Lithium battery temperature field reconstruction method, apparatus and device, and storage medium

Abstract

The invention discloses a lithium battery temperature field reconstruction method, device and equipment and a storage medium, and relates to the field of lithium battery temperature predication, and the method comprises the steps: obtaining the discrete temperature measurement point data of the surface of a battery and the operation parameters of the battery; constructing a multi-objective optimization function according to the discrete temperature measurement point data; performing iterative solution on the multi-objective optimization function by adopting an optimization algorithm to obtain a limited number of optimal temperature measurement point positions; acquiring temperature data under various working conditions as first training data at a limited number of optimal temperature measuring point positions; constructing a training data set in combination with the first training data and the battery operation parameters; training a temperature field reconstruction model by using the training data set to obtain a target model; and inputting the temperature data collected in real time at the limited number of optimal temperature measuring points and the real-time battery operation parameters into the target model to obtain a three-dimensional temperature field of the battery. The method can improve the prediction accuracy of the three-dimensional temperature field of the battery.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of lithium battery temperature prediction, and in particular to a lithium battery temperature field reconstruction method, device, equipment and storage medium. BACKGROUND

[0002] With the rapid development of new energy vehicles and large-scale energy storage power stations, as the core energy storage unit, the thermal safety and thermal management of lithium ion batteries have become the key to determining the performance and operation reliability of the system. The temperature distribution of the battery directly affects its charging and discharging efficiency, cycle life and safety, and local overheating may cause thermal runaway and cause serious safety accidents. Therefore, real-time, full-range and accurate monitoring and reconstruction of the battery temperature field is an important technical foundation for ensuring the long-term stable operation of the battery system and improving energy utilization efficiency.

[0003] At present, the monitoring means for battery temperature field mainly includes two types: one is a full-range temperature measurement scheme based on dense sensor arrangement, which directly obtains multi-point temperature data by arranging a large number of temperature sensors on the surface or inside the battery; the other is a temperature field reconstruction method based on limited measurement points, which usually uses traditional interpolation algorithm or simple regression model to deduce the overall temperature distribution from a small amount of measurement point data.

[0004] However, the above existing scheme has low prediction accuracy for the three-dimensional temperature field of the battery under the condition of limited measurement points. SUMMARY

[0005] The present application provides a lithium battery temperature field reconstruction method, device, equipment and storage medium, which can improve the prediction accuracy of the three-dimensional temperature field of the battery.

[0006] To achieve the above purpose, the present application adopts the following technical scheme: In a first aspect, the present application provides a lithium battery temperature field reconstruction method, comprising: obtaining discrete temperature measurement point data on the surface of the battery and battery operating parameters; constructing a multi-objective optimization function according to the discrete temperature measurement point data; using an optimization algorithm to iteratively solve the multi-objective optimization function to obtain a limited number of optimal temperature measurement point positions; collecting temperature data under multiple working conditions at the limited number of optimal temperature measurement point positions as first training data; combining the first training data and the battery operating parameters to construct a training data set; training a temperature field reconstruction model using the training data set to obtain a target model; inputting the real-time temperature data collected at the limited number of optimal temperature measurement points and the real-time battery operating parameters into the target model to obtain the three-dimensional temperature field of the battery.

[0007] Optionally, the temperature field reconstruction model is obtained by the following method: Based on the conditional generative adversarial network (CGAN) and the Transformer architecture, the temperature field reconstruction model is obtained.

[0008] Optionally, the multi-objective optimization function is constructed according to the discrete temperature measurement point data, including: A multi-objective optimization function is constructed with the dual objectives of minimizing the deviation of the discrete temperature measurement point data from the global real temperature data and maximizing the spatial coverage of the measurement points.

[0009] Optionally, the temperature field reconstruction model is trained using the training data set to obtain a target model, including: The temperature field reconstruction model is iteratively trained using the training data set. When the reconstruction error of the temperature field reconstruction model on the validation set meets the preset convergence condition, the target model is obtained, and the preset convergence condition is that the reconstruction error does not decrease in consecutive multiple rounds of validation.

[0010] Optionally, the multi-objective optimization function is iteratively solved using an optimization algorithm to obtain a limited number of optimal temperature measurement point positions, including: The multi-objective optimization function is iteratively solved using a differential evolution algorithm, and when the multi-objective optimization function value converges to a stable state or reaches a preset maximum number of iterations, the sensor position distribution corresponding to the individual with the highest fitness in the current population is output as the limited number of optimal temperature measurement point positions.

[0011] Optionally, the multiple working conditions include uniform heat generation, non-uniform heat generation, and thermal runaway.

[0012] Optionally, after obtaining the three-dimensional temperature field of the battery, it further includes: The three-dimensional temperature field is de-normalized to obtain first temperature field data. The invalid zero value region in the first temperature field data is smoothed and optimized using a mask Gaussian smoothing algorithm, and the target temperature field distribution is output.

[0013] In a second aspect, the present application provides a lithium battery temperature field reconstruction device, comprising: An acquisition module is configured to acquire discrete temperature measurement point data on the surface of a battery and battery operating parameters. The processing module is configured to construct a multi-objective optimization function according to the discrete temperature measurement point data, iteratively solve the multi-objective optimization function by using an optimization algorithm to obtain a limited number of optimal temperature measurement point positions, and collect temperature data under multiple working conditions at the limited number of optimal temperature measurement point positions as first training data. The training module is configured to train a temperature field reconstruction model by using the training data set to obtain a target model. The output module is configured to input temperature data of the limited number of optimal temperature measurement points and real-time battery operating parameters into the target model to obtain a three-dimensional temperature field of the battery.

[0014] In a third aspect, the present application provides a computing device including a memory and a processor. The memory stores one or more computer programs including instructions, and when the instructions are executed by the processor, the computing device performs the method of any one of the first aspect.

[0015] In a fourth aspect, the present application provides a computer readable storage medium for storing a computer program for executing the method of any one of the first aspect.

[0016] From the above technical solutions, the present application has at least the following beneficial effects: In the present application, by constructing a multi-objective optimization function that minimizes temperature deviation and maximizes spatial coverage, and combining with a differential evolution algorithm to obtain optimal measurement point positions, the cost and heat dissipation interference of dense sensors are avoided, and the problem of insufficient data representation of traditional limited measurement point schemes is solved. Then, through a model fused with CGAN and Transformer, accurate reconstruction from limited measurement point data to three-dimensional temperature field is realized, which breaks through the limitation of low prediction accuracy of traditional interpolation / regression models under complex working conditions, and effectively captures risk features such as local overheating of the battery.

[0017] Further, the training data covers multiple working conditions such as uniform heat generation, non-uniform heat generation and thermal runaway, so that the model adapts to the full operating scenario of the battery. The real-time output three-dimensional temperature field can accurately reflect the global temperature distribution after denormalization and smoothing optimization, provide timely and comprehensive decision basis for the thermal management system, reduce the risk of thermal runaway caused by local overheating, and ensure long-term stable operation of the battery system.

[0018] Further, only a limited number of optimal measurement points are needed to achieve global temperature monitoring, greatly reducing the number of sensor layouts and costs, while the model training and real-time reconstruction process is simple and efficient, easy to integrate with existing battery management systems, improving energy utilization efficiency, and adapting to the practical application needs of new energy vehicles, energy storage power stations and other scenarios.

[0019] It should be understood that the description of technical features, technical solutions, beneficial effects or similar language in this application does not imply that all features and advantages can be achieved in any single embodiment. On the contrary, it can be understood that the description of a feature or beneficial effect means that the specific technical feature, technical solution or beneficial effect is included in at least one embodiment. Therefore, the description of technical features, technical solutions or beneficial effects in this specification does not necessarily refer to the same embodiment. Further, the technical features, technical solutions and beneficial effects described in this embodiment can be combined in any appropriate manner. Those skilled in the art will understand that the embodiments can be implemented without one or more specific technical features, technical solutions or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects can be identified in specific embodiments that do not embody all embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A flowchart of a lithium battery temperature field reconstruction method provided by an embodiment of the present application; Figure 2 A schematic diagram of a lithium battery temperature field reconstruction device provided by an embodiment of the present application; Figure 3 A schematic diagram of a computing device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0021] The terms "first", "second" and "third" and the like in the specification and the drawings of the present application are used to distinguish different objects, and are not intended to limit a specific order.

[0022] In the embodiments of the present application, the words "exemplary" or "for example" are used to mean serving as an example, instance, or illustration. Any embodiment or design presented as "exemplary" or "for example" in the embodiments of the present application should not be construed as being more preferred or advantageous than other embodiments or designs. Rather, the use of "exemplary" or "for example" is intended to present concepts in a concrete manner. The words "first", "second", and "third" and the like in the specification and the drawings of the present application are used to distinguish different objects, and are not intended to limit a specific order.

[0023] For the sake of clear and concise description of the following embodiments, first give a brief introduction of the related art: The battery temperature field refers to the temperature distribution state of the surface and internal regions of the battery during operation, which is an index reflecting the thermal state of the battery and is directly related to the efficiency, life and safety of the battery.

[0024] With the rapid development of new energy vehicles and large-scale energy storage power stations, the thermal safety and thermal management of lithium ion batteries have become the core elements determining the system performance and operation reliability. The key technical problem currently faced by the industry is that under the condition of limited measurement points, the monitoring accuracy of the three-dimensional temperature field of the battery is insufficient, it is difficult to achieve real-time, global and reliable temperature distribution perception, and it cannot provide accurate decision basis for the thermal management system, which not only restricts the improvement of battery charging and discharging efficiency and cycle life, but also poses a safety hazard of local overheating leading to thermal runaway for system operation.

[0025] The main cause of the above technical problem lies in the design defects of the existing monitoring scheme. On the one hand, the global temperature measurement scheme with densely arranged sensors can guarantee accuracy, but there are practical problems such as high cost, limited installation space, and interference with battery heat dissipation, making it difficult to be applied on a large scale. On the other hand, the traditional limited measurement point temperature field reconstruction scheme does not optimize the measurement point position scientifically, resulting in insufficient representativeness of the collected discrete data, and the interpolation algorithm or simple regression model relied on cannot effectively capture the spatio-temporal correlation characteristics of temperature changes of the battery under different working conditions, ultimately resulting in low accuracy of three-dimensional temperature field prediction, which cannot meet the needs of refined thermal management and safety prevention and control.

[0026] Therefore, the embodiments of the present application provide a lithium battery temperature field reconstruction method, which can be executed by a processing device. The processing device can be a terminal or a server. The terminal includes, but is not limited to, a smart phone, a tablet computer, a notebook computer, a personal digital assistant, or a smart wearable device, etc. The server can be a cloud server, such as a central server in a central cloud computing cluster or an edge server in an edge cloud computing cluster. Of course, the server can also be a server in a local data center. The local data center refers to a data center directly controlled by the user.

[0027] In the method, in order to solve the problem of insufficient monitoring accuracy of the three-dimensional temperature field of the battery under the condition of limited measurement points, the present application first solves the representativeness problem of the data source by scientifically optimizing the measurement point position, and then solves the accurate conversion problem of discrete data to global temperature field by using an efficient fusion model, so as to finally balance the cost and accuracy. Specifically, a multi-objective optimization function considering the minimization of temperature deviation and the maximization of spatial coverage is constructed based on the discrete temperature measurement point data of the battery, and an optimization algorithm is used to select the best measurement point position to ensure that the limited measurement points can collect the most valuable temperature information. At the same time, the rationality of temperature field generation is ensured by using the adversarial learning mechanism of conditional generative adversarial network (CGAN), and the time-series-spatial correlation characteristics of temperature changes are captured by using the Transformer architecture to construct a fusion model for high-precision reconstruction. Finally, the model is trained by using multi-working-condition data, and the reconstructed results are optimized by post-processing, so as to achieve the goal of real-time, global and accurate monitoring of the three-dimensional temperature field of the battery.

[0028] In order to make the technical solutions of the present application more clear and easy to understand, the following describes a lithium battery temperature field reconstruction method provided by an embodiment of the present application with reference to the accompanying drawings. As shown in the drawings, the drawings are flowcharts of a lithium battery temperature field reconstruction method provided by an embodiment of the present application. The method comprises the following steps. Figure 1 S201, a processing device acquires discrete temperature measurement point data of a battery surface and battery operation parameters.

[0029] The discrete temperature measurement point data of the battery surface refers to a local temperature value set collected by a temperature sensor through selection of a plurality of discrete specific points on the battery surface. Discrete means that the measurement points do not continuously cover the whole domain of the battery, and only reflect the temperature information of each point, which is the basic data for deducing the whole temperature field.

[0030] The battery operation parameters refer to parameters representing the working state of the battery, including but not limited to charging and discharging current, voltage, remaining capacity SOC, charging and discharging state, and environmental temperature, etc. Such parameters directly affect the heat generation efficiency and temperature variation law of the battery.

[0031] The processing device acquires the temperature values of the discrete points on the battery surface through a preset temperature sensing device, and reads the real-time operation parameters of the battery from the battery management system, charging and discharging equipment and other associated devices through a communication interface. The synchronous acquisition of the two types of data provides complete and necessary original input for subsequent construction of a multi-objective optimization function, screening of the best measurement point position and training of a temperature field reconstruction model, and is a prerequisite for accurate temperature field reconstruction.

[0032] S202, the processing device constructs a multi-objective optimization function according to the discrete temperature measurement point data.

[0033] Specifically, the processing device constructs a multi-objective optimization function with the dual objectives of minimizing the deviation of the discrete temperature measurement point data from the whole real temperature data and maximizing the spatial coverage of the measurement points.

[0034] The whole real temperature data refers to real temperature distribution data covering all regions on the battery surface and inside, which can be acquired by calibration of a high-precision whole domain temperature measurement device, and serves as a reference benchmark for measuring the accuracy of the discrete measurement point data.

[0035] The spatial coverage of the measurement points refers to the coverage degree of the discrete temperature measurement points on the whole domain of the battery. The higher the coverage, the more the measurement point data can reflect the overall temperature distribution characteristics of the battery, avoiding the loss of local temperature information due to uneven distribution of the measurement points.

[0036] ​A multi-objective optimization function is a mathematical function that contains two or more optimization objectives. By solving the function, an optimal solution that takes into account multiple objectives can be obtained, which is different from a function that only takes a single index as the optimization direction.

[0037] The processing device constructs a multi-objective optimization function around two objectives, providing a mathematical basis for solving subsequent optimization algorithms: the first objective is to minimize the deviation between discrete temperature measurement point data and the real temperature data of the entire domain, ensuring the accuracy of the measurement point data; the second objective is to maximize the spatial coverage of the measurement points, ensuring the comprehensiveness of the measurement point data. Through the coupled optimization of the two objectives, the final selected measurement point locations can accurately reflect the temperature characteristics and fully cover the entire battery domain.

[0038] Multi-objective optimization function The corresponding calculation expression is:

[0039] in, The location variable represents the discrete temperature measurement point data; The function representing the deviation between discrete temperature measurement point data and the true temperature data over the entire area is expressed using the root mean square error, and its expression is:

[0040] in, Indicates the number of measurement points. Indicates the number of sampling times. Indicates the first The first measuring point The true global temperature value at any given moment. Indicates the first The first measuring point The temperature value collected at any given time.

[0041] The spatial coverage objective function is expressed as:

[0042] in, This represents the total surface area of ​​the battery. Indicates the number of effective areas covered by the measuring points. Indicates the first The area of ​​each effective region. This is to transform the goal of maximizing spatial coverage into a unified direction of minimization optimization.

[0043] S203. The processing equipment uses an optimization algorithm to iteratively solve the multi-objective optimization function to obtain a finite number of optimal temperature measurement point locations.

[0044] Specifically, the processing device iteratively solves the multi-objective optimization function by the differential evolution algorithm, and outputs the sensor position distribution corresponding to the individual with the highest fitness in the current population as the limited optimal temperature measurement point positions when the multi-objective optimization function value converges to a stable state or reaches a preset maximum iteration number.

[0045] In the present example, the parameters of the differential evolution algorithm are set as follows: population size 50, global exploration iteration number 3, local development iteration number 10, total iteration number 200, mutation factor F1=0.9, and crossover probability CR=0.9.

[0046] The differential evolution algorithm is a heuristic global optimization algorithm based on population evolution, which realizes iterative optimization through mutation, crossover and selection operations on individuals in the population. It is suitable for solving multi-objective and nonlinear complex optimization problems, and has the characteristics of fast convergence speed and strong robustness.

[0047] Iterative solution refers to a process of starting from an initial solution, repeatedly executing mutation-crossover-selection calculation process multiple times, constantly updating candidate solutions, and gradually approaching the optimal solution.

[0048] Convergence of the multi-objective optimization function value to a stable state means that the variation amplitude of the two objective values (temperature deviation value and spatial coverage value) of the multi-objective optimization function is less than a preset threshold, for example, 0.001, and remains stable for multiple consecutive rounds, which means that the current candidate solution is close to the optimal state.

[0049] The preset maximum iteration number is a calculation termination condition set to avoid the algorithm falling into infinite iteration, and is a hard constraint to ensure the calculation efficiency of the algorithm, for example, the preset maximum iteration number is 200.

[0050] The population is a set of multiple candidate solutions in the differential evolution algorithm, and each candidate solution corresponds to a set of sensor position distribution schemes.

[0051] The individual with the highest fitness refers to the candidate solution in the population that optimally satisfies the two objectives (minimizing temperature deviation and maximizing spatial coverage) of the multi-objective optimization function, and its corresponding sensor position distribution scheme is the current optimal scheme.

[0052] The optimal temperature measurement point positions refer to a limited number of sensor deployment points that balance the accuracy of temperature data and the comprehensiveness of spatial coverage after optimization by the algorithm.

[0053] The processing device calls the differential evolution algorithm, takes the multi-objective optimization function constructed previously as the optimization target of the algorithm, and initializes a group of sensor position candidate solutions to form an initial population. Then the algorithm performs mutation, crossover and selection operations on each individual in the population in turn: the mutation operation generates new candidate solutions by the difference between individuals in the population, expanding the search range; the crossover operation fuses the mutated solution with the original individual, retaining the excellent features; the selection operation compares the fitness of the new solution and the original individual, and selects the better individual into the next generation population. Through such a round of iteration after another, the sensor position scheme corresponding to the individuals in the population is continuously optimized, and the multi-objective optimization function value also gradually approaches the better direction.

[0054] After each iteration, the processing device judges the current optimization state. As long as any one of the two conditions is met, the iteration will stop: one is that the multi-objective optimization function value converges to a stable state, that is, the change amplitudes of the temperature deviation and the spatial coverage in continuous multiple iterations are less than the preset threshold, which means that the optimal scheme close to the optimal scheme has been found; the second is that the number of iterations reaches the preset maximum number of iterations, which is to avoid excessive iteration of the algorithm to consume too much computing resources and ensure the solving efficiency.

[0055] After the iteration stops, the processing device evaluates the fitness of all individuals in the final population and selects the individual with the highest fitness. The sensor position distribution scheme corresponding to the individual can maximize the spatial coverage of the measuring points and minimize the deviation between the discrete measuring point data and the global real temperature data, fully meeting the needs of multi-objective optimization. The processing device determines the point corresponding to the scheme as the position of the limited number of best temperature measuring points, providing accurate point basis for subsequent collection of multi-condition temperature data and construction of training data set.

[0056] The whole process solves the problem of strong subjectivity and insufficient representativeness of traditional measuring point arrangement scheme through the global optimization ability of the differential evolution algorithm, ensures that the limited measuring points can collect the most valuable temperature information, and is the premise of realizing subsequent high-precision temperature field reconstruction.

[0057] S204、In the limited number of best temperature measuring point positions, collect temperature data under multiple working conditions as the first training data.

[0058] The multiple working conditions refer to different working states of the battery in the actual operation process, and include a uniform heat production condition, a non-uniform heat production condition, and a thermal runaway condition. The uniform heat production condition refers to the operation of the battery in a normal and stable charging and discharging state, the internal chemical reaction rate is uniform, the heat is evenly distributed in the whole battery, and there is no obvious local overheating area, which is the normal working condition of the battery. The non-uniform heat production condition refers to the state that the local area produces heat at a higher rate than other areas due to factors such as charging and discharging rate fluctuations, local electrode aging, or external heat dissipation condition differences. At this time, the battery temperature distribution is uneven, and potential overheating risk points are easily formed. The thermal runaway condition refers to the dangerous state that the battery produces heat at a rate far exceeding the heat dissipation capacity due to internal short circuit, overcharging or overdischarging, or serious heat dissipation failure, and the temperature rises sharply and triggers a chain heat release reaction, which is a fault condition leading to battery fire and explosion.

[0059] The first training data refers to the multi-condition temperature data collected at the optimal measurement point position, which is the basic data for constructing the training data set of the temperature field reconstruction model, and provides support for the model to learn the temperature distribution law under different conditions.

[0060] Specifically, the processing device will arrange sensors based on the selected optimal temperature measurement point position, and then simulate or collect temperature values of the battery under multiple typical operating conditions such as uniform heat production, non-uniform heat production, and thermal runaway. These temperature data will be arranged into a structured data set, i.e. the first training data. The purpose of this step is to obtain temperature information covering all conditions through optimal measurement points, to ensure that the model trained later can adapt to different actual operating states of the battery, to avoid insufficient model generalization ability due to single training data scene, and to lay a data foundation for ultimately realizing high-precision three-dimensional temperature field reconstruction.

[0061] S205, the processing device constructs a training data set in combination with the first training data and the battery operating parameters.

[0062] The training data set refers to a structured data set constructed for training the temperature field reconstruction model, which contains input feature data and corresponding label data, and can enable the model to learn the mapping relationship between the input data and the whole battery temperature distribution.

[0063] Specifically, the processing device will associate and integrate the first training data (optimal measurement point temperature data under multiple conditions) with the corresponding battery operating parameters, match the current, voltage, SOC, and other operating parameters for each group of temperature data; at the same time, the integrated data is preprocessed, including removing outliers, data normalization, adding time stamps, etc., to finally construct a training data set containing multi-condition measurement point temperature and operating parameter input features.

[0064] The purpose of this step is to make the training data not only contain temperature information, but also contain the working condition background information that affects the temperature change, to ensure that the subsequent trained model can more accurately learn the temperature distribution law of the battery under different operating conditions, and improve the generalization ability and reconstruction accuracy of the model.

[0065] In S206, the processing device trains the temperature field reconstruction model using the training data set to obtain a target model.

[0066] Based on the conditional generative adversarial network CGAN and the Transformer architecture, a temperature field reconstruction model is obtained.

[0067] Firstly, the model takes the conditional generative adversarial network (CGAN) as the generation framework. CGAN is an improved version of the generative adversarial network (GAN). Its characteristic is to introduce explicit conditional constraint information in the training process, such as battery operating parameters and measurement point temperature data in this application. Compared with traditional unconstrained GAN, it has stronger directional generation ability and can guide the model to generate temperature field data that meets the current battery operating conditions. CGAN mainly includes two components: generator and discriminator, which are optimized through adversarial training: CGAN generator construction: U-Net improved structure is adopted, which is divided into input fusion layer, dense connection layer, encoder, residual block and decoder five parts, forming a complete generation process of input fusion-feature extraction-feature enhancement-size reduction. Specifically, firstly, the input fusion layer splices the latent noise vector (providing random generation power) and the condition vector (providing directional constraints) into a unified fusion input, taking into account randomness and pertinence; then, after 2 layers of dense connection layer processing, BatchNorm layer (stabilizing the training process, accelerating convergence) and LeakyReLU activation function (relieving gradient vanishing problem) are matched, and the output features are reshaped into a feature map of a specific size; the encoder contains 3 convolutional layers, each of which is connected with BatchNorm layer and LeakyReLU activation function, and realizes feature down-sampling through convolution operation, while accurately extracting local detail features of the temperature field; the residual block is connected at the end of the encoder, which strengthens the feature representation ability through shortcut connection and avoids gradient attenuation in deep network training; the decoder contains 2 layers of transpose convolutional layers, which realize feature up-sampling through transpose convolution, while aligning with the feature map of the corresponding layer of the encoder and realizing the jump connection (effectively preserving the low-level detail features), and finally output the preliminary feature map through the pixel shuffling block and 3x3 convolutional layer, and then match the original input size through the cropping layer and flatten to obtain the preliminary generated temperature field features.

[0068] Discriminator construction: PatchGAN structure is adopted to focus on strengthening the authenticity discrimination ability of local features of the temperature field and providing supervision for generator optimization. First, a conditional fusion module is designed to convert the conditional vector into a feature map consistent with the size of the temperature field feature map through a dense connection layer, and then the temperature field feature map is spliced in the channel dimension, so that the discriminator can refer to both the temperature field data and the corresponding working condition conditions, avoiding one-sidedness of relying solely on temperature field data for discrimination, and improving discrimination accuracy; the main part of the discriminator includes 2 convolution layers, each connected with a LeakyReLU activation function (enhancing non-linear fitting ability, adapting to complex feature differences) and a dropout layer (suppressing overfitting and improving generalization ability); finally, the extracted features are processed through a flattening layer, and the authenticity discrimination probability (close to 1 for real temperature field, close to 0 for generated temperature field) in the 0-1 interval is output through a single-channel dense connection layer, providing a clear supervision signal for parameter adjustment of the generator.

[0069] Secondly, to make up for the shortcoming of traditional CGAN that convolutional operations are difficult to capture long-range spatiotemporal correlations, the model introduces a Transformer architecture as a feature enhancement module, embedded between the encoder and decoder of the CGAN generator, forming a connection logic of encoding local features, enhancing global correlations with Transformers, and restoring temperature fields with decoders. Transformer is based on self-attention mechanism, which has the advantage of accurately capturing long-range dependencies in data. The specific construction process is as follows: First, flatten the 256-channel feature map output by the encoder into a sequence feature to adapt to the input format of the Transformer; then build a multi-head attention mechanism (with 8 attention heads), calculate the correlation weights of each position in the sequence, deeply mine the potential correlations of temperature data at different points and different times, and accurately capture the long-range spatiotemporal dependencies of the temperature field; then configure a feedforward neural network to perform nonlinear conversion on the attention-enhanced sequence features, further refining effective features and strengthening feature expression; finally, reshape the enhanced feature sequence into a feature map and input it into the decoder for subsequent upsampling process, so that the generated temperature field not only retains local details but also has reasonable global correlations.

[0070] In summary, the temperature field reconstruction model uses CGAN as the framework, and through the adversarial training of the generator and the discriminator to ensure the rationality and authenticity of the generated temperature field; uses Transformer as a feature enhancement module to make up for the limitations of local feature extraction and improve the ability to capture global correlations. The two work together to build a reconstruction model that takes into account directionality, authenticity, and high precision, effectively realizing the conversion from limited discrete measurement point data to global three-dimensional temperature field.

[0071] The temperature field reconstruction model is a deep learning model built by fusing a CGAN and a Transformer architecture, taking battery measurement point temperature data and operating parameters as input, and outputting an accurate battery three-dimensional temperature field as the target.

[0072] Specifically, the processing device iteratively trains the temperature field reconstruction model using a training data set; when the reconstruction error of the temperature field reconstruction model on the validation set meets the preset convergence condition, the target model is obtained, and the preset convergence condition is that the reconstruction error does not decrease in continuous multiple rounds of validation.

[0073] Iterative training refers to inputting the training data set into the model in batches, and through the process of continuously calculating the error between the model output and the true value, and optimizing the model parameters through back propagation, the optimal performance of the deep learning model is approached.

[0074] The validation set is a data set divided from the original data and independent of the training set, used to evaluate the generalization ability of the model during training to avoid overfitting of the model.

[0075] The reconstruction error refers to the degree of deviation between the reconstructed temperature field output by the model and the true temperature field, which is an index for measuring the reconstruction accuracy of the model, quantified by indicators such as root mean square error (RMSE).

[0076] The convergence condition is a threshold condition for determining whether the model training has reached the optimal state, which in this application is specifically manifested as the reconstruction error not decreasing in continuous multiple rounds of validation, meaning that the model performance has stabilized and further training cannot significantly improve the accuracy.

[0077] The target model is the temperature field reconstruction model that meets the convergence condition and has the optimal performance, which can be directly used to input real-time data to output the battery three-dimensional temperature field.

[0078] Specifically, first, a model is built by fusing a conditional generative adversarial network (CGAN) and a Transformer architecture, using the directional generation capability of CGAN to ensure the rationality of the temperature field output, and using the self-attention mechanism of Transformer to capture the spatio-temporal correlation features of the temperature data, and combining the two to form a temperature field reconstruction model with excellent performance; then the processing device inputs the constructed training data set into the model for iterative training, after each round of training, an independent validation set is used to evaluate the reconstruction error of the model, and the error change trend is continuously monitored; when the reconstruction error on the validation set does not decrease for multiple rounds in a row, it is determined that the model has reached the convergence state, at which point the training is stopped and the model at the current state is saved as the target model. The entire process ensures that the target model has both accurate temperature field reconstruction capability and good generalization performance by fusing two advanced model structures and setting a scientific convergence condition, and can adapt to the temperature field monitoring needs of the battery under different operating conditions.

[0079] S207, the processing device inputs the temperature data collected in real time at a limited number of optimal temperature measurement points and the real-time battery operation parameters into the target model to obtain a three-dimensional temperature field of the battery.

[0080] The three-dimensional temperature field of the battery refers to a complete temperature distribution state covering all regions on the surface and inside the battery and containing spatial three-dimensional coordinates and corresponding temperature values, which can intuitively and comprehensively reflect the global temperature condition of the battery and is a basis for accurate regulation and control of the thermal management system.

[0081] Specifically, during actual operation of the battery, the processing device synchronously completes two data collection tasks: one is to capture local temperature data of each measurement point in real time through sensors arranged at a limited number of optimal temperature measurement points, to ensure that the data accurately reflect the current temperature characteristics of the battery; and the other is to read the operation parameters of the battery in real time through a communication interface with associated devices such as a battery management system (BMS) and a charging and discharging device, to obtain working condition background information affecting temperature change.

[0082] Subsequently, the processing device standardizes the two types of real-time data according to the input format requirements (such as data dimension matching and consistent feature order) in the target model training stage, and then synchronously inputs them into the target model that has been trained. The target model quickly operates and deduces the input real-time discrete data by means of the directional generation capability of CGAN and the ability of the Transformer architecture to capture the spatiotemporal correlation features of temperature data, and expands the local temperature data of the limited measurement points to three-dimensional temperature field data covering the entire battery domain through the feature extraction, correlation analysis and temperature field generation logic inside the model.

[0083] The entire process realizes efficient connection of real-time data input, model rapid operation and global temperature field output, and provides accurate and comprehensive data support for the battery thermal management system to grasp the thermal state of the battery in real time and timely warn of local overheating risks.

[0084] Based on the above description, the present application has the following beneficial effects: In the present application, the optimal measurement point positions are obtained by constructing a multi-objective optimization function that minimizes temperature deviation and maximizes spatial coverage, and combining a differential evolution algorithm, which avoids the cost and heat interference of dense sensors and solves the problem of insufficient data representativeness of traditional limited measurement point schemes; and the model fused with CGAN and Transformer realizes accurate reconstruction from limited measurement point data to a three-dimensional temperature field, breaks through the limitation of low prediction accuracy of traditional interpolation / regression models under complex working conditions, and can effectively capture risk features such as local overheating of the battery.

[0085] Further, the training data covers multiple working conditions including uniform heat generation, non-uniform heat generation and thermal runaway, so that the model is adapted to the whole operation scenario of the battery; after the real-time output three-dimensional temperature field is de-normalized and smoothed, the global temperature distribution can be accurately reflected, so as to provide timely and comprehensive decision basis for the thermal management system, reduce the risk of thermal runaway caused by local overheating, and ensure long-term stable operation of the battery system.

[0086] Further, only a limited number of optimal measurement points are needed to realize global temperature monitoring, which greatly reduces the number and cost of sensor layout, and the model training and real-time reconstruction process is simple and efficient, easy to integrate with the existing battery management system, which improves the energy utilization efficiency and meets the actual application requirements of new energy vehicles, energy storage power stations and other scenarios.

[0087] The above Figure 1 The lithium battery temperature field reconstruction method provided by the embodiments of the application is described in detail, and the device and equipment provided by the embodiments of the application will be introduced in combination with the drawings.

[0088] As Figure 2 shown, the figure is a schematic diagram of a lithium battery temperature field reconstruction device provided by an embodiment of the application, which comprises: The acquisition module 301 is configured to acquire discrete temperature measurement point data on the surface of the battery and battery operation parameters. The processing module 302 is configured to construct a multi-objective optimization function according to the discrete temperature measurement point data, and to obtain a limited number of optimal temperature measurement point positions by iteratively solving the multi-objective optimization function using an optimization algorithm; temperature data under multiple working conditions are collected at the limited number of optimal temperature measurement point positions as first training data; and a training data set is constructed in combination with the first training data and the battery operation parameters. The training module 303 is configured to train a temperature field reconstruction model using the training data set to obtain a target model. The output module 304 is configured to input the temperature data of the limited number of optimal temperature measurement points and real-time battery operation parameters collected in real time into the target model to obtain a three-dimensional temperature field of the battery.

[0089] Optionally, the processing module 302 is specifically configured to obtain a temperature field reconstruction model based on a conditional generative adversarial network (CGAN) and a Transformer architecture.

[0090] Optionally, the processing module 302 is specifically configured to construct a multi-objective optimization function with the dual objectives of minimizing the deviation of the discrete temperature measurement point data from the global real temperature data and maximizing the spatial coverage of the measurement points.

[0091] Optionally, the training module 303 is specifically used to iteratively train the temperature field reconstruction model using the training dataset; when the reconstruction error of the temperature field reconstruction model on the validation set meets a preset convergence condition, the target model is obtained, wherein the preset convergence condition is that the reconstruction error does not decrease in multiple consecutive validation rounds.

[0092] Optionally, the processing module 302 is specifically used to iteratively solve the multi-objective optimization function using the differential evolution algorithm. When the value of the multi-objective optimization function converges to a stable state or reaches the preset maximum number of iterations, it outputs the sensor position distribution corresponding to the individual with the highest fitness in the current population as the finite number of optimal temperature measurement point positions.

[0093] Optionally, the processing module 302 is specifically used to perform inverse normalization processing on the three-dimensional temperature field to obtain the first temperature field data; and to use a mask Gaussian smoothing algorithm to smooth and optimize the invalid zero-value regions in the first temperature field data, and output the target temperature field distribution.

[0094] The lithium battery temperature field reconstruction device according to the embodiments of this application can correspondingly execute the method described in the embodiments of this application, and the other operations and / or functions of each module / unit of the lithium battery temperature field reconstruction device are respectively for realizing Figure 1 For the sake of brevity, the corresponding processes of each method in the illustrated embodiments will not be described in detail here.

[0095] This application also provides a computing device. For example... Figure 3 As shown in the figure, this is a schematic diagram of a computing device provided in an embodiment of this application. The computing device 700 includes a bus 701, a processor 702, a communication interface 703, and a memory 704. The processor 702, the memory 704, and the communication interface 703 communicate with each other via the bus 701.

[0096] The 701 bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0097] The processor 702 can be any one or more of a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP), etc.

[0098] The communication interface 703 is configured to communicate with the outside.

[0099] The memory 704 can include a volatile memory, such as a random access memory (RAM), and / or a non-volatile memory, such as a read-only memory (ROM), a floppy disk, a hard disk, or a solid state drive (SSD).

[0100] The memory 704 stores executable code, and the processor 702 executes the executable code to perform the aforementioned lithium battery temperature field reconstruction method.

[0101] Specifically, in the case of implementing the embodiment shown in the figure, and Figure 2 In the case of implementing the embodiment shown in the figure, and Figure 2 In the case of implementing the embodiment shown in the figure, and Figure 2 In the case of implementing the embodiment shown in the figure, and

[0102] The embodiments of the present application also provide a computer readable storage medium. The computer readable storage medium can be any available medium or data storage device that can store data which can be accessed by a computing device, or a data center containing one or more available media or data storage devices. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk), etc. The computer readable storage medium includes instructions indicating the computing device to execute the aforementioned lithium battery temperature field reconstruction method.

[0103] The embodiments of the present application further provide a computer program product including one or more computer instructions. When the computer instructions are loaded and executed on a computing device, all or part of the processes or functions described in the embodiments of the present application are generated.

[0104] The computer instructions can be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions can be transmitted from one website, computer or data center to another website, computer or data center through wired (for example, coaxial cable, optical fiber, digital subscriber line) or wireless (for example, infrared, wireless, microwave, etc.).

[0105] When the computer program product is executed by a computer, the computer executes any of the above-mentioned lithium battery temperature field reconstruction methods. The computer program product can be a software installation package, and when any of the above-mentioned lithium battery temperature field reconstruction methods is needed, the computer program product can be downloaded and executed on the computer.

[0106] The description of the corresponding processes or structures of each of the above-mentioned figures has its own emphasis, and the parts not described in detail in a certain process or structure can be referred to the related description of other processes or structures.

[0107] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited to this, any change or replacement within the technical scope disclosed in the present application should be covered in the protection scope of the present application.

Claims

1. A method for reconstructing the temperature field of a lithium battery, characterized in that, The method includes: Acquire discrete temperature measurement data and battery operating parameters on the battery surface; Based on the discrete temperature measurement data, a multi-objective optimization function is constructed; An optimization algorithm is used to iteratively solve the multi-objective optimization function to obtain a finite number of optimal temperature measurement point locations. Temperature data under various operating conditions are collected at the finite number of optimal temperature measurement points as the first training data. By combining the first training data and the battery operating parameters, a training dataset is constructed; The temperature field reconstruction model is trained using the training dataset to obtain the target model; The temperature data and real-time battery operating parameters collected at the finite number of optimal temperature measuring points are input into the target model to obtain the three-dimensional temperature field of the battery.

2. The method according to claim 1, characterized in that, The temperature field reconstruction model is obtained in the following way: A temperature field reconstruction model is obtained based on the conditional generative adversarial network CGAN and the Transformer architecture.

3. The method according to claim 1, characterized in that, The step of constructing a multi-objective optimization function based on the discrete temperature measurement point data includes: A multi-objective optimization function is constructed with the dual objectives of minimizing the deviation between discrete temperature measurement point data and the real temperature data of the whole domain and maximizing the spatial coverage of the measurement points.

4. The method according to claim 1, characterized in that, The step of training the temperature field reconstruction model using the training dataset to obtain the target model includes: The temperature field reconstruction model is iteratively trained using the training dataset. The target model is obtained when the reconstruction error of the temperature field reconstruction model on the validation set meets the preset convergence condition. The preset convergence condition is that the reconstruction error does not decrease in multiple consecutive validation rounds.

5. The method according to claim 1, characterized in that, The optimization algorithm is used to iteratively solve the multi-objective optimization function to obtain a finite number of optimal temperature measurement point locations, including: The multi-objective optimization function is solved iteratively by differential evolution algorithm. When the value of the multi-objective optimization function converges to a stable state or reaches the preset maximum number of iterations, the sensor position distribution corresponding to the individual with the highest fitness in the current population is output as the finite number of optimal temperature measurement point positions.

6. The method according to claim 1, characterized in that, The various operating conditions include uniform heat generation, non-uniform heat generation, and thermal runaway.

7. The method according to claim 1, characterized in that, After obtaining the three-dimensional temperature field of the battery, the method further includes: The three-dimensional temperature field is inversely normalized to obtain the first temperature field data; The masked Gaussian smoothing algorithm is used to smooth and optimize the invalid zero-value regions in the first temperature field data, and the target temperature field distribution is output.

8. A lithium battery temperature field reconstruction device, characterized in that, The device includes: The acquisition module is used to acquire discrete temperature measurement point data on the battery surface and battery operating parameters; The processing module is used to construct a multi-objective optimization function based on the discrete temperature measurement point data; iteratively solve the multi-objective optimization function using an optimization algorithm to obtain a finite number of optimal temperature measurement point locations; collect temperature data under various operating conditions at the finite number of optimal temperature measurement point locations as first training data; and construct a training dataset by combining the first training data and the battery operating parameters. The training module is used to train the temperature field reconstruction model using the training dataset to obtain the target model; The output module is used to input the temperature data of the finite number of optimal temperature measurement points collected in real time and the real-time battery operating parameters into the target model to obtain the three-dimensional temperature field of the battery.

9. A computing device, characterized in that, Including memory and processor; The memory stores one or more computer programs, the one or more computer programs including instructions; when the instructions are executed by the processor, the computing device performs the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program for performing the method as described in any one of claims 1 to 7.

Images