Battery failure prediction method, model training method, device, system, and medium

Abstract

Embodiments of the present disclosure relate to the technical field of battery management, and specifically relate to a battery fault prediction method, a model training method, equipment, a system and a medium. The main steps of the foregoing method include: obtaining a training data set and a fault data set carrying multiple battery state parameters from battery historical operation data, performing data processing on the training data set and the fault data set, extracting feature data corresponding to the battery state parameters to obtain a training feature set and a fault feature set, performing feature correlation analysis on each feature in the fault feature set to obtain feature correlations of each feature corresponding to a fault type, inputting the training feature set as training sample data into an initial model constructed based on a recurrent neural network and a Transformer to obtain an initial prediction model, adjusting parameters of the initial prediction model according to the feature correlations, and obtaining a battery fault prediction model. The foregoing method can improve the accuracy of the battery fault prediction model in predicting and identifying battery faults.

Description

Technical Field

[0001] This disclosure relates to the field of battery management technology, and in particular to a battery fault prediction method, model training method, device, system and medium. Background Technology

[0002] With the rapid development of electric vehicles, energy storage systems, and other fields, the safety and reliability of batteries, as core energy components, have become a major concern. However, batteries may malfunction during use due to overcharging, over-discharging, abnormal temperatures, or other reasons, potentially leading to serious safety accidents. Therefore, accurate prediction of battery malfunctions is crucial for ensuring the safe operation of battery systems.

[0003] Currently, battery fault prediction technology mainly relies on traditional statistical methods and simple machine learning models. For example, threshold detection methods based on battery voltage, current and other operating data, or predictions based on shallow machine learning models, while simple to implement, are difficult to fully capture the complex characteristics of battery faults, resulting in inaccurate battery fault prediction results. Summary of the Invention

[0004] Based on this, it is necessary to provide a battery fault prediction method, model training method, device, system, and medium to improve the accuracy of battery fault prediction, addressing at least one of the aforementioned technical problems.

[0005] In a first aspect, embodiments of this disclosure provide a battery fault prediction model training method, which may include the following steps: obtaining a training dataset and a fault dataset carrying multiple battery state parameters from historical battery operating data; performing data processing on the training dataset and the fault dataset to extract feature data corresponding to the battery state parameters, obtaining a training feature set and a fault feature set; performing feature correlation analysis on each feature in the fault feature set to obtain the feature correlation of each feature corresponding to the fault type; using the training feature set as training sample data and inputting it into an initial model constructed based on a recurrent neural network and a Transformer to obtain an initial prediction model; and adjusting the parameters of the initial prediction model according to the feature correlation to obtain a battery fault prediction model.

[0006] The training dataset includes multiple normal data records of normal battery operation and multiple fault data records of battery failure. Each fault data record includes a fault type label to indicate the fault type, and feature correlation is used to indicate the frequency or intensity of abnormal battery state parameters corresponding to the feature.

[0007] By constructing an initial model using recurrent neural networks and Transformers, and inputting the training dataset into this initial model for training, an initial prediction model is obtained. This approach combines the advantages of recurrent neural networks and Transformers to capture short-term and long-term dependencies in the sequence data of the training dataset, and provides a comprehensive analysis of the time-series data. This allows for better capture of the dynamic changes in the feature data corresponding to battery state parameters, improving the accuracy of the prediction results. Furthermore, by adjusting the parameters of the initial prediction model through feature correlation, a battery fault prediction model is obtained, which further optimizes the model's training effect and improves its performance.

[0008] In some embodiments, the feature data includes at least voltage feature data, current feature data, temperature feature data, and internal resistance feature data.

[0009] Considering multiple feature data during model training can improve the accuracy of model training.

[0010] In some embodiments, performing feature correlation analysis on each feature in the fault feature set may include the steps of: calculating the Pearson correlation coefficient or performing heatmap analysis on each feature in the fault feature set.

[0011] The Pearson correlation coefficient can quantify the correlation between features. Heatmap analysis can intuitively show the correlation between features through color intensity, which has the effect of visually displaying feature relationships. By performing feature correlation analysis, redundant features can be removed, model complexity can be reduced, and the computational cost and time of model training can also be reduced.

[0012] In some embodiments, performing feature correlation analysis on each feature in the fault feature set may include the following steps: calculating the Pearson correlation coefficient or performing heatmap analysis on each feature in the fault feature set to obtain the preliminary feature correlation of each feature, and optimizing the preliminary feature correlation based on preset fault mechanism data.

[0013] Among them, the fault mechanism data is used to represent the change process of various battery state parameters corresponding to various battery faults.

[0014] By further utilizing fault mechanism data to optimize the results of the preliminary feature correlation analysis, the feature correlation situation can be made more consistent with the actual situation, thereby improving the accuracy and generalization ability of the model.

[0015] In some embodiments, the structure of the initial model may include an input layer, an intermediate layer, a mapping layer, and an output layer.

[0016] The intermediate layer includes a Transformer sublayer and an LSTM sublayer, and the mapping layer includes a pooling layer or a fully connected layer.

[0017] The input layer receives the training sample data and converts it into vector data to meet the model's data requirements. The intermediate layer includes an LSTM sub-layer and a Transformer sub-layer, using a cascaded structure. The LSTM sub-layer processes the time-series data in the vector data, extracting local temporal features and data features. The Transformer sub-layer uses a self-attention mechanism to capture global dependencies between the local temporal features and data features, obtaining perceptual sequence features. The mapping layer maps the perceptual sequence features to obtain prediction data. The output layer processes the prediction data and outputs the corresponding prediction results.

[0018] By employing a cascaded structure of multi-layer LSTM and Transformer, the dynamic changes of battery state parameters in the battery capture training sample data, i.e., the battery's historical operating data, can be better captured, thus improving the accuracy of battery fault prediction.

[0019] In some embodiments, the battery fault prediction model training method may further include the following steps: obtaining a test dataset from historical battery operation data, processing the test dataset to obtain a test feature set, inputting the test feature set into the battery fault prediction model to obtain test results, obtaining an evaluation index of the battery fault prediction model based on the test results, and adjusting the number of training iterations, learning rate, and / or the number of network layers within the model based on the evaluation index to optimize model quality.

[0020] By testing the accuracy of the trained battery fault prediction model using test data, the model's effectiveness can be verified, thereby improving the accuracy of the model's prediction results.

[0021] In some embodiments, the training feature set is used as training sample data and input into an initial model constructed by a recurrent neural network and a Transformer to obtain an initial prediction model. This may include the steps of: iteratively training the training sample data using the initial model, calculating the error using the cross-entropy loss function, and updating the model parameter weights using the Adam optimizer to obtain the initial prediction model.

[0022] The cross-entropy loss function effectively measures the difference between the predicted probability distribution and the true distribution by calculating the error. Furthermore, the gradient calculation of the cross-entropy loss function is simple and stable, which helps the model converge quickly. The momentum mechanism and adaptive learning rate of the Adam optimizer can improve the stability of model training, thereby enhancing model performance.

[0023] In some embodiments, the battery fault prediction model training method may further include the steps of: performing data cleaning and data transformation on the training dataset and the fault dataset.

[0024] By cleaning and transforming the running data in the training and failure datasets, the data quality of the training and failure datasets can be improved, ensuring data consistency and increasing model training speed.

[0025] In a second aspect, embodiments of this disclosure provide a battery fault prediction method, which may include the following steps: acquiring the operating data of the battery under test, inputting the operating data of the battery under test into a battery fault prediction model, and obtaining the fault prediction result of the battery under test.

[0026] The battery fault prediction model is trained by the battery fault prediction model training method provided in any embodiment of the first aspect of this disclosure.

[0027] By inputting the operating data of the battery under test into a pre-trained battery fault prediction model for fault prediction, highly accurate battery fault prediction results can be obtained, enabling timely detection and management of battery faults and improving the safety of battery operation.

[0028] In some embodiments, the fault prediction result includes predicted fault type information, and the fault prediction method may further include: sending the fault prediction result to the BMS unit so that the BMS unit can control the operating state of the battery under test according to the predicted fault type information.

[0029] Sending the fault prediction results to the BMS unit enables the BMS unit to control the battery operating status in a timely manner based on battery faults, thereby improving the safety of battery operation and the timeliness of battery fault handling.

[0030] In a third aspect, embodiments of this disclosure provide a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the battery fault prediction model training method provided in any embodiment of the first aspect of this disclosure.

[0031] In a fourth aspect, embodiments of the present disclosure provide a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the battery failure prediction method provided in any embodiment of the second aspect of the present disclosure.

[0032] In a fifth aspect, embodiments of this disclosure provide a battery fault prediction system, which includes a BMS unit and a model running unit.

[0033] The battery fault prediction model deployed in the model running unit is trained by the battery fault prediction model training method provided in any embodiment of the first aspect of this disclosure.

[0034] The BMS unit communicates with the model running unit to send the operating data of the battery under test to the model running unit and controls the operating status of the battery under test based on the fault prediction results fed back by the model running unit.

[0035] The model execution unit is used to acquire the operating data of the battery under test, input the operating data of the battery under test into the battery fault prediction model, and obtain the fault prediction results.

[0036] The battery fault prediction model is deployed in the model running unit and does not occupy the system resources of the BMS unit, which can improve the response speed and real-time performance of battery fault prediction.

[0037] In some embodiments, the battery fault prediction system further includes a remote upgrade unit, which is communicatively connected to the model running unit, for remotely upgrading the battery fault prediction model deployed in the model running unit.

[0038] The remote upgrade unit enables remote upgrades and updates to the battery fault prediction model, making model upgrades more convenient, improving the efficiency of fault prediction model updates, and avoiding poor model performance due to untimely model updates.

[0039] In a sixth aspect, embodiments of the present disclosure provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the battery fault prediction model training method provided in any embodiment of the first aspect of the present disclosure.

[0040] In a seventh aspect, embodiments of the present disclosure provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the battery failure prediction method provided in any embodiment of the second aspect of the present disclosure.

[0041] The aforementioned battery fault prediction method, model training method, device, system, and medium utilize Recurrent Neural Networks (RNNs) and Transformers to construct an initial model. A training dataset including both normal and fault data is input into this initial model for training, resulting in an initial prediction model. This approach combines the advantages of RNNs and Transformers, capturing short-term and long-term dependencies in the sequence data of the training dataset and providing a comprehensive analysis of the time-series data. This allows for better capture of the dynamic changes in the characteristic data corresponding to battery state parameters, improving the accuracy of the prediction results. Furthermore, by adjusting the parameters of the initial prediction model through feature correlation, a battery fault prediction model is obtained, further optimizing the model's training effect and improving its performance. Attached Figure Description

[0042] Figure 1 This is a diagram illustrating the application environment of the battery fault prediction model training method in some embodiments.

[0043] Figure 2 This is a flowchart illustrating the battery fault prediction model training method in some embodiments;

[0044] Figure 3 This is a flowchart illustrating the feature correlation analysis steps in some embodiments;

[0045] Figure 4 This is a flowchart illustrating the model testing steps in some embodiments;

[0046] Figure 5 This is a flowchart illustrating a battery failure prediction method in some embodiments;

[0047] Figure 6 These are internal structural diagrams of the computer device in some embodiments;

[0048] Figure 7 This is a schematic diagram of a battery failure prediction system in some embodiments. Detailed Implementation

[0049] To make the technical solutions and advantages of this disclosure clearer, the embodiments and related technical content of this disclosure will be further described in detail below with reference to the accompanying drawings and text description. It should be understood that the embodiments described below are only used to explain the technical solutions of the embodiments of this disclosure and are not intended to limit more possible implementations of this disclosure.

[0050] It should be noted that relational terms such as "first" and "second" appearing in this document are used only to distinguish things, states, or actions, and do not necessarily indicate or imply relative importance or order. The terms "including," "comprising," or any other variations thereof are used to indicate non-exclusive inclusion, and the included objects may not be limited to those listed in this document. The terms "multiple" or other variations are used to indicate that the number of objects is two or more.

[0051] In a first aspect, embodiments of this disclosure provide a method for training a battery fault prediction model. This method can be applied to, for example... Figure 1 The application environment shown includes a first device 101 and a second device 102. The first device 101, used for training a battery fault prediction model, can be a local computer, a server, or other device capable of performing battery fault prediction model training. The second device 102 can be a device running a Battery Management System (BMS), a storage device for storing historical battery operating data, or other device used for collecting or recording historical battery operating data. The first device 101 can communicate with the second device 102 via a network or other means, receiving historical battery operating data from the second device 102.

[0052] Applying battery fault prediction model training methods to Figure 1 Taking the first device 101 as an example, in some embodiments, such as Figure 2 As shown, the battery fault prediction model training method includes steps S201 to S205 that can be executed by the first device 101. Each step is described in detail below.

[0053] Step S201: Obtain a training dataset and a fault dataset carrying multiple battery state parameters from the battery's historical operating data.

[0054] The training dataset includes multiple records of normal battery operation and multiple records of battery failure. Both the training and failure datasets contain historical battery operation data with various battery state parameters. Each failure data entry includes a failure type label indicating the type of failure. These battery state parameters can include parameters representing battery health such as voltage, current, temperature, and internal resistance, as well as battery state of charge, health status, differential pressure, and gas levels. The failure type label indicates the type of failure corresponding to the current failure data, and can include internal short circuits, overcharge, over-discharge, and thermal runaway. The historical battery operation data can be historical operation data from batteries in battery compartments of completed energy storage projects, such as historical operation data from energy storage devices used in industrial, commercial, and grid-scale applications.

[0055] In some specific examples, the training dataset could be 70% of the total historical battery running data. The fault dataset could be all or part of the historical battery running data.

[0056] Specifically, the first device 101 obtains normal operation data and fault operation data from the battery's historical operation data as a training dataset, and obtains fault data from the battery's historical operation data as a fault dataset. The battery's historical operation data can be obtained directly from the BMS unit that records the operation data, or it can be obtained from the battery database or memory that stores the battery's historical operation data. The source of the operation data is not particularly restricted here.

[0057] In some specific embodiments, the battery fault prediction model training method may further include grouping the running data in the fault dataset according to the fault type label to obtain multiple fault data groups. The grouped fault dataset may include an overcharge fault data group, an over-discharge fault data group, a short-circuit fault data group, and a thermal runaway fault data group.

[0058] In some specific embodiments, after obtaining the battery's training dataset and fault dataset, a training database and a fault database can be built separately so that they can be directly called when there is a need for model training of other batteries in the future.

[0059] Step S202: Process the training dataset and the fault dataset to extract feature data corresponding to the battery state parameters, and obtain the training feature set and the fault feature set.

[0060] The feature data may include voltage feature data, current feature data, temperature feature data, and internal resistance feature data corresponding to each battery state parameter.

[0061] Specifically, preliminary feature engineering is performed on the running data in the training dataset and the fault dataset to extract the features corresponding to each battery state parameter and obtain feature data.

[0062] Step S203: Perform feature correlation analysis on each feature data in the fault feature set to obtain the feature correlation of each feature corresponding to the fault type.

[0063] Among them, feature correlation is used to represent the frequency or intensity of abnormalities in the battery state parameters corresponding to a feature.

[0064] Abnormal battery status parameters refer to anomalies in various parameters indicating the battery's health status. These include abnormally high or low voltage, voltage exceeding the normal range, voltage dropping to zero, abnormal current fluctuations, current dropping to zero, increased temperature, decreased internal resistance, increased internal resistance, and internal resistance failure. Different anomalies or combinations of anomalies correspond to different fault types. For example, when characteristic data includes a sudden voltage drop, abnormal current fluctuations, increased temperature, and decreased internal resistance, the corresponding fault type is an internal short circuit. Conversely, when characteristic data includes zero voltage, zero current, and a sharp increase in temperature, the corresponding fault type is thermal runaway.

[0065] In some specific embodiments, feature correlation analysis is performed on the feature data in the fault feature set. This can involve statistically analyzing the probability or intensity percentage of the occurrence of features corresponding to various fault types in the fault data. The feature correlation of each feature can be expressed as a percentage. Based on the feature correlation of each feature corresponding to different fault types, the strength of the correlation between each feature and the corresponding fault type can be determined.

[0066] Step S204: Use the training feature set as training samples and input it into the initial model constructed by the recurrent neural network and Transformer to train and obtain the initial prediction model.

[0067] During the model training phase, the training feature set is input into the initial model constructed by recurrent neural networks (RNNs) and Transformers (a sequence model based on an attention mechanism) for training to obtain the initial prediction model.

[0068] The initial model includes a Transformer layer and a recurrent neural network layer, where the recurrent neural network layer can be a Long Short-Term Memory (LSTM) network.

[0069] Step S205: Adjust the parameters of the initial prediction model according to the feature correlation to obtain the battery failure prediction model.

[0070] The parameters of the initial prediction model adjusted based on feature relevance can be the feature weight coefficients corresponding to each feature, or the judgment thresholds corresponding to the classification results in the initial prediction model.

[0071] In some specific embodiments, the feature weight coefficients corresponding to each feature in the initial prediction model are adjusted according to the feature relevance. When the feature relevance is high, the weight coefficient corresponding to that feature is increased to strengthen the influence of that feature on the model prediction results.

[0072] In some specific embodiments, the judgment threshold of the initial prediction model is adjusted according to the feature correlation, and the range of the judgment threshold is adjusted according to the feature correlation to improve the accuracy of the model prediction results.

[0073] By utilizing Recurrent Neural Networks (RNNs) and Transformers to construct an initial model, and inputting a training dataset including both normal and fault data into this initial model for training, an initial prediction model is obtained. This approach combines the advantages of RNNs and Transformers, capturing both short-term and long-term dependencies in the sequence data of the training dataset, and comprehensively analyzing the time-series data. This allows for better capture of the dynamic changes in the feature data corresponding to battery state parameters, improving the accuracy of the prediction results. Furthermore, by adjusting the parameters of the initial prediction model through feature correlation, a battery fault prediction model is obtained, which further optimizes the model's training effect and improves its performance.

[0074] In some embodiments, the battery fault prediction method may further include inputting a training dataset into an initial model for training, and adjusting or training the judgment threshold in the initial model using feature correlation to obtain a battery fault prediction model.

[0075] In some embodiments, the feature data includes at least voltage feature data, current feature data, temperature feature data, and internal resistance feature data. The feature data may also include differential pressure feature data, gas feature data, and insulation failure feature data, etc. Considering multiple feature data during model training can improve the accuracy of model training.

[0076] In some embodiments, performing feature correlation analysis on each feature in the fault feature set may include: calculating the Pearson correlation coefficient or performing heatmap analysis on each feature data in the fault feature set.

[0077] The Pearson correlation coefficient can quantify the correlation between features. Heatmap analysis can visually display the correlation between features through color intensity, providing an intuitive display of feature relationships. Furthermore, by performing feature correlation analysis, highly correlated feature pairs can be found, redundant features can be removed, model complexity can be reduced, and the computational cost and time of model training can also be lowered.

[0078] In some embodiments, performing feature correlation analysis on each feature in the fault feature set may include steps S301 and S302.

[0079] Step S301: Calculate the Pearson correlation coefficient or perform heat map analysis on each feature in the fault feature set to obtain the preliminary feature correlation.

[0080] Step S302: Optimize the preliminary feature correlation based on the preset fault mechanism data.

[0081] The fault mechanism data is used to represent the changes in various battery state parameters corresponding to different types of battery faults. This fault mechanism data can include feature data corresponding to various types of battery faults. The feature correlations in the initial feature correlations that do not conform to the fault mechanism data are adjusted to obtain optimized feature correlations.

[0082] By further utilizing fault mechanism data to optimize the results of the preliminary feature correlation analysis, the feature correlation situation can be made more consistent with the actual situation, thereby improving the accuracy and generalization ability of the model.

[0083] In some embodiments, feature correlation may include high correlation, medium correlation, and low correlation. The correlation of features can be classified into high correlation, medium correlation, and low correlation based on the magnitude of the calculated Pearson correlation number or the color intensity of the heatmap.

[0084] In some embodiments, feature correlation can be expressed as a percentage, and the feature correlation of each feature corresponding to each fault type can be shown in Table 1.

[0085] Table 1: Correlation of Various Fault Characteristics

[0086]

[0087] Table 1 shows the feature data and feature correlation of each feature corresponding to each type of battery fault. The battery fault types in the table include short circuit, overcharge, over-discharge and thermal runaway. The percentages in the table are the feature correlation of each feature data.

[0088] In some embodiments, the structure of the initial model may include an input layer, an intermediate layer, a mapping layer, and an output layer.

[0089] The intermediate layer includes a Transformer sublayer and an LSTM sublayer, and the mapping layer includes a pooling layer or a fully connected layer.

[0090] The input layer receives the training sample data and converts it into vector data to meet the model's data requirements. The intermediate layer includes an LSTM sub-layer and a Transformer sub-layer, using a cascaded LSTM and Transformer structure. The LSTM sub-layer processes the time-series data in the vector data, extracting local temporal features and data features. The Transformer sub-layer uses a self-attention mechanism to capture global dependencies between the local temporal features and data features, obtaining perceptual sequence features. The mapping layer maps the perceptual sequence features to obtain prediction data. The output layer processes the prediction data and outputs the corresponding prediction results.

[0091] The Transformer and LSTM sublayers can each have multiple layers. The training sample data includes multi-dimensional features such as voltage, current, and temperature to represent the battery's health state. By using multiple Transformer layers stacked with LSTM, the dynamic changes of battery state parameters in the historical battery operation data (i.e., the battery's training sample data) can be better captured, improving the accuracy of battery fault prediction. The context-aware sequence features are generated by the Transformer sublayer through a self-attention mechanism, capturing global dependencies from local temporal features and data features to create a context-aware feature representation.

[0092] In some embodiments, such as Figure 4 As shown, the battery fault prediction model training method also includes steps S401 to S404.

[0093] Step S401: Obtain the test dataset from the battery's historical operating data.

[0094] Step S402: Process the test dataset to obtain the test feature set.

[0095] Step S403: Input the test feature set into the battery fault prediction model to obtain the test results.

[0096] Step S404: Obtain the evaluation index of the battery fault prediction model based on the test results, and adjust the number of training iterations, learning rate and / or the number of network layers within the model according to the evaluation index to optimize the model quality.

[0097] The test dataset can be 30% of the total historical battery operation data. This historical battery operation data can be recorded and stored by the BMS unit or stored in a specific database. Evaluation metrics can include accuracy and recall, as well as evaluation metrics for classification tasks such as precision and F1 score.

[0098] Data processing operations performed on the running data in the test dataset may include data cleaning, data transformation, and preliminary feature engineering to extract features from the test dataset and obtain a test feature set.

[0099] When the obtained evaluation index is lower than the preset pass standard, the hyperparameters of the model are tuned. For example, the number of training iterations, learning rate, and number of Transformer layers can be adjusted accordingly to improve the accuracy of the model's prediction results.

[0100] By testing the accuracy of the trained battery fault prediction model using test data, the model's effectiveness can be verified, thereby improving the accuracy of the model's prediction results.

[0101] In some embodiments, step S204 may include: iteratively training the training sample data using the initial model, calculating the error using the cross-entropy loss function, updating the model parameter weights using the Adam optimizer (Adaptive Moment Estimation, an optimization algorithm based on gradient descent), and obtaining the initial prediction model.

[0102] The cross-entropy loss function effectively measures the difference between the predicted probability distribution and the true distribution by calculating the error. Furthermore, the gradient calculation of the cross-entropy loss function is simple and stable, which helps the model converge quickly. The momentum mechanism and adaptive learning rate of the Adam optimizer can improve the stability of model training, thereby enhancing model performance.

[0103] In some embodiments, the battery fault prediction model training method may further include data cleaning and data transformation of the running data in the training dataset and the fault dataset.

[0104] Data cleaning can include handling missing values ​​in the data, such as adding or deleting missing data. Data transformation can involve standardizing or normalizing continuous values ​​in the data so that different data features have the same dimensions.

[0105] By cleaning and transforming the running data in the training and failure datasets, the data quality of the training and failure datasets can be improved, ensuring data consistency and increasing model training speed.

[0106] It should be understood that, although Figures 2 to 4 The steps in the flowchart are shown sequentially according to the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Figures 2 to 4 Unless otherwise expressly stated herein, the steps illustrated and other steps involved in the embodiments are not subject to strict order restrictions and may be performed in other orders. Furthermore, at least some steps in the foregoing embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but may be performed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but may be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0107] In a second aspect, embodiments of this disclosure provide a battery fault prediction method, such as... Figure 5 As shown, the battery fault prediction method includes steps S501 and S502.

[0108] Step S501: Obtain the operating data of the battery under test.

[0109] Step S502: Input the operating data of the battery under test into the battery fault prediction model to obtain the battery fault prediction result.

[0110] The battery fault prediction model is trained by the battery fault prediction model training method provided in any embodiment of the first aspect of this disclosure.

[0111] The battery under test can be a battery in a large energy storage device, such as a battery stack or a single battery cell in a power station, or a power battery pack in a new energy vehicle.

[0112] Specifically, the operational data can be battery status data collected in real time by the battery status monitoring unit, including battery voltage, current, temperature, internal resistance, and other data. The operational data of the battery under test can be obtained by the first device 101 from the operational data recorded by the BMS unit.

[0113] By inputting the operating data of the battery under test into a pre-trained battery fault prediction model for fault prediction, highly accurate battery fault prediction results can be obtained, enabling timely detection and management of battery faults and improving the safety of battery operation.

[0114] In some embodiments, the fault prediction result may include predicted fault type information, which represents the predicted fault type. The predicted fault type may include short circuit, overcharge, over-discharge, and thermal runaway. The predicted fault type can be set according to actual conditions and is not particularly limited here. The battery fault prediction method may also include sending the fault prediction result to the BMS unit so that the BMS unit can control the operating state of the battery under test based on the predicted fault type information.

[0115] Sending the fault prediction results to the BMS unit enables the BMS unit to control the battery operating status in a timely manner based on battery faults, thereby improving the safety of battery operation and the timeliness of battery fault handling.

[0116] Specifically, the BMS unit inputs the operational data of the battery under test into the battery fault prediction model to perform fault prediction, and feeds back the fault prediction results to the BMS unit. The BMS unit determines and executes the corresponding fault response operation based on the fault type in the fault prediction results and the preset fault response strategy. In some specific examples, the fault response strategy can include the following three categories:

[0117] (1) Level 1 fault: Charging and discharging are prohibited, and the entire battery system shuts down;

[0118] (2) Level 2 fault: Charging and discharging are prohibited, and the battery system partially shuts down;

[0119] (3) Level 3 fault: issue an alarm notification.

[0120] In some embodiments, the fault prediction result may include fault level information, which is used to represent the fault level. The battery fault prediction method may also include sending the fault prediction result to the BMS unit so that the BMS unit can respond to the fault according to the fault level.

[0121] The fault levels can be set according to actual conditions. For example, thermal runaway and short circuit faults are classified as Level 1 faults, while overcharge and over-discharge faults can be classified as Level 1, Level 2, or Level 3 faults respectively, depending on the specific voltage threshold. Specifically, different fault levels can be determined by setting the percentage range within which the overcharge or over-discharge value exceeds the standard voltage value. Different levels correspond to different fault response strategies. For example, when the fault level is Level 1, the BMS unit's fault response strategy may include issuing a warning notification, controlling the entire battery system to shut down, and activating the fire suppression system. When the fault level is Level 2, the fault response strategy may include issuing a warning notification and controlling a partial shutdown of the battery system. When the fault level is Level 3, the fault response strategy may include issuing a warning notification. Warning notifications can take the form of text notifications, sound alarms, etc., and can be sent via message prompts on the monitoring platform, pop-up reminders, or other methods.

[0122] In a third aspect, embodiments of this disclosure provide a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the battery fault prediction model training method provided in any embodiment of the first aspect of this disclosure.

[0123] In a fourth aspect, embodiments of the present disclosure provide a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the battery failure prediction method provided in any embodiment of the second aspect of the present disclosure.

[0124] In some embodiments, the computer device described above may be a server, and its internal structure diagram may be as follows: Figure 6As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores historical battery operating data. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements the battery fault prediction model training method in any embodiment of this document, or, when executed by the processor, implements the battery fault prediction method in any embodiment of this document.

[0125] Those skilled in the art will understand that Figure 6 The structures shown are merely block diagrams of some structures related to the embodiments of this disclosure and do not constitute a limitation on the computer devices to which the embodiments of this disclosure are applied. Specific computer devices may include more or fewer components than those shown in the figures, or combine certain components, or have different component arrangements.

[0126] In a fifth aspect, embodiments of this disclosure provide a battery failure prediction system, such as... Figure 7 As shown, the battery fault prediction system 700 includes a BMS unit 710 and a model running unit 720. The battery fault prediction model deployed in the model running unit is trained using the battery fault prediction model training method provided in any embodiment of the first aspect of this disclosure.

[0127] The BMS unit 710 is communicatively connected to the model operation unit 720, and is used to send the operating data of the battery under test to the model operation unit, and control the operating status of the battery under test according to the fault prediction results fed back by the model operation unit.

[0128] The model running unit 720 is used to acquire the operating data of the battery under test, input the operating data of the battery under test into the battery fault prediction model, and obtain the fault prediction results.

[0129] The battery fault prediction model is deployed independently in the model running unit, without occupying the system resources of the BMS unit, which can improve the response speed and real-time performance of battery fault prediction.

[0130] In some embodiments, the system further includes a remote upgrade unit, which is communicatively connected to the model running unit, for remotely upgrading the battery fault prediction model deployed in the model running unit 720.

[0131] The remote upgrade unit enables remote upgrades and updates to the battery fault prediction model, making model upgrades more convenient, improving the efficiency of fault prediction model updates, and avoiding poor model performance due to untimely model updates.

[0132] In a sixth aspect, embodiments of the present disclosure provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the battery fault prediction model training method provided in any embodiment of the first aspect of the present disclosure.

[0133] The computer-readable storage medium may be Figure 6 The computer-readable storage medium in the computer device shown.

[0134] In a seventh aspect, embodiments of the present disclosure provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the battery failure prediction method provided in any embodiment of the second aspect of the present disclosure.

[0135] The computer-readable storage medium may be Figure 6 The computer-readable storage medium in the computer device shown.

[0136] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The aforementioned computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments of this disclosure can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0137] 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 in 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 disclosure.

[0138] The above embodiments merely illustrate several implementation methods of this disclosure, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of protection of this disclosure. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this disclosure, and these all fall within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the appended claims.

Claims

1. A method for training a battery fault prediction model, characterized in that, The method includes: Obtain training and fault datasets carrying multiple battery state parameters from historical battery operation data. Each fault dataset includes a fault type label to indicate the fault type. Data processing is performed on the training dataset and the fault dataset to extract feature data corresponding to the battery state parameters, thereby obtaining the training feature set and the fault feature set. Feature correlation analysis is performed on each feature in the fault feature set to obtain the feature correlation degree of each feature corresponding to the fault type. The feature correlation degree is used to represent the frequency or intensity of abnormal battery state parameters corresponding to the feature. The step of performing feature correlation analysis on each feature in the fault feature set includes: calculating the Pearson correlation coefficient or performing heatmap analysis on each feature in the fault feature set to obtain the preliminary feature correlation of each feature; fault mechanism data is used to represent the change process of various battery state parameters corresponding to various battery faults, and the fault mechanism data may include feature data corresponding to various battery faults; the feature correlation of features that do not conform to the fault mechanism data in the preliminary feature correlation of the feature data is adjusted to obtain the optimized feature correlation; The training feature set is used as training sample data and input into the initial model constructed by the recurrent neural network and the Transformer to obtain the initial prediction model; The parameters of the initial prediction model are adjusted based on the feature correlation to obtain a battery failure prediction model.

2. The method according to claim 1, characterized in that, The characteristic data includes at least voltage characteristic data, current characteristic data, temperature characteristic data, and internal resistance characteristic data.

3. The method according to claim 1, characterized in that, The initial model includes: The input layer is used to receive the training sample data and convert the training sample data into vector data. The intermediate layer includes an LSTM sub-layer and a Transformer sub-layer. The LSTM sub-layer is used to process the time series data in the vector data and extract local temporal features and data features. The Transformer sub-layer is used to capture global dependencies between the local temporal features and the data features through a self-attention mechanism to obtain perceptual sequence features. The mapping layer, including a pooling layer or a fully connected layer, is used to map the features of the perceived sequence to obtain prediction data; The output layer is used to process the predicted data and output the prediction results.

4. The method according to claim 1, characterized in that, The method also includes, Obtain the test dataset from the battery's historical operating data; The test dataset is processed to obtain a test feature set; The test feature set is input into the battery fault prediction model to obtain the test results; The evaluation index of the battery fault prediction model is obtained based on the test results. The number of training iterations, learning rate and / or the number of network layers in the model are adjusted according to the evaluation index to optimize the model quality.

5. The method according to claim 1, characterized in that, The step of using the training feature set as training sample data and inputting it into the initial model built based on recurrent neural networks and Transformers to obtain the initial prediction model includes: The initial model is iteratively trained using the training sample data; The error is calculated using the cross-entropy loss function; The initial prediction model is obtained by updating the model parameter weights using the Adam optimizer.

6. The method according to claim 1, characterized in that, The method further includes: Data cleaning and transformation are performed on the training dataset and the fault dataset.

7. A battery fault prediction method, characterized in that, The method includes: Acquire the operating data of the battery under test; The operating data of the battery under test is input into the battery fault prediction model to obtain the fault prediction result of the battery under test, wherein the battery fault prediction model is trained by the training method of any one of claims 1 to 6.

8. The method according to claim 7, characterized in that, The fault prediction result includes information on the predicted fault type, and the method further includes: The fault prediction results are sent to the BMS unit so that the BMS unit can control the operating status of the battery under test based on the predicted fault type information.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

10. A battery fault prediction system, characterized in that, The system includes a BMS unit and a model running unit, wherein the battery fault prediction model deployed in the model running unit is trained by the training method described in any one of claims 1 to 6; The BMS unit is communicatively connected to the model running unit and is used to send the operating data of the battery under test to the model running unit, and control the operating status of the battery under test according to the fault prediction results fed back by the model running unit. The model running unit is used to acquire the operating data of the battery under test, input the operating data of the battery under test into the battery fault prediction model, and obtain the fault prediction result.

11. The system according to claim 10, characterized in that, The system also includes a remote upgrade unit, which is communicatively connected to the model running unit and is used to remotely upgrade the battery fault prediction model deployed in the model running unit.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 8.

Images