Electric vehicle connector temperature rise prediction method based on GA-BP algorithm optimized by LHS

Abstract

This application relates to a method for predicting the temperature rise of electric vehicle connectors based on an LHS-optimized GA-BP algorithm. The method includes obtaining optimization instructions; constructing an optimization dataset based on a preset sampling density and a preset LHS sampling algorithm; optimizing a pre-constructed BP neural network model using the optimization dataset and a preset GA algorithm; when a prediction instruction is received, inputting the thermal simulation parameter values ​​contained in the prediction instruction into the BP neural network model and outputting the prediction result; based on the response output of the BP neural network model to each data interval during the optimization process, determining the first response sensitivity and sampling density of each data interval; defining multi-level sampling densities for each thermal simulation parameter in the high-sensitivity interval; sampling the thermal simulation parameters in the high-sensitivity interval step-by-step according to the corresponding multi-level sampling densities; determining the second response sensitivity and sampling density of each thermal simulation parameter in the high-sensitivity interval; and updating the sampling density. This application improves the accuracy and stability of temperature rise prediction.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of temperature rise prediction of electric vehicle connectors, in particular to an electric vehicle connector temperature rise prediction method based on an LHS optimized GA-BP algorithm. BACKGROUND

[0002] In recent years, the electric vehicle industry in China has achieved rapid growth, but range anxiety is still a major concern for users, which has largely limited the further expansion of the electric vehicle market. Direct current charging technology can provide fast energy supply for electric vehicles, and is expected to significantly alleviate users' range anxiety. In a direct current charging system, the direct current charging connector is not only a bridge between the power supply device and the electric vehicle, but also a weak link in the entire charging system, and is a key factor in direct current charging connectivity and safety.

[0003] Under the background of the rapid development of the current electric vehicle industry, in order to meet the urgent need for temperature rise suppression technology, various measures have been taken in the industry, including changing the connection method, optimizing the terminal structure, improving the plating process, and introducing a liquid cooling system. However, these measures involve a wide range of fields and require a long time for testing and verification, which is in conflict with the requirement of the electric vehicle industry for rapid product iteration. In order to ensure development progress while ensuring quality, accurate temperature prediction analysis must be carried out during the product design phase to shorten the development cycle. It is crucial to develop a high-precision, fast and adaptable temperature prediction method to predict performance at an early design stage and reduce time and cost losses due to design adjustments, thereby meeting the industry's demand for rapid response to the market.

[0004] The traditional BP neural network learns and trains through forward signal transmission and backward error propagation, and the model is easily trapped in a local optimal solution due to the randomness of the initial weight and threshold, ultimately resulting in limited stability and accuracy of the prediction result, so it needs to be improved. SUMMARY

[0005] In order to optimize the prediction stability and accuracy of the electric vehicle connector temperature rise data, the application provides an electric vehicle connector temperature rise prediction method based on an LHS optimized GA-BP algorithm.

[0006] In a first aspect, the application provides an electric vehicle connector temperature rise prediction method based on an LHS optimized GA-BP algorithm, which adopts the following technical solution:

[0007] An optimization instruction is obtained, and based on a preset sampling density and a preset LHS sampling algorithm, an optimization data set formed by the thermal simulation parameter values of the electric vehicle connector is constructed. The pre-constructed BP neural network model is optimized using the optimization data set and a preset GA algorithm.

[0008] When receiving the prediction instruction, input the thermal simulation parameter value contained in the prediction instruction as an input parameter into the latest optimized BP neural network model, and output a prediction result with temperature data through the BP neural network model;

[0009] Determine the first response sensitivity and the sampling density of each data interval based on the response output result of the BP neural network model corresponding to the data set of each data interval in the optimization process; wherein the data interval refers to a plurality of numerical intervals formed by dividing the thermal simulation parameters when sampling through the LHS sampling algorithm; the first response sensitivity is used to represent the influence significance of the input variable of the corresponding data interval on the output result of the BP neural network model, and the higher the response sensitivity and the larger the sampling density are, the more significant the influence is;

[0010] Define a multi-order sampling density for each thermal simulation parameter in the high sensitivity interval according to the first response sensitivity of each data interval, and determine the second response sensitivity and the sampling density of each thermal simulation parameter in the high sensitivity interval after sampling the thermal simulation parameters in the high sensitivity interval according to the corresponding multi-order sampling density; wherein the high sensitivity interval refers to the data interval with a first response sensitivity higher than a preset sensitivity; the second response sensitivity is used to represent the influence significance of the sampling density of the thermal simulation parameters in the high sensitivity interval on the output result of the BP neural network model;

[0011] Update the preset sampling density of the non-high sensitivity interval and the sampling density corresponding to each thermal simulation parameter in the high sensitivity interval by using the determined sampling density.

[0012] By using the above technical solution, the multi-element thermal simulation parameters are sampled based on the Latin hypercube sampling algorithm (LHS algorithm), ensuring the uniformity of sampling; then the BP neural network model is trained by using the sampling data, so that the BP neural network model continuously adjusts the weights and thresholds of the network through its own learning mechanism, reduces the prediction error, and further, the genetic algorithm (GA algorithm) is used to optimize the BP neural network model, and finally the temperature rise prediction of the electric vehicle connector is realized through the optimized BP neural network model, improving the prediction accuracy and prediction efficiency; in addition, the sampling density is adjusted according to the influence significance of the thermal simulation parameter value on the prediction result, the sampling density of the sample interval with significant influence is increased, and the sampling density of the sample interval with insignificant influence is reduced, so as to reasonably allocate the computing resources, appropriately reduce unnecessary sampling and training operations on the basis of ensuring the training effect of the BP neural network model.

[0013] Optionally, the method further comprises:

[0014] The charging information is acquired in real time based on a pre-communication connection of an electric vehicle on-board system, and the charging efficiency of an electric vehicle connector is determined according to the charging information, and when the charging efficiency meets a preset triggering condition, an optimization instruction is triggered;

[0015] The BP neural network model after each optimization and the corresponding charging efficiency are recorded, and the effective prediction range of the BP neural network model is determined;

[0016] When the prediction instruction is received, the thermal simulation parameter value contained in the prediction instruction is input as an input parameter into the latest optimized BP neural network model, and the prediction result with temperature data is output by the BP neural network model, including:

[0017] When the prediction instruction is received, whether the prediction time contained in the prediction instruction is within the effective prediction range of the current optimized BP neural network model is determined;

[0018] If yes, the thermal simulation parameter value contained in the prediction instruction is input as an input parameter into the latest optimized BP neural network model, and the prediction result with temperature data is output by the BP neural network model; if no, the prediction result with temperature data is output based on a preset proxy model.

[0019] By adopting the above technical solution, when the same electric vehicle connector inputs the same thermal simulation data as an input parameter value into the BP neural network model to obtain an output result at different stages, the corresponding output result may be different, which may be caused by the performance decline of the electric vehicle connector due to the reduction of service life, the aging of physical materials, and the like, thereby inversely affecting the temperature rise data. Therefore, the application proposes to use the charging efficiency to represent the performance change of the electric vehicle connector, and to construct the corresponding relationship between the charging efficiency and the BP neural network model, so as to select the corresponding BP neural network model at different stages of the electric vehicle (i.e. the stages corresponding to different charging efficiencies), thereby realizing adaptive adjustment of the temperature rise prediction technology according to the performance change of the electric vehicle connector, and ensuring the accuracy of temperature rise prediction.

[0020] Optionally, when the prediction instruction is received, whether the prediction time contained in the prediction instruction is within the effective prediction range of the current optimized BP neural network model is determined, including:

[0021] The change trend of the charging efficiency over time is acquired, and the model effective period corresponding to the charging efficiency corresponding to the current optimized BP neural network model is predicted, the model effective period being a time period during which the corresponding charging efficiency is maintained;

[0022] If the prediction time is within the model effective period, it is considered to be within the effective prediction range of the current optimized BP neural network model, otherwise it is considered not to be within the effective prediction range of the current optimized BP neural network model.

[0023] By adopting the above technical solution, the period that each charging efficiency can maintain is predicted by combining the operation and use state of the electric vehicle connector itself (i.e. the change trend of charging efficiency over time), so as to determine the model effective period corresponding to the current optimized BP neural network model. It is believed that the temperature data predicted by the BP neural network model will exceed the fault tolerance range, i.e. the error is large, after exceeding the model effective period. Therefore, the application determines whether the prediction time is within the model effective period, as a basis for determining whether the corresponding prediction instruction can be predicted and executed by the current optimized BP neural network model.

[0024] Optionally, the method further comprises:

[0025] Whenever an optimized BP neural network model is obtained, determine the temperature deviation of the optimized BP neural network model corresponding to different times, and train the pre-constructed proxy model with the charging efficiency difference as the input parameter and the temperature deviation as the output parameter;

[0026] If not, predict the prediction result with temperature data based on the pre-set proxy model, comprising:

[0027] If not, the thermal simulation parameter value contained in the prediction instruction is input as an input parameter into the latest optimized BP neural network model, and the initial temperature data is output by the BP neural network model;

[0028] According to the prediction time, the corresponding charging efficiency is predicted, and the charging efficiency difference between the charging efficiency corresponding to the prediction time and the charging efficiency corresponding to the current BP neural network model is determined, and the temperature deviation is output based on the charging efficiency difference by the proxy model;

[0029] The initial temperature data and the temperature deviation are fused to obtain temperature data, and the prediction result with the temperature data is output.

[0030] By adopting the technical scheme, the application utilizes the agent model to predict the prediction error (i.e., temperature deviation) of the prediction instruction beyond the effective prediction range of the current BP neural network model, and fuses the prediction result (i.e., initial temperature data) obtained by predicting the prediction instruction beyond the effective prediction range of the current BP neural network model with the temperature deviation to finally predict the temperature data. The application periodically trains the agent model by taking the temperature deviation of the BP neural network model optimized at different times as an input parameter and the temperature deviation as an output parameter, improves the prediction accuracy of the agent model, reduces the deviation between the temperature data finally predicted and the actual temperature data, and improves the prediction accuracy.

[0031] Optionally, the thermal simulation parameter value included in the prediction instruction at least includes a charging duration; and the prediction time is a charging period corresponding to the start charging time and the charging duration.

[0032] If not, the thermal simulation parameter value included in the prediction instruction is taken as an input parameter and input into the latest optimized BP neural network model, and the initial temperature data is output by the BP neural network model; the charging efficiency corresponding to the prediction time is predicted, and the charging efficiency difference between the charging efficiency corresponding to the prediction time and the charging efficiency corresponding to the current BP neural network model is determined, the temperature deviation is output by the agent model based on the charging efficiency difference, the initial temperature data is fused with the temperature deviation to obtain the temperature data, and the prediction result with the temperature data is output, including:

[0033] If not, the initial temperature data is output by the current BP neural network model based on the thermal simulation parameter value included in the prediction instruction.

[0034] It is determined whether there is an intersection between the prediction time and the model effective period corresponding to the current BP neural network model.

[0035] If there is no intersection, the first charging efficiency difference between the first charging efficiency corresponding to the prediction time and the charging efficiency corresponding to the current BP neural network model is determined based on the first charging efficiency corresponding to the prediction time, the first temperature deviation is output by the agent model based on the first charging efficiency difference, the initial temperature data is fused with the first temperature deviation to obtain the temperature data, and the prediction result with the temperature data is output.

[0036] If there is an intersection, the time period corresponding to the intersection part in the prediction time is taken as an effective prediction time period, and the time period of the non-effective prediction time period is taken as a deviation prediction time period.

[0037] The initial temperature data is used as the temperature data for the effective prediction period; the second charging efficiency predicted for the deviation prediction period is determined, as well as the second charging efficiency difference between the second charging efficiency and the charging efficiency corresponding to the current BP neural network model. The second temperature deviation is output based on the second charging efficiency difference through the surrogate model. The initial temperature data and the second temperature deviation are fused to obtain the temperature data corresponding to the deviation prediction period. The prediction result with the effective prediction period, the deviation prediction period and its corresponding temperature data is output.

[0038] By adopting the above technical solution, this solution takes into account the situation where the charging efficiency changes abruptly during the charging process, triggering optimization instructions and causing a sudden increase in the prediction error of the current BP neural network model for the temperature rise from the moment of the change to the end of charging. Therefore, it performs segmented analysis on the prediction time based on the charging duration to achieve segmented prediction and minimize the prediction error as much as possible.

[0039] Optionally, the step of triggering an optimization command when the charging efficiency meets a preset trigger condition includes:

[0040] Based on the real-time acquired charging efficiency, when the charging efficiency changes and the change is not transient, it is determined whether the change meets the preset triggering conditions. If so, an optimization instruction is triggered.

[0041] By adopting the above technical solution, the charging efficiency of electric vehicle connectors may suddenly change due to unstable connection during the charging process, which may cause a brief interruption of charging. However, this change usually disappears after the connection is re-stabilized. That is, this change is temporary and does not represent a change in the performance of electric vehicle connectors. Therefore, before analyzing whether the change in charging efficiency meets the preset triggering conditions, this application first excludes the aforementioned situation of temporary change.

[0042] Optionally, determining the second response sensitivity and sampling density of each thermal simulation parameter within the high-sensitivity range includes:

[0043] All thermal simulation parameters within the high-sensitivity interval are randomly combined to form parameter combinations. Each parameter combination within the high-sensitivity interval is sampled step-by-step according to the corresponding multi-order sampling density to determine the third response sensitivity of each parameter combination. The third response sensitivity is used to characterize the significance of the sampling density of the parameter combination in the high-sensitivity interval on the output result of the BP neural network model.

[0044] Based on the second response sensitivity of each of the thermal simulation parameters and the third response sensitivity of each of the parameter combinations, the sampling density of each thermal simulation parameter within the high-sensitivity range is determined.

[0045] By adopting the above technical solution, all thermal simulation parameters within the high-sensitivity range are randomly combined to determine the influence of different sampling densities on the prediction results of the BP neural network model when sampling all thermal simulation parameters contained in a single parameter combination according to the corresponding sampling density. If the influence becomes more significant as the sampling density increases, it can be considered that the corresponding parameter combination synergistically affects the prediction results. Thus, the sampling density of the thermal simulation parameters contained in the parameter combination can be defined by this sampling density, achieving accurate determination of the sampling density corresponding to different thermal simulation parameters in different data ranges.

[0046] Secondly, this application provides an electric vehicle connector temperature rise prediction system based on the LHS-optimized GA-BP algorithm, comprising:

[0047] The prediction model optimization module is used to obtain optimization instructions, construct an optimization dataset based on the thermal simulation parameter values ​​of electric vehicle connectors, and optimize the pre-constructed BP neural network model using the optimization dataset and the preset GA algorithm.

[0048] The temperature rise data prediction module is used to take the thermal simulation parameter values ​​contained in the prediction instruction as input parameters and input them into the latest optimized BP neural network model whenever a prediction instruction is received, and output the prediction result with temperature data through the BP neural network model.

[0049] The sampling density update module is used to determine the first response sensitivity and sampling density of each data interval based on the response output of the BP neural network model to the dataset corresponding to each data interval during the optimization process. The data interval refers to several sampling intervals formed by the LHS sampling algorithm according to the thermal simulation parameter values. The first response sensitivity is used to characterize the significance of the input variable of the corresponding data interval on the output result of the BP neural network model. The more significant the influence, the higher the corresponding response sensitivity and the greater the sampling density.

[0050] The sampling density update module is further configured to define a multi-level sampling density for each thermal simulation parameter in a high-sensitivity interval based on the first response sensitivity of each data interval, and after sampling the thermal simulation parameters in the high-sensitivity interval step by step according to the corresponding multi-level sampling density, determine the second response sensitivity and sampling density of each thermal simulation parameter in the high-sensitivity interval; wherein, the high-sensitivity interval refers to the data interval in which the first response sensitivity is higher than a preset sensitivity; the second response sensitivity is used to characterize the significance of the influence of the sampling density of the thermal simulation parameters in the high-sensitivity interval on the output result of the BP neural network model;

[0051] The sampling density update module is also used to update the preset sampling density of the non-high-sensitivity range and the sampling density corresponding to each thermal simulation parameter in the high-sensitivity range using the determined sampling density.

[0052] Thirdly, this application provides an electric vehicle connector temperature rise prediction device based on the LHS-optimized GA-BP algorithm, including a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed as any of the methods described in the first aspect.

[0053] Fourthly, this application provides a computer-readable storage medium storing a computer program that can be loaded by a processor and executed as in any of the methods described in the first aspect.

[0054] In summary, this application includes the following beneficial technical effects:

[0055] Traditional backpropagation (BP) neural networks learn and train through forward signal propagation and backward error propagation. However, the model is prone to getting trapped in local optima due to the randomness of initial weights and thresholds, leading to limited stability and accuracy in prediction results. To overcome this limitation, an LHS-optimized GA-BP neural network is introduced. This network optimizes the weight and threshold configuration of the BP neural network using a genetic algorithm and employs the LHS method to improve the initial diversity and global coverage of the population. The LHS method generates a globally covering initial population by uniformly partitioning and randomly sampling within the parameter space, thus avoiding the trap of local optima in early iterations. The optimized GA-BP neural network model further iterates using the selection, crossover, and mutation processes of the genetic algorithm, gradually improving the network's search ability and convergence performance. Compared to traditional methods, the LHS-optimized GA-BP neural network significantly improves the model's global search ability and prediction accuracy, demonstrating superior application value in electric vehicle connector temperature management. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 This is a flowchart illustrating a method for predicting the temperature rise of an electric vehicle connector based on the LHS-optimized GA-BP algorithm disclosed in an embodiment of this application.

[0058] Figure 2 This is a schematic diagram illustrating the contact points of the connection terminals in an embodiment of this application.

[0059] Figure 3 This is a schematic diagram of the optimization logic of the LHS-based optimized GA-BP algorithm disclosed in the embodiments of this application.

[0060] Figure 4 This is a structural block diagram of an electric vehicle connector temperature rise prediction system based on the LHS-optimized GA-BP algorithm according to an embodiment of this application.

[0061] Figure labeling: 201, Prediction model optimization module; 202, Temperature rise data prediction module; 203, Sampling density update module. Detailed Implementation

[0062] This application discloses a method for predicting the temperature rise of electric vehicle connectors based on the LHS-optimized GA-BP algorithm (hereinafter referred to as the temperature rise prediction method). The temperature rise prediction method uses sample data obtained through LHS uniform sampling to train a BP neural network model optimized by the GA algorithm. Finally, the temperature rise data of the electric vehicle connector is predicted through the aforementioned trained and optimized BP neural network model, improving prediction efficiency and accuracy. The execution entity of the temperature rise prediction method is an electric vehicle connector temperature rise prediction system based on the LHS-optimized GA-BP algorithm (hereinafter referred to as the temperature rise prediction system). The following will refer to the attached... Figures 1-3 This section details the specific implementation steps of the temperature rise prediction system for temperature rise prediction methods.

[0063] S101, Obtain optimization instructions, construct an optimization dataset based on the preset sampling density and preset LHS sampling algorithm, using the optimized dataset and preset GA algorithm to optimize the pre-constructed BP neural network model.

[0064] S102: Whenever a prediction command is received, the thermal simulation parameter values ​​contained in the prediction command are used as input parameters and input into the latest optimized BP neural network model. The prediction result with temperature data is output through the BP neural network model.

[0065] In implementation, the optimization command can be triggered manually or automatically by the temperature rise prediction system under preset conditions. When the optimization command is triggered (i.e., when the temperature rise system obtains the optimization command), the temperature rise prediction system will use the predicted LHS sampling algorithm and sample according to a preset sampling density. The sampling object is the specific value of the thermal simulation parameters of the electric vehicle connector. In this embodiment, the thermal simulation parameters include at least the charging time, the resistance value of the charging gun end connection area, the resistance value of the charging socket end connection area, the contact resistance value of the terminal mating area, the ambient temperature, and the temperature value of the connection terminal contact point. Specifically, there can be multiple connection terminal contact points (e.g., ...). Figure 2The temperature value of the corresponding connection terminal contact point is the temperature value at each connection terminal contact point.

[0066] The temperature rise prediction system will generate an optimized dataset from the thermal simulation parameter values ​​obtained from the above sampling. The optimized dataset contains several data groups, each of which includes input parameter values ​​and output parameter values. The input parameter values ​​are charging time, resistance value of the charging gun end connection area, resistance value of the charging socket end connection area, contact resistance value of the terminal mating area, and ambient temperature. The output parameter value is the temperature value of the connection terminal contact point.

[0067] Accordingly, refer to Figure 3 The pre-constructed BP neural network model in this application specifically includes an input layer, a hidden layer, and an output layer. The input layer comprises 5 neurons, corresponding to the 5 input parameter values ​​(i.e., charging time, resistance value of the charging gun connection area, resistance value of the charging socket connection area, contact resistance value of the terminal mating area, and ambient temperature). The hidden layer has 10 neurons and uses the sigmoid activation function tansig. The output layer has a number of neurons, with the number of neurons corresponding to the number of output parameter values ​​(i.e., the number of connection terminal contact points). In this embodiment, the output layer has 7 neurons, corresponding to the temperature values ​​of the 7 different connection terminal contact points of the electric vehicle charging connector, and uses the purelin linear transfer function. The hidden layer output signal is: In the formula, This represents the output of the k-th node in the hidden layer; Let be the weights from the i-th node in the input layer to the k-th node in the hidden layer. It is the input value of the i-th node in the input layer. This represents the bias of the k-th node in the hidden layer. The formula for calculating the output layer is: In the formula, For the j-th output signal, Let J be the weight from node j to node k. The threshold for node k. The output signal of node k. It should also be noted that the BP neural network model pre-constructed in this application is a BP neural network model that has been pre-trained, whose error has converged to a preset threshold or no longer significantly reduced, and whose prediction ability meets the standard.

[0068] Based on a pre-constructed BP neural network model, a GA algorithm is performed, with an initial population size of 100. The fitness value of each individual is calculated, and the termination condition is determined by the error. In this embodiment, the fitness function is defined as: In the formula, Let be the fitness value of the j-th individual, u be the predicted output of the BP network, and v be the expected output.

[0069] To improve population diversity, a mutation operation is introduced into the GA-BP algorithm, randomly adjusting the gene values ​​of weights and biases. The crossover operation uses gene fragment exchange, with a crossover probability of 0.6 and a mutation probability of 0.02 to increase the genetic diversity of individuals with high fitness, ensuring convergence to the optimal solution. The individual with the highest fitness value is selected as the offspring, and a roulette wheel selection method is used to select individuals with excellent fitness from the population, incorporating parental characteristics for genetic evolution. Through mutation and crossover, the individual fitness is optimized generation by generation to enhance global search capabilities. The weights and biases obtained from GA optimization are used as the initial parameters of the BP neural network, which is then retrained until the error between the model output and the training samples meets the preset target. At this point, the LHS-optimized GA-BP neural network model can be used for prediction. The probability calculation formula for selecting individuals with high fitness from the population using the roulette wheel selection method is as follows: In the formula, Let be the reciprocal of the fitness of the j-th individual, c be a constant, and m be the total number of individuals. To determine the selection probability, a real-number encoded crossover method is used to generate new individuals during the crossover operation. For individuals x and y, the new individual after crossover is calculated as follows:

[0070]

[0071] In the formula, It is a random number within the interval [0, 1]. After mutation and generating a new population, the individual gene values ​​can be adjusted according to the following rules:

[0072]

[0073] In the formula, and These are the upper and lower bounds of the gene, respectively. , As a random factor, is a random number within the interval, and g is the current generation number. During the genetic evolution process, individuals with high fitness are retained and individuals with low fitness are replaced to update the population. The above operation steps are repeated until the fitness exceeds the predetermined threshold or the number of iterations reaches the set upper limit. The iteration terminates and the current optimal weight threshold is recorded.

[0074] S103, based on the response output of the BP neural network model to the dataset corresponding to each data interval during the optimization process, determine the first response sensitivity and sampling density of each data interval; where, the data interval refers to several numerical intervals formed by dividing the thermal simulation parameters when sampling through the LHS sampling algorithm; the first response sensitivity is used to characterize the significance of the influence of the input variable of the corresponding data interval on the output result of the BP neural network model, and the more significant the influence, the higher the corresponding response sensitivity and the greater the sampling density.

[0075] S104, based on the first response sensitivity of each data interval, define a multi-level sampling density for each thermal simulation parameter in the high-sensitivity interval. After sampling the thermal simulation parameters in the high-sensitivity interval step by step according to the corresponding multi-level sampling density, determine the second response sensitivity and sampling density of each thermal simulation parameter in the high-sensitivity interval. Here, the high-sensitivity interval refers to the data interval where the first response sensitivity is higher than the preset sensitivity. The second response sensitivity is used to characterize the significance of the influence of the sampling density of the thermal simulation parameters in the high-sensitivity interval on the output result of the BP neural network model.

[0076] S105, update the preset sampling density of the non-high-sensitivity region and the sampling density corresponding to each thermal simulation parameter in the high-sensitivity region using the determined sampling density.

[0077] In implementation, the temperature difference between the temperature rise data predicted by the BP neural network model and the preset temperature threshold can be used as the criterion for determining the significance of the impact. The first response sensitivity and the sampling density for each data interval are determined based on a pre-stored table of the correspondence between the temperature difference range, the first response sensitivity value, and the sampling density. It is assumed that the smaller the temperature difference, the larger the corresponding first response sensitivity value, the more significant the impact, and the higher the required sampling density. That is, the sampling density is increased for data intervals where the predicted temperature rise data is close to the preset temperature threshold, thereby improving the BP neural network model's prediction sensitivity and accuracy within that data interval.

[0078] Furthermore, the difference between the temperature rise data output by the BP neural network model after predicting the input parameters belonging to the same data interval (i.e., the temperature rise data difference, hereinafter referred to as the first difference) can also be used as the criterion for determining the significance of the influence. The first response sensitivity and the sampling density used for each data interval are determined according to the pre-stored correspondence table between the temperature rise data difference range, the first response sensitivity value, and the sampling density. It is assumed that: the larger the first difference, the larger the corresponding first response sensitivity value, the more significant the corresponding influence, and the larger the required sampling density. That is, the temperature rise data predicted by the corresponding data interval has greater variability. By increasing the sampling density of such data intervals, the predictive sensitivity and accuracy of the BP neural network for that data interval can be enhanced.

[0079] Accordingly, after determining the first response sensitivity and the corresponding sampling density for each data interval, for data intervals where the first response sensitivity is higher than the preset sensitivity (i.e., high-sensitivity intervals), this application proposes to further refine and determine its sampling density (hereinafter referred to as the initial sampling density). Specifically, the sampling density of each thermal simulation parameter (in this article, the input parameter value) within the high-sensitivity interval is redefined. Finally, the sampling density is replaced by the sampling density of the non-high-sensitivity intervals and the sampling density determined for each thermal simulation parameter in each high-sensitivity interval. This is used as the sampling density for each input parameter when triggering optimization instructions in the future.

[0080] Specifically, the sampling density of each thermal simulation parameter within the high-sensitivity range is determined as follows:

[0081] Based on the initial sampling density of the high-sensitivity interval, a multi-level sampling density is defined for all thermal simulation parameters (i.e., input parameter values) within the high-sensitivity interval. The initial sampling density is used as the lowest-order sampling density. Each order of sampling density has a different value, with higher-order sampling densities having larger values. Then, using all the sampling density values ​​contained in the multi-order sampling density, the thermal simulation parameters are sampled order by order until all thermal simulation parameters (i.e., input parameter values) are sampled according to all the sampling density values. The input parameters for each data set in the resulting optimized dataset can be specifically represented as: (charging time) Resistance value of the charging gun connection area Resistance value of the charging dock connection area Contact resistance value of terminal mating area (Ambient temperature); where a, b, c, and d represent the sampling density of the corresponding input parameters in the data set. At the same sampling time, the sampling density values ​​used for different thermal simulation parameters can be the same or different.

[0082] The temperature rise prediction system determines the second response sensitivity and sampling density of the input parameter, charging time, based on the changes in temperature rise data predicted by the BP neural network under the same bcd (heat rate and temperature). The larger the difference (hereinafter referred to as the second difference) between the predicted results (temperature rise data) obtained after sampling the charging time with different sampling densities (i.e., a), the greater the significance of the effect. The second response sensitivity and corresponding sampling density of the charging time can be determined from the aforementioned correspondence based on the pre-stored range of several differences, the second response sensitivity, and the sampling density, and based on the maximum second temperature difference. In this way, the sampling density of all input parameters within the high-sensitivity range can be determined, ensuring that the sampling density of input parameters other than the target input parameter remains unchanged. By switching the sampling density of the target input parameter, and then using the input parameter value obtained at that sampling density to input into the BP neural network for prediction, the second response sensitivity and sampling density are determined based on the pre-defined second difference; here, the target input parameter refers to any input parameter. Ultimately, this operation is used to further identify the input parameters that significantly affect the prediction results from the high-sensitivity interval and increase their sampling density, thereby helping to improve the prediction accuracy of the BP neural network model.

[0083] Optionally, determining the second response sensitivity and sampling density for each thermal simulation parameter within the high-sensitivity range further includes:

[0084] All thermal simulation parameters within the high-sensitivity interval are randomly combined to form parameter combinations. Each parameter combination within the high-sensitivity interval is sampled step-by-step according to the corresponding multi-order sampling density to determine the third response sensitivity of each parameter combination. The third response sensitivity is used to characterize the significance of the sampling density of the parameter combination in the high-sensitivity interval on the output result of the BP neural network model.

[0085] Based on the second response sensitivity of each of the thermal simulation parameters and the third response sensitivity of each of the parameter combinations, the sampling density of each thermal simulation parameter within the high-sensitivity range is determined.

[0086] In implementation, all thermal simulation parameters within the high-sensitivity range are randomly combined to form a parameter combination. This parameter combination contains one or more arbitrary input parameters. Similarly, the sampling density of all input parameters other than those included in this parameter combination is kept constant. The sampling density of the input parameters included in the target parameter combination is switched to sample the corresponding input parameters. Then, the sampled input parameter values ​​(such as charging time) are used to determine the optimal parameters. Resistance value of the charging gun connection area (where a and b can be the same or different). The system uses a BP neural network model to predict output temperature rise data. The significance of the influence on the predicted temperature rise data is assessed based on the difference in temperature rise data obtained at different sampling densities (hereinafter referred to as the third difference). The temperature rise prediction system can pre-store different ranges of the third difference, the corresponding third response sensitivity value for each range, and the sampling density corresponding to each third response sensitivity value. The larger the third difference, the greater the corresponding third response sensitivity. When the third response sensitivity is higher than a preset sensitivity threshold, the sampling density of the input parameters included in the defined parameter combination is used based on the current sampling density. This operation is used to find parameter combinations that significantly affect the temperature rise data from the high-sensitivity range, and the prediction accuracy of the BP neural network model is optimized by adjusting their sampling density.

[0087] Optional methods for predicting temperature rise also include:

[0088] The system acquires charging information in real time based on the electric vehicle onboard system with a pre-connected communication connection, and determines the charging efficiency of the electric vehicle connector based on the charging information. Based on the real-time acquired charging efficiency, when the charging efficiency changes and the change is not transient, it determines whether the change meets the preset triggering conditions. If so, it triggers an optimization command.

[0089] Record the BP neural network model after each optimization, along with the corresponding charging efficiency, and determine the effective prediction range of the BP neural network model.

[0090] S102 further includes the following sub-steps:

[0091] S1021, Whenever a prediction instruction is received, determine whether the prediction time is within the effective prediction range of the current optimized BP neural network model based on the prediction time contained in the prediction instruction.

[0092] S1022, if yes, the thermal simulation parameter values ​​contained in the prediction instruction are used as input parameters and input into the latest optimized BP neural network model, and the prediction result with temperature data is output through the BP neural network model; if no, the prediction result with temperature data is output based on the preset surrogate model.

[0093] In implementation, the temperature rise prediction system is pre-connected to the electric vehicle's onboard system. This onboard system can monitor the battery level of the electric vehicle in real time. Therefore, when the electric vehicle connector connects to the electric vehicle for charging, the temperature rise prediction system can communicate with the corresponding onboard system in real time to obtain the charging efficiency of the electric vehicle connector, which can be considered as the amount of charge per unit time. Since the performance of the charging connector degrades due to aging or poor contact after prolonged use, and these conditions also affect the temperature rise of the electric vehicle connector, this application uses a change in charging efficiency as the criterion for determining whether the electric vehicle connector's performance has changed. When a change occurs and meets preset trigger conditions, an optimization command is triggered to re-optimize the BP neural network model based on the collected thermal simulation parameters after the change in the electric vehicle connector's performance. This allows the BP neural network model to adaptively adjust its predictive ability according to the changes in the electric vehicle connector's performance, ensuring prediction accuracy.

[0094] Specifically, the temperature rise prediction system will acquire charging efficiency in real time. If the difference between the charging efficiency acquired at the target time and any charging efficiency acquired during a specified period before the target time is greater than a preset difference, then the charging efficiency is considered to have changed at the target time and this change meets a preset trigger condition. Furthermore, if the number of times the charging efficiency corresponding to the target time occurs within a specified period after the target time exceeds a preset number, then the change in charging efficiency at the target time is considered not a transient change. Here, "target time" refers to any given moment.

[0095] Whenever an optimization command is triggered and optimization is completed, the temperature rise prediction system stores the optimized BP neural network model and assigns it a corresponding charging efficiency. This charging efficiency is represented by a numerical range. For example, the range is defined as [tm, t+m], where m is a specific value, and this range is also the effective prediction range of the corresponding BP neural network model.

[0096] The temperature rise prediction system is used to determine whether the prediction time contained in the prediction instruction is within the effective prediction range of the latest optimized BP neural network model each time a prediction instruction is received. The prediction time is specifically a time range and is used to represent the charging period. The corresponding prediction instruction is to request the prediction of the temperature rise data of the electric vehicle connector during the charging period. That is, the prediction time refers to a certain period in the future.

[0097] The specific determination scheme is as follows, and correspondingly, S1021 also includes the following steps:

[0098] S10211, Obtain the trend of charging efficiency over time, and predict the effective period of the current optimized BP neural network model for the corresponding charging efficiency. The effective period of the model refers to the time period during which the corresponding charging efficiency is maintained.

[0099] S10212, if the prediction time is within the effective period of the model, it is considered to be within the effective prediction range of the current optimized BP neural network model; otherwise, it is considered not to be within the effective prediction range of the current optimized BP neural network model.

[0100] In implementation, the temperature rise prediction system is used to acquire the real-time trend of charging efficiency over time, i.e., the rate of change of charging efficiency. Based on the charging efficiency corresponding to the optimized BP neural network model, it predicts the corresponding effective period of the model. The effective period of the model is the time period in the future when the corresponding charging efficiency is the charging efficiency of the current BP neural network. Then, it is determined whether the predicted time is within the effective period of the model. If it is, the predicted time is considered to be within the effective prediction range of the current optimized BP neural network model, and can be used for prediction by the current optimized BP neural network model. Otherwise, it is not, and from the perspective of prediction accuracy, the temperature rise data predicted by the current optimized BP neural network model cannot be used as the final prediction result.

[0101] For prediction commands whose prediction time is within the effective prediction range of the current optimized BP neural network model, the temperature rise prediction system uses the thermal simulation parameter values ​​contained in the prediction command as input parameters into the current optimized BP neural network, and then outputs the prediction result with temperature data through the BP neural network model.

[0102] For prediction instructions whose prediction time is outside the effective prediction range of the current optimized BP neural network model, the prediction result with temperature data is output based on the preset surrogate model. The specific prediction scheme is as follows:

[0103] The temperature rise prediction method also includes the following steps:

[0104] After each optimization of the BP neural network model, the temperature deviation of the optimized BP neural network model at different times is determined, and the pre-built surrogate model is trained with the charging efficiency difference as the input parameter and the temperature deviation as the output parameter.

[0105] If not in S1022, then based on the preset surrogate model, the prediction result with temperature data is output, including the following sub-steps:

[0106] If not, the initial temperature data will be predicted and output based on the thermal simulation parameter values ​​contained in the prediction instruction using the current BP neural network model.

[0107] Determine whether there is any overlap between the predicted time and the effective time period corresponding to the current BP neural network model;

[0108] If there is no intersection, based on the first charging efficiency predicted at the predicted time, determine the first charging efficiency difference between the first charging efficiency and the charging efficiency corresponding to the current BP neural network model, output the first temperature deviation based on the first charging efficiency difference through the surrogate model, fuse the initial temperature data and the first temperature deviation to obtain temperature data, and output the prediction result with the temperature data.

[0109] If there is an intersection, the time period corresponding to the intersection part in the prediction time will be the effective prediction time period, and the time period of the non-effective prediction time period will be the deviation prediction time period.

[0110] The initial temperature data is used as the temperature data for the effective prediction period; the predicted second charging efficiency corresponding to the deviation prediction period is determined, as well as the second charging efficiency difference between the second charging efficiency and the charging efficiency corresponding to the current BP neural network model. A surrogate model outputs a second temperature deviation based on the second charging efficiency difference. The initial temperature data and the second temperature deviation are fused to obtain the temperature data corresponding to the deviation prediction period. A prediction result containing the effective prediction period, the deviation prediction period, and their corresponding Wendy's data is output.

[0111] In implementation, the temperature rise prediction system is also used to record the optimization time after each optimization to obtain a new BP neural network model. Then, the same input parameter values ​​are input into the BP neural network models corresponding to different optimization times. The difference in temperature rise data output by any two BP neural network models (i.e., temperature deviation) and the difference in charging efficiency corresponding to these two BP neural network models (i.e., charging efficiency difference) are calculated. The charging efficiency difference is used as the input parameter and the temperature deviation is used as the output parameter to train a pre-built surrogate model. That is, the surrogate model is used to predict the output temperature deviation based on the charging efficiency difference, i.e., the deviation of the prediction results of the BP neural network models corresponding to different optimization times.

[0112] Accordingly, when a proxy model is needed to predict the temperature rise data corresponding to the current prediction command, the temperature rise prediction system will first use the current BP neural network model to predict the temperature rise data (i.e., the initial temperature data) based on the thermal simulation parameters contained in the prediction command. Then, it will determine whether there is an intersection between the prediction time in the prediction command and the effective time period corresponding to the current BP neural network model. If there is no intersection, it will predict the charging efficiency corresponding to the prediction time (i.e., the first charging efficiency) based on the charging efficiency change trend over time, as described above. The difference between the first charging efficiency and the charging efficiency corresponding to the current BP neural network model (i.e., the first charging efficiency difference) will then be input into the proxy model. The proxy model will output the first temperature deviation, and the initial temperature data will be summed with the first temperature deviation to obtain the temperature data. Finally, the prediction result containing this temperature data will be output, completing the prediction.

[0113] If there is an overlap, the portion of the prediction time that overlaps is taken as the effective prediction period, and the portion of the prediction time that is not an effective prediction period is taken as the deviation prediction period. The initial temperature data is the temperature rise data predicted for the effective prediction period. For the deviation prediction period, based on the aforementioned trend of charging efficiency over time, the charging efficiency corresponding to the deviation prediction period (i.e., the second charging efficiency) is predicted. The difference between the second charging efficiency and the charging efficiency corresponding to the current BP neural network model (i.e., the second charging efficiency difference) is then input into the surrogate model. The surrogate model outputs the second temperature deviation. The initial temperature data and the second temperature deviation are summed to obtain the temperature data, which is the temperature data predicted for the deviation prediction period. Finally, the prediction result with the effective prediction period, the deviation prediction period, and their corresponding temperature data is output.

[0114] This application also discloses an electric vehicle connector temperature rise prediction system based on the LHS-optimized GA-BP algorithm. (Refer to...) Figure 4 Electric vehicle connector temperature rise prediction based on LHS-optimized GA-BP algorithm includes:

[0115] The prediction model optimization module 201 is used to obtain optimization instructions, construct an optimization dataset based on the thermal simulation parameter values ​​of electric vehicle connectors, and optimize the pre-constructed BP neural network model using the optimization dataset and the preset GA algorithm.

[0116] The temperature rise data prediction module 202 is used to take the thermal simulation parameter values ​​contained in the prediction instruction as input parameters and input them into the latest optimized BP neural network model whenever a prediction instruction is received, and output the prediction result with temperature data through the BP neural network model.

[0117] The sampling density update module 203 is used to determine the first response sensitivity and sampling density of each data interval based on the response output of the BP neural network model to the dataset corresponding to each data interval during the optimization process. Here, the data interval refers to several sampling intervals formed by the LHS sampling algorithm according to the thermal simulation parameter values. The first response sensitivity is used to characterize the significance of the input variable of the corresponding data interval on the output result of the BP neural network model. The more significant the influence, the higher the corresponding response sensitivity and the greater the sampling density.

[0118] The sampling density update module 203 is also used to define a multi-level sampling density for each thermal simulation parameter in a high-sensitivity interval based on the first response sensitivity of each data interval, and to determine the second response sensitivity and sampling density of each thermal simulation parameter in the high-sensitivity interval after sampling the thermal simulation parameters in the high-sensitivity interval according to the corresponding multi-level sampling density; wherein, the high-sensitivity interval refers to the data interval in which the first response sensitivity is higher than the preset sensitivity; the second response sensitivity is used to characterize the significance of the influence of the sampling density of the thermal simulation parameters in the high-sensitivity interval on the output result of the BP neural network model;

[0119] The sampling density update module 203 is also used to update the preset sampling density of the non-high-sensitivity range and the sampling density corresponding to each thermal simulation parameter in the high-sensitivity range using the determined sampling density.

[0120] Optionally, it also includes an optimization triggering module, which is used to obtain charging information in real time based on the electric vehicle on-board system with a pre-connected communication, and determine the charging efficiency of the electric vehicle connector based on the charging information. When the charging efficiency meets the preset triggering conditions, it triggers an optimization command. It is also used to record the BP neural network model after each optimization, as well as the corresponding charging efficiency, and determine the effective prediction range of the BP neural network model.

[0121] The temperature rise data prediction module 202 is also used to determine whether the prediction time is within the effective prediction range of the current optimized BP neural network model whenever a prediction instruction is received, based on the prediction time contained in the prediction instruction; if so, the thermal simulation parameter value contained in the prediction instruction is used as the input parameter and input into the latest optimized BP neural network model, and the prediction result with temperature data is output through the BP neural network model; if not, the prediction result with temperature data is output based on the preset surrogate model.

[0122] Optionally, the temperature rise data prediction module 202 is also used to obtain the trend of charging efficiency over time and predict the effective period of the current optimized BP neural network model for the charging efficiency. The effective period of the model refers to the time period during which the corresponding charging efficiency is maintained. If the predicted time is within the effective period of the model, it is considered to be within the effective prediction range of the current optimized BP neural network model; otherwise, it is considered not to be within the effective prediction range of the current optimized BP neural network model.

[0123] Optionally, it also includes a surrogate model training module, which is used to determine the temperature deviation of the optimized BP neural network model at different times after each optimization of the BP neural network model, and train the pre-built surrogate model with the charging efficiency difference as the input parameter and the temperature deviation as the output parameter.

[0124] The temperature rise data prediction module 202 is also used to, if not, take the thermal simulation parameter values ​​contained in the prediction instruction as input parameters and input them into the latest optimized BP neural network model, and output the initial temperature data through the BP neural network model; it is also used to predict the corresponding charging efficiency based on the prediction time, and determine the charging efficiency difference between the charging efficiency corresponding to the prediction time and the charging efficiency corresponding to the current BP neural network model, and output the temperature deviation based on the charging efficiency difference through the surrogate model; it is also used to fuse the initial temperature data and the temperature deviation to obtain temperature data, and output the prediction result with temperature data.

[0125] The temperature rise data prediction module 202 is also used to: if not, predict and output initial temperature data based on the thermal simulation parameter values ​​contained in the prediction instruction using the current BP neural network model; determine whether there is an intersection between the prediction time and the effective time period corresponding to the current BP neural network model; if there is no intersection, determine the first charging efficiency difference between the first charging efficiency and the charging efficiency corresponding to the current BP neural network model based on the first charging efficiency predicted at the prediction time, output the first temperature deviation based on the first charging efficiency difference through the proxy model, fuse the initial temperature data and the first temperature deviation to obtain temperature data, and output the prediction result with temperature data; if there is an intersection, take the time period corresponding to the intersection part in the prediction time as the effective prediction time period, and take the time period of the non-effective prediction time period as the deviation prediction time period;

[0126] It is also used to use the initial temperature data as the temperature data for the effective prediction period; to determine the second charging efficiency predicted for the deviation prediction period, and the second charging efficiency difference between the second charging efficiency and the charging efficiency corresponding to the current BP neural network model; to output the second temperature deviation based on the second charging efficiency difference through the surrogate model; to fuse the initial temperature data and the second temperature deviation to obtain the temperature data corresponding to the deviation prediction period; and to output the prediction result with the effective prediction period, the deviation prediction period and the corresponding temperature data.

[0127] Optionally, the optimization trigger module is also used to determine whether the change meets the preset trigger conditions when the charging efficiency changes based on the real-time acquired charging efficiency and the change is not a transient change. If so, an optimization instruction is triggered.

[0128] Optionally, the sampling density update module 203 is further configured to randomly combine all thermal simulation parameters within the high-sensitivity interval to form parameter combinations, and to sample each parameter combination within the high-sensitivity interval step by step according to the corresponding multi-order sampling density to determine the third response sensitivity of each parameter combination; wherein, the third response sensitivity is used to characterize the significance of the influence of the sampling density of the parameter combination in the high-sensitivity interval on the output result of the BP neural network model; and is further configured to determine the sampling density of each thermal simulation parameter within the high-sensitivity interval based on the second response sensitivity of each thermal simulation parameter and the third response sensitivity of each parameter combination.

[0129] This application also discloses an electric vehicle connector temperature rise prediction device based on the LHS-optimized GA-BP algorithm. The electric vehicle connector temperature rise prediction device based on the LHS-optimized GA-BP algorithm includes a memory and a processor. The memory stores a computer program that can be loaded by the processor and executed as described above for the electric vehicle connector temperature rise prediction method based on the LHS-optimized GA-BP algorithm.

[0130] This application also discloses a computer-readable storage medium that stores a computer program that can be loaded by a processor and executed as described above for the electric vehicle connector temperature rise prediction method based on the LHS-optimized GA-BP algorithm. The computer-readable storage medium includes, for example, various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0131] It should be noted that in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0132] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit the scope of protection of the application. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on these embodiments, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

Claims

1. A method for predicting the temperature rise of electric vehicle connectors based on the LHS-optimized GA-BP algorithm, characterized in that, include: Obtain optimization instructions and construct an optimization dataset based on the preset sampling density and preset LHS sampling algorithm, using the thermal simulation parameter values ​​of the electric vehicle connector. The pre-constructed BP neural network model is optimized using the optimized dataset and the preset GA algorithm; Whenever a prediction instruction is received, the thermal simulation parameter values ​​contained in the prediction instruction are used as input parameters and input into the latest optimized BP neural network model. The prediction result with temperature data is output through the BP neural network model. Based on the response output of the BP neural network model to the dataset corresponding to each data interval during the optimization process, the first response sensitivity and sampling density of each data interval are determined; wherein, the data interval refers to several numerical intervals formed by dividing the thermal simulation parameters when sampling by the LHS sampling algorithm; the first response sensitivity is used to characterize the significance of the influence of the input variable of the corresponding data interval on the output result of the BP neural network model, and the more significant the influence, the higher the corresponding response sensitivity and the greater the sampling density. Based on the first response sensitivity of each data interval, a multi-level sampling density is defined for each thermal simulation parameter in the high-sensitivity interval. After sampling the thermal simulation parameters in the high-sensitivity interval step by step according to the corresponding multi-level sampling density, the second response sensitivity and sampling density of each thermal simulation parameter in the high-sensitivity interval are determined. The high-sensitivity interval refers to the data interval in which the first response sensitivity is higher than the preset sensitivity. The second response sensitivity is used to characterize the significance of the influence of the sampling density of the thermal simulation parameters in the high-sensitivity interval on the output result of the BP neural network model. The preset sampling density in the non-high-sensitivity region and the sampling density corresponding to each thermal simulation parameter in the high-sensitivity region are updated using the determined sampling density.

2. The method for predicting the temperature rise of electric vehicle connectors based on the LHS-optimized GA-BP algorithm according to claim 1, characterized in that, The method further includes: The system acquires charging information in real time based on the electric vehicle onboard system with a pre-connected communication connection, determines the charging efficiency of the electric vehicle connector based on the charging information, and triggers an optimization command when the charging efficiency meets the preset trigger conditions. Record the BP neural network model after each optimization, along with the corresponding charging efficiency, and determine the effective prediction range of the BP neural network model. Whenever a prediction instruction is received, the thermal simulation parameter values ​​contained in the prediction instruction are used as input parameters and input into the latest optimized BP neural network model. The BP neural network model then outputs a prediction result with temperature data, including: Whenever a prediction instruction is received, the prediction time contained in the prediction instruction is used to determine whether the prediction time is within the effective prediction range of the current optimized BP neural network model. If yes, the thermal simulation parameter values ​​contained in the prediction instruction are used as input parameters and input into the latest optimized BP neural network model, and the prediction result with temperature data is output through the BP neural network model; if no, the prediction result with temperature data is output based on the preset surrogate model.

3. The method for predicting the temperature rise of electric vehicle connectors based on the LHS-optimized GA-BP algorithm according to claim 2, characterized in that, The step of determining whether the prediction time is within the effective prediction range of the currently optimized BP neural network model based on the prediction time contained in the prediction instruction whenever a prediction instruction is received includes: The charging efficiency trend over time is obtained, and the effective period of the current optimized BP neural network model is predicted. The effective period of the model refers to the time period during which the corresponding charging efficiency is maintained. If the prediction time falls within the effective period of the model, it is considered to be within the effective prediction range of the current optimized BP neural network model; otherwise, it is considered not to be within the effective prediction range of the current optimized BP neural network model.

4. The method for predicting the temperature rise of electric vehicle connectors based on the LHS-optimized GA-BP algorithm according to claim 3, characterized in that, The method further includes: After each optimization of the BP neural network model, the temperature deviation of the optimized BP neural network model at different times is determined, and the pre-built surrogate model is trained with the charging efficiency difference as the input parameter and the temperature deviation as the output parameter. If not, then based on a preset surrogate model, a prediction result with temperature data is output, including: If not, the thermal simulation parameter values ​​contained in the prediction instruction are used as input parameters and input into the latest optimized BP neural network model, and the initial temperature data is output through the BP neural network model. The charging efficiency is predicted based on the predicted time, and the difference between the charging efficiency corresponding to the predicted time and the charging efficiency corresponding to the current BP neural network model is determined. The temperature deviation is output based on the charging efficiency difference through the surrogate model. The initial temperature data and the temperature deviation are fused to obtain temperature data, and the prediction result containing the temperature data is output.

5. The method for predicting the temperature rise of electric vehicle connectors based on the LHS-optimized GA-BP algorithm according to claim 3, characterized in that, The thermal simulation parameter values ​​included in the prediction instruction include at least the charging duration; the prediction time is the charging period corresponding to the combination of the start charging time and the charging duration. If not, the thermal simulation parameter values ​​contained in the prediction instruction are used as input parameters and input into the latest optimized BP neural network model, and the initial temperature data is output through the BP neural network model; the charging efficiency corresponding to the prediction time is predicted, and the charging efficiency difference between the charging efficiency corresponding to the prediction time and the charging efficiency corresponding to the current BP neural network model is determined. The temperature deviation is output through the surrogate model based on the charging efficiency difference. The initial temperature data and the temperature deviation are fused to obtain temperature data, and the prediction result with the temperature data is output, including: If not, the initial temperature data will be predicted and output based on the thermal simulation parameter values ​​contained in the prediction instruction using the current BP neural network model. Determine whether there is any overlap between the predicted time and the effective time period corresponding to the current BP neural network model; If there is no intersection, based on the first charging efficiency predicted at the predicted time, determine the first charging efficiency difference between the first charging efficiency and the charging efficiency corresponding to the current BP neural network model, output the first temperature deviation based on the first charging efficiency difference through the surrogate model, fuse the initial temperature data and the first temperature deviation to obtain temperature data, and output the prediction result with the temperature data. If there is an intersection, the time period corresponding to the intersection part in the prediction time will be the effective prediction time period, and the time period of the non-effective prediction time period will be the deviation prediction time period. The initial temperature data is used as the temperature data for the effective prediction period; the second charging efficiency predicted for the deviation prediction period is determined, as well as the second charging efficiency difference between the second charging efficiency and the charging efficiency corresponding to the current BP neural network model. The second temperature deviation is output based on the second charging efficiency difference through the surrogate model. The initial temperature data and the second temperature deviation are fused to obtain the temperature data corresponding to the deviation prediction period. The prediction result with the effective prediction period, the deviation prediction period and its corresponding temperature data is output.

6. The method for predicting the temperature rise of electric vehicle connectors based on the LHS-optimized GA-BP algorithm according to claim 2, characterized in that, The step of triggering an optimization command when the charging efficiency meets a preset trigger condition includes: Based on the real-time acquired charging efficiency, when the charging efficiency changes and the change is not transient, it is determined whether the change meets the preset triggering conditions. If so, an optimization instruction is triggered.

7. The method for predicting the temperature rise of electric vehicle connectors based on the LHS-optimized GA-BP algorithm according to claim 1, characterized in that, The determination of the second response sensitivity and sampling density for each thermal simulation parameter within the high-sensitivity range includes: All thermal simulation parameters within the high-sensitivity interval are randomly combined to form parameter combinations. Each parameter combination within the high-sensitivity interval is sampled step-by-step according to the corresponding multi-order sampling density to determine the third response sensitivity of each parameter combination. The third response sensitivity is used to characterize the significance of the sampling density of the parameter combination in the high-sensitivity interval on the output result of the BP neural network model. Based on the second response sensitivity of each of the thermal simulation parameters and the third response sensitivity of each of the parameter combinations, the sampling density of each thermal simulation parameter within the high-sensitivity range is determined.

8. A temperature rise prediction system for electric vehicle connectors based on the LHS-optimized GA-BP algorithm, characterized in that, include: The prediction model optimization module (201) is used to obtain optimization instructions, construct an optimization dataset based on the thermal simulation parameter values ​​of electric vehicle connectors, and optimize the pre-constructed BP neural network model using the optimization dataset and the preset GA algorithm. The temperature rise data prediction module (202) is used to take the thermal simulation parameter values ​​contained in the prediction instruction as input parameters and input them into the latest optimized BP neural network model whenever a prediction instruction is received, and output the prediction result with temperature data through the BP neural network model. The sampling density update module (203) is used to determine the first response sensitivity and sampling density of each data interval based on the response output of the BP neural network model to the dataset corresponding to each data interval during the optimization process; wherein, the data interval refers to several sampling intervals formed by the LHS sampling algorithm according to the thermal simulation parameter values; the first response sensitivity is used to characterize the significance of the input variable of the corresponding data interval on the output result of the BP neural network model, and the more significant the influence, the higher the corresponding response sensitivity and the greater the sampling density; The sampling density update module (203) is further configured to define a multi-level sampling density for each thermal simulation parameter in a high-sensitivity interval based on the first response sensitivity of each data interval, and after sampling the thermal simulation parameters in the high-sensitivity interval step by step according to the corresponding multi-level sampling density, determine the second response sensitivity and sampling density of each thermal simulation parameter in the high-sensitivity interval; wherein, the high-sensitivity interval refers to a data interval in which the first response sensitivity is higher than a preset sensitivity; the second response sensitivity is used to characterize the significance of the influence of the sampling density of the thermal simulation parameters in the high-sensitivity interval on the output result of the BP neural network model; The sampling density update module (203) is also used to update the preset sampling density of the non-high-sensitivity range and the sampling density corresponding to each thermal simulation parameter in the high-sensitivity range using the determined sampling density.

9. A device for predicting the temperature rise of an electric vehicle connector based on the LHS-optimized GA-BP algorithm, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer program is stored that can be loaded by a processor and executed as described in any one of claims 1 to 7.

Images