Charging pile power panel overload protection method, device and equipment and storage medium
Through the analysis of thermal characteristics of the charging pile power PCBA board and multi-level PWM control, combined with machine learning and model prediction control, the problems of overload and overheating during charging are solved, the protection reliability and system efficiency are improved, and misjudgment is reduced and missed judgment rate.
Patent Information
- Application Number
- CN202411500555.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-25
- Publication Date
- 2024-11-26
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The PCBA board of the charging pile power supply is prone to overload and overheating during charging. The traditional overload protection method cannot dynamically adjust the protection parameters, resulting in insufficient protection or excessive protection. It is difficult for single-point temperature monitoring to accurately reflect the overall thermal state, and it is impossible to predict and prevent in advance. Potential overload risk, and PWM control has large switching losses and electromagnetic interference.
By performing thermal characteristics analysis on PCBA boards, dividing thermal calculation nodes and establishing heat transfer equations, calculating temperature response and dynamic temperature rise characteristic curves, solving the maximum safe current value with component parameters, analyzing historical data using machine learning algorithms, and establishing overload The risk assessment model uses a multi-level PWM control strategy to dynamically adjust the duty cycle and frequency at the zero-current switching point, and optimize the control parameters based on the model prediction control and sliding time window algorithm.
Improve the reliability of overload risk determination, reduce the misjudgment rate and leakage rate, significantly reduce switching losses and electromagnetic interference, improve the energy conversion efficiency and real-time control capabilities of the system, and avoid the protection blind spots and temperature monitoring of traditional methods. limitation.
Smart Images

Figure CN119017981A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power board overload protection, and in particular to a charging pile power board overload protection method, device, equipment and storage medium. Background Art
[0002] With the rapid development of the electric vehicle industry, the safe and stable operation of charging piles has become a focus of attention. During the charging process, the charging pile power PCBA board is under the working pressure of high current and high power, and overload and overheating problems occur frequently, which seriously threatens the service life and operation safety of the charging pile.
[0003] Traditional charging pile overload protection methods mainly rely on preset fixed thresholds, and are unable to dynamically adjust protection parameters according to the actual working environment and load conditions, resulting in problems of insufficient or excessive protection. At the same time, due to the uneven temperature distribution and obvious hot spot effect during the charging process, single-point temperature monitoring is difficult to accurately reflect the overall thermal state of the PCBA board, which can easily cause local overheating damage. In addition, the existing overload protection strategy lacks in-depth analysis of historical data and intelligent early warning mechanisms, and is unable to predict and prevent potential overload risks in advance. In terms of PWM control, simple switch control methods will produce large switching losses and electromagnetic interference, affecting charging efficiency and grid quality. Summary of the invention
[0004] The present invention provides a charging pile power board overload protection method, device, equipment and storage medium, which are used to improve the reliability of overload risk judgment and reduce the misjudgment rate and missed judgment rate.
[0005] In a first aspect, the present invention provides a charging pile power board overload protection method, the charging pile power board overload protection method comprising: Perform thermal characteristic analysis on the PCBA board of the charging pile power supply, divide the PCBA board into multiple thermal calculation nodes, and establish the corresponding heat transfer equations; According to the heat transfer equation group, the temperature response of the heat calculation node is calculated, and convergence is determined according to the temperature difference between two adjacent iterations being less than a preset threshold, so as to obtain a dynamic temperature rise characteristic curve; The dynamic temperature rise characteristic curve is combined with the rated parameters of the components to solve the current threshold under the temperature limit condition, and obtain the maximum safe current value of each thermal calculation node; The charging pile test platform is used to collect the original multidimensional data set, and cluster analysis and overload feature recognition are performed to obtain the target overload feature vector; Inputting the target overload feature vector into an overload risk assessment model to perform overload risk assessment to obtain an overload risk determination result; Based on the maximum safe current value and the overload risk determination result, a multi-level PWM control strategy is adopted to dynamically adjust the PWM duty cycle and frequency at the zero current switching point to output a charging control signal.
[0006] In a second aspect, the present invention provides a charging pile power board overload protection device, the charging pile power board overload protection device comprising: Establish a module for performing thermal characteristic analysis on the PCBA board of the charging pile power supply, dividing the PCBA board into multiple thermal calculation nodes, and establishing a corresponding set of heat transfer equations; A calculation module, used to calculate the temperature response of the thermal calculation node according to the heat transfer equation group, and determine convergence according to the temperature difference between two adjacent iterations being less than a preset threshold, so as to obtain a dynamic temperature rise characteristic curve; A solution module, used for combining the dynamic temperature rise characteristic curve with the rated parameters of the components, solving the current threshold under the temperature limit condition, and obtaining the maximum safe current value of each thermal calculation node; The recognition module is used to collect the original multidimensional data set using the charging pile test platform, and perform cluster analysis and overload feature recognition to obtain the target overload feature vector; An evaluation module, used for inputting the target overload feature vector into an overload risk evaluation model to perform overload risk evaluation and obtain an overload risk determination result; The output module is used to dynamically adjust the PWM duty cycle and frequency at the zero current switching point based on the maximum safe current value and the overload risk determination result, and output a charging control signal by adopting a multi-level PWM control strategy.
[0007] The third aspect of the present invention provides a charging pile power board overload protection device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the charging pile power board overload protection device executes the above-mentioned charging pile power board overload protection method.
[0008] A fourth aspect of the present invention provides a computer-readable storage medium, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes the above-mentioned charging pile power board overload protection method.
[0009] In the technical solution provided by the present invention, by establishing a multi-node thermal network model and a group of heat transfer equations, accurate calculation and dynamic monitoring of the temperature field of the PCBA board are achieved, the limitations of the traditional single-point temperature measurement method are avoided, and the accuracy of temperature monitoring is improved; the Lagrange multiplier method is used to solve the maximum safe current value, and the current threshold is accurately calculated in combination with the rated parameters of the components and the real-time temperature rise characteristics, avoiding the protection blind spot caused by the fixed threshold; a machine learning algorithm is introduced to analyze and extract features of historical data, and a multi-dimensional overload risk assessment model is established, so that the system has intelligent early warning capabilities; a multi-level PWM control strategy is designed, and the switching loss and electromagnetic interference are significantly reduced through zero current switching and dynamic parameter optimization, thereby improving the energy conversion efficiency of the system; the model predictive control and sliding time window algorithm are adopted to achieve early prediction of the temperature change trend and dynamic optimization of the control parameters, thereby enhancing the real-time control capability of the system; the Bayesian fusion algorithm is used to integrate multi-source information, thereby improving the reliability of overload risk judgment and reducing the misjudgment rate and missed judgment rate.
[0010] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.
[0011] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0012] Figure 1 A schematic diagram of an embodiment of a method for overload protection of a charging pile power board in an embodiment of the present invention; Figure 2 A schematic diagram of an embodiment of an overload protection device for a charging pile power board in an embodiment of the present invention; Figure 3 It is a schematic diagram of an embodiment of the overload protection device of the charging pile power board in an embodiment of the present invention. DETAILED DESCRIPTION
[0013] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0014] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device end including a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or device ends.
[0015] To facilitate understanding of this embodiment, a charging pile power board overload protection method disclosed in an embodiment of the present invention is first described in detail. Figure 1 As shown, the method comprises the following steps: 101. Perform thermal characteristic analysis on the PCBA board of the charging pile power supply, divide the PCBA board into multiple thermal calculation nodes, and establish the corresponding heat transfer equations; It is understandable that the execution subject of the present invention may be a charging pile power board overload protection device, or a terminal or a server, which is not specifically limited here. The embodiment of the present invention is described by taking a server as an execution subject as an example.
[0016] Specifically, the heat source position on the PCBA board is accurately located, including the main heat-generating components such as power devices, transformers and inductors. The positions of all key components on the PCBA board are scanned by high-precision scanning equipment to obtain the specific layout coordinate data of each component. According to the component layout coordinate data, the PCBA board is divided into N×M grid units to obtain the initial grid division result. The PCBA board is evenly divided into several basic calculation areas so that each grid unit can reflect the thermal characteristics of the local area. The grid units containing key components in the initial grid division result are subdivided to more accurately characterize the thermal distribution characteristics of the components in the thermal calculation and obtain the thermal calculation nodes. The material parameters of each layer of the PCBA board and the thickness data of each layer are collected. The PCBA board is composed of multiple layers of different materials, and the thermal conductivity between different layers is significantly different. According to the material properties and thickness data of each layer, the thermal conductivity matrix between PCB layers is established to accurately describe the heat transfer between the layers inside the board. Based on the thermal conductivity matrix, the heat transfer equation between the thermal calculation nodes is established to describe the flow relationship of heat between the thermal calculation nodes. These equations are used to simulate the internal heat flow transfer of the entire PCBA board. Based on the surface area and convection heat transfer coefficient of the thermal calculation node, the convection heat transfer equation between the thermal calculation node and the environment is established. This equation describes the heat exchange between the node and the outside world due to the flow of fluids such as air. In addition, the radiation heat transfer characteristics of the thermal calculation node are considered. According to the radiation coefficient and surface temperature of each thermal calculation node, the radiation heat transfer equation of the thermal calculation node is established to describe the heat transfer between the node and the surrounding environment through electromagnetic waves. The heat transfer equation between nodes, the convection heat transfer equation between nodes and the environment, and the radiation heat transfer equation of nodes are combined to construct a complete set of heat transfer equations. By solving the heat transfer equations, the temperature changes of the PCBA board under different working conditions can be accurately simulated, and the temperature response of each thermal calculation node can be analyzed.
[0017] 102. According to the heat transfer equations, the temperature response of the thermal calculation node is calculated, and the convergence is determined when the temperature difference between two adjacent iterations is less than a preset threshold, so as to obtain a dynamic temperature rise characteristic curve; Specifically, according to the heat transfer equations, a Newton iteration matrix is constructed, so that the heat transfer equations can be quickly solved in each calculation process, and the temperature evolution process can be effectively simulated. The initial temperature is assigned to each heat calculation node to obtain the temperature initial value vector, which serves as the starting point of the entire calculation process and provides the initial temperature state. The initial temperature is set according to the ambient temperature or experimental data so that the subsequent calculations are more in line with the actual situation. Based on the temperature initial value vector, the heat generated by the heat conduction, convection and radiation between each heat calculation node is calculated to obtain the node heat flow matrix, which reflects the heat flow between each node under the current temperature state. Through the node heat flow matrix, the temperature change rate of each heat calculation node is calculated to obtain the temperature gradient vector, which describes the temperature change trend of each node over time and reflects the speed and direction of heat propagation in the entire system. The temperature gradient vector is integrated to obtain the temperature value of each heat calculation node under the current iteration step. The integration process simulates the cumulative effect of heat in the system over time and obtains the temperature distribution at the next moment. Through the above calculations, the temperature value of each node in the current iteration step is obtained. The node temperature value of the current iteration step is calculated by the difference with the node temperature value of the previous iteration to obtain the temperature iteration difference. The temperature iteration difference is compared with the preset threshold value for judgment. If the difference is greater than the preset threshold, it means that the current temperature change has not reached a stable state, and there is still a significant temperature change. It is necessary to return to the previous step, recalculate the node heat flow matrix, and continue to iterate. The iterative process is repeated until the temperature iteration difference is less than the preset threshold, indicating that the system has reached a stable state and outputs the final temperature distribution. The final temperature distribution output by each iteration is arranged in a time series, and a time series temperature data table is constructed to reflect the entire temperature rise process of the system from the initial state to the stable state. The time series temperature data table is interpolated and smoothed to eliminate the fluctuations or errors caused by numerical calculations, so that the temperature change curve is more physically meaningful and readable. The dynamic temperature rise characteristic curve is obtained to describe the temperature change trend of each node over time.
[0018] 103. Combine the dynamic temperature rise characteristic curve with the rated parameters of the components to solve the current threshold that meets the temperature limit condition and obtain the maximum safe current value of each thermal calculation node; Specifically, extract the key parameters of each component from the PCBA board, including the maximum operating temperature , power consumption coefficient And the safety margin factor These parameters together determine the safe operating range of components. Based on these parameters, a temperature constraint equation is constructed. , , Substitute into the formula , which describes the temperature constraint of each thermal computing node under different current conditions. represents the current variable, represents the temperature function of the corresponding thermal computing node. For each thermal computing node, the calculation process is repeated to construct The temperature constraint equations reflect the temperature limit conditions of each node in the system. The data of the temperature-current corresponding points are extracted from the dynamic temperature rise characteristic curve to reflect the temperature rise behavior of the thermal calculation node under different current conditions. In order to better describe the relationship between temperature and current, polynomial fitting is performed on the temperature-current corresponding point data, and the least squares method is used to reduce the error to obtain the temperature-current function of each thermal calculation node to describe the impact of current changes on the node temperature. According to the temperature-current function of each thermal calculation node, the current is set as the objective function, with the aim of maximizing the current that can pass through each node and improving the working efficiency of the system. The objective function is compared with the constructed The Lagrangian function is constructed by combining the temperature constraint equations with the Lagrangian multiplier. The Lagrangian function can jointly solve the maximization of the objective function and each constraint condition to ensure that the maximum safe current value is found when all temperature constraints are met. In order to solve the Lagrangian function, partial derivatives are calculated for each variable, and all partial derivatives are set to 0, and n+1 equations are formed together. These equations constitute the KKT condition equation group. The KKT condition is a necessary condition for the optimization problem, especially when there are constraints. The solution can ensure that the optimality condition and the constraint condition are met. The equation group is solved by the iterative method. Each time the updated value is substituted into the equation group, the difference between the updated value and the previous iteration value is judged. If the difference is less than the set threshold, it is considered that the iteration has converged, the iteration is stopped, and the current solution is obtained. The obtained current value is substituted into all constraint equations for verification to check whether all constraints are met. If all constraints are met, it is considered to be a valid solution. If any constraint is not met, it is necessary to return to the previous step, adjust the initial value and solve again to ensure that the final solution meets all temperature constraints. When a valid solution is confirmed, substitute its current value into the temperature-current function of each thermal calculation node, calculate the corresponding node temperature value, and compare and verify with the temperature limit of each node to ensure that the obtained current does not cause the node temperature to exceed its safe operating range. Output the maximum safe current value that meets all constraints. The maximum safe current value represents the maximum current that can pass through the system under overload protection conditions, thereby ensuring that the charging pile power board will not cause the component temperature to be too high due to overcurrent during operation.
[0019] 104. Use the charging pile test platform to collect the original multidimensional data set, perform cluster analysis and overload feature recognition, and obtain the target overload feature vector; Specifically, the sampling frequency of the charging pile test platform is set to capture the electrical parameters and thermal states of different dimensions of the charging pile when it is working. For the sampling frequency of current, it is set to 20kHz to obtain the rapidly changing current signal with high precision. For voltage, the sampling frequency is set to 10kHz to fully capture the voltage fluctuation characteristics. As an important monitoring parameter of thermal characteristics, the sampling frequency of temperature is set to 1kHz. Through the above settings, an original multidimensional data set containing different dimensions such as current, voltage, and temperature is obtained. The original multidimensional data set is normalized to eliminate the numerical differences between different physical quantities due to different dimensions, and a standardized data matrix is obtained. Based on this matrix, the covariance matrix between each dimension is calculated. The covariance matrix can reflect the correlation between the variables. The covariance matrix is decomposed by eigenvalue to obtain the principal component transformation matrix, through which the data is projected into a new coordinate system to achieve data dimensionality reduction. The standardized data matrix is multiplied by the principal component transformation matrix to obtain the feature data set after dimensionality reduction. The dimensionality reduction process can effectively reduce the redundant information of the data, making the subsequent analysis more efficient and helping to focus on the most important change features in the data. The feature data set after dimensionality reduction is input into the K-means clustering algorithm. In order to realize cluster analysis, the number of cluster centers is set to K. K-means clustering obtains the final data clustering result by iteratively calculating the Euclidean distance from the sample point to each cluster center and updating the position of the cluster center until the position of the cluster center no longer changes significantly. Based on the data clustering result, the statistical characteristics of each cluster cluster are extracted to form a cluster feature matrix, which reflects the distribution of data in different cluster clusters and the internal characteristics of each cluster cluster, such as the mean and variance of the cluster. In order to identify the overload feature, the cluster feature matrix is input into the support vector machine classifier for classification analysis. The radial basis kernel function is used for nonlinear mapping of the feature space, so that the originally linearly inseparable features become linearly separable in the high-dimensional space, so that the classification boundary of the overload feature can be effectively found. According to the classification boundary of the overload feature, the feature parameters of the key dimension are extracted by analyzing the classification boundary. The feature parameters of the key dimension are integrated to construct the target overload feature vector.
[0020] 105. Input the target overload feature vector into the overload risk assessment model to perform overload risk assessment and obtain an overload risk determination result; Specifically, the target overload feature vector is input into the first-level feature detection network of the overload risk assessment model, which consists of three parallel 1D convolutional neural networks (1D-CNN) with convolution kernel sizes of 3, 5, and 7, respectively. These convolution kernels of different sizes are used to capture the feature information of the feature vector at different scales and obtain multi-scale feature maps. Each 1D-CNN convolution operation is followed by a batch normalization layer to stabilize the network training, and then a nonlinear transformation is introduced through the ReLU activation function, so that the model can better express complex features. The multi-scale feature map is subjected to a maximum pooling operation, with the pooling window size set to 2 and the step size set to 1, to reduce the size of the feature map while retaining the most important features and enhance the anti-interference ability of the model. After pooling, the results of all feature maps are concatenated to form an integrated feature representation, and the dimension is reduced through the fully connected layer to obtain the first-level feature vector. The first-level feature vector is input into the second-level integrated network of the overload risk assessment model. The second-level integrated network consists of five decision tree classifiers, each with a maximum depth of 5 and a minimum number of sample splits of 10. The Gini coefficient is used as the splitting criterion. Through this configuration, the decision tree can effectively perform feature division while controlling the complexity of the model and generate five primary judgment results. The advantage of integrating multiple decision trees is that the bias of a single model can be reduced through a diversified judgment mechanism, and the robustness to overload risk can be improved. The five primary judgment results are weighted averaged to obtain the intermediate judgment score. The weight coefficient is dynamically adjusted by the Adam optimizer, so that the network can better adapt to the data distribution and improve the accuracy and stability of the intermediate judgment score. The intermediate judgment score is input into the third-level correction network of the overload risk assessment model. The correction network adopts a three-layer feedforward neural network structure, and the number of neurons in its hidden layer is 64, 32 and 16 respectively. In order to prevent the model from overfitting, the Dropout mechanism is applied in each hidden layer to enhance the generalization ability of the model by randomly discarding some neurons. Through the correction network, the intermediate judgment score is optimized to obtain the corrected risk score. Based on the corrected risk score, the overload risk is graded. The high, medium and low risk thresholds are set to 0.8, 0.5 and 0.3 respectively. The corrected risk score is compared with the threshold to preliminarily determine the risk level. Perform time-series correlation analysis on the preliminary risk level. By introducing historical overload data, the trend weight is calculated using the exponentially weighted moving average method to obtain the risk trend factor. The risk trend factor can capture the evolution trend of risk over time. Combine the preliminary risk level with the risk trend factor, calculate the final risk probability through the Bayesian fuser, and obtain the judgment result of the overload risk. The Bayesian fuser can comprehensively consider the current risk level and historical risk trend, and evaluate the overload risk currently faced by the system in a probabilistic way.
[0021] 106. Based on the maximum safe current value and overload risk determination results, a multi-level PWM control strategy is adopted to dynamically adjust the PWM duty cycle and frequency at the zero current switching point to output a charging control signal.
[0022] Specifically, the maximum safe current value and the overload risk determination result are input into a multi-level PWM controller, which consists of a fast response layer, a predictive control layer, and an optimization control layer. These layers work together to ensure that the system can operate safely and efficiently under different load conditions. The multi-level PWM controller obtains the initial control parameters by comprehensively processing the input information. The charging current is sampled based on the initial control parameters, and the sampling frequency is set to 50kHz. The zero-crossing detection circuit is used to identify the zero-crossing point of the current, and the zero-current switching moment is obtained to minimize the loss of the switching device, thereby improving the efficiency of the entire system and reducing heat. The current waveform before and after the zero-current switching moment is analyzed to calculate the switching loss and the electromagnetic interference (EMI) index caused by the switching action. Through these indicators, the optimization objective function of PWM is established, the purpose is to find the best PWM parameter optimization interval while ensuring the minimum switching loss and EMI, and to achieve precise control of the charging process. The PWM parameter optimization interval is input into the fast response layer, and the fast response layer uses the proportional integral control algorithm to calculate the adjustment amount of the PWM duty cycle to obtain the reference PWM waveform. The PI control algorithm enables the controller to respond quickly when an error occurs to minimize the system deviation. The reference PWM waveform is input into the prediction control layer, which calculates the temperature change trend within the next 500 microseconds based on the temperature prediction algorithm of the sliding time window. Based on the temperature change trend, the frequency adjustment coefficient is determined to dynamically adjust the PWM frequency. The frequency adjustment range is set to 20kHz to 40kHz to achieve the optimal operating frequency under different load conditions. At the same time, the duty cycle of the reference PWM waveform is appropriately compensated to ensure that the system can maintain a stable output power even in the case of frequency adjustment. The compensated PWM waveform is input into the optimization control layer, and the optimal switching time is calculated by the model prediction control algorithm, and the phase of the PWM waveform is finely adjusted to obtain the target PWM control signal. The model prediction control algorithm can predict the future state based on the dynamic model of the system and make real-time optimal control decisions, so that the entire system can maintain the best performance under variable working conditions. By fine-tuning the phase of the PWM waveform, switching losses can be reduced, energy efficiency can be improved, and electromagnetic interference caused by rapid changes in current and voltage can be reduced. Based on the target PWM control signal, the MOSFET switch tube is driven to output the final charging control signal. As a switching device, the accuracy of the MOSFET control signal determines the efficiency and safety of the entire charging process. Through multi-level PWM control strategies, the charging pile system can respond flexibly under different overload conditions, dynamically adjust output parameters, and achieve refined control of the charging process.
[0023] In the embodiment of the present invention, by establishing a multi-node thermal network model and a group of heat transfer equations, accurate calculation and dynamic monitoring of the temperature field of the PCBA board are achieved, the limitations of the traditional single-point temperature measurement method are avoided, and the accuracy of temperature monitoring is improved; the Lagrange multiplier method is used to solve the maximum safe current value, and the current threshold is accurately calculated in combination with the rated parameters of the components and the real-time temperature rise characteristics, avoiding the protection blind spot caused by the fixed threshold; a machine learning algorithm is introduced to analyze and extract features of historical data, and a multi-dimensional overload risk assessment model is established, so that the system has intelligent early warning capabilities; a multi-level PWM control strategy is designed, and the switching loss and electromagnetic interference are significantly reduced through zero current switching and dynamic parameter optimization, thereby improving the energy conversion efficiency of the system; the model predictive control and sliding time window algorithm are adopted to achieve early prediction of the temperature change trend and dynamic optimization of the control parameters, thereby enhancing the real-time control capability of the system; the Bayesian fusion algorithm is used to integrate multi-source information, thereby improving the reliability of overload risk judgment and reducing the misjudgment rate and missed judgment rate.
[0024] In a specific embodiment, the process of executing step 101 may specifically include the following steps: Scan the positions of power devices, transformers, and inductors on the PCBA board to obtain component layout coordinate data, and divide the PCBA board into N×M grid units according to the component layout coordinate data to obtain the initial grid division result; Subdivide the mesh units containing key components in the initial mesh division results to obtain thermal calculation nodes; Collect material parameters and thickness data of each layer of PCBA board, establish thermal conductivity matrix between PCB layers, and establish heat transfer equation between thermal calculation nodes based on thermal conductivity matrix between PCB layers; Based on the surface area and convection heat transfer coefficient of the thermal calculation node, the convection heat transfer equation between the thermal calculation node and the environment is established, and based on the radiation coefficient and surface temperature of the thermal calculation node, the radiation heat transfer equation of the thermal calculation node is established; The heat transfer equations are constructed by combining the conduction heat transfer equation, the convection heat transfer equation and the radiation heat transfer equation.
[0025] Specifically, the positions of components such as power devices, transformers, and inductors on the PCBA board are scanned to obtain their specific layout coordinate data. Using high-precision automatic optical inspection (AOI) equipment or other similar scanning technologies, through visual recognition technology, combined with computer image processing algorithms, the physical positions of components on the PCBA board can be identified and recorded. The coordinate data of each component is obtained by scanning to form a component layout coordinate table, which contains the coordinates of each component. Position coordinates and corresponding size information. According to the component layout coordinate data, the entire PCBA board is divided into The entire PCBA area is discretized to facilitate thermal calculation and simulation. The initial meshing forms a basic thermal analysis unit mesh, which is used to calculate the thermal distribution of each area. The mesh units containing key components in the initial meshing results are subdivided to obtain thermal calculation nodes, which helps to perform more accurate thermal simulation of local high-temperature areas. Collect material parameters and thickness data of each layer of the PCBA board. PCBA is a multi-layer structure composed of different conductive materials and insulating materials. The material properties of each layer, including thermal conductivity , layer thickness All of these factors will affect the heat transfer within the board. Through these data, the thermal conductivity matrix between PCB layers is established. The thermal conductivity matrix is used to describe the heat conduction characteristics between PCB layers. For two adjacent thermal calculation nodes and , heat transfer is described by Fourier's law, and the heat conduction equation is: ; in, Indicates the heat from the node Pass to node The rate of Representation Node and nodes The equivalent thermal conductivity between is the effective heat transfer area between nodes, and The nodes are and nodes The temperature, is the distance between two nodes. Through the above formula, the heat transfer equation between the heat calculation nodes is established to describe the heat exchange between different nodes inside the PCB. Consider the heat exchange between the node and the surrounding environment. Each heat calculation node undergoes convective heat exchange with the external environment through its surface area. The intensity of convective heat transfer is related to the node surface temperature, ambient temperature, and convective heat transfer coefficient. According to Newton's law of cooling, the convection heat transfer equation is expressed as: ; in, is the convection heat flow between the node and the environment, is the convective heat transfer coefficient, is the surface area of the node, is the surface temperature of the node, is the ambient temperature. In this way, the convection heat transfer equation between the thermal computing node and the external environment is established to describe the heat dissipation of the node. In addition to convection heat transfer, the node also exchanges heat with the surrounding environment through radiation. Radiation heat transfer mainly depends on the radiation coefficient of the node And the surface temperature, according to the Stefan-Boltzmann law, the radiation heat transfer equation of the node is established: ; in, is the heat flow of radiation heat transfer, is the radiation coefficient of the node, is the Stefan-Boltzmann constant ( W / ), is the radiation surface area of the node, is the absolute temperature of the node, is the absolute temperature of the environment. The heat transfer equation, convection heat transfer equation and radiation heat transfer equation are combined to construct the heat transfer equation group of the entire PCBA board, which is used to describe the temperature change and heat exchange behavior of each thermal calculation node in the system, and realize the comprehensive simulation of the thermal characteristics of the PCBA board.
[0026] In a specific embodiment, the process of executing step 102 may specifically include the following steps: According to the heat transfer equations, the Newton iteration matrix is constructed to assign the initial temperature of the thermal calculation node and obtain the temperature initial value vector; Based on the temperature initial value vector, the heat conduction, convection and radiation heat between the thermal calculation nodes are calculated to obtain the node heat flow matrix. Based on the node heat flow matrix, the temperature change rate of the thermal calculation node is calculated to obtain the temperature gradient vector. Integrate the temperature gradient vector to obtain the node temperature value of the current iteration step; Calculate the difference between the node temperature value of the current iteration step and the node temperature value of the previous iteration to obtain the temperature iteration difference; The temperature iteration difference is compared with the preset threshold. If it is greater than the preset threshold, the calculation node heat flux matrix step is returned to continue the iteration. If it is less than the preset threshold, the final temperature distribution is output. The final temperature distribution is arranged in time series, a time series temperature data table is constructed, and the time series temperature data table is interpolated and smoothed to obtain a dynamic temperature rise characteristic curve.
[0027] Specifically, the corresponding Newton iteration matrix is constructed based on the heat transfer equations. The heat transfer equations describe the heat transfer between thermal calculation nodes and their heat exchange with the environment. The purpose of constructing the Newton iteration matrix is to gradually approach the temperature convergence state of each thermal calculation node through an iterative method. For the initial temperature of the thermal calculation node, each node is assigned a value to obtain the temperature initial value vector ,in Represents the temperature state of each node at the beginning of the iteration. These initial temperatures are based on experimental measurements or empirical settings to ensure rapid convergence of the iteration process. Based on the initial temperature vector, the heat transfer between the thermal calculation nodes is calculated by conduction, convection, and radiation. The heat transfer of the conduction part is described by Fourier's law, while the convection and radiation heat transfer are described by Newton's cooling law and Stefan-Boltzmann's law. The heat flow between nodes is described by the node heat flow matrix Indicates that each element Representation Node and nodes The heat flow rate between nodes. The construction of the node heat flow matrix requires point-by-point calculation of the temperature of all nodes. By calculating the sum of the heat flow of each node, the total heat flow of the node is obtained. , and its total heat flow is expressed as: ; in, Representation Node The total heat flow, Representation Node and The thermal conductivity between is the effective heat transfer area between nodes, and The nodes are and nodes The temperature, is the distance between two nodes, is the convective heat transfer coefficient between the node and the environment, is the node surface area, is the ambient temperature, is the radiation coefficient of the node, is the Stefan-Boltzmann constant. The node heat flux matrix is obtained by the above formula , used to calculate the temperature change rate of the node. Based on the node heat flux matrix, the temperature change rate of each heat calculation node is calculated to obtain the temperature gradient vector The rate of temperature change through the heat capacity of the node And the heat flux of the node is calculated. The temperature change rate is expressed as: ; in, Representation Node The rate of change of temperature with time, is the total heat flux at the node, is the heat capacity of the node. The temperature gradient vector Contains the temperature change rate information of all nodes. For the temperature gradient vector Perform an integral operation to obtain the node temperature value of the current iteration step The integral operation is realized by numerical integration, for example, Euler's method is used for simple numerical integration, namely: ; in, is the temperature vector for the next step, is the current temperature vector, is the temperature gradient vector, is the time step. Through integration operation, the temperature value of each thermal calculation node in the current iteration step is obtained. The node temperature value of the current iteration step is calculated by difference with the node temperature value of the previous iteration to obtain the temperature iteration difference. The iterative difference is used to determine whether the system temperature has converged. Greater than the preset threshold , it means that the system has not reached a stable state, and it is necessary to return to the previous step, continue to calculate the node heat flow matrix and iterate; if the iteration difference is less than the preset threshold, it is considered that the system has reached thermal equilibrium and output the final temperature distribution. Arrange the final temperature distribution in time series to construct a time series temperature data table. The time series temperature data table records the temperature changes during the entire iteration process. Interpolate and smooth the time series temperature data table. Interpolation uses linear interpolation or spline interpolation methods to fill in the data gaps in the time series; and smoothing uses moving average or other filtering methods to remove high-frequency noise and obtain a smooth dynamic temperature rise characteristic curve, which reflects the temperature change trend of the entire system under different load conditions.
[0028] In a specific embodiment, the process of executing step 103 may specifically include the following steps: S1: Extract the maximum operating temperature Tmax, power consumption coefficient K and safety margin coefficient S of the components in the PCBA board; substitute Tmax, K and S into the formula f(i)≤Tmax / K×S to construct the temperature constraint equation, where i is the current variable and f(i) is the temperature constraint equation group corresponding to the thermal calculation node; repeat the calculation for each thermal calculation node to obtain n temperature constraint equation groups; S2: extract the temperature-current corresponding point data in the dynamic temperature rise characteristic curve, use the least square method to perform polynomial fitting on the temperature-current corresponding point data, and obtain the temperature-current function of each thermal calculation node; S3: according to the temperature-current function of each thermal calculation node, the current is set as the objective function, and the objective function is combined with the n temperature constraint equations through Lagrangian multipliers to obtain the Lagrangian function; S4: Take partial derivatives of the Lagrangian function, set all partial derivatives equal to 0, and combine n+1 equations to form the KKT conditional equation system; S5: Solve the equation group using an iterative method, substitute the updated value into the equation group to calculate the updated value; determine the difference between the updated value and the last iterative value; stop iterating when the difference is less than a set threshold; S6: Substitute the obtained current value into all constraint equations; check whether all constraint conditions are satisfied; if all are satisfied, confirm it as a valid solution; if there is any non-satisfactory condition, return to S5 to adjust the initial value and solve again; S7: Substitute the current value of the effective solution into the temperature-current function of each thermal calculation node, calculate the corresponding node temperature value, and compare and verify it with the node temperature limit; output the maximum safe current value that meets all constraints.
[0029] Specifically, extract the maximum operating temperature of each component in the PCBA board , power consumption coefficient and safety margin factor These parameters determine the maximum temperature that the components can withstand under the action of current and the relationship between current and temperature. Based on these parameters, the temperature constraint equation of each thermal calculation node is constructed to ensure that the temperature of each component does not exceed its rated safety range. , and Substituting into the formula: ; in, Indicates that each thermal calculation node has a current The temperature below is the maximum operating temperature of the component, is the power consumption coefficient, is the safety margin factor. These constraint equations ensure that the temperature of all nodes does not exceed its maximum temperature under given current conditions. Repeat this process for each thermal calculation node and obtain There are a set of temperature constraint equations, each of which represents the upper temperature limit of a hot node under current conditions. In order to describe the relationship between current and temperature, the temperature-current data points in the dynamic temperature rise characteristic curve are extracted to describe the temperature changes of each node under different current conditions. In order to obtain the temperature-current function of each node, the data of the corresponding points of temperature-current are fitted with polynomials, and the least squares method is used to reduce the fitting error. Assume that the temperature-current relationship obtained by fitting is: ; in, is the temperature, is the current, are the coefficients of the polynomial fit, is the order of the polynomial. The least squares method is used to find these coefficients so that the fitting curve is as close as possible to the actual measured data points, thereby accurately reflecting the relationship between temperature and current. In order to find the maximum current of each thermal calculation node under all temperature constraints, the current is set as the objective function, and the objective function is combined with the temperature constraint equations of each thermal calculation node through the Lagrange multiplier method. For each thermal node, find a current Make the node temperature not exceed its limit, and the current is as large as possible. The Lagrangian function is expressed as: ; in, is the Lagrangian function, is the target current, is the Lagrange multiplier, Hot computing node The temperature-current function, , , Node The maximum operating temperature, power consumption coefficient and safety margin coefficient of the system are calculated. By constructing the Lagrangian function, the optimization problem is converted into an optimization problem under equality constraints. The Lagrangian function is used to calculate the current i and each Lagrangian multiplier. Perform partial derivative operations and set all partial derivatives equal to 0 to obtain a set of equations, namely the KKT conditional equations. Combine these equations to form a system of n+1 equations, which are used to solve the optimal current value under temperature constraints. Use the iterative method to solve the KKT equations. Use Newton's method or other numerical optimization methods to iteratively solve the equations, each time substituting into the equations to obtain new current and Lagrange multiplier values, and calculate the difference between these updated values and the last iteration value. If the difference is less than the preset threshold, stop the iteration and consider the obtained current value to be a converged solution. Otherwise, continue to iterate until the convergence condition is met. Substitute the current value into all constraint equations for verification to check whether all temperature constraints are met. If all constraints are met, it is confirmed that this is a valid solution, indicating that the current value is feasible under the given temperature limit. If any constraint condition is not met, it is necessary to return to step S5, adjust the initial value and solve again to find a current solution that meets the conditions. Substitute the obtained effective current value into the temperature-current function of each thermal calculation node, calculate the corresponding node temperature value, and compare and verify it with the node temperature limit. In this way, it is ensured that the temperature of each node does not exceed the maximum allowable temperature under the corresponding current conditions. The obtained current value is the maximum safe current value that meets all constraints.
[0030] In a specific embodiment, the process of executing step 104 may specifically include the following steps: The sampling frequency of the charging pile test platform is set to 20kHz for current, 10kHz for voltage, and 1kHz for temperature, to obtain the original multidimensional data set; Normalize the original multidimensional data set to obtain a standardized data matrix, and based on the standardized data matrix, calculate the covariance matrix of each dimensional data, perform eigenvalue decomposition on the covariance matrix, and obtain the principal component transformation matrix; The standardized data matrix is multiplied by the principal component transformation matrix to obtain the feature data set after dimensionality reduction, and the feature data set is clustered by K-means. The number of cluster centers is set to K, and the distance between the sample point and the cluster center is calculated by Euclidean distance. The position of the cluster center is iteratively updated to obtain the data clustering result. Based on the data clustering results, the statistical characteristics of each cluster are extracted to obtain the clustering feature matrix, which is then input into the support vector machine classifier. The radial basis kernel function is used for feature space mapping to obtain the overload feature classification boundary. According to the overload feature classification boundary, the feature parameters of the key dimensions are extracted to construct the target overload feature vector.
[0031] Specifically, the sampling frequency of the charging pile test platform is set to ensure that important current, voltage and temperature data can be fully captured during system operation. The sampling frequency of current is set to 20kHz. For voltage, the sampling frequency is set to 10kHz. For temperature, since it changes relatively slowly, a sampling frequency of 1kHz is selected to effectively balance accuracy and data volume, and the original multidimensional data set is obtained, which includes multi-dimensional measurements such as current, voltage and temperature. The original multidimensional data set is normalized to eliminate the scale differences between different physical quantities and ensure the fairness of subsequent calculations. Normalization is achieved by subtracting the mean of each data dimension and dividing it by the standard deviation. The formula is: ; in, is the original data, is the mean of the data, is the standard deviation of the data, is the normalized data. The standardized data matrix is obtained by calculation, and each dimension in the matrix has zero mean and unit standard deviation. Based on the standardized data matrix, the covariance matrix between the data of each dimension is calculated. The covariance matrix is used to describe the correlation between different dimensions, and its elements are Representation Dimension and dimensions The covariance between is defined as: ; in, are the elements of the covariance matrix, and Respectively The sample in and The values in the dimensions, and Respectively and The mean of the dimensions, is the total number of samples. Through the eigenvalue decomposition of the covariance matrix, the eigenvalues and eigenvectors of the covariance matrix are obtained, where the eigenvectors constitute the principal component transformation matrix, which is used to project the original data into a new coordinate system to achieve data dimensionality reduction. Multiply the standardized data matrix with the principal component transformation matrix to obtain the reduced-dimensional feature data set. Assume that the principal component transformation matrix is , the standardized data matrix is , then the feature dataset after dimensionality reduction It is expressed as: ; in, is the feature dataset after dimensionality reduction, is the standardized data matrix, is the principal component transformation matrix. Through projection, the original high-dimensional data is compressed into a lower-dimensional feature space while retaining most of the change information in the data. K-means clustering analysis is performed on the feature data set after dimensionality reduction, and the number of cluster centers is set to , divide the data into Clusters are formed to identify different feature patterns. K-means clustering minimizes the sum of squared errors within the cluster by iteratively updating the location of the cluster center. Initial cluster centers are obtained, and then for each data point, the distance between it and each cluster center is calculated based on the Euclidean distance, and it is assigned to the cluster with the closest distance. The position of the cluster center is updated by calculating the mean of all data points in each cluster, and this process is repeated until the position of the cluster center no longer changes significantly, and the final data clustering result is obtained. Based on the data clustering results obtained by K-means clustering, the statistical features of each cluster, such as the mean and variance of each cluster, are extracted to form a cluster feature matrix to describe the characteristics of each cluster. The cluster feature matrix is input into the support vector machine classifier to realize the classification and identification of overload features. In order to deal with the nonlinear relationship in the data, the radial basis kernel function is used to perform nonlinear mapping of the feature space, and the features that are originally linearly inseparable in the low-dimensional space are mapped to the high-dimensional space, so that they become linearly separable in the high-dimensional space. The support vector machine classifier finds an optimal classification hyperplane in the high-dimensional space, which effectively separates the features of different categories and obtains the classification boundary of the overload feature. According to the classification boundary of overload characteristics, the characteristic parameters of key dimensions are extracted to describe the significant characteristics of the overload state and construct the target overload feature vector.
[0032] In a specific embodiment, the process of executing step 105 may specifically include the following steps: The target overload feature vector is input into the first-level feature detection network of the overload risk assessment model. The first-level feature detection network contains three parallel 1D-CNNs with convolution kernel sizes of 3, 5, and 7 respectively. Each 1D-CNN is followed by a BatchNorm layer and a ReLU activation function to obtain a multi-scale feature map. Perform the maximum pooling operation on the multi-scale feature map, with a pooling window size of 2 and a step size of 1. Concatenate the pooling results and reduce the dimension through the fully connected layer to obtain the first-level feature vector. The first-level feature vector is input into the second-level integrated network of the overload risk assessment model. The second-level integrated network contains 5 decision tree classifiers, the maximum depth of the tree is 5, the minimum number of sample splits is 10, and the Gini coefficient is used as the splitting criterion to obtain 5 primary judgment results; The five primary judgment results are weighted averaged, and the weight coefficient is dynamically adjusted through the Adam optimizer to obtain the intermediate judgment score; The intermediate judgment score is input into the third-level correction network of the overload risk assessment model. The third-level correction network adopts a 3-layer feedforward neural network structure with the number of hidden layer neurons being [64, 32, 16]. Dropout is used to prevent overfitting, and the corrected risk score is obtained. Based on the revised risk score, the high, medium and low risk thresholds are set at 0.8, 0.5 and 0.3 respectively to divide the risk levels and obtain the preliminary risk levels. The preliminary risk levels are then analyzed for time series correlation with historical overload data, and the trend weights are calculated using the exponentially weighted moving average method to obtain the risk trend factor. The preliminary risk level is combined with the risk trend factor, and the final risk probability is calculated through the Bayesian fuser to obtain the overload risk determination result.
[0033] Specifically, the target overload feature vector is input into the first-level feature detection network of the overload risk assessment model. In the first-level feature detection network, three parallel 1D convolutional neural networks (1D-CNNs) are included, and their convolution kernel sizes are 3, 5, and 7, respectively, so as to capture the different scale features of the feature vector under different receptive fields. Using convolution kernels of multiple sizes helps the model better understand the characteristics of the input signal at different frequencies and modes. Each 1D-CNN convolution is followed by a batch normalization layer to standardize the output of each batch to accelerate network training and stabilize the model. The output of each convolution layer is nonlinearly transformed by the ReLU activation function. ReLU can set negative values to zero and only retain positive values, thereby enhancing the network's ability to express nonlinear features. After parallel processing of these three 1D-CNN pathways, multi-scale feature maps are obtained to capture the details and global characteristics of the input feature vector. The multi-scale feature map is subjected to a maximum pooling operation. The window size of the maximum pooling is set to 2 and the step size is 1. The purpose of the maximum pooling is to reduce the size of the feature map, thereby reducing the computational complexity, while making the features more robust and reducing the impact of local noise. After the pooling operation, all pooling results are spliced together to form a more comprehensive feature representation. Then, the feature dimension is reduced through a fully connected layer to obtain a first-level feature vector. The first-level feature vector is input into the second-level integrated network of the overload risk assessment model. The integrated network contains five decision tree classifiers. The maximum depth of each decision tree is 5, the minimum number of sample splits is 10, and the Gini coefficient is used as the splitting criterion. The decision tree is a model suitable for processing nonlinear relationships. It builds a tree by continuously performing binary splits on the input features to classify the data. The Gini coefficient is used to measure the impurity of the data set. The decision tree selects the splitting method that minimizes the Gini coefficient at each split, thereby improving the purity. Through the integration of five decision trees, the deviation of a single decision tree is effectively reduced, and the stability and robustness of the model are enhanced. Each decision tree will give a primary judgment result for the input feature vector, and five primary judgment results are obtained. The five primary judgment results are weighted averaged to obtain an intermediate judgment score. The weight coefficient of the weighted average is dynamically adjusted through the Adam optimizer, which is an optimization algorithm with adaptive learning rate. It dynamically adjusts the weight by calculating the first-order moment and second-order moment of the gradient, and finds the optimal solution faster during the training process. In this way, it is ensured that the judgment results of each decision tree can be reasonably distributed according to their accuracy when weighted, so as to obtain a more reliable intermediate judgment score. The intermediate judgment score is input into the third-level correction network of the overload risk assessment model, which adopts a three-layer feedforward neural network structure, and the number of hidden layer neurons is 64, 32 and 16 respectively. The role of the feedforward neural network is to adjust and optimize the intermediate judgment score to obtain a more accurate risk assessment.In order to prevent the model from overfitting, a Dropout mechanism is added to each hidden layer to randomly discard some neurons, increase the generalization ability of the model, and avoid the problem of good performance on training data but poor performance on test data. The final corrected risk score is obtained by correcting the network processing. Based on the corrected risk score, the thresholds for high, medium and low risks are set to 0.8, 0.5 and 0.3 respectively. The corrected risk scores are compared according to these thresholds, and the risk level of the system is preliminarily divided into high risk, medium risk and low risk. The preliminary risk level reflects the instantaneous risk status of the current system. The preliminary risk level is analyzed for time series correlation with historical overload data. The risk trend weight is calculated using the exponentially weighted moving average (EWMA) method, and the formula is:. ; in, Indicates the risk trend at the current moment, is the current preliminary risk level. is the risk trend at the last moment, is a smoothing coefficient, which usually takes a value between 0 and 1. Through EWMA, the trend of risk level changes over time is captured to obtain a risk trend factor, which is used to describe the direction of risk change of the system. The preliminary risk level is combined with the risk trend factor, and the final risk probability is calculated through the Bayesian fuser. The role of the Bayesian fuser is to combine the current risk level and historical trends, and conduct risk assessment in a probabilistic way to obtain a more comprehensive and reliable overload risk determination result. Use the Bayesian theorem to update the risk probability, that is, obtain the posterior probability through the prior probability and the likelihood function: ; in, Indicates that in known events In case of occurrence, the event Probability of occurrence; is the likelihood function, which means that in the event Under the conditions of occurrence, the event Probability of occurrence; is the prior probability, is the marginal probability. Through Bayesian fusion, the current measurement results are integrated with historical information to obtain a more stable risk assessment result.
[0034] In a specific embodiment, the process of executing step 106 may specifically include the following steps: The maximum safe current value and the overload risk determination result are input into a multi-level PWM controller, which includes a fast response layer, a prediction control layer and an optimization control layer, to obtain initial control parameters; Based on the initial control parameters, the charging current is sampled at a sampling frequency of 50kHz, and a zero-crossing detection circuit is used to identify the current zero-crossing point to obtain the zero-current switching moment; Analyze the current waveform before and after the zero current switching moment, calculate the switching loss and electromagnetic interference index, establish the PWM optimization objective function, and obtain the PWM parameter optimization range; Input the PWM parameter optimization interval into the fast response layer, use the proportional-integral control algorithm to calculate the adjustment amount of the PWM duty cycle, and obtain the reference PWM waveform; Input the reference PWM waveform into the prediction control layer, calculate the temperature change trend within the next 500us based on the temperature prediction algorithm of the sliding time window, and obtain the frequency adjustment coefficient; The PWM frequency is dynamically adjusted according to the frequency adjustment coefficient, the frequency range is 20kHz to 40kHz, and the duty cycle of the reference PWM waveform is compensated to obtain a compensated PWM waveform; The compensated PWM waveform is input into the optimization control layer, the optimal switching time is calculated through the model predictive control algorithm, the phase of the PWM waveform is fine-tuned, and the target PWM control signal is obtained; Based on the target PWM control signal, the MOSFET switch is driven to output a charging control signal.
[0035] Specifically, the maximum safe current value and the overload risk determination result are input into the multi-level PWM controller, which includes a fast response layer, a predictive control layer, and an optimization control layer. These layers work together to ensure that the charging system can operate efficiently within a safe range and obtain the initial control parameters. These control parameters are the basic data of the multi-level controller in each level of control stage, and are used to dynamically adjust the system output to cope with changing load conditions. Based on the initial control parameters, the charging current is sampled, and the sampling frequency is set to 50kHz. A zero-crossing detection circuit is used to identify the zero-crossing point of the current, and determine the moment when the current changes from positive to negative or from negative to positive, that is, the zero-current switching moment. Zero-crossing detection can effectively reduce switching losses. Through zero-current switching operation, the MOSFET switches when the current is close to zero, thereby reducing switching losses and electromagnetic interference (EMI) caused by switching. After identifying the zero-current switching moment, the current waveform before and after the zero-current switching moment is analyzed, focusing on calculating the indicators of switching losses and electromagnetic interference. Switching loss Usually expressed as: ; in, is the drain-source voltage of the MOSFET, is the switching current, is the on-time of the switch, is the switching frequency. The electromagnetic interference index is mainly related to the sharp change of current and the sudden change of voltage during the switching process. By analyzing the current waveform, the EMI index of the system is estimated. The analysis results are used to establish the optimization objective function of PWM. The purpose is to optimize the PWM parameters under the condition of minimizing switching loss and EMI, and obtain the PWM parameter optimization interval. The PWM parameter optimization interval is input into the fast response layer, and the fast response layer uses the proportional integral (PI) control algorithm to calculate the adjustment amount of the PWM duty cycle. The purpose of proportional integral control is to quickly calculate the adjustment amount to eliminate the deviation when a deviation is detected, so as to ensure the stability and response speed of the system. The PI control algorithm is expressed as: ; in, To control the amount, is the proportional gain, is the integral gain, is the current error value, is time. Through Pl control, a reference PWM waveform is obtained to keep the system in a basically stable working state. The reference PWM waveform is input into the predictive control layer. The predictive control layer calculates the temperature change trend within the next 500 microseconds based on the temperature prediction algorithm of the sliding time window, and predicts the temperature rise of the system under the current load conditions, so as to make timely adjustments to the frequency and duty cycle of PWM. Temperature prediction uses the sliding time window method to estimate future temperature changes by recording temperature change data in the recent period of time. For example, temperature prediction is expressed as: ; in, is the predicted temperature, is the current temperature, is the weighting coefficient, The first The temperature change value, is the size of the sliding window. Based on the predicted temperature change trend, the frequency adjustment coefficient is calculated to dynamically adjust the PWM frequency under different load and temperature conditions to ensure that the system always works in the optimal thermal state. According to the frequency adjustment coefficient, the PWM frequency is dynamically adjusted, and the frequency range is set to 20kHz to 40kHz. The dynamic adjustment of the frequency helps to optimize the system performance under different workloads and temperature conditions. The change in frequency will also affect the duty cycle of the PWM waveform. The duty cycle of the reference PWM waveform is compensated accordingly to ensure that the output power of the system is not affected by the frequency change, and the compensated PWM waveform is obtained. The compensated PWM waveform is input to the optimization control layer, and the optimization control layer uses the model predictive control algorithm to calculate the optimal switching time. Model predictive control calculates the system state in the future by establishing a prediction model of the system, and determines the optimal control action on this basis. Through the model predictive control algorithm, the optimization control layer calculates the optimal switching time of the PWM waveform and fine-tunes the phase of the PWM waveform to reduce switching losses and ensure system stability. The target PWM control signal is obtained. Based on the target PWM control signal, the MOSFET switch tube is driven to output the final charging control signal. The driving signal of the MOSFET switch tube determines the working state of the charging system, and the frequency and duty cycle of the control signal affect the output power of the power supply and the energy efficiency of the system. Through multi-level control strategies, including fast response, predictive control and optimized control, the system can remain stable under changing working conditions, respond to load changes in a timely manner, and ensure safety in overload conditions.
[0036] The above describes the overload protection method of the charging pile power board in the embodiment of the present invention. The following describes the overload protection device of the charging pile power board in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, an overload protection device for a charging pile power board includes: Establishing module 201, for performing thermal characteristic analysis on the PCBA board of the charging pile power supply, dividing the PCBA board into a plurality of thermal calculation nodes, and establishing a corresponding heat transfer equation group; The calculation module 202 is used to calculate the temperature response of the thermal calculation node according to the heat transfer equation group, and determine the convergence according to the temperature difference between two adjacent iterations being less than a preset threshold, so as to obtain a dynamic temperature rise characteristic curve; A solution module 203 is used to combine the dynamic temperature rise characteristic curve with the rated parameters of the components to solve the current threshold that meets the temperature limit condition and obtain the maximum safe current value of each thermal calculation node; The identification module 204 is used to collect the original multi-dimensional data set using the charging pile test platform, and perform cluster analysis and overload feature identification to obtain a target overload feature vector; An evaluation module 205 is used to input the target overload feature vector into the overload risk evaluation model to perform overload risk evaluation and obtain an overload risk determination result; The output module 206 is used to dynamically adjust the PWM duty cycle and frequency at the zero current switching point based on the maximum safe current value and the overload risk determination result, and output a charging control signal by adopting a multi-level PWM control strategy.
[0037] Through the collaborative cooperation of the above-mentioned components, by establishing a multi-node thermal network model and a group of heat transfer equations, the accurate calculation and dynamic monitoring of the temperature field of the PCBA board are realized, avoiding the limitations of the traditional single-point temperature measurement method and improving the accuracy of temperature monitoring; the Lagrange multiplier method is used to solve the maximum safe current value, and the current threshold is accurately calculated in combination with the rated parameters of the components and the real-time temperature rise characteristics, avoiding the protection blind spot caused by the fixed threshold; the machine learning algorithm is introduced to analyze and extract features of historical data, and a multi-dimensional overload risk assessment model is established, so that the system has intelligent early warning capabilities; a multi-level PWM control strategy is designed, which significantly reduces switching losses and electromagnetic interference through zero-current switching and dynamic parameter optimization, and improves the energy conversion efficiency of the system; the model predictive control and sliding time window algorithm are used to realize the early prediction of temperature change trends and the dynamic optimization of control parameters, enhancing the real-time control capability of the system; the Bayesian fusion algorithm is used to integrate multi-source information, improve the reliability of overload risk judgment, and reduce the misjudgment rate and missed judgment rate.
[0038] above Figure 2 The overload protection device of the charging pile power board in the embodiment of the present invention is described in detail from the perspective of modular functional entity. The overload protection device of the charging pile power board in the embodiment of the present invention is described in detail from the perspective of hardware processing.
[0039] Figure 3It is a structural schematic diagram of a charging pile power board overload protection device provided by an embodiment of the present invention. The charging pile power board overload protection device 300 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 can be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the charging pile power board overload protection device 300. Furthermore, the processor 310 can be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the charging pile power board overload protection device 300 to implement the steps of the above-mentioned charging pile power board overload protection method.
[0040] The charging pile power board overload protection device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It can be understood by those skilled in the art that Figure 3 The charging pile power board overload protection device structure shown does not constitute a limitation on the charging pile power board overload protection device provided by the present invention, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.
[0041] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the charging pile power board overload protection method.
[0042] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0043] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.
[0044] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A charging pile power board overload protection method, characterized in that: The method comprises: Perform thermal characteristic analysis on the PCBA board of the charging pile power supply, divide the PCBA board into multiple thermal calculation nodes, and establish the corresponding heat transfer equations; According to the heat transfer equation group, the temperature response of the heat calculation node is calculated, and convergence is determined according to the temperature difference between two adjacent iterations being less than a preset threshold, so as to obtain a dynamic temperature rise characteristic curve; The dynamic temperature rise characteristic curve is combined with the rated parameters of the components to solve the current threshold under the temperature limit condition, and obtain the maximum safe current value of each thermal calculation node; The charging pile test platform is used to collect the original multidimensional data set, and cluster analysis and overload feature recognition are performed to obtain the target overload feature vector; Inputting the target overload feature vector into an overload risk assessment model to perform overload risk assessment to obtain an overload risk determination result; Based on the maximum safe current value and the overload risk determination result, a multi-level PWM control strategy is adopted to dynamically adjust the PWM duty cycle and frequency at the zero current switching point to output a charging control signal.
2. The charging pile power board overload protection method according to claim 1, characterized in that: The thermal characteristics of the PCBA board of the charging pile power supply are analyzed, the PCBA board is divided into multiple thermal calculation nodes, and the corresponding heat transfer equations are established, including: Scan the positions of power devices, transformers, and inductors on the PCBA board to obtain component layout coordinate data, and divide the PCBA board into N×M grid units according to the component layout coordinate data to obtain an initial grid division result; Subdividing the grid cells containing key components in the initial grid division result to obtain the thermal calculation nodes; Collecting material parameters and thickness data of each layer of the PCBA board, establishing a PCB interlayer thermal conductivity matrix, and establishing a heat transfer equation between the thermal calculation nodes according to the PCB interlayer thermal conductivity matrix; Based on the surface area and convection heat transfer coefficient of the heat calculation node, a convection heat transfer equation between the heat calculation node and the environment is established, and according to the radiation coefficient and surface temperature of the heat calculation node, a radiation heat transfer equation of the heat calculation node is established; The heat transfer equation, the convection heat transfer equation and the radiation heat transfer equation are combined to construct a heat transfer equation group.
3. The charging pile power board overload protection method according to claim 2, characterized in that: The step of calculating the temperature response of the heat calculation node according to the heat transfer equation group, and determining convergence according to the temperature difference between two adjacent iterations being less than a preset threshold, to obtain a dynamic temperature rise characteristic curve includes: According to the heat transfer equations, a Newton iteration matrix is constructed, and the initial temperature of the heat calculation node is assigned to obtain an initial temperature vector; Based on the temperature initial value vector, the heat conduction, convection and radiation heat between the thermal calculation nodes are calculated to obtain a node heat flow matrix, and according to the node heat flow matrix, the temperature change rate of the thermal calculation node is calculated to obtain a temperature gradient vector; Performing an integration operation on the temperature gradient vector to obtain a node temperature value of a current iteration step; Calculate the difference between the node temperature value of the current iteration step and the node temperature value of the previous iteration to obtain a temperature iteration difference; Compare and judge the temperature iteration difference with a preset threshold value, if it is greater than the preset threshold value, return to the step of calculating the node heat flow matrix to continue iteration, if it is less than the preset threshold value, output the final temperature distribution; The final temperature distribution is arranged in time series to construct a time series temperature data table, and the time series temperature data table is interpolated and smoothed to obtain a dynamic temperature rise characteristic curve.
4. The charging pile power board overload protection method according to claim 3, characterized in that: The method combines the dynamic temperature rise characteristic curve with the rated parameters of the components to solve the current threshold under the temperature limit condition and obtain the maximum safe current value of each thermal calculation node, including: S1: Extract the maximum operating temperature Tmax, power consumption coefficient K and safety margin coefficient S of the components in the PCBA board; substitute Tmax, K and S into the formula f(i)≤Tmax / K×S to construct a temperature constraint equation, where i is the current variable and f(i) is the temperature constraint equation group corresponding to the thermal calculation node; repeat the calculation for each thermal calculation node to obtain n temperature constraint equation groups; S2: extracting the temperature-current corresponding point data in the dynamic temperature rise characteristic curve, and performing polynomial fitting on the temperature-current corresponding point data using the least square method to obtain the temperature-current function of each thermal calculation node; S3: according to the temperature-current function of each thermal calculation node, the current is set as the objective function, and the objective function is combined with the n temperature constraint equations through Lagrangian multipliers to obtain the Lagrangian function; S4: Take partial derivatives of the Lagrangian function, set all partial derivatives equal to 0, and combine n+1 equations to form the KKT conditional equation system; S5: Solve the equation group using an iterative method, substitute the updated value into the equation group to calculate the updated value; determine the difference between the updated value and the last iterative value; stop iterating when the difference is less than a set threshold; S6: Substitute the obtained current value into all constraint equations; check whether all constraint conditions are satisfied; if all are satisfied, confirm it as a valid solution; if there is any non-satisfactory condition, return to S5 to adjust the initial value and solve again; S7: Substitute the current value of the effective solution into the temperature-current function of each thermal calculation node, calculate the corresponding node temperature value, and compare and verify it with the node temperature limit; output the maximum safe current value that meets all constraints.
5. The charging pile power board overload protection method according to claim 4, characterized in that: The charging pile test platform is used to collect the original multidimensional data set, and cluster analysis and overload feature recognition are performed to obtain the target overload feature vector, including: The sampling frequency of the charging pile test platform is set to 20kHz for current, 10kHz for voltage, and 1kHz for temperature, to obtain the original multidimensional data set; Normalizing the original multidimensional data set to obtain a standardized data matrix, and based on the standardized data matrix, calculating the covariance matrix of each dimensional data, performing eigenvalue decomposition on the covariance matrix to obtain a principal component transformation matrix; The standardized data matrix is multiplied by the principal component transformation matrix to obtain a feature data set after dimensionality reduction, and K-means clustering is performed on the feature data set, the number of cluster centers is set to K, the distance between the sample point and the cluster center is calculated by Euclidean distance, the position of the cluster center is iteratively updated, and the data clustering result is obtained; Based on the data clustering results, the statistical features of each cluster cluster are extracted to obtain a cluster feature matrix, and the cluster feature matrix is input into a support vector machine classifier. The radial basis kernel function is used to perform feature space mapping to obtain an overload feature classification boundary. According to the overload feature classification boundary, the feature parameters of the key dimensions are extracted to construct a target overload feature vector.
6. The charging pile power board overload protection method according to claim 5, characterized in that: The step of inputting the target overload feature vector into an overload risk assessment model to perform overload risk assessment and obtain an overload risk determination result includes: Inputting the target overload feature vector into the first-level feature detection network of the overload risk assessment model, the first-level feature detection network includes three parallel 1D-CNNs, the convolution kernel sizes are 3, 5, and 7 respectively, and each 1D-CNN is followed by a BatchNorm layer and a ReLU activation function to obtain a multi-scale feature map; Performing a maximum pooling operation on the multi-scale feature map, with a pooling window size of 2 and a step size of 1, concatenating the pooling results, and reducing the dimension through a fully connected layer to obtain a first-level feature vector; Inputting the primary feature vector into the second-level integrated network of the overload risk assessment model, the second-level integrated network includes 5 decision tree classifiers, the maximum depth of the tree is 5, the minimum number of sample splits is 10, and the Gini coefficient is used as the splitting criterion to obtain 5 primary judgment results; The five primary judgment results are weighted averaged, and the weight coefficient is dynamically adjusted through the Adam optimizer to obtain an intermediate judgment score; Input the intermediate judgment score into the third-level correction network of the overload risk assessment model, wherein the third-level correction network adopts a 3-layer feedforward neural network structure, the number of neurons in the hidden layer is [64, 32, 16], and Dropout is used to prevent overfitting, so as to obtain a corrected risk score; Based on the revised risk score, high, medium and low risk thresholds are set to 0.8, 0.5 and 0.3 respectively, and risk levels are divided to obtain preliminary risk levels. The preliminary risk levels are analyzed for time series correlation with historical overload data, and the trend weights are calculated using the exponentially weighted moving average method to obtain the risk trend factor. The preliminary risk level is combined with the risk trend factor, and the final risk probability is calculated through a Bayesian fuser to obtain an overload risk determination result.
7. The charging pile power board overload protection method according to claim 6, characterized in that: The method adopts a multi-level PWM control strategy based on the maximum safe current value and the overload risk determination result, dynamically adjusts the PWM duty cycle and frequency at the zero current switching point, and outputs a charging control signal, including: Inputting the maximum safe current value and the overload risk determination result into a multi-level PWM controller, wherein the multi-level PWM controller comprises a fast response layer, a prediction control layer and an optimization control layer, to obtain initial control parameters; Based on the initial control parameters, the charging current is sampled at a sampling frequency of 50 kHz, and a zero-crossing detection circuit is used to identify the current zero-crossing point to obtain the zero-current switching moment; Analyze the current waveform before and after the zero current switching moment, calculate the switching loss and electromagnetic interference index, establish a PWM optimization objective function, and obtain the PWM parameter optimization interval; The PWM parameter optimization interval is input into the fast response layer, and a proportional integral control algorithm is used to calculate the adjustment amount of the PWM duty cycle to obtain a reference PWM waveform; The reference PWM waveform is input into the prediction control layer, and the temperature change trend within the next 500us is calculated based on the temperature prediction algorithm of the sliding time window to obtain the frequency adjustment coefficient; Dynamically adjusting the PWM frequency according to the frequency adjustment coefficient, with the frequency range being 20kHz to 40kHz, and compensating the duty cycle of the reference PWM waveform to obtain a compensated PWM waveform; The compensated PWM waveform is input into the optimization control layer, the optimal switching time is calculated by the model predictive control algorithm, the phase of the PWM waveform is fine-tuned, and the target PWM control signal is obtained; Based on the target PWM control signal, the MOSFET switch tube is driven to output a charging control signal.
8. A charging pile power board overload protection device, characterized in that: The device is used to perform the charging pile power board overload protection method according to any one of claims 1 to 7, and comprises: Establish a module for performing thermal characteristic analysis on the PCBA board of the charging pile power supply, dividing the PCBA board into multiple thermal calculation nodes, and establishing a corresponding set of heat transfer equations; A calculation module, used to calculate the temperature response of the thermal calculation node according to the heat transfer equation group, and determine convergence according to the temperature difference between two adjacent iterations being less than a preset threshold, so as to obtain a dynamic temperature rise characteristic curve; A solution module, used for combining the dynamic temperature rise characteristic curve with the rated parameters of the components, solving the current threshold under the temperature limit condition, and obtaining the maximum safe current value of each thermal calculation node; The identification module is used to collect the original multidimensional data set using the charging pile test platform, and perform cluster analysis and overload feature identification to obtain the target overload feature vector; An evaluation module, used for inputting the target overload feature vector into an overload risk evaluation model to perform overload risk evaluation and obtain an overload risk determination result; The output module is used to dynamically adjust the PWM duty cycle and frequency at the zero current switching point based on the maximum safe current value and the overload risk determination result, and output a charging control signal by adopting a multi-level PWM control strategy.
9. A charging pile power board overload protection device, characterized in that: The charging pile power board overload protection device comprises: a memory and at least one processor, wherein the memory stores instructions; The at least one processor calls the instruction in the memory so that the charging pile power board overload protection device executes the charging pile power board overload protection method according to any one of claims 1-7.
10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instruction is executed by the processor, the charging pile power board overload protection method as described in any one of claims 1-7 is implemented.
Citation Information
Cited By
Overload protection device and method for ring main unit
CN119482272A
High-speed blender overload dynamic control method based on multi-source signal fusion
CN121035920A
