Dual-mode heat dissipation charger control method and system based on intelligent algorithm

Abstract

The present application relates to the technical field of charger heat dissipation control, and particularly relates to a dual-mode heat dissipation charger control method and system based on an intelligent algorithm, which obtains temperature and power data, constructs a three-dimensional temperature field and identifies a heat flow convergence point, establishes a heat diffusion dynamics model in combination with charging power, predicts temperature evolution and calculates a time difference at which a temperature peak of the heat flow convergence point appears, starts an active heat dissipation mode when the time difference is less than a threshold value, and reversely calculates a required convective heat transfer coefficient according to a difference between the peak and a safety limit value and the time difference, and then controls an active heat dissipation unit to operate, so that the present application realizes accurate prediction and active intervention on an internal thermal state of the charger, and improves heat dissipation efficiency and charging safety.

Description

Technical Field

[0001] This invention relates to the field of charger heat dissipation control technology, and in particular to a dual-mode heat dissipation charger control method and system based on intelligent algorithms. Background Technology

[0002] In existing technologies, heat dissipation control, especially for high-power chargers, is crucial for ensuring their safe and stable operation. Conventional heat dissipation control schemes typically rely on simple temperature threshold judgments. Specifically, these schemes place one or a few temperature sensors at key locations inside the charger to monitor the local temperature in real time. When the monitored temperature exceeds a preset safety threshold, the control system activates active cooling devices such as fans for forced cooling; when the temperature drops below a lower threshold, the active cooling devices are shut down, relying solely on natural heat dissipation.

[0003] However, this conventional control method based on point-based temperature monitoring and fixed thresholds has significant limitations. Relying on temperature information from only a few discrete points, the system cannot comprehensively and accurately perceive the overall temperature distribution inside the charger. The heat sources inside the charger are unevenly distributed, and the heat diffusion process is dynamic and spatially correlated. Localized high-temperature points may not be detected by the sensors in time due to lag in heat conduction, leading to system response delays. Conversely, the temperature at the sensor's location may also change rapidly due to factors such as local airflow, failing to accurately reflect the overall thermal situation. This could cause the cooling system to start or stop prematurely or too late.

[0004] Furthermore, since the rate and peak of temperature rise cannot be predicted, the system struggles to optimize the activation timing and workload of the active cooling unit, often operating at maximum power to ensure safety. This results in unnecessary energy consumption and noise, reducing energy efficiency and user experience. Summary of the Invention

[0005] This invention provides a dual-mode heat dissipation charger control method and system based on intelligent algorithms, which can solve the problems in the prior art.

[0006] A first aspect of the present invention provides a dual-mode heat dissipation charger control method based on intelligent algorithms, comprising: Acquire temperature sensor array data and charging power timing data of the charger; Spatial interpolation is performed on the temperature sensor array data to construct a three-dimensional temperature field distribution map inside the charger, and the temperature gradient vector field is calculated by the gradient operator to identify the heat flow convergence point. The three-dimensional temperature field distribution map is divided into multiple grid cells using the finite element method. A volume heat source term is added to the corresponding grid cell according to the charging power time series data. The grid cell where the heat flow convergence point is located is weighted to establish a thermal diffusion dynamics model based on partial differential equations. By iteratively solving the thermal diffusion dynamics model, the temperature evolution sequence of each grid cell in the future multiple time steps is predicted, the temperature peak value and occurrence time of the grid cell where the heat flow convergence point is located are extracted, and the time difference between the occurrence time and the current time is calculated. When the time difference is less than the lead threshold, the active heat dissipation mode is activated, and based on the difference between the peak amplitude and the safe temperature limit, the convective heat transfer coefficient that the active heat dissipation unit needs to apply at the boundary is calculated by inversely solving the thermal diffusion dynamics model in combination with the time difference. The operating parameters of the active heat dissipation unit are controlled to achieve the convective heat transfer coefficient, and the temperature field data after heat dissipation is collected to update the thermal diffusion dynamics model.

[0007] Spatial interpolation is performed on the temperature sensor array data to construct a three-dimensional temperature field distribution map inside the charger. A temperature gradient vector field is calculated using a gradient operator to identify heat flow convergence points, including: Using the spatial coordinates of each temperature sensor in the temperature sensor array data as sampling points, the spatial distance between any two sampling points is calculated, and a semi-variogram function is established based on the spatial distance. The spatial autocorrelation weight matrix between each sampling point is calculated using the semi-variogram function. The inverse matrix of the spatial autocorrelation weight matrix is ​​then solved. The inverse matrix is ​​multiplied by the temperature measurement value of each sampling point to obtain the interpolation coefficient vector. A regular grid structure is established inside the charger, the spatial distance between each grid node and each sampling point is calculated, the spatial distance is substituted into the semi-variogram function to obtain the correlation weight, and the weighted sum is performed with the interpolation coefficient vector to obtain the temperature estimate, and a three-dimensional temperature field distribution map is constructed. The first-order temperature derivatives of each grid node in the three-dimensional temperature field distribution map in the three coordinate axes are calculated using the central difference method and combined into a temperature gradient vector to generate the temperature gradient vector field. Calculate the gradient vector direction angle of each grid node in the temperature gradient vector field, and count the proportion of gradient vectors pointing to each grid node within the neighborhood. When the proportion exceeds the convergence threshold, the grid node is identified as a heat flow convergence point.

[0008] The finite element method is used to divide the three-dimensional temperature field distribution map into multiple grid cells. Based on the charging power time-series data, a volume heat source term is added to the corresponding grid cell. Weights are applied to the grid cells containing the heat flow convergence point to establish a thermal diffusion dynamics model based on partial differential equations, including: The three-dimensional temperature field distribution map is spatially discretized, and the internal space of the charger is divided into multiple hexahedral grid units. Each grid unit is assigned a unit number and node coordinates, and a topological mapping relationship between the grid unit and the spatial location is established. The charging power value at each moment in the charging power time series data is obtained. Based on the spatial distribution of the heating elements inside the charger, the grid cell number corresponding to each heating element is determined. The charging power value is converted into a unit volume power density and used as the volume heat source item of the corresponding grid cell. Extract the spatial coordinates of the heat flow convergence point, determine the grid cell to which the heat flow convergence point belongs, calculate the absolute value of the divergence of the grid cell to which each heat flow convergence point belongs and normalize it to obtain the weight coefficient of the corresponding grid cell. A partial differential equation is constructed based on Fourier's law of heat conduction, and the volumetric heat source term is added to the source term of the partial differential equation. The temperature change rate term of the grid cell where the heat flow converges is multiplied by the weighting coefficient to form a thermal diffusion dynamics model.

[0009] A partial differential equation is constructed based on Fourier's law of heat conduction, and the volumetric heat source term is added to the source term of the partial differential equation. The temperature change rate term of the grid cell where the heat flow converges is located is multiplied by the weighting coefficient to form a thermal diffusion dynamics model, including: Based on the topological mapping relationship between grid cells and spatial locations, the material region of each grid cell is determined, and the thermal conductivity, density and specific heat capacity of the corresponding material are extracted from the material property library to assign thermophysical parameters to each grid cell. A partial differential equation is established based on the principle of heat conservation and Fourier's law of heat conduction. The left side of the equation is the product of the rate of change of temperature with respect to time, density, and specific heat capacity. The right side of the equation is the combination of the spatial temperature gradient and thermal conductivity. The volume heat source term is superimposed on the right side of the equation as an independent source term. The spatial coordinates of the heat flow convergence points are retrieved through the topological mapping relationship to determine the grid cell number to which each heat flow convergence point belongs, the corresponding weight coefficients are extracted, and a correspondence between the grid cell number and the weight coefficients is established. Locate the equation corresponding to the grid cell where the heat flow convergence point is located in the partial differential equation, extract the temperature change rate term on the left side of the equation and multiply it with the weighting coefficient, replace the original temperature change rate term with the product, and keep the terms on the right side of the equation unchanged to form a thermal diffusion dynamics model.

[0010] By iteratively solving the thermal diffusion kinetics model, the temperature evolution sequence of each grid cell over multiple time steps is predicted, including: Set the start and end times of the simulation, determine the total number of iterations based on the time span and preset time step, obtain the initial temperature data of each grid cell of the charger at the start time, identify the boundary grid cells, and establish boundary heat exchange conditions for the boundary grid cells based on the ambient temperature and convective heat transfer coefficient. The heat diffusion dynamics model is discretized in the time domain. The temperature-time partial derivative, spatial partial derivative, and volume heat source terms in the partial differential equations corresponding to each grid cell are extracted. The partial derivatives are converted into difference form using the finite difference method. An algebraic equation system is constructed with the current time step temperature as the known quantity and the next time step temperature as the unknown quantity. Initialize the iteration counter and start the iteration. In each iteration, read the density, specific heat capacity, thermal conductivity and volume heat source term values ​​of each grid cell, extract the current temperature value of each grid cell and its adjacent grid cells, and substitute them into the corresponding algebraic equation to solve for the temperature value of the next time step. Apply the boundary heat exchange condition constraint to the boundary grid cells. After completing the calculation of each grid cell, the temperature field of the next time step is formed. The iteration counter is incremented and it is determined whether the total number of iterations has been reached. If not, the temperature field of the next time step is used as the current temperature field to continue the iteration. If the total number of iterations has been reached, the iteration ends and the temperature fields of each time step are output to form a temperature evolution sequence.

[0011] Based on the difference between the peak amplitude and the safe temperature limit, and combined with the time difference, the convective heat transfer coefficient that the active cooling unit needs to apply at the boundary is calculated by inversely solving the thermal diffusion dynamics model, including: The target temperature drop is obtained by calculating the difference between the peak amplitude and the safe temperature limit. Based on the target temperature drop and the specific heat capacity, density and volume of the charger shell, the total heat to be released is calculated, and the average heat dissipation power requirement is obtained by combining the time difference. The boundary grid cells are located from the thermal diffusion dynamics model, and the thermal flux boundary conditions of the boundary grid cells are set in reverse. The boundary heat flux density is expressed as the product of the convective heat transfer coefficient and the difference between the boundary temperature and the ambient temperature. The thermal diffusion dynamics model is discretized in reverse time. The temperature field at the current moment is used as the initial condition, and the temperature change trajectory within the time difference is used as the constraint condition. An inverse optimization function is established with the convective heat transfer coefficient as the optimization variable and the temperature field satisfying the safe temperature limit as the objective. The gradient descent method is used to solve the inverse optimization function. In each iteration, the temperature field evolution under the current convective heat transfer coefficient is calculated, the deviation between the temperature field and the safe temperature limit is evaluated, and the sensitivity is calculated. The convective heat transfer coefficient value is updated according to the sensitivity. The iteration terminates when the temperature field meets the safe temperature limit, outputs the converged convective heat transfer coefficient value, and sends a heat dissipation control command containing the convective heat transfer coefficient value to the active heat dissipation unit.

[0012] The gradient descent method is used to solve the inverse optimization function. In each iteration, the temperature field evolution under the current convective heat transfer coefficient is calculated, the deviation of the temperature field from the safe temperature limit is evaluated, and the sensitivity is calculated. The convective heat transfer coefficient value is updated based on the sensitivity, including: Initialize the convective heat transfer coefficient and set the learning rate and gradient convergence threshold. In each iteration, use the current convective heat transfer coefficient as the boundary parameter to calculate the temperature field evolution process within the time difference. Extract the highest temperature value in the temperature field at the termination time and subtract it from the safe temperature limit to obtain the current temperature deviation. Apply positive and negative small perturbations to the current convective heat transfer coefficient respectively, calculate the temperature field evolution under positive and negative perturbation conditions, extract the highest temperature values ​​of positive and negative perturbations, calculate the difference between the two and the safe temperature limit, divide the difference between the positive and negative perturbation temperature deviations by the total perturbation amount, and obtain the sensitivity by the central difference method. The sensitivity value is multiplied by the learning rate to obtain the update step size. The update step size is subtracted from the current convective heat transfer coefficient to obtain the new convective heat transfer coefficient value. The absolute value of the sensitivity is calculated and it is determined whether it is less than the gradient convergence threshold. If it is less than the threshold, the iteration is terminated and the new convective heat transfer coefficient value is output. Otherwise, the new convective heat transfer coefficient value is used as the current convective heat transfer coefficient for the next iteration and the iteration continues.

[0013] A second aspect of the present invention provides a dual-mode heat dissipation charger control system based on intelligent algorithms, comprising: The data acquisition unit is used to acquire temperature sensor array data and charging power timing data of the charger; The temperature field construction unit is used to perform spatial interpolation processing on the temperature sensor array data, construct a three-dimensional temperature field distribution map inside the charger, and calculate the temperature gradient vector field through the gradient operator to identify heat flow convergence points. The heat source modeling unit is used to divide the three-dimensional temperature field distribution map into multiple grid cells using the finite element method, and add volume heat source terms to the corresponding grid cells according to the charging power time series data, and perform weighted summation on the grid cells where the heat flow convergence point is located to establish a thermal diffusion dynamics model based on partial differential equations. The temperature prediction unit is used to predict the temperature evolution sequence of each grid cell in the future multiple time steps by iteratively solving the heat diffusion dynamics model, extract the temperature peak and occurrence time of the grid cell where the heat flow convergence point is located, and calculate the time difference between the occurrence time and the current time. The heat dissipation decision unit is used to activate the active heat dissipation mode when the time difference is less than the advance threshold, and to back-calculate the convective heat transfer coefficient that the active heat dissipation unit needs to apply at the boundary by inversely solving the thermal diffusion dynamics model based on the difference between the peak amplitude and the safe temperature limit and the time difference. The heat dissipation control unit is used to control the operating parameters of the active heat dissipation unit to achieve the convective heat transfer coefficient, and to collect temperature field data after heat dissipation to update the thermal diffusion dynamics model.

[0014] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0015] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0016] This method enables accurate modeling and dynamic prediction of the internal temperature field of a charger. By constructing a three-dimensional temperature field distribution map and calculating the temperature gradient vector field, heat flow convergence points can be accurately identified, thereby locating potential overheating areas. By using the finite element method to divide the grid cells and adding a volumetric heat source term based on charging power time-series data, a high-fidelity thermal diffusion dynamics model can be established. By weighting the grid cells where heat flow convergence points are located, the model can focus on the temperature evolution of key regions.

[0017] By iteratively solving the thermal diffusion dynamics model, the temperature change sequence of each grid cell within multiple future time steps can be predicted. The temperature peak and its occurrence time at the heat flow convergence point can be extracted, and the time difference between the peak and the current time can be calculated, providing a basis for early intervention. When an overheating risk is predicted to occur, the required boundary heat dissipation intensity can be accurately calculated based on the time difference and the gap between the peak value and the safety limit, realizing a shift from "passive response" to "active prevention".

[0018] By inversely solving the thermal diffusion kinetics model to deduce the required convective heat transfer coefficient, the activation timing and operating intensity of the active cooling unit are based on precise model calculations, rather than simple threshold judgments. This ensures the accuracy and efficiency of heat dissipation intervention, effectively preventing overheating while avoiding unnecessary energy consumption and fan noise. Controlling the active cooling unit to achieve the calculated convective heat transfer coefficient enables targeted cooling of heat flow convergence points, significantly improving heat dissipation efficiency. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the dual-mode heat dissipation charger control method based on intelligent algorithms according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the method for establishing a thermal diffusion dynamics model based on partial differential equations in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0022] Figure 1 This is a flowchart illustrating the dual-mode heat dissipation charger control method based on intelligent algorithms according to an embodiment of the present invention. Figure 1 As shown, the dual-mode heat dissipation charger control method based on intelligent algorithms includes: Acquire temperature sensor array data and charging power timing data of the charger; Spatial interpolation is performed on the temperature sensor array data to construct a three-dimensional temperature field distribution map inside the charger, and the temperature gradient vector field is calculated by the gradient operator to identify the heat flow convergence point. The three-dimensional temperature field distribution map is divided into multiple grid cells using the finite element method. A volume heat source term is added to the corresponding grid cell according to the charging power time series data. The grid cell where the heat flow convergence point is located is weighted to establish a thermal diffusion dynamics model based on partial differential equations. By iteratively solving the thermal diffusion dynamics model, the temperature evolution sequence of each grid cell in the future multiple time steps is predicted, the temperature peak value and occurrence time of the grid cell where the heat flow convergence point is located are extracted, and the time difference between the occurrence time and the current time is calculated. When the time difference is less than the lead threshold, the active heat dissipation mode is activated, and based on the difference between the peak amplitude and the safe temperature limit, the convective heat transfer coefficient that the active heat dissipation unit needs to apply at the boundary is calculated by inversely solving the thermal diffusion dynamics model in combination with the time difference. The operating parameters of the active heat dissipation unit are controlled to achieve the convective heat transfer coefficient, and the temperature field data after heat dissipation is collected to update the thermal diffusion dynamics model.

[0023] Figure 2 This is a flowchart illustrating the method for establishing a thermal diffusion dynamics model based on partial differential equations in an embodiment of the present invention. In one optional implementation, spatial interpolation processing is performed on the temperature sensor array data to construct a three-dimensional temperature field distribution map inside the charger, and the temperature gradient vector field is calculated using a gradient operator to identify heat flow convergence points, including: Using the spatial coordinates of each temperature sensor in the temperature sensor array data as sampling points, the spatial distance between any two sampling points is calculated, and a semi-variogram function is established based on the spatial distance. The spatial autocorrelation weight matrix between each sampling point is calculated using the semi-variogram function. The inverse matrix of the spatial autocorrelation weight matrix is ​​then solved. The inverse matrix is ​​multiplied by the temperature measurement value of each sampling point to obtain the interpolation coefficient vector. A regular grid structure is established inside the charger, the spatial distance between each grid node and each sampling point is calculated, the spatial distance is substituted into the semi-variogram function to obtain the correlation weight, and the weighted sum is performed with the interpolation coefficient vector to obtain the temperature estimate, and a three-dimensional temperature field distribution map is constructed. The first-order temperature derivatives of each grid node in the three-dimensional temperature field distribution map in the three coordinate axes are calculated using the central difference method and combined into a temperature gradient vector to generate the temperature gradient vector field. Calculate the gradient vector direction angle of each grid node in the temperature gradient vector field, and count the proportion of gradient vectors pointing to each grid node within the neighborhood. When the proportion exceeds the convergence threshold, the grid node is identified as a heat flow convergence point.

[0024] In this specific embodiment, the spatial coordinates of each temperature sensor in the temperature sensor array data are used as sampling points. In practical applications, multiple temperature sensors are typically arranged inside the charger, and these sensors are distributed near key heat-generating components such as transformers, power transistors, and rectifier bridges. If N temperature sensors are arranged inside the charger, the spatial position of each sensor can be represented by three-dimensional coordinates p. i =(x i y i , z i The values ​​are represented as follows: ), where i = 1, 2, ..., N, and stored in the control system's configuration file. Each sensor simultaneously measures the temperature value T. i The sampling frequency is usually set to 1Hz to 10Hz to meet the needs of real-time monitoring. To ensure the accuracy of spatial interpolation, the sensor layout should follow the principle of uniform distribution to avoid a situation where sensors are too dense in one area and sparse in other areas.

[0025] Calculate the spatial distance between any two sampling points, for sampling point p. i and p j Its spatial distance d ij The calculation formula is Traverse all sampling point pairs to construct a distance matrix D. This matrix is ​​a symmetric matrix with zero diagonal elements. The distance matrix reflects the geometric characteristics of the sensor spatial distribution and is the basic data for establishing the semivariogram. In the calculation process, the distance unit is usually millimeters to match the actual internal size of the charger.

[0026] A semivariogram is established based on spatial distance. The semivariogram describes the decay of the correlation between temperatures at different spatial locations with distance. A spherical model is used to construct the semivariogram. Its expression is when hour, ;when hour, ,in For spatial distance, For the variance of the gold nugget, For sill values, The range parameter, nugget variance, reflects the effects of measurement error and microscale variations, and is typically taken as 5% to 15% of the total variance. The sill value equals the total variance of the data minus the nugget variance. The range parameter determines the rate of correlation decay. Based on the internal thermal conduction characteristics of the charger, the range is set to 1 / 3 to 1 / 2 of the charger's maximum size. The optimal values ​​of these three parameters are determined by fitting an empirical semivariogram to the measured temperature data.

[0027] The spatial autocorrelation weight matrix between each sampling point is calculated using a semi-variogram. The distance between all pairs of sampling points is then calculated. Substituting into the semimutation function, we obtain the semimutated value. Spatial autocorrelation weight matrix The element is defined as This definition method results in greater weights for sampling points that are closer together. Matrix for A 1D symmetric positive definite matrix whose diagonal elements are all equal to The weight matrix reflects the spatial correlation strength between temperature values ​​at different sampling points and is the core of the Kriging interpolation method.

[0028] To find the inverse matrix of the spatial autocorrelation weight matrix, the Cholesky decomposition method is used to invert the weight matrix, yielding the inverse matrix. Since the weight matrix has the property of being symmetric and positive definite, Cholesky decomposition can guarantee numerical stability and improve computational efficiency. The weight matrix is ​​decomposed into... ,in Given a lower triangular matrix, the inverse matrix is ​​obtained by solving a system of linear equations through forward and backward substitutions. In actual programming implementation, to avoid storing the complete inverse matrix, sparse matrix techniques and iterative solving algorithms can be used to reduce memory usage.

[0029] Perform matrix multiplication on the inverse matrix and the temperature measurement values ​​from each sampling point to construct a column vector of temperature measurement values. Perform matrix multiplication This yields the interpolation coefficient vector. The dimension of this vector is Its elements Represents the first The contribution coefficients of each sampling point to the temperature estimation at any spatial location, and the interpolation coefficient vector comprehensively consider the temperature values ​​of all sampling points and their spatial distribution relationship, are key parameters for achieving unbiased optimal interpolation. The computational complexity of matrix multiplication is O(n log n). When there are a large number of sensors, parallel computing can be used to accelerate the process.

[0030] A regular grid structure is established within the charger's internal space. Based on the charger's physical dimensions, the coordinate range in three-dimensional space is defined. , , Divided into three directions respectively , , There are 1 equally spaced grid, and the total number of grid nodes is . The choice of grid spacing needs to balance computational accuracy and efficiency, and is typically set to 1 mm to 5 mm. Each grid node... The coordinates are automatically generated according to the grid index rules. An index mapping relationship is established so that the three-dimensional coordinates... Can be converted into one-dimensional node numbers This facilitates data storage and access. The establishment of the grid structure provides a unified framework for the subsequent spatial discretization of the temperature field.

[0031] Computational grid nodes and sampling points The spatial distance between them is given by the formula: Iterate through all combinations of grid nodes and sampling points to construct a distance matrix. Its dimensions are Since the number of grid nodes is usually much larger than the number of sampling points, the calculation process can be accelerated using spatial indexing techniques, such as KD trees or octrees, to quickly find sampling points near each grid node, prune invalid calculations that are too far away, and immediately substitute the distance value into the semi-mutation function for subsequent processing after the distance calculation is completed.

[0032] Substituting spatial distance into the semi-mutation function yields the correlation weights for grid nodes. With sampling points Distance between Calculate the semivariogram. Thus, the relevance weights are obtained. Construct the weight matrix , its first Line number Column elements are The weight matrix reflects the spatial correlation between each grid node and each sampling point. The closer the sampling point is, the greater its contribution to the temperature estimation of the grid node. To improve numerical stability, the weight values ​​are normalized so that the sum of the weight vectors corresponding to each grid node is a unit value.

[0033] The temperature estimate is obtained by weighted summation with the interpolation coefficient vector for each grid node. The correlation weights of these weights with each sampling point are determined. With the corresponding interpolation coefficients Multiply and sum to obtain the temperature estimate. Traverse all grid nodes to form a sequence of temperature estimates. This estimated sequence constitutes the discrete temperature distribution of the internal space of the charger.

[0034] A three-dimensional temperature field distribution map is constructed by integrating the temperature estimates of all grid nodes. Using 3D visualization technology, the temperature estimates are represented as a heatmap, with colors ranging from blue to red corresponding to low to high temperature regions. To improve visualization, a bilinear interpolation algorithm is used to smooth the grid temperatures, eliminating abrupt changes at grid boundaries. The temperature field distribution map can simultaneously display multiple cross-sectional and longitudinal views, facilitating observation of the temperature status of different parts inside the charger. In practical applications, the temperature field is updated in real-time at a frequency of 0.5Hz to 1Hz, ensuring that the monitoring system can promptly reflect temperature change trends.

[0035] The central difference method is used to calculate the first-order temperature derivatives of each grid node in the three-dimensional temperature field distribution map along the three coordinate axes. For the internal grid node (i, j, k), the formulas for calculating the first-order temperature derivatives in the x, y, and z directions are as follows: ∂T / ∂x(i, j, k) = [T(i+1, j, k) - T(i-1, j, k)] / (2Δx), ∂T / ∂y(i, j, k) = [T(i, j+1, k) - T(i, j-1, k)] / (2Δy), ∂T / ∂z(i, j, k) = [T(i, j, k ..., k+1, k) - T(i, j, k+1, k)] / (2Δy), ∂T / ∂z(i, j, k) = [T(i, j, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k+1, k )-T(i,j,k-1)] / (2Δz), where Δx, Δy, and Δz are the grid spacings in the x, y, and z directions, respectively. For boundary grid nodes, forward or backward difference is used instead of center difference to ensure that all nodes can calculate derivative values. The forward difference formula is ∂T / ∂x(i,j,k)=[T(i+1,j,k)-T(i,j,k)] / Δx, and the backward difference formula is ∂T / ∂x(i,j,k)=[T(i,j,k)-T(i-1,j,k)] / Δx. The y and z directions are treated similarly.

[0036] After calculating the first derivative of temperature, the derivatives in the three directions are combined into a temperature gradient vector: ∇T(i,j,k)=[∂T / ∂x(i,j,k), ∂T / ∂y(i,j,k), ∂T / ∂z(i,j,k)]. This gradient vector represents the direction and magnitude of the fastest temperature change, and its magnitude |∇T(i,j,k)| represents the rate of temperature change, pointing in the direction of the fastest temperature increase. To ensure numerical stability, the temperature field is preprocessed with Gaussian filtering before calculating the derivatives to suppress noise. The accuracy of the gradient calculation directly affects the accuracy of subsequent heat flow analysis; therefore, sufficient spatial resolution must be ensured during mesh generation, typically requiring the mesh spacing to be less than 1 / 5 of the heat source size.

[0037] When generating the temperature gradient vector field, the gradient vectors of all grid nodes are stored in a unified three-dimensional array structure. Each grid node corresponds to a three-dimensional gradient vector. To visually display the gradient field distribution, an arrow graphic can be used to represent the direction and magnitude of the gradient vector. The length of the arrow is proportional to the gradient magnitude and points in the same direction as the gradient. At the same time, a gradient intensity heat map is generated based on the magnitude of the gradient magnitude to identify areas with drastic temperature changes. The gradient vector field is updated at the same frequency as the temperature field to achieve dynamic monitoring of heat flow.

[0038] Calculate the gradient vector orientation angle for each grid node in the temperature gradient vector field. For node (i, j, k), the angles between its gradient vector and the x-axis, y-axis, and z-axis are respectively: θx = arccos(∂T / ∂x(i, j, k) / |∇T(i, j, k)|), θy = arccos(∂T / ∂y(i, j, k) / |∇T(i, j, k)|), θz = arccos(∂T / ∂z(i, j, k) / |∇T(i, j, k)|). When the gradient magnitude |∇T(i, j, k)| is close to zero, the orientation angle calculation result is unstable. In this case, the node is marked as an isothermal point and does not participate in the subsequent heat flow convergence analysis. The orientation angle calculation result is used to determine the orientation relationship between adjacent gradient vectors and is the basic data for identifying heat flow convergence points.

[0039] The percentage of gradient vectors pointing to each grid node within the neighborhood is calculated. For each grid node (i, j, k), its neighborhood is defined as all nodes within a radius r, where r is typically set to 2 to 3 grid spacings. The gradient vector of each neighboring node (p, q, s) is determined to point to the center node (i, j, k) based on the following criterion: the angle between the gradient vector of node (p, q, s) and the direction vector connecting the two nodes is less than a threshold angle α, typically set to 30° to 45°. The number of neighboring nodes meeting this condition is recorded, and their proportion to the total number of neighboring nodes is calculated to obtain the pointing percentage.

[0040] When the pointing percentage exceeds the convergence threshold, the grid node is identified as a heat flow convergence point. The convergence threshold is determined based on the charger's structural characteristics and is typically set to 60% to 75%. Setting the convergence threshold too high will lead to missed detections, while setting it too low will result in false detections. Heat flow convergence points indicate the accumulation of heat energy at that location, making them potential hotspots that require close monitoring. All identified heat flow convergence points are sorted in descending order of pointing percentage and highlighted with special markers on the 3D temperature field distribution map. Their coordinates and current temperature values ​​are also recorded.

[0041] In one optional implementation, the three-dimensional temperature field distribution map is divided into multiple grid cells using the finite element method, and a volume heat source term is added to the corresponding grid cell according to the charging power time series data. The grid cell containing the heat flow convergence point is weighted, and a thermal diffusion dynamics model based on partial differential equations is established, including: The three-dimensional temperature field distribution map is spatially discretized, and the internal space of the charger is divided into multiple hexahedral grid units. Each grid unit is assigned a unit number and node coordinates, and a topological mapping relationship between the grid unit and the spatial location is established. The charging power value at each moment in the charging power time series data is obtained. Based on the spatial distribution of the heating elements inside the charger, the grid cell number corresponding to each heating element is determined. The charging power value is converted into a unit volume power density and used as the volume heat source item of the corresponding grid cell. Extract the spatial coordinates of the heat flow convergence point, determine the grid cell to which the heat flow convergence point belongs, calculate the absolute value of the divergence of the grid cell to which each heat flow convergence point belongs and normalize it to obtain the weight coefficient of the corresponding grid cell. A partial differential equation is constructed based on Fourier's law of heat conduction, and the volumetric heat source term is added to the source term of the partial differential equation. The temperature change rate term of the grid cell where the heat flow converges is multiplied by the weighting coefficient to form a thermal diffusion dynamics model.

[0042] In establishing the thermal diffusion dynamics model, the three-dimensional temperature field distribution map needs to be spatially discretized. The continuous space inside the charger is converted into a computable discrete structure using the finite element method. Hexahedral mesh elements are used as the basic unit of division, as they offer better numerical stability and computational accuracy compared to tetrahedral elements in temperature field solving. Mesh density parameters are set along the length, width, and height directions of the charger during partitioning. The mesh is appropriately densified in areas with dense heat-generating components to improve the computational accuracy of the local temperature field. For example, the mesh size is set to 2mm to 3mm in the power conversion circuit board area, while the mesh size can be increased to 5mm to 8mm in the outer casing area away from the heat source. Each hexahedral mesh element is defined by 8 nodes, each assigned a unique integer number arranged from bottom to top, left to right, and front to back. The three-dimensional spatial coordinates of each node are recorded, with the origin set at a fixed corner of the charger casing. This establishes a topological mapping relationship between the mesh element number, node number, and actual spatial location. This mapping relationship is stored in a data structure for easy retrieval of the corresponding mesh element based on the spatial coordinates.

[0043] After spatial discretization, the charging power time-series data needs to be converted into volumetric heat source terms for each grid cell. This data records the real-time power values ​​of the charger at different times, typically with a time resolution of 100ms to 500ms. The main heat-generating components inside the charger include the rectifier bridge, power switch, transformer, and inductor coil, each occupying a specific area in physical space. The three-dimensional spatial coordinate range of each heat-generating component is obtained, and the grid cells covered by each component are determined through coordinate matching. A heat-generating component may span multiple adjacent grid cells. Therefore, the heat source intensity needs to be allocated based on the component's volume proportion within each cell. For a given charging power value, the proportion of power loss borne by each heat-generating component is calculated based on the circuit topology and device loss model. For example, the rectifier bridge might bear 30% to 40% of the total power loss, the power switch 25% to 35%, and the transformer 20% to 30%. The power loss allocated to a heat-generating component is divided by the total volume occupied by that component to obtain the power density value, expressed in watts per cubic meter. This power density value is the volumetric heat source term of the grid cell covered by the element. If a grid cell contains part of the volume of multiple heat-generating elements, the volumetric heat source term of that cell is the sum of the contributions of each element. The value of the volumetric heat source term is dynamically adjusted with the change of charging power. During the fast charging stage, the power density can reach tens of thousands of watts per cubic meter, while during the trickle charging stage, it drops to hundreds of watts per cubic meter.

[0044] For the grid cells containing heat flow convergence points, special weighting is required. The spatial coordinates of the heat flow convergence points have already been obtained from the temperature gradient vector field analysis, and the grid cells containing each convergence point are determined through coordinate mapping. The absolute value of the divergence of each heat flow convergence point cell is calculated. Divergence characterizes the degree of divergence or convergence of the temperature gradient vector at that point. The divergence calculation involves the temperature gradient vectors of the cell and its neighboring cells, obtained by spatially differentiating and summing the components of the gradient vector using the finite difference method. A positive divergence value indicates that the heat flow is diverging outward from that point, while a negative value indicates that the heat flow is converging towards that point. The larger the absolute value, the stronger the convergence or divergence. The absolute values ​​of the divergence of all heat flow convergence point cells are normalized by dividing each absolute divergence value by the maximum value of the absolute divergence values ​​of all convergence points, so that the weight coefficient is limited to the range of 0 to 1. The normalized value is the weight coefficient of that grid cell. Cells with a weight coefficient close to 1 represent the strongest heat flow convergence and need to be given higher attention in the dynamic model.

[0045] The thermal diffusion kinetic model is based on Fourier's law of heat conduction, which describes the heat conduction process within a substance. Its basic form states that heat flux density is proportional to the temperature gradient. For the three-dimensional temperature field inside the charger, the thermal diffusion process follows a partial differential equation of heat conduction, the general form of which is: ,in Indicates the density of the material. Indicates specific heat capacity. Indicates temperature. Indicates time, Indicates thermal conductivity, This represents the internal heat source per unit volume. For each grid cell, the equation above... The term is filled by the previously calculated volumetric heat source term. The charger contains various materials: the thermal conductivity of the circuit board substrate is approximately 0.3 watts per meter per Kelvin, the thermal conductivity of the copper heat sink is approximately 400 watts per meter per Kelvin, and the thermal conductivity of the plastic shell is approximately 0.2 watts per meter per Kelvin. Therefore, the material parameters of different mesh elements differ. During discretization, the finite volume method is used to spatially discretize the partial differential equations, treating the temperature of each mesh element as an independent variable. The heat flow between elements is coupled through interfacial thermal conduction.

[0046] For the grid cell where the heat flow convergence point is located, the temperature change rate term in the partial differential equation needs to be weighted and corrected. The standard temperature change rate term is: The convergence point unit is corrected to Here, w is the weighting coefficient corresponding to the unit. The physical meaning of this correction is to enhance the sensitivity of the temperature response in the convergence point region, allowing the model to allocate more computational resources to the temperature changes at these key locations during the solution process. When the weighting coefficient is greater than 1, it is equivalent to increasing the effective heat capacity of the unit, making its temperature change more gradual. However, since the weighting coefficient is normalized to the range of 0 to 1 in practical applications, the actual effect is to reduce the effective heat capacity and accelerate the temperature response speed of the unit. The partial differential equations after weighting form a complete thermal diffusion dynamics model. The number of equations in this model is equal to the total number of grid cells. For a typical charger discretization model, the number of grid cells is between 5000 and 20000, and the corresponding equation set has the same scale. The units in the equation set are coupled through thermal conduction boundary conditions, forming a sparse matrix structure. The non-zero elements of the matrix are mainly distributed on the diagonal and its vicinity, reflecting the heat exchange relationship between adjacent units. The model's boundary conditions include natural convection heat transfer between the charger casing and ambient air. The convection heat transfer coefficient is dynamically calculated based on the temperature difference between the casing surface and the ambient temperature, typically ranging from 5 watts per square meter per Kelvin to 15 watts per square meter per Kelvin. The effect of active cooling units such as fans or heat pipes is also reflected in the boundary conditions, simulating forced convection by modifying the corresponding boundary convection heat transfer coefficients.

[0047] The established thermal diffusion kinetic model requires numerical verification and parameter calibration. By comparing the actual measured temperature sensor data with the model calculation results, uncertain parameters such as material parameters and contact thermal resistance in the model are adjusted to ensure that the root mean square error between the model output and the measured value is controlled within 2 degrees Celsius. The model's time discretization adopts an implicit difference scheme, with a time step set from 1 second to 5 seconds, ensuring both computational stability and meeting the speed requirements for real-time prediction. The complete kinetic model provides the physical basis and computational framework for subsequent temperature evolution prediction and active heat dissipation control strategies.

[0048] In one optional implementation, a partial differential equation is constructed based on Fourier's law of heat conduction, and the volume heat source term is added to the source term of the partial differential equation. The temperature change rate term of the grid cell where the heat flow converges is multiplied by the weighting coefficient to form a heat diffusion dynamics model, including: Based on the topological mapping relationship between grid cells and spatial locations, the material region of each grid cell is determined, and the thermal conductivity, density and specific heat capacity of the corresponding material are extracted from the material property library to assign thermophysical parameters to each grid cell. A partial differential equation is established based on the principle of heat conservation and Fourier's law of heat conduction. The left side of the equation is the product of the rate of change of temperature with respect to time, density, and specific heat capacity. The right side of the equation is the combination of the spatial temperature gradient and thermal conductivity. The volume heat source term is superimposed on the right side of the equation as an independent source term. The spatial coordinates of the heat flow convergence points are retrieved through the topological mapping relationship to determine the grid cell number to which each heat flow convergence point belongs, the corresponding weight coefficients are extracted, and a correspondence between the grid cell number and the weight coefficients is established. Locate the equation corresponding to the grid cell where the heat flow convergence point is located in the partial differential equation, extract the temperature change rate term on the left side of the equation and multiply it with the weighting coefficient, replace the original temperature change rate term with the product, and keep the terms on the right side of the equation unchanged to form a thermal diffusion dynamics model.

[0049] In this specific embodiment, the material region of each grid cell needs to be determined based on the topological mapping relationship between grid cells and spatial locations. The charger typically contains multiple materials, such as circuit board substrate, copper wires, semiconductor packaging materials, heat sink materials, and air regions. A pre-established material distribution model maps each grid cell in three-dimensional space to a specific material type. Specifically, the center coordinates of the grid cell are substituted into a material distribution function. This function, based on the charger's structural design drawings, returns the material identifier corresponding to that coordinate point. For example, the center coordinates of a certain grid cell are (x... i y j z k If the material distribution function determines that the location is in the copper wire region, then the mesh cell is marked as copper material.

[0050] After material type labeling is completed, the corresponding thermophysical parameters of the material are extracted from the material property database. The material property database stores the thermal conductivity of various materials at different temperatures. ,density and specific heat capacity Because the thermophysical parameters of some materials change significantly with temperature, interpolation calculations are required based on the current temperature value of the grid cells. For example, the thermal conductivity of copper is approximately 398 W / (m·K) at room temperature, but it decreases slightly with increasing temperature. Accurate thermal conductivity values ​​are extracted from the material property library based on the current temperature of the grid cells using linear or polynomial interpolation methods. Similarly, the density and specific heat capacity of the material corresponding to that grid cell are obtained. For the air domain inside the charger, since the thermophysical parameters of air are greatly affected by temperature and pressure, dynamic calculations are required based on the ideal gas law and empirical formulas. The extracted thermophysical parameters are then assigned to each grid cell, forming a spatially distributed thermophysical parameter field.

[0051] Based on the principle of heat conservation and Fourier's law of heat conduction, a partial differential equation is established. For any infinitesimal volume within the charger, the rate of change of its internal energy over time equals the heat inflow through the boundary minus the heat outflow, plus the heat generated by the internal heat source. Fourier's law of heat conduction states that the heat flux density vector is proportional to the temperature gradient, and the proportionality constant is negative for thermal conductivity. Based on these fundamental principles, a partial differential equation describing the evolution of the temperature field is established. The left side of the equation represents the rate of change of internal energy per unit volume within the grid cell, expressed as the partial derivative of temperature with respect to time. With material density and specific heat capacity The product of, i.e. The right-hand side of the equation describes the combined effect of heat conduction and heat source terms, including the divergence term of the temperature space gradient and the volume heat source term. The divergence term of the temperature space gradient is expressed as... ,in Represents the temperature gradient vector. Let be the thermal conductivity. For isotropic materials, this term can be expanded as follows: .

[0052] Body heat source item This represents the heat generation power density per unit volume per unit time, calculated from charging power time-series data. For the grid cell containing the power device, its volume heat source term equals the device's heat generation power divided by the volume occupied by the device. For example, a power MOSFET with a power loss of 15W has a package volume of... m³, then the volume heat source term of the corresponding grid element is approximately W / m³. For grid cells where non-heat-generating devices reside, the bulk heat source term is set to zero. The bulk heat source term is then superimposed as an independent source term onto the right-hand side of the partial differential equation to form the complete heat conduction equation. This equation describes the spatiotemporal evolution of the internal temperature field of the charger under the combined effects of heat conduction and internal heat sources.

[0053] To highlight the importance of heat flow convergence points, special processing is required for the grid cells containing these key locations. Using the previously calculated spatial coordinates of these convergence points, a topological mapping is employed to determine the grid cell number to which each convergence point belongs. This topological mapping is established through a spatial discretization scheme, dividing the continuous three-dimensional space into discrete grid cells, each with a unique number. For the spatial coordinates of the heat flow convergence points... Calculate the index of the grid cell in which it is located, assuming the grid is in , , The dimensions of the directions are respectively , , The grid cell index can be obtained by dividing the coordinate by the grid size and rounding it down. After determining the grid cell index, the corresponding unique number is extracted from the pre-established grid number mapping table.

[0054] Weighting coefficients are assigned to the grid cells containing each heat flow convergence point based on the magnitude of the gradient. A larger gradient magnitude indicates a more drastic temperature change and a higher degree of heat accumulation at that location, resulting in a larger corresponding weighting coefficient. The weighting coefficient can be calculated using a normalized gradient magnitude, which is obtained by dividing the gradient magnitude of each heat flow convergence point by the largest gradient magnitude among all heat flow convergence points, yielding a weight value between 0 and 1. To enhance the distinguishability of the weights, a nonlinear transformation, such as squaring or cubicing, can be applied to the normalized values. A correspondence between grid cell numbers and weighting coefficients is established, storing the grid cell number of each heat flow convergence point and its corresponding weighting coefficient in an associative data structure for easy subsequent retrieval and application.

[0055] Based on the established partial differential equations, the equations corresponding to the grid cells where the heat flow convergence points are located are modified. The corresponding discretized equations are located by the grid cell number, and the temperature change rate term on the left side of the equations is extracted. This item is then compared with its corresponding weighting coefficient. Multiplying them together yields the weighted rate of temperature change term. Replace the temperature change rate term on the left side of the original equation with this product, and replace the heat conduction term on the right side of the equation with this product. and body heat source item The equation remains unchanged. The corrected form is: The physical meaning of this correction is that when the weight coefficient is less than 1, it is equivalent to enhancing the thermal inertia of the grid cell, which suppresses the rate of temperature change, thus making the prediction of the temperature peak more conservative and safer in the numerical solution process; when the weight coefficient is greater than 1, it enhances the response speed of temperature changes, making the model more sensitive to the temperature evolution of the heat flow convergence point.

[0056] For grid cells where there are no heat flow convergence points, the original partial differential equations are kept unchanged, i.e., the weighting coefficients are set to 1 by default. The discretized equations corresponding to all grid cells are combined to form a thermal diffusion dynamics model describing the evolution of the entire charger's temperature field. When this model is represented in matrix form, the left side is the product of the mass matrix (including weight corrections) and the temperature change rate vector, and the right side is the product of the stiffness matrix and the temperature vector plus the heat source vector. The diagonal elements of the mass matrix correspond to the values ​​of each grid cell. ,in The volume represents the mesh element. The elements of the stiffness matrix are determined by the thermal conductivity and mesh geometry parameters, describing the thermal conduction coupling relationship between adjacent mesh elements. The elements of the heat source vector are the product of the volumetric heat source term and the volume of each mesh element.

[0057] In one optional implementation, the temperature evolution sequence of each grid cell over multiple time steps is predicted by iteratively solving the thermal diffusion kinetics model, including: Set the start and end times of the simulation, determine the total number of iterations based on the time span and preset time step, obtain the initial temperature data of each grid cell of the charger at the start time, identify the boundary grid cells, and establish boundary heat exchange conditions for the boundary grid cells based on the ambient temperature and convective heat transfer coefficient. The heat diffusion dynamics model is discretized in the time domain. The temperature-time partial derivative, spatial partial derivative, and volume heat source terms in the partial differential equations corresponding to each grid cell are extracted. The partial derivatives are converted into difference form using the finite difference method. An algebraic equation system is constructed with the current time step temperature as the known quantity and the next time step temperature as the unknown quantity. Initialize the iteration counter and start the iteration. In each iteration, read the density, specific heat capacity, thermal conductivity and volume heat source term values ​​of each grid cell, extract the current temperature value of each grid cell and its adjacent grid cells, and substitute them into the corresponding algebraic equation to solve for the temperature value of the next time step. Apply the boundary heat exchange condition constraint to the boundary grid cells. After completing the calculation of each grid cell, the temperature field of the next time step is formed. The iteration counter is incremented and it is determined whether the total number of iterations has been reached. If not, the temperature field of the next time step is used as the current temperature field to continue the iteration. If the total number of iterations has been reached, the iteration ends and the temperature fields of each time step are output to form a temperature evolution sequence.

[0058] After obtaining the thermal diffusion dynamics model based on partial differential equations, the model needs to be numerically solved to predict the temperature change trend at various locations inside the charger in the future. Before starting the simulation calculation, the time range needs to be determined, and the simulation start time is set to the current time t0, and the end time is set to t0. end The time span between the two is Based on the charger's thermal response characteristics and calculation accuracy requirements, a preset time step is determined. This step size is typically between 0.1 and 1 second; too large a step size can lead to numerical instability, while too small a step size can increase the computational burden. The total number of iterations is calculated by the ratio of the time span to the time step size. The results are rounded up to ensure complete coverage of the prediction period.

[0059] During the initialization phase, the measured temperature values ​​of each grid cell of the charger at the initial moment are obtained from the temperature sensor array to form the initial temperature distribution. subscript Indicates the first Each grid cell is assigned a number, and all grid cells are traversed. Boundary grid cells are identified by determining whether they are located on the charger casing surface or in direct contact with the external environment. The identified set of boundary grid cells is then processed. Based on the current ambient temperature and initial convective heat transfer coefficient Establish boundary heat exchange conditions. These conditions describe the heat exchange relationship between the boundary grid cells and the environment, expressed as the heat flux density transferred per unit area via convection being proportional to the temperature difference.

[0060] To achieve numerical solutions, the partial differential equations in the continuous-time domain need to be transformed into algebraic equations in the discrete-time domain. This involves extracting the partial differential equations corresponding to each grid cell in the thermal diffusion kinetics model, which contains three key terms: a temperature-time partial derivative term characterizing the rate of temperature change with time, a spatial partial derivative term describing the temperature gradient in various spatial directions, and a volumetric heat source term reflecting the heat generated by the power devices inside the charger. These partial derivative terms are discretized using the finite difference method, transforming the temperature-time partial derivative terms into a difference form. superscript Indicates the current time step, superscript Indicates the next time step. For the spatial partial derivatives, respectively in Cartesian coordinates... , , Central difference discretization is performed in three directions, with the first... Each grid cell in Taking the second-order partial derivative of the direction as an example, it can be expressed as follows: ,in for Grid spacing in the direction, and These respectively represent the mesh cell in Adjacent grids in the direction.

[0061] Through the discretization process described above, the original partial differential equation is transformed into a equation with respect to the temperature of each grid cell at the current time step. Given a known quantity, the temperature at the next time step. This is a system of algebraic equations with unknowns. The coefficient matrix of this system is determined by the material's thermal properties, mesh geometry, and time step. To improve numerical stability, an implicit difference scheme is used, meaning that the temperature value at the next time step is simultaneously influenced by the temperature of the cell itself and the temperatures of its adjacent cells at the next time step. This coupling relationship requires obtaining the updated values ​​of the entire temperature field through matrix solving methods.

[0062] Initialize the iteration counter The time-progression loop begins. In the [number]th... In each iteration, the physical properties of each grid cell are read, including material density. Specific heat capacity and thermal conductivity These parameters change with temperature, so it is necessary to refer to the current temperature value. Retrieve updated values ​​from the material property database. Simultaneously read the volumetric heat source items for each mesh element. This value is calculated by mapping the charging power timing data to the power density of the heating elements within each grid cell. Extracting the first... The current temperature values ​​of each grid cell and its six adjacent grid cells in six directions in three-dimensional space are denoted as follows: , , , , , and The three subscripts correspond to respectively , , Grid index of direction.

[0063] Substituting the above values ​​into the discretized algebraic equation corresponding to the grid cell, the temperature value for the next time step is calculated using numerical methods. For internal mesh elements, the right-hand side of the equation includes the temperature term for the current time step, the spatial difference term for the temperatures of adjacent mesh elements, and the contribution of the volume heat source term divided by the material heat capacity. For boundary mesh elements, in addition to the basic equations, a constraint on the boundary heat exchange condition must be applied. This constraint introduces the convective heat transfer between the boundary mesh element and the environment into the right-hand side of the equation as an additional heat flux term, with the magnitude of the correction term being... ,in The exposed area of ​​the boundary grid cell. The grid cell volume is defined. This treatment ensures that the temperature evolution at the boundary locations is simultaneously influenced by both internal thermal diffusion and external convective heat dissipation.

[0064] After completing the temperature calculation for all grid cells, the temperature of each grid cell will be... The values ​​are assembled into the complete temperature field distribution for the next time step. Increment the iteration counter And determine whether the current iteration count has reached the preset total number of iterations. ,like Then the temperature field of the next time step that was just calculated will be... As the new current temperature field Returning to the beginning of the iteration loop, continue calculating the temperature component for subsequent time steps. In each iteration, the volumetric heat source term... The data is updated based on the corresponding time values ​​of the charging power timing data to reflect the dynamic changes in the heat generation of power devices under the actual working conditions of the charger.

[0065] If the judgment finds Achieved If the iteration process terminates, then the time from the starting time has been obtained. until the end time The temperature field data for all discrete time steps between the given points are arranged in chronological order to form the temperature evolution sequence for each grid cell. This sequence clearly shows the temperature change trajectory at each location over time, providing complete predictive data support for the subsequent extraction of temperature peaks at heat flow convergence points.

[0066] After obtaining the temperature evolution sequence, the grid cells containing previously identified heat flow convergence points are analyzed in detail. The temperature time series of this grid cell is traversed, and extreme points in the sequence are identified by comparing the temperature values ​​of adjacent time steps. When the temperature value at a certain time step is greater than the temperature values ​​of the preceding and following time steps, that moment is determined to be a local temperature peak. The largest local peak is selected as the temperature peak of that grid cell within the prediction time period. At the same time, record the time when the peak occurs. Calculate the time difference between the peak occurrence time and the current time. This time difference represents the remaining warning time from the current state until the temperature reaches its maximum value, providing crucial time window information for initiating the heat dissipation control strategy. The entire iterative solution process transforms the complex partial differential equation of heat conduction into progressively advancing algebraic operations, enabling accurate prediction of the future evolution trend of the internal temperature field of the charger. This allows the control system to take proactive heat dissipation measures in advance before temperature anomalies occur.

[0067] In one optional implementation, based on the difference between the peak amplitude and the safe temperature limit, and in conjunction with the time difference, the convective heat transfer coefficient that the active cooling unit needs to apply at the boundary is calculated by inversely solving the thermal diffusion dynamics model, including: The target temperature drop is obtained by calculating the difference between the peak amplitude and the safe temperature limit. Based on the target temperature drop and the specific heat capacity, density and volume of the charger shell, the total heat to be released is calculated, and the average heat dissipation power requirement is obtained by combining the time difference. The boundary grid cells are located from the thermal diffusion dynamics model, and the thermal flux boundary conditions of the boundary grid cells are set in reverse. The boundary heat flux density is expressed as the product of the convective heat transfer coefficient and the difference between the boundary temperature and the ambient temperature. The thermal diffusion dynamics model is discretized in reverse time. The temperature field at the current moment is used as the initial condition, and the temperature change trajectory within the time difference is used as the constraint condition. An inverse optimization function is established with the convective heat transfer coefficient as the optimization variable and the temperature field satisfying the safe temperature limit as the objective. The gradient descent method is used to solve the inverse optimization function. In each iteration, the temperature field evolution under the current convective heat transfer coefficient is calculated, the deviation between the temperature field and the safe temperature limit is evaluated, and the sensitivity is calculated. The convective heat transfer coefficient value is updated according to the sensitivity. The iteration terminates when the temperature field meets the safe temperature limit, outputs the converged convective heat transfer coefficient value, and sends a heat dissipation control command containing the convective heat transfer coefficient value to the active heat dissipation unit.

[0068] In this specific embodiment, to deduce the convective heat transfer coefficient required by the active cooling unit based on the difference between the peak amplitude and the safe temperature limit, as well as the time difference, it is necessary to determine the target temperature drop amplitude. This is achieved by calculating the difference between the peak temperature of the heat flow convergence point extracted from the temperature evolution sequence and the preset safe temperature limit, thus obtaining the temperature excess that must be eliminated. This temperature excess is the target temperature drop amplitude. Safe temperature limits are typically set based on the temperature tolerance of the electronic components inside the charger. For power semiconductor devices, these limits are generally between 85°C and 105°C. For temperature-sensitive components such as electrolytic capacitors, the limits are more stringent. When the predicted peak temperature exceeds this limit, active cooling is required to bring the temperature back to a safe range.

[0069] When calculating the total heat to be released, the thermophysical parameters of the charger casing and its main internal heat-capacity components must be considered. Charger casings are typically made of aluminum alloy or engineering plastics. Aluminum alloy has a specific heat capacity of approximately 900 J / (kg·K) and a density of approximately 2700 kg / m³, while engineering plastics have a specific heat capacity of approximately 1200 J / (kg·K) and a density of approximately 1100 kg / m³. The casing volume can be calculated by measuring or consulting the geometric dimensions of the charger casing. The shell mass can then be obtained based on the material density. The total heat that needs to be released Through formula The calculation shows that, among which The specific heat capacity is given by the outer casing material. Considering that there are also components with certain heat capacity inside the charger, such as the PCB board and transformer core, the equivalent heat capacity of these components needs to be taken into account in the actual calculation. The total equivalent heat capacity is formed by weighted summation.

[0070] Average heat dissipation power requirement This is obtained by dividing the total heat by the available time difference, i.e. ,in The average heat dissipation power requirement represents the time difference between the predicted peak temperature and the current time. It represents the heat removal rate that the active cooling unit must maintain within the time difference. In practical applications, in order to cope with the uncertainty of the prediction model and the delay in the response of the cooling unit, a safety factor of 1.1 to 1.3 is usually multiplied on the calculated average heat dissipation power to ensure that the heat dissipation capacity has a certain redundancy.

[0071] Locating boundary mesh elements from the thermal diffusion kinetics model is a key step in the reverse solution. Boundary mesh elements refer to mesh elements that are in direct contact with the charger casing or form heat dissipation channels. The temperature changes of these elements are directly affected by external convective heat transfer. In 3D mesh generation, boundary mesh elements are usually located on the outer surface of the model. By traversing all mesh elements and determining whether their node coordinates are located on the predefined boundary surface, all boundary mesh elements can be filtered out. For chargers with forced convection cooling by fans, boundary mesh elements are mainly concentrated on the surface of the heat dissipation fins and near the ventilation holes. For natural convection cooling, the mesh elements on the entire casing surface need to be treated as boundary elements.

[0072] When reversing the setting of the heat flux boundary conditions for the boundary grid cells, the heat flux density at the boundary is... Expressed as convective heat transfer coefficient The product of the temperature difference, i.e. ,in The temperature of the boundary grid cells, The ambient temperature is used as the reference value. This representation conforms to Newton's law of cooling, simplifying the complex convective heat transfer process into a linear relationship. It's important to note that the convective heat transfer coefficient... It is not a constant; its value is affected by various factors such as fluid velocity, fluid properties, and boundary geometry. Under natural convection conditions, Typical values ​​range from 5 to 25 W / (m²·K), and can reach 50 to 250 W / (m²·K) under forced convection conditions.

[0073] Time-reverse discretization transforms the temperature evolution process in the continuous time domain into iterative calculations in discrete time steps. Unlike the forward solution, which progresses from the initial moment to the future, the reverse solution needs to backtrack from the predicted peak occurrence time to the current moment. The time difference... Divided into There are discrete time steps, each with a time step size of [number]. The temperature field at the current moment. As the initial condition for the reverse solution, the temperature at the heat flow convergence point must not exceed the safe temperature limit at the time of peak occurrence. This constitutes the terminal constraint. After time is discretized in reverse, the partial differential equation of heat diffusion is solved numerically using an implicit difference scheme to ensure numerical stability during the backtracking process.

[0074] When establishing the inverse optimization function, the convective heat transfer coefficient is... As an optimization variable, the objective is to ensure that, after applying the specified convective heat transfer coefficient, the evolution of the temperature field over the time difference meets the requirements of the safe temperature limit. The optimization function can be expressed as minimizing the objective function. ,in This refers to the number of critical grid cells located at and around the heat flow convergence point. For the first The grid cell in the first... The predicted temperature at each time step (i.e., the moment when the peak occurs). The objective function is essentially to minimize the sum of squares of temperature exceedances. When the temperature of all critical grid cells does not exceed the safe limit, the objective function value approaches zero.

[0075] When solving the inverse optimization function using the gradient descent method, it is necessary to calculate the gradient of the objective function with respect to the convective heat transfer coefficient. Due to the complex nonlinear coupling relationship between the temperature field and the convective heat transfer coefficient, direct analytical differentiation is difficult; therefore, the finite difference method is typically used for numerical gradient calculation. In the... In this iteration, the current convective heat transfer coefficient is By applying a small perturbation to this value (usually taken) Calculate (0.1% to 1%) respectively. and Corresponding objective function value and The gradient is approximately equal to .

[0076] Each iteration requires a complete solution to the thermal diffusion kinetics model, calculating the current convective heat transfer coefficient. This describes the evolution of the temperature field from the current moment to the moment of peak occurrence under the influence of the heat transfer coefficient. In practice, the current convective heat transfer coefficient is substituted into the boundary conditions, and the calculation is performed step-by-step according to the time step size to obtain the temperature distribution of each grid cell at all time steps. Particular attention is paid to the temperature values ​​of the heat flow convergence point and surrounding grid cells at the moment of peak occurrence. These temperature values ​​are compared with the safe temperature limit to calculate the temperature deviation. The magnitude of the temperature deviation reflects the applicability of the current convective heat transfer coefficient. If the deviation is positive and large, it indicates insufficient heat dissipation capacity, requiring an increase in the convective heat transfer coefficient; if the deviation is negative, it indicates excessive heat dissipation, and the convective heat transfer coefficient can be appropriately reduced to save energy.

[0077] Sensitivity calculations are used to assess the impact of small changes in the convective heat transfer coefficient on the temperature field. Sensitivity is defined as the ratio of the temperature change to the change in the convective heat transfer coefficient, i.e. High sensitivity means that adjusting the convective heat transfer coefficient can significantly change the temperature distribution, allowing for a larger iteration step size to accelerate convergence; low sensitivity requires careful adjustment to avoid numerical oscillations. Based on the calculated gradient and sensitivity, a gradient descent update rule is used to adjust the convective heat transfer coefficient, with the update formula being: ,in The learning rate is typically set initially to 0.01 to 0.1, and can be adjusted using an adaptive strategy during iteration.

[0078] The iteration termination condition is set as follows: the temperature field meets the safe temperature limit. Specifically, the criteria are that the predicted temperature of all key grid cells at the peak time does not exceed the safe temperature limit, and the objective function value is less than a preset convergence threshold, which is generally set to 0.01℃ to 0.1℃. A maximum iteration limit is also set to prevent infinite iterations due to improper model parameter settings or unattainable physical conditions. When the termination condition is met, the convective heat transfer coefficient obtained in the current iteration is output as the final result.

[0079] A heat dissipation control command containing the convective heat transfer coefficient value is sent to the active cooling unit. The active cooling unit is an integrated system consisting of a fan, heat pipes, and heat dissipation fins. The fan has a diameter of 40mm, a maximum speed of 6000r / min, and a flow rate of 16m³ / min. 3The heat pipes are made of pure copper, with a diameter of 6 mm, and the internal working fluid is water. The heat dissipation fins are made of aluminum alloy, with a thickness of 0.5 mm and a spacing of 1.5 mm. Cooling control commands are transmitted to the controller of the active cooling unit via an internal bus using a serial communication protocol. After receiving the convection heat transfer coefficient value, the cooling controller converts it into specific execution parameters, establishing a mapping relationship between the convection heat transfer coefficient and fan speed. For example, when the convection heat transfer coefficient is 40 W / (m²·K), the fan speed is set to 4000 r / min. A fan start-up acceleration curve is set to avoid current surges and sudden noise changes; the acceleration time is typically 0.5 s to 1 s. The fan speed is controlled by a pulse width modulation signal with a frequency of 25 kHz and an accuracy of 0.5%. Feedback monitoring of the cooling effect is achieved in real-time through a temperature sensor at a frequency of 5 Hz. The actual temperature drop rate is compared with the expected target; when the deviation exceeds 15%, the convection heat transfer coefficient is automatically adjusted, forming a closed-loop control.

[0080] In one optional implementation, the gradient descent method is used to solve the inverse optimization function. In each iteration, the temperature field evolution under the current convective heat transfer coefficient is calculated, the deviation of the temperature field from the safe temperature limit is evaluated, and the sensitivity is calculated. The convective heat transfer coefficient value is updated based on the sensitivity, including: Initialize the convective heat transfer coefficient and set the learning rate and gradient convergence threshold. In each iteration, use the current convective heat transfer coefficient as the boundary parameter to calculate the temperature field evolution process within the time difference. Extract the highest temperature value in the temperature field at the termination time and subtract it from the safe temperature limit to obtain the current temperature deviation. Apply positive and negative small perturbations to the current convective heat transfer coefficient respectively, calculate the temperature field evolution under positive and negative perturbation conditions, extract the highest temperature values ​​of positive and negative perturbations, calculate the difference between the two and the safe temperature limit, divide the difference between the positive and negative perturbation temperature deviations by the total perturbation amount, and obtain the sensitivity by the central difference method. The sensitivity value is multiplied by the learning rate to obtain the update step size. The update step size is subtracted from the current convective heat transfer coefficient to obtain the new convective heat transfer coefficient value. The absolute value of the sensitivity is calculated and it is determined whether it is less than the gradient convergence threshold. If it is less than the threshold, the iteration is terminated and the new convective heat transfer coefficient value is output. Otherwise, the new convective heat transfer coefficient value is used as the current convective heat transfer coefficient for the next iteration and the iteration continues.

[0081] After obtaining the mathematical expression of the inverse optimization function, a numerical iterative algorithm is needed for practical solution. Gradient descent, as an efficient and stable optimization method, is particularly suitable for boundary inversion problems involving continuous variables such as the convective heat transfer coefficient. In practical applications, a reasonable initial value for the convective heat transfer coefficient needs to be set before the algorithm begins. Typically, a typical value of the natural convective heat transfer coefficient can be chosen as the initial guess, such as 15 W / (m²·K) in air. At the same time, the learning rate is set to 0.02, and the gradient convergence threshold is set to 0.001 K / (W / (m²·K)). The setting of these parameters directly affects the convergence speed and final accuracy of subsequent iterations. If the initial value is too large, the algorithm will oscillate in the solution space; if it is too small, the convergence speed will be significantly reduced.

[0082] After entering the iterative process, the current convective heat transfer coefficient is used as the boundary parameter and substituted into the established thermal diffusion dynamics model for numerical solution. The temperature field evolution calculation at this point covers the entire time interval from the current moment to the predicted peak temperature, which is the previously calculated time difference. Under the finite element method (FEM) framework, the energy conservation equation of the boundary grid element includes a convective heat transfer term, which is expressed as the difference between the boundary temperature and the ambient temperature multiplied by the convective heat transfer coefficient. This term is applied as a boundary flux condition to the discrete equation of the boundary element. Through time stepping, the temperature field distribution at each time node is calculated progressively until the predicted moment is reached. At the final temperature moment, the temperature values ​​of all grid elements are traversed, and the maximum value is extracted as the highest temperature value of the current iteration. This temperature value is compared with the system's preset safe temperature limit, and the difference between the two is the current temperature deviation. This deviation directly reflects the gap between the current boundary heat dissipation capacity and the actual demand.

[0083] To accurately assess the impact of minute changes in the convective heat transfer coefficient on the temperature field, a sensitivity calculation mechanism is needed. Sensitivity essentially represents the partial derivative of the objective function with respect to the design variables; in this scenario, it is the rate of change of the convective heat transfer coefficient with respect to the temperature deviation. A numerical differential method is used to calculate this by applying a small positive perturbation to the current convective heat transfer coefficient. The perturbation magnitude is typically set to 1% to 3% of the current value. For example, if the current convective heat transfer coefficient is 50 W / (m²·K), the positive perturbation can be set to 1 W / (m²·K), bringing the perturbed value to 51 W / (m²·K). The temperature field evolution is then recalculated using the boundary parameters after this perturbation, covering the entire time difference interval. The highest temperature value of the temperature field at the termination time is extracted as the highest temperature value of the positive perturbation, and the difference between this and the safe temperature limit is calculated to obtain the positive perturbation temperature deviation.

[0084] Similarly, a small negative perturbation is applied to the current convective heat transfer coefficient. The perturbation amplitude is the same as the positive perturbation but in the opposite direction, reducing the convective heat transfer coefficient to 49 W / (m²·K). A complete temperature field evolution calculation is then performed again, and the highest temperature value at the termination point is extracted as the highest temperature value of the negative perturbation. The difference between this value and the safe temperature limit is calculated to obtain the negative perturbation temperature deviation. The sensitivity is then calculated using the central difference method, with the specific expression as follows: , in, Indicates sensitivity. This indicates a positive disturbance temperature deviation. This indicates a negative disturbance temperature deviation. This represents the one-sided disturbance, and the total disturbance is... The central difference method has higher numerical accuracy than the one-sided difference method, which can effectively reduce the influence of numerical truncation error and make the sensitivity calculation results more accurate and reliable.

[0085] After obtaining the sensitivity value, the convective heat transfer coefficient can be updated. The sensitivity value is multiplied by the preset learning rate to obtain the update step size for this iteration. The learning rate controls the aggressiveness of the parameter update; a larger learning rate can accelerate convergence but may cause the solution to oscillate around the optimal value, while a smaller learning rate, although stable, will prolong the number of iterations. In applications with certain real-time requirements, such as charger heat dissipation control, a learning rate value between 0.01 and 0.05 is typically chosen. After calculating the update step size, this step size value is subtracted from the current convective heat transfer coefficient to obtain the new convective heat transfer coefficient value. Subtraction is used here because the gradient descent method searches along the negative direction of the objective function's gradient. When the sensitivity is positive, it means that increasing the convective heat transfer coefficient will increase the temperature deviation, therefore the convective heat transfer coefficient needs to be decreased; conversely, the opposite is also true.

[0086] After updating the parameters, it is necessary to determine whether the algorithm has converged to the vicinity of the optimal solution. The convergence criterion is based on the absolute value of the sensitivity. The absolute value of the sensitivity obtained in the current iteration is calculated and compared with the preset gradient convergence threshold. When the absolute value of the sensitivity is less than the gradient convergence threshold, it indicates that the slope of the objective function at the current position has approached zero, the parameters are in a locally flat region, and the improvement brought by continuing the iteration is extremely limited. At this time, the algorithm is determined to have converged and the iteration process is terminated. The current new convection heat transfer coefficient value is output as the final optimization result. This value is the target boundary heat transfer capacity that the active heat dissipation unit needs to achieve. If the absolute value of the sensitivity is still greater than or equal to the gradient convergence threshold, it means that the parameters have not yet reached the optimal position, and the objective function still has obvious room for decrease. At this time, the newly calculated convection heat transfer coefficient value is used as the current convection heat transfer coefficient for the next iteration, and the process returns to the starting step of the iteration process to recalculate the temperature field evolution, evaluate the temperature deviation, and perform sensitivity analysis.

[0087] In practical engineering implementation, it is also necessary to set a maximum iteration limit for the iterative process to prevent the algorithm from getting stuck in an infinite loop under certain ill-conditioned conditions. Typically, the maximum number of iterations is set to 500 to 1000. If convergence is not achieved after reaching this limit, the algorithm is forcibly terminated, the current result is output, and an anomaly log is recorded for subsequent analysis. Furthermore, to improve numerical stability, a limit can be set on the update magnitude of the convective heat transfer coefficient. For example, the update magnitude should not exceed 30% of the current value in a single update to avoid excessive parameter jumps that could lead to computational divergence. During the temperature field evolution calculation, the time step must satisfy the Courant-Friedrichs-Lewy stability condition to ensure the numerical stability of the explicit time integration scheme. The time step is usually automatically adjusted based on the mesh size and the material's thermal diffusivity, with typical values ​​between 0.01 seconds and 0.1 seconds.

[0088] Through the complete gradient descent iterative process described above, the convective heat transfer coefficient that meets safety temperature requirements can be solved quickly while ensuring computational accuracy. This value is then converted into actual control parameters for the active cooling unit, such as fan speed or water pump flow rate, driving the cooling system to achieve the expected heat transfer capacity, thereby realizing active temperature control of the charger under peak power operation. The entire reverse solution process combines thermodynamic physical models with numerical optimization algorithms, ensuring both theoretical rigor and engineering feasibility, providing reliable technical support for the intelligent thermal management of chargers.

[0089] A second aspect of the present invention provides a dual-mode heat dissipation charger control system based on intelligent algorithms, comprising: The data acquisition unit is used to acquire temperature sensor array data and charging power timing data of the charger; The temperature field construction unit is used to perform spatial interpolation processing on the temperature sensor array data, construct a three-dimensional temperature field distribution map inside the charger, and calculate the temperature gradient vector field through the gradient operator to identify heat flow convergence points. The heat source modeling unit is used to divide the three-dimensional temperature field distribution map into multiple grid cells using the finite element method, and add volume heat source terms to the corresponding grid cells according to the charging power time series data, and perform weighted summation on the grid cells where the heat flow convergence point is located to establish a thermal diffusion dynamics model based on partial differential equations. The temperature prediction unit is used to predict the temperature evolution sequence of each grid cell in the future multiple time steps by iteratively solving the heat diffusion dynamics model, extract the temperature peak and occurrence time of the grid cell where the heat flow convergence point is located, and calculate the time difference between the occurrence time and the current time. The heat dissipation decision unit is used to activate the active heat dissipation mode when the time difference is less than the advance threshold, and to back-calculate the convective heat transfer coefficient that the active heat dissipation unit needs to apply at the boundary by inversely solving the thermal diffusion dynamics model based on the difference between the peak amplitude and the safe temperature limit and the time difference. The heat dissipation control unit is used to control the operating parameters of the active heat dissipation unit to achieve the convective heat transfer coefficient, and to collect temperature field data after heat dissipation to update the thermal diffusion dynamics model.

[0090] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0091] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0092] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A dual-mode heat dissipation charger control method based on intelligent algorithms, characterized in that, include: Acquire temperature sensor array data and charging power timing data of the charger; Spatial interpolation is performed on the temperature sensor array data to construct a three-dimensional temperature field distribution map inside the charger, and the temperature gradient vector field is calculated by the gradient operator to identify the heat flow convergence point. The three-dimensional temperature field distribution map is divided into multiple grid cells using the finite element method. A volume heat source term is added to the corresponding grid cell according to the charging power time series data. The grid cell where the heat flow convergence point is located is weighted to establish a thermal diffusion dynamics model based on partial differential equations. By iteratively solving the thermal diffusion dynamics model, the temperature evolution sequence of each grid cell in the future multiple time steps is predicted, the temperature peak value and occurrence time of the grid cell where the heat flow convergence point is located are extracted, and the time difference between the occurrence time and the current time is calculated. When the time difference is less than the lead threshold, the active heat dissipation mode is activated, and based on the difference between the peak amplitude and the safe temperature limit, the convective heat transfer coefficient that the active heat dissipation unit needs to apply at the boundary is calculated by inversely solving the thermal diffusion dynamics model in combination with the time difference. The operating parameters of the active heat dissipation unit are controlled to achieve the convective heat transfer coefficient, and the temperature field data after heat dissipation is collected to update the thermal diffusion dynamics model.

2. The method according to claim 1, characterized in that, Spatial interpolation is performed on the temperature sensor array data to construct a three-dimensional temperature field distribution map inside the charger. A temperature gradient vector field is calculated using a gradient operator to identify heat flow convergence points, including: Using the spatial coordinates of each temperature sensor in the temperature sensor array data as sampling points, the spatial distance between any two sampling points is calculated, and a semi-variogram function is established based on the spatial distance. The spatial autocorrelation weight matrix between each sampling point is calculated using the semi-variogram function. The inverse matrix of the spatial autocorrelation weight matrix is ​​then solved. The inverse matrix is ​​multiplied by the temperature measurement value of each sampling point to obtain the interpolation coefficient vector. A regular grid structure is established inside the charger, the spatial distance between each grid node and each sampling point is calculated, the spatial distance is substituted into the semi-variogram function to obtain the correlation weight, and the weighted sum is performed with the interpolation coefficient vector to obtain the temperature estimate, and a three-dimensional temperature field distribution map is constructed. The first-order temperature derivatives of each grid node in the three-dimensional temperature field distribution map in the three coordinate axes are calculated using the central difference method and combined into a temperature gradient vector to generate the temperature gradient vector field. Calculate the gradient vector direction angle of each grid node in the temperature gradient vector field, and count the proportion of gradient vectors pointing to each grid node within the neighborhood. When the proportion exceeds the convergence threshold, the grid node is identified as a heat flow convergence point.

3. The method according to claim 1, characterized in that, The finite element method is used to divide the three-dimensional temperature field distribution map into multiple grid cells. Based on the charging power time-series data, a volume heat source term is added to the corresponding grid cell. Weights are applied to the grid cells containing the heat flow convergence point to establish a thermal diffusion dynamics model based on partial differential equations, including: The three-dimensional temperature field distribution map is spatially discretized, and the internal space of the charger is divided into multiple hexahedral grid units. Each grid unit is assigned a unit number and node coordinates, and a topological mapping relationship between the grid unit and the spatial location is established. The charging power value at each moment in the charging power time series data is obtained. Based on the spatial distribution of the heating elements inside the charger, the grid cell number corresponding to each heating element is determined. The charging power value is converted into a unit volume power density and used as the volume heat source item of the corresponding grid cell. Extract the spatial coordinates of the heat flow convergence point, determine the grid cell to which the heat flow convergence point belongs, calculate the absolute value of the divergence of the grid cell to which each heat flow convergence point belongs and normalize it to obtain the weight coefficient of the corresponding grid cell. A partial differential equation is constructed based on Fourier's law of heat conduction, and the volumetric heat source term is added to the source term of the partial differential equation. The temperature change rate term of the grid cell where the heat flow converges is multiplied by the weighting coefficient to form a thermal diffusion dynamics model.

4. The method according to claim 3, characterized in that, A partial differential equation is constructed based on Fourier's law of heat conduction, and the volumetric heat source term is added to the source term of the partial differential equation. The temperature change rate term of the grid cell where the heat flow converges is located is multiplied by the weighting coefficient to form a thermal diffusion dynamics model, including: Based on the topological mapping relationship between grid cells and spatial locations, the material region of each grid cell is determined, and the thermal conductivity, density and specific heat capacity of the corresponding material are extracted from the material property library to assign thermophysical parameters to each grid cell. A partial differential equation is established based on the principle of heat conservation and Fourier's law of heat conduction. The left side of the equation is the product of the rate of change of temperature with respect to time, density, and specific heat capacity. The right side of the equation is the combination of the spatial temperature gradient and thermal conductivity. The volume heat source term is superimposed on the right side of the equation as an independent source term. The spatial coordinates of the heat flow convergence points are retrieved through the topological mapping relationship to determine the grid cell number to which each heat flow convergence point belongs, the corresponding weight coefficients are extracted, and a correspondence between the grid cell number and the weight coefficients is established. Locate the equation corresponding to the grid cell where the heat flow convergence point is located in the partial differential equation, extract the temperature change rate term on the left side of the equation and multiply it with the weighting coefficient, replace the original temperature change rate term with the product, and keep the terms on the right side of the equation unchanged to form a thermal diffusion dynamics model.

5. The method according to claim 1, characterized in that, By iteratively solving the thermal diffusion kinetics model, the temperature evolution sequence of each grid cell over multiple time steps is predicted, including: Set the start and end times of the simulation, determine the total number of iterations based on the time span and preset time step, obtain the initial temperature data of each grid cell of the charger at the start time, identify the boundary grid cells, and establish boundary heat exchange conditions for the boundary grid cells based on the ambient temperature and convective heat transfer coefficient. The heat diffusion dynamics model is discretized in the time domain. The temperature-time partial derivative, spatial partial derivative, and volume heat source terms in the partial differential equations corresponding to each grid cell are extracted. The partial derivatives are converted into difference form using the finite difference method. An algebraic equation system is constructed with the current time step temperature as the known quantity and the next time step temperature as the unknown quantity. Initialize the iteration counter and start the iteration. In each iteration, read the density, specific heat capacity, thermal conductivity and volume heat source term values ​​of each grid cell, extract the current temperature value of each grid cell and its adjacent grid cells, and substitute them into the corresponding algebraic equation to solve for the temperature value of the next time step. Apply the boundary heat exchange condition constraint to the boundary grid cells. After completing the calculation of each grid cell, the temperature field of the next time step is formed. The iteration counter is incremented and it is determined whether the total number of iterations has been reached. If not, the temperature field of the next time step is used as the current temperature field to continue the iteration. If the total number of iterations has been reached, the iteration ends and the temperature fields of each time step are output to form a temperature evolution sequence.

6. The method according to claim 1, characterized in that, Based on the difference between the peak amplitude and the safe temperature limit, and combined with the time difference, the convective heat transfer coefficient that the active cooling unit needs to apply at the boundary is calculated by inversely solving the thermal diffusion dynamics model, including: The target temperature drop is obtained by calculating the difference between the peak amplitude and the safe temperature limit. Based on the target temperature drop and the specific heat capacity, density and volume of the charger shell, the total heat to be released is calculated, and the average heat dissipation power requirement is obtained by combining the time difference. The boundary grid cells are located from the thermal diffusion dynamics model, and the thermal flux boundary conditions of the boundary grid cells are set in reverse. The boundary heat flux density is expressed as the product of the convective heat transfer coefficient and the difference between the boundary temperature and the ambient temperature. The thermal diffusion dynamics model is discretized in reverse time. The temperature field at the current moment is used as the initial condition, and the temperature change trajectory within the time difference is used as the constraint condition. An inverse optimization function is established with the convective heat transfer coefficient as the optimization variable and the temperature field satisfying the safe temperature limit as the objective. The gradient descent method is used to solve the inverse optimization function. In each iteration, the temperature field evolution under the current convective heat transfer coefficient is calculated, the deviation between the temperature field and the safe temperature limit is evaluated, and the sensitivity is calculated. The convective heat transfer coefficient value is updated according to the sensitivity. The iteration terminates when the temperature field meets the safe temperature limit, outputs the converged convective heat transfer coefficient value, and sends a heat dissipation control command containing the convective heat transfer coefficient value to the active heat dissipation unit.

7. The method according to claim 6, characterized in that, The gradient descent method is used to solve the inverse optimization function. In each iteration, the temperature field evolution under the current convective heat transfer coefficient is calculated, the deviation of the temperature field from the safe temperature limit is evaluated, and the sensitivity is calculated. The convective heat transfer coefficient value is updated based on the sensitivity, including: Initialize the convective heat transfer coefficient and set the learning rate and gradient convergence threshold. In each iteration, use the current convective heat transfer coefficient as the boundary parameter to calculate the temperature field evolution process within the time difference. Extract the highest temperature value in the temperature field at the termination time and subtract it from the safe temperature limit to obtain the current temperature deviation. Apply positive and negative small perturbations to the current convective heat transfer coefficient respectively, calculate the temperature field evolution under positive and negative perturbation conditions, extract the highest temperature values ​​of positive and negative perturbations, calculate the difference between the two and the safe temperature limit, divide the difference between the positive and negative perturbation temperature deviations by the total perturbation amount, and obtain the sensitivity by the central difference method. The sensitivity value is multiplied by the learning rate to obtain the update step size. The update step size is subtracted from the current convective heat transfer coefficient to obtain the new convective heat transfer coefficient value. The absolute value of the sensitivity is calculated and it is determined whether it is less than the gradient convergence threshold. If it is less than the threshold, the iteration is terminated and the new convective heat transfer coefficient value is output. Otherwise, the new convective heat transfer coefficient value is used as the current convective heat transfer coefficient for the next iteration and the iteration continues.

8. A dual-mode heat dissipation charger control system based on intelligent algorithms, used to implement the method as described in any one of claims 1-7, characterized in that, include: The data acquisition unit is used to acquire temperature sensor array data and charging power timing data of the charger; The temperature field construction unit is used to perform spatial interpolation processing on the temperature sensor array data, construct a three-dimensional temperature field distribution map inside the charger, and calculate the temperature gradient vector field through the gradient operator to identify heat flow convergence points. The heat source modeling unit is used to divide the three-dimensional temperature field distribution map into multiple grid cells using the finite element method, and add volume heat source terms to the corresponding grid cells according to the charging power time series data, and perform weighted summation on the grid cells where the heat flow convergence point is located to establish a thermal diffusion dynamics model based on partial differential equations. The temperature prediction unit is used to predict the temperature evolution sequence of each grid cell in the future multiple time steps by iteratively solving the heat diffusion dynamics model, extract the temperature peak and occurrence time of the grid cell where the heat flow convergence point is located, and calculate the time difference between the occurrence time and the current time. The heat dissipation decision unit is used to activate the active heat dissipation mode when the time difference is less than the advance threshold, and to back-calculate the convective heat transfer coefficient that the active heat dissipation unit needs to apply at the boundary by inversely solving the thermal diffusion dynamics model based on the difference between the peak amplitude and the safe temperature limit and the time difference. The heat dissipation control unit is used to control the operating parameters of the active heat dissipation unit to achieve the convective heat transfer coefficient, and to collect temperature field data after heat dissipation to update the thermal diffusion dynamics model.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images