high and low categories. Specifically, the K-means algorithm randomly initializes two cluster centers, calculates the distance from each data point to the center, assigns the point to the nearest cluster, updates the center to the mean of the points within the cluster, and repeats this process until convergence. In practical applications, this is used to extract features from the effective temperature difference values ​​corresponding to the marked out-of-limit locations. For example, if the effective values ​​of the out-of-limit temperature difference are 9℃, 11℃, 10℃, and 13℃, and the number of clusters is set to two, the algorithm might cluster 9℃ and 10℃ into one cluster with a center of approximately 9.5℃, while 11℃ and 13℃ might cluster into another with a center of approximately 12℃. This process arises from the need to distinguish between mild and severe out-of-limit temperatures, and results in more accurate identification of overheating risks. Through fitting, cluster labels help classify the data points. In one possible implementation, when labeling the cluster with the highest value as overheating based on its cluster center value, the values ​​of the two centers can be directly compared. For example, if one center is 9.5℃ and the other is 12℃, then the cluster with the 12℃ value is labeled as overheating, reflecting that areas with higher effective temperature difference values ​​are more likely to correspond to overheating problems. Specifically, this labeling is based on the statistical characteristics of the clustering results. In the thermal management of power modules, it helps to specifically address groups with higher cluster center values.

[0072] In one possible implementation, when obtaining the corresponding overheating location points from the overheating classification and mapping them to the physical arrangement positions of the sensor array, it is necessary to trace the cluster labels back to the original matrix coordinates. For example, the effective temperature difference values ​​in the overheating cluster correspond to T2(3,2) and T2(4,3), and these (i,j) are the physical arrangement positions such as row 3 column 2 and row 4 column 2, thus determining the coordinates of the overheating hot spot.

[0073] It is worth noting that in step S4, temperature data of adjacent collection points are obtained based on the coordinates of the hot spot.

[0074] The coordinates of the hot spot, the temperature data of the hot spot, the coordinates of the sampling points adjacent to the hot spot, and the temperature data of the sampling points adjacent to the hot spot are combined to generate a complete temperature dataset containing the coordinates of the hot spot, the temperature data of the hot spot, the coordinates of the sampling points adjacent to the hot spot, and the temperature data of the sampling points adjacent to the hot spot.

[0075] For the complete temperature dataset, a temperature grid array was created using NumPy;

[0076] Fill the temperature grid array with initial temperature values ​​from the complete temperature dataset;

[0077] For the temperature grid array, iterative temperature update calculations are performed to obtain a new temperature grid array, wherein the temperature update calculation is as follows: ,in Indicates the new temperature value. The old temperature value is represented by dt, the time step by dt, and the thermal conductivity by k. c represents density, and c represents specific heat capacity. express Spatial second derivative along the x-direction, express The spatial second derivative along the y-direction; where the x-direction is consistent with the row direction of the temperature spatial distribution feature matrix or the temperature distribution matrix, and the y-direction is consistent with the column direction of the temperature spatial distribution feature matrix or the temperature distribution matrix, and the NumPy finite difference method is used for each temperature in the temperature grid array. Calculate the spatial second derivatives along the x-direction and along the y-direction;

[0078] The NumPy finite difference method is used to analyze each temperature in the temperature grid array. Calculate the spatial second derivatives along the x-direction and along the y-direction;

[0079] Temperature The spatial second derivative along the x-direction multiplied by the thermal conductivity is used as the diffusion component of that temperature along the x-direction. The temperature in the temperature grid array is then... The spatial second derivative along the y-direction multiplied by the thermal conductivity is taken as the diffusion component of that temperature along the y-direction.

[0080] The heat diffusion trend vector is obtained by fusing the diffusion direction components along the x-direction and the diffusion direction components along the y-direction using SciPy vector addition.

[0081] In one possible implementation, when obtaining temperature data of adjacent acquisition points based on the coordinates of the hot spot, it is necessary to first locate the specific position of the hot spot in the sensor array. Assuming the coordinates of the hot spot are (3, 4), then the adjacent acquisition points include sensors in the four directions of up, down, left, and right, such as (2, 4), (4, 4), (3, 3), and (3, 5). By reading the temperature data of these points, for example, if the temperature of the hot spot is 85°C, the adjacent points are 82°C, 88°C, 80°C, and 84°C, respectively. The reason for this acquisition process is that hot spots are often affected by local current density, leading to heat accumulation. The consequence is that ignoring adjacent data may not be able to fully assess the heat generation risk. By integrating this data, a more comprehensive thermal distribution basis can be provided for subsequent analysis. Specifically, when combining the coordinates of hotspots and the temperature data of their adjacent collection points to generate a complete temperature dataset, these data can be organized into a structured list or dictionary. For example, the coordinates (3,4) can be bound to 85°C, and the adjacent points can be bound accordingly to form a complete temperature dataset such as {(3,4):85, (2,4):82, (4,4):88, (3,3):80, (3,5):84}. This combination helps to expand isolated hotspot data into regional heat maps.

[0082] When creating a temperature grid array using NumPy for a complete temperature dataset, first determine the array size, such as 5x5, to cover the hotspots and their surroundings. Then, initialize a zero-based array and fill it with values ​​at corresponding positions based on the complete temperature dataset. For example, if the hotspot is located at the center, the corresponding elements in the array after filling are set to temperature data. This creation process can transform discrete data into a continuous grid, facilitating mathematical calculations. Specifically, when filling the temperature grid array with initial temperature values ​​from the complete temperature dataset, the dataset is iterated through, and the temperature of each coordinate is assigned to the corresponding index in the array. For example, 85°C at (3, 4) is placed in the array (assuming a zero-based index). In business applications, this step is necessary because a unified data structure is needed to simulate the thermal field. The consequence is that improper filling may lead to calculation errors. By accurately filling, it is ensured that the grid reflects the true heat distribution.

[0083] When iteratively performing temperature update calculations on a temperature grid array to obtain a new temperature grid array, the temperature update calculation is based on the finite difference approximation of the discretized heat conduction equation, involving the time step dt, thermal conductivity k, and density. In the thermal management of electric vehicle charging stations, parameters such as specific heat capacity (c) are used, for example, selecting dt as 0.1 seconds and k as 150 W / m·K. The value is 2700 kg / m³, and c is 900 J / kg·K. This iteration is repeated many times, such as 10 times, to simulate the thermal equilibrium process. The reason is that the initial grid only represents the static state, and the consequence is that the thermal evolution trend can be predicted after iteration.

[0084] Specifically, the NumPy finite difference method is used for each temperature in the temperature grid array. When calculating the spatial second derivatives along the x and y directions, the central difference scheme is used. For example, for the x direction, the derivative is approximately... Where dx is the grid spacing in the x-direction, and similarly, for the y-direction, the derivative is approximately... Where dy is the grid spacing in the y direction; the NumPy finite difference method is used to analyze each temperature in the temperature grid array. When calculating the spatial second derivatives along the x and y directions, the central difference scheme is used. For example, for the x direction, the derivative is approximately... Where dx is the grid spacing in the x-direction, and similarly, for the y-direction, the derivative is approximately... , where dy is the grid spacing in the y direction. Specifically, when using the NumPy difference method to analyze each temperature in the temperature grid array... When calculating the spatial second derivatives along the x and y directions, or using the NumPy finite difference method for each temperature in the temperature grid array... When calculating the spatial second derivatives along the x and y directions, boundary values ​​are preset based on expert advice. Specifically, the values ​​of T_old(i-1) and T_new(i-1) are preset when i=1, and the values ​​of T_old(i+1) and T_new(i+1) are preset when i is at its maximum value, ensuring the formulas can be directly applied. In the simulation of gallium nitride power modules, this calculation quantifies the curvature of the thermal gradient, thus revealing the direction of heat flow. In one possible implementation, the temperature... The spatial second derivative along the x-direction multiplied by the thermal conductivity is used as the diffusion component kX of the temperature along the x-direction. Similarly, the diffusion component kY of the temperature along the y-direction is obtained by processing the y-direction. Through this component calculation, the anisotropic contribution of heat propagation can be accurately described.

[0085] Finally, when the heat diffusion trend vector is obtained by fusing the diffusion direction components along the x-direction and the diffusion direction components along the y-direction using SciPy vector addition, the x and y components are added as vector elements to form a vector such as (kX component, kY component). The cause of this fusion is the multidimensionality of heat diffusion, and the consequence is that the vector can guide the adjustment of cooling design. In this way, a comprehensive understanding of the thermal trend can be achieved.

[0086] Preferably, in step S5, the magnitude of the heat diffusion trend vector is calculated as the temperature change amplitude, and the temperature rise risk score of each hot spot is calculated based on the temperature change amplitude, wherein the formula for the temperature rise risk score is S=ΔTw, where S represents the risk score, ΔT represents the temperature change amplitude, and w represents the preset weight.

[0087] For the risk score of temperature rise, a preset risk threshold is used for comparison. If the score exceeds the preset risk threshold, the corresponding hot spot is selected as a high-risk point.

[0088] High-risk points and their adjacent collection points are integrated into high-risk areas, and all high-risk areas are combined to generate a high-risk list.

[0089] The heat diffusion trend vector represents the direction and intensity of heat flow. This vector is obtained by fusing the x and y directional components in a previous step, and its magnitude, i.e., the length of the vector, quantifies the overall magnitude of heat diffusion. For example, assuming a heat diffusion trend vector is (2.5, 3.0), its magnitude is calculated as follows: The value is approximately 3.9. This temperature change reflects the potential rate of temperature change from the hot spot to the surrounding area. The cause is that uneven heat diffusion may lead to local overheating. The consequence is that through the modulus assessment, it can provide basic data for risk quantification, thereby helping to predict the risk of thermal runaway in business.

[0090] Specifically, when calculating the temperature rise risk score for each hotspot based on the magnitude of temperature change, ΔT in the formula S=ΔTw is the aforementioned modulus, and w is a preset weight, such as 1.2, used to adjust the risk sensitivity. For example, if ΔT is 3.9 and w is 1.2, then S is approximately 4.68. This score integrates the relationship between the magnitude and the weight through multiplication. The principle is that the weight can be determined based on expert advice, thus ensuring that the calculation process reflects the actual potential for temperature rise.

[0091] In one possible implementation, after calculating S for multiple hot spots, a comparison is made and it is found that S=4.68 is below the threshold and is not screened, while another hot spot S=6.2 exceeds the threshold and is selected as a high-risk point. The reason for this comparison is that it is necessary to distinguish between ordinary hot spots and potential dangerous points. The consequence is that after screening, resources can be allocated to monitor high-risk areas first, thereby optimizing the overall thermal management efficiency.

[0092] When integrating high-risk points and their adjacent collection points into a high-risk area, it involves expanding the surrounding area of ​​the hotspot, such as including neighboring points above, below, left, and right to form an area set. For example, a high-risk point with coordinates (5,6) integrates its neighboring points (4,6), (6,6), (5,5), and (5,7) to generate the area {(5,6), (4,6), (6,6), (5,5), (5,7)}. This integration process is based on the principle of neighborhood influence due to heat conduction, ensuring that the area covers potentially affected points. When combining all high-risk areas to generate a high-risk list, multiple areas can be stored using a list structure, such as [Area 1, Area 2]. The combined list facilitates subsequent intervention. The reason for this generation is that scattered areas need to be managed uniformly, and the consequence is that the list supports fast querying and response, improving system security.

[0093] Specifically, in step S6, the current power value of each gallium nitride chip in the high-risk area is extracted and averaged to obtain the high-risk average power; the high-risk average power is calculated by subtracting a preset power threshold to obtain the difference, and then multiplied by the number of gallium nitride chips in the high-risk area to obtain the remaining power share.

[0094] Each chip outside the high-risk area and each non-gallium nitride chip within the high-risk area are considered as non-high-risk chips. The current power value of each non-high-risk chip is obtained, and the sum of the current power values ​​of all non-high-risk chips is calculated as the non-high-risk power sum.

[0095] The first power allocation weight for each non-high-risk chip is determined by the ratio of its current power value to the total power of all non-high-risk chips. The power increase value for each non-high-risk chip is obtained by multiplying the remaining power share by the first power allocation weight. The power adjustment value for each non-high-risk chip is obtained by adding the current power value to the power increase value.

[0096] The sum of the current power values ​​of all gallium nitride chips in the high-risk area is calculated as the power sum of the first high-risk area; the ratio of the current power value of each gallium nitride chip in the high-risk area to the power sum of the first high-risk area is used as the second power allocation weight of the corresponding gallium nitride chip.

[0097] The difference between the current total output power of the power module and the sum of the power adjustment values ​​of all non-high-risk chips is used as the power sum of the second high-risk area. The power adjustment value of the corresponding gallium nitride chip in the high-risk area is obtained by multiplying the power sum of the second high-risk area by the second power allocation weight.

[0098] The power adjustment values ​​of chips in non-high-risk areas and the power adjustment values ​​of gallium nitride chips in high-risk areas are used as the output configuration of the power module.

[0099] In one possible implementation, when extracting the current power value of each gallium nitride chip in a high-risk area and averaging it to obtain the high-risk average power, it is first necessary to obtain real-time data of these chips from the power monitoring system. For example, the high-risk area may include gallium nitride chips such as high-voltage gallium nitride power switches. Assuming there are 3 gallium nitride chips in the high-risk area with current power values ​​of 50W, 60W and 70W respectively, a simple average calculation yields approximately 60W as the high-risk average power. The reason for this process is that high-risk areas often face overheating problems and require quantification of power levels. The consequence is that the average value can provide a benchmark for subsequent difference calculations, thereby helping to optimize energy distribution in business operations to prevent thermal runaway.

[0100] Specifically, the difference between the average power of high-risk areas and a preset power threshold is calculated. This difference is then multiplied by the number of gallium nitride chips in the high-risk area to obtain the remaining power share. The preset power threshold can be set according to the device specifications. For example, if the threshold for the charger is 55W, the difference is 5W. Multiplying this by 3 gives 15W as the remaining power share. This calculation is based on the principle of power redundancy to ensure that the remaining share reflects the potential for redistributable energy.

[0101] In one possible implementation, when calculating the total non-high-risk power, for example in a power module application, the non-high-risk chips may include silicon-based chips and peripheral circuits. Assuming the total is 200W, obtaining this total involves summarizing the monitoring data of all relevant chips. The reason is to distinguish risk types to achieve precise management, and the consequence is to facilitate subsequent weight calculations and support the overall system stability.

[0102] Specifically, when calculating the power increase, in the aforementioned data center scenario, for example, if a chip has a power of 40W and a ratio of 0.2, it is multiplied by 15W to obtain a 3W increase. The principle of this weighting mechanism is based on proportional allocation, ensuring that the power increase is fair and based on contribution. Then, the current power value of each non-high-risk chip is added to the power increase value as the power adjustment value. This adjustment process can improve the load capacity of non-high-risk parts. The technical goal is to balance the overall power distribution and avoid overload in a single area.

[0103] Specifically, when calculating the total power of the first high-risk area, this total reflects the total energy consumption within the high-risk area. The reason for this is that high-risk areas need special attention, and the consequence is to lay the foundation for weight allocation. Then, the power adjustment value of the corresponding chip is calculated based on the second power allocation weight. The reason for this mechanism is the need for dynamic redistribution, and the consequence is to achieve thermal balance in the business and improve module efficiency.

[0104] When the power adjustment values ​​of chips in non-high-risk areas and gallium nitride chips in high-risk areas are used as the output configuration of the power module, this configuration can generate the final power mapping table. The technical effect of the output configuration is to reduce the overheating risk of each gallium nitride chip in the high-risk area and ensure the long-term reliable operation of the system.

[0105] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and these variations still fall within the protection scope of the present invention.

Claims

1. A high-efficiency power transfer control method for a gallium nitride power module, characterized in that, Includes the following steps: S1: Collect temperature data from multiple sampling points on the surface of the power module to obtain the temperature distribution matrix on the surface of the power module; S2: Calculate the effective value of the temperature difference between adjacent sampling points in the temperature distribution matrix, and combine all the effective values ​​of temperature difference to generate a temperature spatial distribution feature matrix; In step S2, temperature data between adjacent sampling points are obtained from the temperature distribution matrix, and the initial temperature difference of the current sampling point is determined by the difference calculation method of subtracting the temperature data of the previous sampling point from the temperature data of the current sampling point; the Euclidean norm of all the effective values ​​of the initial temperature difference of the current sampling point is calculated to obtain the effective value of the temperature difference. For the effective value of the temperature difference, the sensor array is mapped into a two-dimensional matrix according to its physical arrangement on the surface of the gallium nitride power module to obtain the temperature spatial distribution feature matrix. The effective value of the temperature difference in the temperature spatial distribution feature matrix is ​​represented by T2(i,j), and the physical arrangement position of the sensor array on the surface of the gallium nitride power module is represented by (i,j), where i represents the i-th row of the sensor array, j represents the j-th column of the sensor array, i is a positive integer, and j is a positive integer. S3: If there is a valid temperature difference value in the temperature spatial distribution feature matrix that exceeds the preset gradient threshold, then the valid temperature difference value exceeding the preset gradient threshold is analyzed by a classification algorithm to obtain the coordinates of the hot spot location; S4: The temperature data of the hot spot and its adjacent collection points are processed by the discretized heat conduction equation to obtain the heat diffusion trend vector; S5: Evaluate the temperature rise risk score of the hot spot based on the heat diffusion trend vector, filter out the hot spots corresponding to the temperature rise risk scores that exceed the preset risk threshold, and regard the hot spots and the adjacent collection points as high-risk areas. S6: With the aim of reducing the power of gallium nitride chips in high-risk areas, the power of all chips in the power module is reallocated, and the updated power of each chip is used as the output configuration of the power module.

2. The high-efficiency energy transfer control method for a gallium nitride power module according to claim 1, characterized in that: In step S1, temperature data is collected by a sensor array with multiple acquisition points deployed on the surface of the gallium nitride power module to generate a temperature distribution matrix on the surface of the power module.

3. The high-efficiency energy transfer control method for a gallium nitride power module according to claim 2, characterized in that: In step S1, the step of generating the temperature distribution matrix on the surface of the power module includes: For the temperature data, the sensor array is mapped into a two-dimensional matrix according to its physical arrangement on the surface of the gallium nitride power module to form a temperature distribution matrix; The temperature data in the temperature distribution matrix is ​​represented by T1(i,j), where (i,j) represents the physical arrangement position of the sensor array on the surface of the gallium nitride power module, i represents the i-th row of the sensor array, j represents the j-th column of the sensor array, and i and j are positive integers.

4. The high-efficiency power transfer control method for a gallium nitride power module according to claim 1, characterized in that: In step S3, the effective value of the temperature difference in the temperature spatial distribution feature matrix is ​​compared with a preset gradient threshold. If the effective value of the temperature difference is greater than the preset gradient threshold, the sampling point corresponding to the effective value of the temperature difference is marked as an out-of-limit location point. The effective temperature difference values ​​corresponding to the out-of-limit locations are extracted, and the effective temperature difference values ​​are processed using the K-means clustering algorithm. The number of clusters is set to two, and the cluster labels and cluster center values ​​are obtained by fitting the data. Based on the cluster center value, the cluster with the highest value is marked as overheated; Obtain the corresponding overheating location points from the overheating classification, map them to the physical arrangement position (i,j) of the sensor array on the power module surface, and determine the physical arrangement position (i,j) as the coordinates of the overheating hot spot.

5. The high-efficiency energy transfer control method for a gallium nitride power module according to claim 1, characterized in that: In step S4, temperature data of adjacent collection points are obtained based on the coordinates of the hot spot. The coordinates of the hot spot, the temperature data of the hot spot, the coordinates of the sampling points adjacent to the hot spot, and the temperature data of the sampling points adjacent to the hot spot are combined to generate a complete temperature dataset containing the coordinates of the hot spot, the temperature data of the hot spot, the coordinates of the sampling points adjacent to the hot spot, and the temperature data of the sampling points adjacent to the hot spot. For the complete temperature dataset, a temperature grid array was created using NumPy; Fill the temperature grid array with initial temperature values ​​from the complete temperature dataset; For the temperature grid array, iterative temperature update calculations are performed to obtain a new temperature grid array, wherein the temperature update calculation is as follows: ,in Indicates the new temperature value. The old temperature value is represented by dt, the time step by dt, and the thermal conductivity by k. c represents density, and c represents specific heat capacity. express Spatial second derivative along the x-direction, express The spatial second derivative along the y-direction is obtained by using the NumPy finite difference method for each temperature in the temperature grid array. Calculate the spatial second derivatives along the x-direction and along the y-direction; The NumPy finite difference method is used to analyze each temperature in the temperature grid array. Calculate the spatial second derivatives along the x-direction and along the y-direction; Temperature The spatial second derivative along the x-direction multiplied by the thermal conductivity is used as the diffusion component of that temperature along the x-direction. The temperature in the temperature grid array is then... The spatial second derivative along the y-direction multiplied by the thermal conductivity is taken as the diffusion component of that temperature along the y-direction. The heat diffusion trend vector is obtained by fusing the diffusion direction components along the x-direction and the diffusion direction components along the y-direction using SciPy vector addition.

6. The high-efficiency power transfer control method for a gallium nitride power module according to claim 1, characterized in that: In step S5, the magnitude of the heat diffusion trend vector is calculated as the temperature change amplitude. Based on the temperature change amplitude, the temperature rise risk score of each hot spot is calculated, where the formula for the temperature rise risk score is S=ΔTw, where S represents the risk score, ΔT represents the temperature change amplitude, and w represents the preset weight. For the risk score of temperature rise, a preset risk threshold is used for comparison. If the score exceeds the preset risk threshold, the corresponding hot spot is selected as a high-risk point. High-risk points and their adjacent collection points are integrated into high-risk areas, and all high-risk areas are combined to generate a high-risk list.

7. The high-efficiency energy transfer control method for a gallium nitride power module according to claim 1, characterized in that: In step S6, the current power value of each gallium nitride chip in the high-risk area is extracted, and the average power of the high-risk area is obtained by averaging. The difference between the average power of high-risk areas and the preset power threshold is calculated, and then multiplied by the number of gallium nitride chips in the high-risk area to obtain the remaining power share. Each chip outside the high-risk area and each non-gallium nitride chip within the high-risk area are considered as non-high-risk chips. The current power value of each non-high-risk chip is obtained, and the sum of the current power values ​​of all non-high-risk chips is calculated as the non-high-risk power sum. The first power allocation weight for each non-high-risk chip is determined by the ratio of its current power value to the total power of all non-high-risk chips. The power increase value for each non-high-risk chip is obtained by multiplying the remaining power share by the first power allocation weight. The power adjustment value for each non-high-risk chip is obtained by adding the current power value to the power increase value. The sum of the current power values ​​of all gallium nitride chips in the high-risk area is calculated as the power sum of the first high-risk area; The ratio of the current power value of each gallium nitride chip in the high-risk area to the total power of the first high-risk area is used as the second power allocation weight for the corresponding gallium nitride chip. The difference between the current total output power of the power module and the sum of the power adjustment values ​​of all non-high-risk chips is used as the power sum of the second high-risk area. The power adjustment value of the corresponding gallium nitride chip in the high-risk area is obtained by multiplying the power sum of the second high-risk area by the second power allocation weight. The power adjustment values ​​of chips in non-high-risk areas and the power adjustment values ​​of gallium nitride chips in high-risk areas are used as the output configuration of the power module.

Images