Early prediction method and system for battery cycle life, electronic equipment and storage medium

Abstract

The invention provides a battery cycle life early prediction method and system, electronic equipment and a storage medium. The battery cycle life early prediction method comprises the following steps: acquiring battery early cycle data; domain knowledge features and circulation global features are extracted from the battery early-stage circulation data; processing the domain knowledge features based on a multilayer perceptron model to obtain domain knowledge feature representation, and processing the loop global features based on a bidirectional attention model to obtain loop global feature representation; fusing the domain knowledge feature representation and the loop global feature representation to obtain a fused feature; and predicting the battery cycle life based on the fused features. The early prediction method for the cycle life of the battery can significantly improve the accuracy of battery life prediction, reduce the battery test cost, shorten the test period of a battery product, and provide more reliable technical support for design optimization, performance evaluation, use strategies and the like of the battery.

Description

Technical Field

[0001] This application relates to the field of battery life prediction technology, and in particular to a method, system, electronic device and storage medium for early prediction of battery cycle life. Background Technology

[0002] Batteries are widely used in electric vehicles, energy storage systems, and consumer electronics due to their high energy density, long cycle life, and environmental friendliness. However, batteries undergo aging phenomena such as capacity decay and increased internal resistance during long-term use, affecting their performance and safety. Therefore, accurately predicting battery life is of great significance.

[0003] Traditional battery life assessment methods primarily rely on full-cycle life testing, which involves subjecting batteries to hundreds to thousands of charge-discharge cycles until their capacity decays to a failure threshold. While this method can accurately reflect the actual lifespan of a battery, it is time-consuming and costly, making it difficult to meet the demands of rapid product iteration. Therefore, how to effectively predict battery life based on limited cycle data is a pressing challenge that needs to be addressed.

[0004] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] Based on this, embodiments of this application provide a method, system, electronic device, and storage medium for early prediction of battery cycle life, which can significantly improve the accuracy of battery life prediction, reduce battery testing costs, shorten the testing cycle of battery products, and provide more reliable technical support for battery design optimization, performance evaluation, and usage strategies.

[0006] According to some embodiments, this application provides a method for predicting the early cycle life of a battery, including:

[0007] Acquire early battery cycle data;

[0008] Extract domain knowledge features and cyclic global features from the early cycle data of the battery;

[0009] The domain knowledge features are processed based on a multilayer perceptron model to obtain a domain knowledge feature representation, and the recurrent global features are processed based on a bidirectional attention model to obtain a recurrent global feature representation.

[0010] The domain knowledge feature representation and the cyclic global feature representation are fused to obtain the fused feature;

[0011] Battery cycle life is predicted based on the fused features.

[0012] In some embodiments, the process of obtaining the domain knowledge feature representation includes the following steps:

[0013] Extract initial domain knowledge features from the early cycle data of the battery;

[0014] The initial domain knowledge features are filtered to remove redundancy, and the multilayer perceptron model is used for feature processing to obtain the domain knowledge feature representation.

[0015] In some embodiments, the cyclic global features are obtained by constructing cyclic differences.

[0016] In some embodiments, the processing of the recurrent global features based on the bidirectional attention model to obtain a recurrent global feature representation includes:

[0017] Based on the bidirectional attention model, the cyclic direction attention and voltage direction attention are used to extract the temporal information of the cyclic global features in the cyclic direction and voltage direction, respectively. The features in the two directions are then fused to obtain a cyclic global feature representation containing bidirectional temporal information.

[0018] In some embodiments, the fusion of features in two directions includes:

[0019] The features of the cyclic direction and the voltage direction are spliced ​​along the channel direction, and the features of the two directions are fused through a bidirectional feature fusion layer.

[0020] In some embodiments, during the fusion of the domain knowledge feature representation and the recurrent global feature representation, a joint learning strategy is introduced to guide the model to learn and fuse features through multi-task learning.

[0021] In some embodiments, the joint learning strategy improves the learning process by introducing an auxiliary prediction task, which includes predicting battery cycle life based solely on the domain knowledge features and predicting battery cycle life based solely on the cyclic global features.

[0022] According to some embodiments, this application also provides a battery cycle life early prediction system for implementing the battery cycle life early prediction method in the above embodiments, including:

[0023] The data acquisition module is configured to acquire early battery cycle data;

[0024] The feature extraction module, connected to the battery early cycle data acquisition module, is configured to extract domain knowledge features and cyclic global features from the battery early cycle data.

[0025] The feature processing module, connected to the feature extraction module, is configured to process the domain knowledge features based on a multilayer perceptron model to obtain a domain knowledge feature representation, and to process the recurrent global features based on a bidirectional attention model to obtain a recurrent global feature representation.

[0026] The fusion output module, connected to the feature processing module, is configured to fuse the domain knowledge feature representation and the cyclic global feature representation to obtain fused features;

[0027] The prediction module, connected to the fusion output module, is configured to predict battery cycle life based on the fusion features.

[0028] According to some embodiments, this application further provides an electronic device, including:

[0029] processor;

[0030] A memory in which executable instructions of the processor are stored;

[0031] The processor is configured to perform the steps of the battery cycle life early prediction method in the above embodiments by executing the executable instructions.

[0032] According to some embodiments, this application further provides a computer-readable storage medium for storing a program that, when executed by a processor, implements the steps of the battery cycle life early prediction method in the above embodiments.

[0033] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application.

[0034] The embodiments of this application may have, or at least have, the following advantages:

[0035] This application acquires early battery cycling data and extracts domain knowledge features and cyclic global features from it. After acquiring the cyclic global features, this application uses a multilayer perceptron model to process the domain knowledge features, learn the deep intrinsic correlations in the battery degradation process, and helps improve the accuracy of the life prediction model. Furthermore, it processes the cyclic global features based on a bidirectional attention model, thereby more fully mining the information in the cyclic global features and overcoming the limitation of only considering cyclic statistical features and losing potential temporal dynamic change features. This allows for a more comprehensive characterization of the early battery degradation process and accurate reflection of the battery's aging behavior.

[0036] After obtaining the recurrent global feature representation, this application integrates the domain knowledge features and the recurrent global feature representation, fully combining the multi-dimensional information in the domain knowledge and the recurrent global features to achieve effective integration of different types of features, thereby ensuring the learning of complementary and high-quality fused features.

[0037] Therefore, predicting battery cycle life based on this fusion feature can significantly improve the accuracy of battery life prediction, thereby reducing battery testing costs, shortening the testing cycle of battery products, and providing more reliable technical support for battery design optimization, performance evaluation, and usage strategies.

[0038] Other advantages, objectives, and features of this application will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from practice of this application. The objectives and other advantages of this application can be realized and obtained through the following description. Attached Figure Description

[0039] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0040] Figure 1 This is a flowchart illustrating the early prediction method for battery cycle life in this application.

[0041] Figure 2 This is a flowchart illustrating the method for predicting early battery cycle life in some embodiments of this application;

[0042] Figure 3 This is a flowchart illustrating the overall workflow of the battery cycle life early prediction method in some embodiments of this application;

[0043] Figure 4 This is a framework diagram of a bidirectional attention model in an early prediction method for battery cycle life according to some embodiments of this application;

[0044] Figure 5 The image shows the prediction results of the early prediction method for battery cycle life in some embodiments of this application.

[0045] Figure 6 This is a schematic diagram of the structure of the early prediction system for battery cycle life in some embodiments of this application;

[0046] Figure 7 This is a schematic diagram of the structure of an electronic device in some embodiments of this application;

[0047] Figure 8 This is a schematic diagram of the structure of a computer-readable storage medium in some embodiments of this application. Detailed Implementation

[0048] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided to make this application more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0049] Furthermore, the accompanying drawings are merely illustrative of this application and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0050] Traditional battery life assessment methods primarily rely on full-cycle life testing, which involves subjecting the battery to hundreds to thousands of charge-discharge cycles until its capacity decays to a failure threshold. While this method can accurately reflect the actual lifespan of a battery, it is time-consuming and costly, making it difficult to meet the demands of rapid product iteration.

[0051] Therefore, this application aims to provide a solution that can address the aforementioned technical problems, significantly improving the accuracy of battery life prediction, reducing battery testing costs, shortening the testing cycle of battery products, and providing more reliable technical support for battery design optimization, performance evaluation, and usage strategies. Details will be elaborated in subsequent embodiments.

[0052] According to some embodiments, this application provides a method for predicting early battery cycle life. Please refer to... Figure 1 The method for predicting the early cycle life of a battery may specifically include the following steps S100 to S500:

[0053] S100: Acquire early battery cycle data.

[0054] S200: Extract domain knowledge features and cyclic global features from early battery cycle data.

[0055] S300: Based on the multilayer perceptron model, the domain knowledge features are processed to obtain the domain knowledge feature representation, and based on the bidirectional attention model, the recurrent global features are processed to obtain the recurrent global feature representation.

[0056] S400: The domain knowledge feature representation and the cyclic global feature representation are fused to obtain the fused feature.

[0057] S500: Predicts battery cycle life based on fused features.

[0058] The aforementioned method for early prediction of battery cycle life acquires early battery cycle data and extracts domain knowledge features and cyclic global features from it. After acquiring the cyclic global features, a multilayer perceptron model is used to process the domain knowledge features, learning the deep intrinsic correlations in the battery degradation process, which helps to improve the accuracy of the life prediction model. Furthermore, the cyclic global features are processed based on a bidirectional attention model, thereby more fully mining the information in the cyclic global features and overcoming the limitation of only considering cyclic statistical features and losing potential temporal dynamic change features. This allows for a more comprehensive characterization of the early battery degradation process and accurate reflection of the battery's aging behavior.

[0059] After obtaining the cyclic global feature representation, the early prediction method for battery cycle life integrates the domain knowledge features and the cyclic global feature representation, fully combining the multi-dimensional information in the domain knowledge and cyclic global features to achieve effective integration of different types of features, thereby ensuring the learning of complementary and high-quality fused features.

[0060] Therefore, predicting battery cycle life based on this fusion feature can significantly improve the accuracy of battery life prediction, thereby reducing battery testing costs, shortening the testing cycle of battery products, and providing more reliable technical support for battery design optimization, performance evaluation, and usage strategies.

[0061] The above is the core idea of ​​this application. The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0062] In step S100, early battery cycle data is acquired, specifically, the early battery cycle data for the first N cycles is acquired. This embodiment does not strictly limit the specific value of N; the value of N can be adaptively adjusted based on factors such as battery type, application scenario, test conditions, and requirements for prediction accuracy.

[0063] As an example, such as Figure 2 As shown, the first 100 cycles can be selected as the range for collecting early cycle data of the battery.

[0064] It should also be noted that this application does not specifically limit the type of early battery cycle data collected. For example, such as... Figure 2 As shown, the voltage, current, discharge capacity, temperature, charging time and / or internal resistance data of the battery in the first 100 cycles can be collected as early cycle data of the battery, but it is not limited to this.

[0065] In practical applications, the specific type of early battery cycle data can be flexibly selected based on factors such as battery type, application scenario, and prediction accuracy requirements. As long as the collected data can effectively reflect the battery degradation information in the early cycle stage, it can be used as the early battery cycle data for subsequent feature extraction, processing, and fusion.

[0066] To more clearly illustrate the early prediction methods for battery cycle life in some embodiments, please refer to the following... Figure 2 and Figure 3 Understand steps S200 to S400 of some embodiments of this application. For example... Figure 2 As shown, a feature fusion network is established, including a domain knowledge feature processing module, a recurrent global feature processing module, and a fusion output module. Figure 3 A framework diagram of the feature fusion network in some embodiments is shown.

[0067] For step S200, in some embodiments, the domain knowledge feature representation can be obtained through the following steps S210 to S220:

[0068] S210: Extract initial domain knowledge features from early battery cycling data.

[0069] S220: The initial domain knowledge features are filtered to remove redundancy, and feature processing is performed using a multilayer perceptron model to obtain the domain knowledge feature representation.

[0070] The following is by Figure 3 The domain knowledge feature processing module shown above is used as an example to illustrate the aforementioned steps.

[0071] Feature F DKF As input, a two-stage feature selection strategy is employed to remove redundant features. First, the Spearman correlation coefficient can be used to evaluate the correlation between features and cycle lifetime, expressed as:

[0072]

[0073] Among them, R(X) i ) and R(Y i Let ρ and Y represent the rank of the i-th data point in vectors X and Y, respectively. A threshold |ρ| ≥ 0.45 is set to filter out multiple most relevant features. For example, filtering out 12 most relevant features is represented as follows:

[0074] Secondly, XGBoost regression is used for recursive feature elimination, ultimately retaining the most discriminative features. For example, retaining eight of the most discriminative features is represented as follows:

[0075] F XGB The input is fed into an encoder E consisting of multiple fully connected layers. DKF (·):

[0076] F latent =E DKF (F XGB )

[0077] in, This represents the latent features obtained after DKF encoding.

[0078] The following example illustrates the process of extracting domain knowledge features by using the collection of discharge capacity and voltage data as early battery cycle data.

[0079] Based on the collected discharge capacity and voltage data, the period-to-period difference of the battery discharge capacity curve Q(V) is calculated and expressed as:

[0080] ΔQ c2-c1 (V)=Q c2 (V)-Q c1 (V)

[0081] Where c1 and c2 represent the cycle numbers. Extract ΔQ 100-10 The variance, mean, minimum, and skewness of (V) are calculated, and multiple domain knowledge features, including charging time characteristics, temperature change characteristics, and / or internal resistance change characteristics, are generated based on information such as charging time, temperature, and internal resistance. These features are represented as follows:

[0082]

[0083] For example, z-score normalization can be used for initialization before performing subsequent steps (such as feature processing using a multilayer perceptron model).

[0084] In some embodiments, the cyclic global features in step S200 can be obtained by constructing cyclic differences.

[0085] By constructing cyclic differences, the temporal nature of cyclic features can be effectively preserved, providing a more reliable data foundation for subsequent feature analysis and lifetime prediction based on time-series information, enabling the model to better capture battery degradation patterns.

[0086] Compared to traditional cyclic statistical features, the aforementioned method of constructing cyclic differences reduces sensitivity to outliers during feature extraction, focusing more on reflecting relative changes between cycles, thus mitigating outlier sensitivity. Furthermore, this method avoids information loss issues arising from considering only cyclic statistical features, contributing to improved accuracy and reliability of lifetime prediction.

[0087] The following example illustrates the process of extracting global cyclic features by taking the first 100 cycles as early battery cycle data.

[0088] Construct the cyclic global feature ΔQ that differs from the second cycle from the third to the 100th cycle. 3:100-2 (V), and resampled into 128 linearly spaced voltage points within a voltage range of 3.6V to 2.0V for alignment. ΔQ 3:100-2 (V) can be calculated using the following formula:

[0089]

[0090] For example, initialization can be performed using min-max normalization before performing subsequent steps (such as inputting a bidirectional attention model).

[0091] The bidirectional attention model involved in step S300, for example Figure 4 As shown, this can also be called the bidirectional attention Transformer model. By using the bidirectional attention Transformer model to process cyclic global features, it is possible to fully mine the temporal information in the cyclic global features and capture the potential features of the cyclic global features in two temporal directions, thereby more comprehensively representing the early degradation process of the battery.

[0092] Combination Figure 3 It is understood that in some embodiments, the recurrent global feature processing module uses the bidirectional attention Transformer model described above to learn and fuse degradation information from two temporal directions from the recurrent global features, generating latent features based on the recurrent global features.

[0093] In some embodiments, based on a bidirectional attention model, cyclic direction attention and voltage direction attention can be used to extract the time-series information of the cyclic global features in the cyclic direction and voltage direction, respectively, thereby simultaneously capturing the early degradation dynamics of the battery in the cyclic direction and voltage direction; and the features in the two directions are fused to obtain a cyclic global feature representation containing bidirectional time-series information.

[0094] The following is from Figure 3 Taking the aforementioned steps as an example, the loop global feature processing module shown above is combined with... Figure 4 An example is provided.

[0095] Cyclic directional attention: using recurrent global features ΔQ 3:100-2 (V) is used as input, and positional information is supplemented using sine and cosine positional encoding methods. The encoded input is represented as... Then, a linear transformation is performed on X to obtain the cyclic directional attention mechanism. and Represented as:

[0096]

[0097] in, This is the learnable weight matrix for the three linear layers in the recurrent direction. Then, the attention in the recurrent direction can be calculated using a dot product:

[0098]

[0099] in, The attention score represents the direction of repetition; the Softmax function is used for normalization; d c =98 is the scaling factor.

[0100] Voltage direction attention: X is transposed and used as input to the voltage direction attention model. To reduce computational redundancy and allow attention scores from both directions to operate in the same value space, thereby ensuring the fusion and supplementation of bidirectional temporal information, in some embodiments, the voltage direction self-attention model is defined by sharing V as follows:

[0101]

[0102] in, It is a learnable weight matrix for two linear layers in the voltage direction. Based on the sharing mechanism, by V c Generate by transposing. Attention indicating the direction of circulation, The attention fraction representing the voltage direction, d v =128 is the scaling factor.

[0103] In some embodiments, the process of fusing features in two directions described above may specifically include splicing features in the cyclic direction and voltage direction along the channel direction, and fusing features in the two directions through a bidirectional feature fusion layer.

[0104] As an example, features from two directions can be fused using a bidirectional feature fusion layer consisting of two 1×1 convolutional layers, as shown below:

[0105]

[0106] Among them, Conv 1x1 This represents a convolutional layer with a 1×1 kernel size, ReLU represents the activation function, and [,] represents a channel-wise cascading operation. It is a mapping matrix, representing the fused output features.

[0107] To enrich the bidirectional temporal information combination, the first 1×1 convolution increases the dimension from 2 to 4, and the second 1×1 convolution expands it back to 1. The learned features can then be obtained through a feedforward network (FFN).

[0108] G FFN =A fusion +FFN(LN(A fusion +X))

[0109] in, This represents the output of the bidirectional attention Transformer model, and LN represents layer normalization.

[0110] Finally, the fused vector is flattened into a high-dimensional vector and input into the encoder E, which consists of multiple fully connected layers. CGF Generate more compact feature representations in (·):

[0111] G laten =E CGF (Flatten(G FFN ))

[0112] in, This represents the latent features obtained after recurrent global feature encoding.

[0113] For step S400, as an example, it can be adopted as follows: Figure 3 The fusion output module shown integrates the features learned by the aforementioned domain knowledge feature processing module and the cyclic global feature processing module to obtain fused features.

[0114] As an example, the F output by the aforementioned domain knowledge feature processing module latent and the output of the global feature processing module G latent By splicing and merging, a new feature is formed.

[0115] F Fusion =Concat(F latent G laten )

[0116] Here, concat(·) represents a join operation.

[0117] Considering F latent and G latent They may have different scales, applying batch normalization (BN) to F Fusion To ensure stable training, the normalized features are then input into a predictor P composed of fully connected layers. Fusion (·):

[0118]

[0119] in, This indicates the predicted battery cycle life.

[0120] In some embodiments, a joint learning strategy can be introduced during the fusion of domain knowledge feature representations and recurrent global feature representations. This strategy guides the model to learn and fuse features through multi-task learning. This helps obtain complementary and high-quality feature representations, thereby effectively improving the model's generalization ability and prediction accuracy.

[0121] In some embodiments, the above-described joint learning strategy can improve the learning process by introducing an auxiliary prediction task.

[0122] The aforementioned joint learning strategy provides more information for model training by utilizing the prediction results and loss function of the auxiliary prediction task. It improves the feature learning process through multi-task joint learning, allowing features learned from each path to participate in the prediction of the corresponding auxiliary prediction task, thereby ensuring that the features from each path retain their own lifetime prediction capability during the fusion process. This helps the model better learn the relationship between input features and battery cycle life.

[0123] For example, the auxiliary prediction task may include predicting battery cycle life based solely on domain knowledge features, and predicting battery cycle life based solely on global cycle features. Exemplary examples are provided below.

[0124] Auxiliary Task 1: F Only latent Used for prediction; Auxiliary Task 2: G only latent Used for prediction.

[0125]

[0126] Among them, P DKF and P CGF Both represent predictors composed of fully connected layers. and This is the corresponding prediction result. The loss function for the above auxiliary prediction task is calculated using the following formula:

[0127]

[0128] Based on this, the mean squared error loss function MSELoss can be expressed as:

[0129]

[0130] The joint loss function, which is a weighted sum of the losses for each task, can be expressed as:

[0131] L Joint =αLFusion +βL DKF +γL CGF

[0132] Here, α, β, and γ represent trade-off parameters. This application does not strictly limit the specific values ​​of the trade-off parameters α, β, and γ in the joint loss function; α, β, and γ can be adaptively adjusted according to factors such as the actual application scenario, battery type, and prediction accuracy requirements. For example, α can be set to 1, β to 1.2, and γ to 0.2.

[0133] To verify the effectiveness of the early prediction method for battery cycle life provided in the embodiments of this application, the following experiments were conducted.

[0134] The experiment was validated using 123 commercially available 18650 lithium iron phosphate / graphite batteries, and the dataset was divided into a training set, a primary test set, and a secondary test set (the secondary test set considered calendar aging factors). Data acquisition and feature generation were performed using the early prediction method for battery cycle life provided in this application. Subsequently, a cycle life prediction model was trained based on the training set, and finally, battery cycle life was predicted using two test sets.

[0135] The predicted battery cycle life results of the above experiments are as follows: Figure 5 As shown, it can be observed that most predicted points are highly consistent with the actual values, closely resembling the y=x line. Further analysis of the error distribution plot reveals that in both test sets, the prediction errors are concentrated only around 0, with only a very few outliers deviating significantly. Figure 5 The prediction results shown strongly demonstrate that the battery cycle life early prediction method provided in this application has excellent effectiveness and generalization performance in the field of battery cycle life early prediction.

[0136] To further illustrate the effectiveness of the embodiments of this application, comparative experiments are conducted below with Parallel Feature Fusion Network (PFFN) and Convolutional Neural Network-Fully Connected (CNN-FC).

[0137] In the comparative experiments, the evaluation metrics used included Mean Relative Error (MAPE) and Root Mean Square Error (RMSE). To ensure the fairness of the experiments, all methods were tested 10 times repeatedly, and the average value was taken as the result. Based on the comparison of prediction accuracy of the three methods shown in Table 1, it is clear that the embodiments of this application outperform the other comparative methods in terms of evaluation metrics on both test sets. This result fully demonstrates the significant advantage of the embodiments of this application in accurately modeling the complex relationship between early cycle characteristics and battery cycle life.

[0138] Table 1 Comparison of experimental results

[0139]

[0140] Based on the same inventive concept, this application also provides a battery cycle life early prediction system according to some embodiments, for implementing the battery cycle life early prediction method in the above embodiments. Therefore, the battery cycle life early prediction system can also achieve the technical effects that the aforementioned battery cycle life early prediction method can achieve, and will not be described in detail here.

[0141] Please see Figure 6 The battery cycle life early prediction system includes a data acquisition module 101, a feature extraction module 102, a feature processing module 103, a fusion output module 104, and a prediction module 105.

[0142] The data acquisition module 101 is configured to acquire early battery cycle data; the feature extraction module 102 is connected to the early battery cycle data acquisition module 101 and is configured to extract domain knowledge features and cyclic global features from the early battery cycle data; the feature processing module 103 is connected to the feature extraction module 102 and is configured to process the domain knowledge features based on a multilayer perceptron model to obtain a domain knowledge feature representation, and process the cyclic global features based on a bidirectional attention model to obtain a cyclic global feature representation; the fusion output module 104 is connected to the feature processing module 103 and is configured to fuse the domain knowledge feature representation and the cyclic global feature representation to obtain a fused feature; the prediction module 105 is connected to the fusion output module 104 and is configured to predict the battery cycle life based on the fused feature.

[0143] The modules in the aforementioned early battery cycle life prediction system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0144] It is understood that the battery cycle life early prediction system of this application also includes other existing functional modules that support the operation of the battery cycle life early prediction system. Figure 6 The battery cycle life early prediction system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0145] This application also provides an electronic device, including a processor; a memory storing executable instructions of the processor; wherein the processor is configured to perform the steps of the battery cycle life early prediction method described in the foregoing embodiments by executing the executable instructions.

[0146] Those skilled in the art will understand that various aspects of this application can be implemented as a system, method, or program product. Therefore, various aspects of this application can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, collectively referred to herein as a "circuit," "module," or "platform."

[0147] The following reference Figure 7 To describe an electronic device 600 according to this embodiment of the present application. Figure 7 The electronic device 600 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0148] like Figure 7 As shown, the electronic device 600 is presented in the form of a general-purpose computing device. The components of the electronic device 600 may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different system components (including storage unit 620 and processing unit 610), a display unit 640, etc.

[0149] The storage unit stores program code that can be executed by the processing unit 610, causing the processing unit 610 to perform the steps described in the above-described section of this specification regarding the method for predicting early battery cycle life according to various exemplary embodiments of this application. For example, the processing unit 610 can perform actions such as... Figure 1 The steps are shown in the figure.

[0150] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 6201 and / or a cache storage unit 6202, and may further include a read-only memory unit (ROM) 6203.

[0151] The storage unit 620 may also include a program / utility 6204 having a set (at least one) program module 6205, such program module 6205 including but not limited to: an operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0152] Bus 630 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0153] Electronic device 600 can also communicate with one or more external devices 700 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 600, and / or with any device that enables electronic device 600 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 650. Furthermore, electronic device 600 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 660. Network adapter 660 can communicate with other modules of electronic device 600 via bus 630. It should be understood that, although... Figure 7 As not shown, other hardware and / or software modules may be used in conjunction with electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0154] In the electronic device, when the program in the memory is executed by the processor, it implements the steps of the battery cycle life early prediction method described in the foregoing embodiments. Therefore, the electronic device can also obtain the technical effects of the aforementioned battery cycle life early prediction method.

[0155] This application also provides a computer-readable storage medium for storing a program that, when executed by a processor, implements the steps of the battery cycle life early prediction method described in the foregoing embodiments. In some possible implementations, various aspects of this application can also be implemented as a program product comprising program code that, when executed on a terminal device, causes the terminal device to perform the steps of the various exemplary embodiments of this application described in the battery cycle life early prediction method section of this specification.

[0156] refer to Figure 8 As shown, a program product 800 for implementing the above-described method according to an embodiment of this application is described. It may employ a portable compact disc read-only memory (CD-ROM) and include program code, and can be executed on a terminal device, such as a personal computer. However, the program product of this application is not limited thereto. In this document, a readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0157] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0158] The computer-readable storage medium may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable storage medium may also be any readable medium other than a readable storage medium, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0159] Program code for performing the operations of this application can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0160] When the program in the computer storage medium is executed by the processor, it implements the steps of the battery cycle life early prediction method. Therefore, the computer storage medium can also achieve the technical effects of the above-mentioned battery cycle life early prediction method.

[0161] In the description of this specification, references to terms such as "some embodiments," "as an example," "exemplarily," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative descriptions of the above terms do not necessarily refer to the same embodiment or example.

[0162] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features of the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0163] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for predicting early battery cycle life, characterized in that, include: Acquire early battery cycle data; Extract domain knowledge features and cyclic global features from the early cycle data of the battery; The domain knowledge features are processed based on a multilayer perceptron model to obtain a domain knowledge feature representation, and the recurrent global features are processed based on a bidirectional attention model to obtain a recurrent global feature representation. The domain knowledge feature representation and the cyclic global feature representation are fused to obtain the fused feature; Battery cycle life is predicted based on the fused features.

2. The method for early prediction of battery cycle life according to claim 1, characterized in that, The process of obtaining the domain knowledge feature representation includes the following steps: Extract initial domain knowledge features from the early cycle data of the battery; The initial domain knowledge features are filtered to remove redundancy, and the multilayer perceptron model is used for feature processing to obtain the domain knowledge feature representation.

3. The method for early prediction of battery cycle life according to claim 1, characterized in that, The cyclic global features are obtained by constructing cyclic differences.

4. The method for early prediction of battery cycle life according to claim 1, characterized in that, The process of processing the recurrent global features based on the bidirectional attention model to obtain the recurrent global feature representation includes: Based on the bidirectional attention model, the cyclic direction attention and voltage direction attention are used to extract the temporal information of the cyclic global features in the cyclic direction and voltage direction, respectively. The features in the two directions are then fused to obtain a cyclic global feature representation containing bidirectional temporal information.

5. The method for early prediction of battery cycle life according to claim 4, characterized in that, The fusion of features from two directions includes: The features of the cyclic direction and the voltage direction are spliced ​​along the channel direction, and the features of the two directions are fused through a bidirectional feature fusion layer.

6. The method for early prediction of battery cycle life according to claim 1, characterized in that, In the process of fusing the domain knowledge feature representation and the recurrent global feature representation, a joint learning strategy is introduced to guide the model to learn and fuse features through multi-task learning.

7. The method for early prediction of battery cycle life according to claim 6, characterized in that, The joint learning strategy improves the learning process by introducing auxiliary prediction tasks, which include predicting battery cycle life based solely on the domain knowledge features and predicting battery cycle life based solely on the cyclic global features.

8. A battery cycle life early prediction system, characterized in that, The method for early prediction of battery cycle life according to any one of claims 1 to 7 includes: The data acquisition module is configured to acquire early battery cycle data; The feature extraction module, connected to the battery early cycle data acquisition module, is configured to extract domain knowledge features and cyclic global features from the battery early cycle data. The feature processing module, connected to the feature extraction module, is configured to process the domain knowledge features based on a multilayer perceptron model to obtain a domain knowledge feature representation, and to process the recurrent global features based on a bidirectional attention model to obtain a recurrent global feature representation. The fusion output module, connected to the feature processing module, is configured to fuse the domain knowledge feature representation and the cyclic global feature representation to obtain fused features; The prediction module, connected to the fusion output module, is configured to predict battery cycle life based on the fusion features.

9. An electronic device, characterized in that, include: processor; A memory in which executable instructions of the processor are stored; The processor is configured to perform the steps of the battery cycle life early prediction method according to any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium for storing a program, characterized in that, When the program is executed by the processor, it implements the steps of the battery cycle life early prediction method according to any one of claims 1 to 7.

Images