Battery status detection methods, devices, systems, electronic equipment and storage media

Abstract

This application discloses a battery state detection method, apparatus, system, electronic device, and storage medium, belonging to the field of battery technology. The battery state detection method includes: acquiring first operating state data of a battery to be detected; inputting the first operating state data into an anomaly diagnosis model to obtain a first battery diagnosis result of the battery to be detected output by the anomaly diagnosis model; wherein the first operating state data and the first battery label corresponding to the first operating state data are used to construct incremental training samples, and the incremental training samples are used to perform online update training of the anomaly diagnosis model. This battery state detection method trains the anomaly diagnosis model through online incremental learning. The anomaly diagnosis model diagnoses the battery state of the battery to be detected based on the operating state data of the battery to be detected, and can adapt to different diagnostic tasks, effectively improving the accuracy and efficiency of battery detection.

Description

Technical Field

[0001] This application belongs to the field of battery technology, and in particular relates to a battery state detection method, device, system, electronic device and storage medium. Background Technology

[0002] Batteries are widely used in various everyday scenarios such as mobile phones, laptops, power tools, and new energy vehicles. Batteries have advantages such as high energy density, long lifespan, low self-discharge rate, light weight, and environmental friendliness. However, they also pose risks of poor safety and explosion under high-temperature conditions. Battery anomaly detection is crucial. By identifying and promptly repairing battery malfunctions, greater losses can be effectively prevented.

[0003] Currently, common battery anomaly detection methods include those based on parameter inconsistency, those based on battery models, and those based on machine learning or deep learning. These methods are typically only applicable to one type, some types, or some operating conditions of battery anomalies. If it is necessary to adapt to different types or operating conditions of battery anomaly detection, it is necessary to reset the threshold, rebuild the battery model, and retrain the machine learning or deep learning model to adapt to the new battery or new operating condition anomaly detection task. The versatility is generally limited, and the re-adaptation workflow is time-consuming. Summary of the Invention

[0004] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes a battery state detection method, apparatus, system, electronic device, and storage medium, which are adapted to the detection of abnormalities in batteries of different types or under different operating conditions, effectively improving the accuracy and efficiency of battery detection.

[0005] In a first aspect, this application provides a battery state detection method, the method comprising:

[0006] Acquire the first operating status data of the battery to be tested;

[0007] The first operating status data is input into the anomaly diagnosis model to obtain the first battery diagnosis result of the battery to be tested output by the anomaly diagnosis model.

[0008] The first operating status data and the first battery label corresponding to the first operating status data are used to construct incremental training samples, which are used to update and train the anomaly diagnosis model online.

[0009] According to the battery state detection method of this application, an anomaly diagnosis model is trained through online incremental learning. The anomaly diagnosis model diagnoses the battery state of the battery under test based on the operating state data of the battery under test. It can be adapted to different diagnostic tasks and effectively improve the accuracy and efficiency of battery detection.

[0010] According to one embodiment of this application, after obtaining the first battery diagnostic result of the battery to be tested output by the anomaly diagnostic model, the method further includes:

[0011] The diagnostic results of the first battery are verified and reviewed to obtain the verification results of the first battery.

[0012] The first battery verification result is input into the anomaly diagnosis model, and the anomaly diagnosis model is updated and trained online.

[0013] According to one embodiment of this application, the step of inputting the first battery verification result into the anomaly diagnosis model and performing online update training on the anomaly diagnosis model includes:

[0014] Based on the verification results of the first battery, the first battery label is corrected;

[0015] The first operating status data and the corrected first battery label are used as incremental training samples to update and train the anomaly diagnosis model online.

[0016] According to one embodiment of this application, correcting the first battery label based on the first battery verification result includes:

[0017] If it is determined that the first corrected label corresponding to the first battery verification result is a historical label, the first battery label is replaced with the first corrected label, and / or, sample data with the same label as the first battery label are removed from the incremental training samples.

[0018] According to one embodiment of this application, correcting the first battery label based on the first battery verification result includes:

[0019] If the second corrected label corresponding to the first battery verification result is determined to be a newly added label, the second corrected label and the first battery label are used as the incremental training samples;

[0020] Alternatively, if it is determined that the second corrected label is a newly added label, the first battery label is replaced with the second corrected label, and the labels of the same type as the first battery label in the incremental training samples are replaced with the second corrected label.

[0021] According to one embodiment of this application, the anomaly diagnosis model is trained through the following steps:

[0022] Historical battery operating status data and historical battery tags are input into the first diagnostic model for model training to obtain the initial anomaly diagnostic model. The historical battery tags are determined based on the historical battery operating status data.

[0023] The incremental training samples are input into the anomaly diagnosis model to perform online iterative updates of the anomaly diagnosis model.

[0024] According to one embodiment of this application, the first battery tag is obtained through the following steps:

[0025] Feature extraction is performed on the first operating state data to obtain the first operating state features;

[0026] The first battery tag is determined based on the first operating state characteristics.

[0027] Secondly, this application provides a battery state detection device, the device comprising:

[0028] The acquisition module is used to acquire the first operating status data of the battery to be tested;

[0029] The processing module is used to input the first operating status data into the anomaly diagnosis model and obtain the first battery diagnosis result of the battery to be tested output by the anomaly diagnosis model.

[0030] The first operating status data and the first battery label corresponding to the first operating status data are used to construct incremental training samples, which are used to update and train the anomaly diagnosis model online.

[0031] According to the battery state detection device of this application, an anomaly diagnosis model is trained through online incremental learning. The anomaly diagnosis model diagnoses the battery state of the battery under test based on the operating state data of the battery under test. It can be adapted to different diagnostic tasks and effectively improve the accuracy and efficiency of battery detection.

[0032] Thirdly, this application provides a battery state detection system, the system comprising:

[0033] The data processing module is used to determine the first battery label corresponding to the first operating state data of the battery to be detected, and the first operating state data and the first battery label are used to construct incremental training samples.

[0034] A diagnostic model module is connected to the data processing module. The diagnostic model module includes an anomaly diagnostic model. The diagnostic model module is used to input the first operating status data into the anomaly diagnostic model to obtain the first battery diagnostic result of the battery to be tested output by the anomaly diagnostic model. The incremental training samples are used to perform online update training on the anomaly diagnostic model.

[0035] According to the battery state detection system of this application, the anomaly diagnosis model is trained through online incremental learning. The anomaly diagnosis model diagnoses the battery state of the battery under test based on the operating state data of the battery under test. It can be adapted to different diagnostic tasks and effectively improve the accuracy and efficiency of battery detection.

[0036] Fourthly, this application provides a battery state detection system, the system comprising:

[0037] Battery Management System (BMS)

[0038] An embedded development board is installed in the battery BMS system and is used to collect battery operating status data of the battery BMS system.

[0039] The cloud is communicatively connected to the embedded development board. The embedded development board is used to transmit the battery operating status data to the cloud. The cloud is used to input the battery operating status data into the anomaly diagnosis model, obtain the battery diagnosis result of the battery BMS system output by the anomaly diagnosis model, and send the battery diagnosis result of the battery BMS system back to the embedded development board.

[0040] The battery operating status data and the corresponding battery tags are used to construct incremental training samples, which are used to update and train the anomaly diagnosis model online.

[0041] According to the battery state detection system of this application, data acquisition and transmission are performed through an embedded development board, and model training and diagnostic inference are completed in the cloud. Only by deploying an embedded development board in the battery BMS system can the battery state detection of the battery BMS system be realized. The main calculation work of diagnosis is performed in the cloud, which is suitable for small-scale battery systems.

[0042] Fifthly, this application provides a battery state detection system, the system comprising:

[0043] Battery Management System (BMS)

[0044] The server is communicatively connected to the battery BMS system. The server is used to acquire battery operating status data of the battery BMS system, input the battery operating status data into the anomaly diagnosis model, obtain the battery diagnosis result of the battery BMS system output by the anomaly diagnosis model, and send the battery diagnosis result of the battery BMS system back to the battery BMS system.

[0045] The battery operating status data and the corresponding battery tags are used to construct incremental training samples, which are used to update and train the anomaly diagnosis model online.

[0046] According to the battery status detection system of this application, data acquisition, data processing, model training and diagnostic reasoning are performed by a server to realize battery status detection of the battery BMS system. The server has a fast data processing speed, which can ensure the timeliness of battery status detection.

[0047] In a sixth aspect, this application provides an electronic 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 battery state detection method as described in the first aspect above.

[0048] In a seventh aspect, this application provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the battery state detection method as described in the first aspect above.

[0049] Eighthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the battery state detection method as described in the first aspect above.

[0050] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0051] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0052] Figure 1 This is one of the flowcharts illustrating the battery state detection method provided in the embodiments of this application;

[0053] Figure 2 This is a second schematic flowchart of the battery state detection method provided in the embodiments of this application;

[0054] Figure 3This is one of the structural schematic diagrams of the battery state detection system provided in the embodiments of this application;

[0055] Figure 4 This is a second schematic diagram of the battery state detection system provided in the embodiments of this application;

[0056] Figure 5 This is the third schematic diagram of the battery state detection system provided in the embodiments of this application;

[0057] Figure 6 This is a schematic diagram of the battery state detection device provided in the embodiments of this application;

[0058] Figure 7 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0059] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0060] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0061] In related technologies, there are multiple ways to perform battery anomaly detection:

[0062] Firstly, a detection method based on inconsistent parameters.

[0063] In this implementation, based on the assumption of consistency between individual battery cells in the battery pack, the batteries in the same battery pack are generally of the same type and under the same operating conditions. The batteries in the same battery pack should have a high degree of consistency. If there is inconsistency among individual cells, it is considered to be an anomaly.

[0064] This method generally involves low computational cost and has good robustness, but it is highly dependent on the selection of calibration cells, and the detection results vary when multiple abnormal cells are present at the same time.

[0065] Secondly, the detection method based on the battery model.

[0066] In this implementation, when the measured parameters differ from the constructed model parameters, the individual cell is considered to be abnormal, based on battery model or parameter identification.

[0067] This approach is highly dependent on the accuracy of the battery model construction, and updating the model parameters to ensure model accuracy under complex operating conditions and battery aging is a problem that needs to be solved.

[0068] Third, detection methods based on machine learning or deep learning.

[0069] In this implementation, a model is trained and predicted based on a large amount of historical battery data to detect battery anomalies.

[0070] This approach requires high-quality and relatively large amounts of data to ensure the accuracy and robustness of the trained model. Furthermore, when operating conditions or battery types change, the model needs to be retrained offline, which affects its generalization ability.

[0071] In addition, there are other detection methods such as self-discharge-based detection and remaining charge capacity-based detection. The former is generally only applicable to offline detection, and the measurement time is long, and the applicability of the working conditions is also obviously limited. The latter is only applicable to charging conditions and is affected by the equalization function.

[0072] All the battery anomaly detection methods mentioned above, whether based on parameter inconsistency, battery model, or machine learning or deep learning, share a common limitation: once the detection model is determined, it can usually only be used for one type, some types, or some operating conditions of battery anomaly detection. If it is necessary to adapt to different types or operating conditions of battery anomaly detection, it is necessary to reset the threshold, rebuild the battery model, and retrain the machine learning or deep learning model to adapt to the new battery or new operating condition anomaly detection task. The versatility of the solution is generally limited, and the re-adaptation workflow is time-consuming.

[0073] The battery state detection method, battery state detection device, battery state detection system, electronic device, and readable storage medium provided in this application will be described in detail below with reference to the accompanying drawings and through specific embodiments and application scenarios.

[0074] Among them, the battery status detection method can be applied to the terminal, and can be executed by the hardware or software in the terminal.

[0075] The terminal includes, but is not limited to, portable communication devices such as mobile phones or tablets with touch-sensitive surfaces (e.g., touchscreen displays and / or touchpads). It should also be understood that, in some embodiments, the terminal may not be a portable communication device, but rather a desktop computer with touch-sensitive surfaces (e.g., touchscreen displays and / or touchpads).

[0076] The following embodiments describe a terminal including a display and a touch-sensitive surface. However, it should be understood that the terminal may include one or more other physical user interface devices such as a physical keyboard, mouse, and joystick.

[0077] The battery state detection method provided in this application embodiment can be executed by an electronic device or a functional module or entity in an electronic device that can implement the battery state detection method. The electronic devices mentioned in this application embodiment include, but are not limited to, mobile phones, tablets, computers, cameras, and wearable devices. The battery state detection method provided in this application embodiment will be described below using an electronic device as the execution subject as an example.

[0078] like Figure 1 As shown, the battery state detection method includes steps 110 and 120.

[0079] Step 110: Obtain the first operating status data of the battery to be tested.

[0080] Among them, the first operating state data of the battery under test is the data that characterizes the actual operating state of the battery under test.

[0081] In actual operation, the first operating state data includes, but is not limited to, the voltage, current, temperature, internal resistance, remaining battery charge percentage (SOC), and battery health (SOH) of the battery under test.

[0082] Step 120: Input the first operating status data into the anomaly diagnosis model to obtain the first battery diagnosis result of the battery to be tested output by the anomaly diagnosis model.

[0083] The first operating state data and the first battery label corresponding to the first operating state data are used to construct incremental training samples, which are used to update and train the anomaly diagnosis model online.

[0084] The first operating status data is input into the anomaly diagnosis model, which detects the actual operating status of the battery under test and obtains the first battery diagnosis result.

[0085] The first battery diagnostic result of the battery under test can be normal, abnormal, internal short circuit, undervoltage, overvoltage, or temperature rise, etc. The type of diagnostic result can be adjusted according to the actual application requirements.

[0086] Before processing the first operating state data and detecting the actual operating state of the battery to be tested, the anomaly diagnosis model has been trained using the previous battery operating state data and battery labels of the battery to be tested. The battery operating state data and battery labels are used to construct incremental training samples and to update and train the anomaly diagnosis model online. The battery labels are used as labels when the anomaly diagnosis model performs battery diagnosis, that is, the battery labels are related to the battery diagnosis results.

[0087] The battery labels used in training the anomaly diagnosis model are used to describe whether the battery state is abnormal or not. Correspondingly, the battery diagnosis results output by the anomaly diagnosis model also characterize whether the battery state is abnormal or not.

[0088] In this embodiment, an anomaly diagnosis model is trained based on the first operating state data and the first battery label. The anomaly diagnosis model includes, but is not limited to, machine learning models, deep learning models, etc. The training method is online incremental training, that is, the first operating state data and the first battery label are used as incremental training samples to update and train the anomaly diagnosis model online.

[0089] In practice, the anomaly diagnosis model can be an optimized model of stochastic gradient descent (SGD) or mini-batch stochastic gradient descent (MiniBatch SGD) to achieve incremental training, such as lightGBM, XGboost, etc.

[0090] Understandably, the anomaly diagnosis model is updated and trained online through incremental learning. In the actual battery detection perception data, the amount of data gradually increases. When faced with new battery data, the trained anomaly diagnosis model is modified through incremental learning to learn the knowledge contained in the new battery data. The time cost of modifying the trained model is much lower than the time cost of retraining the model.

[0091] Incremental learning can be described as follows: whenever new data is added, it is not necessary to rebuild the entire knowledge base. Instead, based on the existing knowledge base, only the changes caused by the new data are updated. Incremental learning is more in line with human thinking principles, mainly in two aspects: firstly, it reduces the storage space occupied because it does not need to save historical data; secondly, incremental learning makes full use of historical training results in the current sample training, thus significantly reducing the time for subsequent training.

[0092] In related technologies, deterministic diagnostic methods or models are used for battery anomaly detection. When the battery's operating conditions, state, or type change, recalibration, modeling, or training is required, which not only takes a long time but also has significant limitations in adaptability.

[0093] In this embodiment, by using an anomaly diagnosis model trained through online incremental learning for battery anomaly detection, the problem of learning new data can be significantly solved. This allows the model to better adapt to the current state of the battery and avoids catastrophic forgetting of previously learned knowledge. When the diagnosis type changes, the model can also learn new content and adjust the previously learned content to adapt to different types and operating conditions of battery anomaly detection. It can also adapt to different detection tasks of the same type of battery, such as detecting abnormal and normal tasks, as well as detecting voltage difference anomalies, current anomalies, temperature anomalies, etc. The detection target can be changed, that is, it can adapt to different diagnostic tasks, effectively improving the accuracy and efficiency of battery detection.

[0094] According to the battery state detection method provided in the embodiments of this application, the anomaly diagnosis model is trained through online incremental learning. The anomaly diagnosis model diagnoses the battery state of the battery under test based on the operating state data of the battery under test. It can be adapted to different diagnostic tasks and effectively improve the accuracy and efficiency of battery detection.

[0095] In some embodiments, after obtaining the first battery diagnostic result of the battery to be tested output by the anomaly diagnostic model in step 120, the battery state detection method may further include:

[0096] The diagnostic results of the first battery are verified and reviewed to obtain the verification results of the first battery.

[0097] The results of the first battery verification are input into the anomaly diagnosis model, and the anomaly diagnosis model is updated and trained online.

[0098] In this embodiment, the first battery diagnostic result of the battery to be tested is verified and reviewed. For battery information that is incorrectly judged by the anomaly diagnostic model, the feedback can be verified by machine or by human. The verification result of the first battery is input into the anomaly diagnostic model for training, which can effectively improve the robustness of the anomaly diagnostic model.

[0099] In actual implementation, the diagnostic results of the first battery are verified and reviewed. The verification results of the first battery are corrections to the diagnostic results of the first battery, which can characterize whether the diagnostic results of the first battery are correct or not. The verification results of the first battery are fed into the anomaly diagnosis model for training, so as to optimize the iterative model and improve the model's adaptability.

[0100] For example, if the first battery diagnosis result is that the battery under test has low voltage, and the first battery diagnosis result is verified, the first battery verification result is also that the battery under test has low voltage. The first battery verification result confirms that the first battery diagnosis result is correct and that the battery under test does indeed have low voltage.

[0101] For example, the first battery diagnosis result is that the battery under test has overvoltage. The first battery diagnosis result is verified and checked. The first battery check result is that the battery under test has a temperature rise. The first battery check result confirms that the first battery diagnosis result is wrong and the battery under test has actually experienced a temperature rise.

[0102] In some embodiments, the first battery verification result is input into the anomaly diagnosis model, and the anomaly diagnosis model is updated and trained online, including:

[0103] Based on the verification results of the first battery, the label of the first battery was corrected;

[0104] The first operating state data and the corrected first battery label are used as incremental training samples to update and train the anomaly diagnosis model online.

[0105] Understandably, the anomaly diagnosis model uses the first running state data and the first battery label as incremental training samples for online update training. After verifying and reviewing the diagnosis results of the first battery, the first battery label is corrected based on the review results of the first battery, thereby correcting the incremental training samples used for training the anomaly diagnosis model.

[0106] It should be noted that if the diagnostic result of the first battery is determined to be incorrect based on the verification result of the first battery, the label of the first battery should be corrected. If the diagnostic result of the first battery is determined to be correct based on the verification result of the first battery, there is no need to correct the label of the first battery.

[0107] In practice, the correction of the first battery label based on the verification results of the first battery can manifest in at least two of the following ways:

[0108] Firstly, corrections were made to existing historical labels.

[0109] In this embodiment, correcting the first battery label based on the first battery verification result may include:

[0110] If the first corrected label corresponding to the first battery verification result is determined to be a historical label, the first battery label is replaced with the first corrected label, and / or, sample data with the same label as the first battery label are removed in the incremental training samples.

[0111] The first corrected label corresponding to the first battery verification result is the battery label after verification. When it is determined that the first corrected label corresponding to the first battery verification result is an existing historical label in the anomaly diagnosis model, the first battery label is corrected based on the first battery verification result, which is a correction to the existing label system.

[0112] In this embodiment, the existing labeling system is modified by directly replacing the first battery label with the first modified label, or by directly removing sample data with the same first battery label from the incremental training samples, thereby improving the optimization effect of the anomaly diagnosis model training.

[0113] Secondly, corrections were made to the newly added tags.

[0114] In this embodiment, correcting the first battery label based on the first battery verification result may include:

[0115] If the second corrected label corresponding to the first battery verification result is determined to be a newly added label, the second corrected label and the first battery label will be used as incremental training samples.

[0116] Alternatively, if the second corrected label is determined to be a newly added label, the first battery label is replaced with the second corrected label, and the labels of the same type as the first battery label in the incremental training samples are replaced with the second corrected label.

[0117] The second corrected label corresponding to the first battery verification result is the newly added label. The newly added label is one that the anomaly diagnosis model did not have before. For the correction of the newly added label, there are two optional optimization iteration methods:

[0118] (1) No replacement method required.

[0119] In this embodiment, if the second corrected label corresponding to the first battery verification result is determined to be a newly added label, the second corrected label and the first battery label are used as incremental training samples.

[0120] For the newly added labeling system, there is no need to correct the historical data that has already been learned. Instead, the battery data is run with multiple labels, that is, the second corrected label and the first battery label are retained as incremental training samples at the same time.

[0121] For example, if the label data of the previous anomaly diagnosis model only had the label of anomaly, now that a new label of undervoltage has been added, and the option of not replacing it is selected, the anomaly diagnosis model can have both anomaly and undervoltage labels at the same time.

[0122] (2) Replacement method.

[0123] In this embodiment, if the second corrected label is determined to be a newly added label, the first battery label is replaced with the second corrected label, and the labels of the same type as the first battery label in the incremental training samples are replaced with the second corrected label.

[0124] For the newly added labeling system, the data already learned by the anomaly diagnosis model is corrected, and the original labels are replaced with new labels.

[0125] To determine which label in the incremental training samples is the same as the first battery label, a similarity calculation or other data selection method can be used. That is, search for the labels in the learned incremental training samples. If the similarity between a certain label and the second corrected label is greater than a preset threshold, then the two labels are determined to be of the same type, and the label is replaced with the second corrected label.

[0126] The first battery diagnostic results are labeled and fed back. By verifying the first battery review results, the historical learning data of the anomaly diagnostic model is corrected and input into the anomaly diagnostic model for training, which further improves the adaptability of the anomaly diagnostic model and enables it to better complete the battery detection work of various anomaly categories.

[0127] In some embodiments, the anomaly diagnosis model can be trained through the following steps:

[0128] Historical battery operating status data and historical battery tags are input into the first diagnostic model for model training to obtain an initial anomaly diagnostic model. Historical battery tags are determined based on historical battery operating status data.

[0129] Incremental training samples are input into the anomaly diagnosis model to perform online iterative updates of the anomaly diagnosis model.

[0130] It should be noted that the training process of the anomaly diagnosis model includes a cold start training phase and an online incremental learning training phase.

[0131] During the cold start training phase, historical battery operating status data is continuously acquired and automatically labeled to obtain historical battery tags. The historical battery operating status data and historical battery tags are then used as historical data for training the first diagnostic model.

[0132] The first diagnostic model is the initial model framework, which can be a machine learning model or a deep learning model.

[0133] In this embodiment, the first diagnostic model is trained using historical data such as historical battery operating status data and historical battery tags to complete the cold start training phase of the abnormal diagnostic model and obtain the initial abnormal diagnostic model.

[0134] It should be noted that the model does not begin diagnosis immediately upon deployment. Instead, it begins diagnosing battery anomalies only after a batch of data (historical battery operating status data and historical battery tags) has been acquired as training samples and the first diagnostic model has been trained to obtain an initial anomaly diagnosis model.

[0135] During the online incremental learning training phase, the anomaly diagnosis model is optimized and iterated during the duration of battery online operation. The method is online incremental learning training. For newly acquired battery operating status data, corresponding battery labels are processed and used as a new batch of training samples for anomaly diagnosis model training.

[0136] If the battery verification result of the abnormal diagnosis model is the same as the battery diagnosis result, the model can be trained using only the battery's operating status data and the corresponding battery labels obtained through processing, resulting in an iteratively optimized abnormal diagnosis model.

[0137] If the battery verification result obtained by verifying the battery diagnosis result output by the anomaly diagnosis model is different from the battery diagnosis result, the battery verification result is used to guide the correction of the battery's operating status data and the processing to obtain the corresponding battery tag. The battery's operating status data, the processed battery tag, and the battery verification result are combined to train the model and obtain the iteratively optimized anomaly diagnosis model.

[0138] In some embodiments, the first battery tag is obtained through the following steps:

[0139] Feature extraction is performed on the first operating state data to obtain the features of the first operating state;

[0140] The first battery tag is determined based on the characteristics of the first operating state.

[0141] In this embodiment, feature extraction is performed on the first operating state data of the battery to be tested to obtain the first operating state features, and the first battery tag is automatically identified based on the first operating state features.

[0142] By extracting features from the first operating state data of the battery under test, we can obtain first operating state features that characterize the actual operating state and estimate the information of the battery under test, including but not limited to the battery's voltage, current, temperature, internal resistance, SOC, SOH, etc.

[0143] The identified battery tags are used to describe whether the battery is abnormal and are used as tags in the anomaly diagnosis model. When the amount of data is small, the anomaly diagnosis model is highly dependent on battery tag identification. As the amount of data increases, the automatic identification weight of battery tags will decrease.

[0144] In this embodiment, the first battery tag is determined based on the characteristics of the first operating state. The battery tag can be automatically identified by inspection methods such as temperature consistency inspection, voltage consistency inspection, and internal resistance consistency inspection, or by learning methods such as clustering or other unsupervised, semi-supervised or self-supervised learning methods.

[0145] The following is an example of an overall implementation architecture for an anomaly diagnosis model.

[0146] like Figure 2 As shown, battery testing using the anomaly diagnosis model is divided into three parts: data processing, diagnosis model, and result verification.

[0147] During the cold start training phase, historical data such as historical battery operating status data and historical battery tags are used to train the first diagnostic model, resulting in an initial anomaly diagnostic model. Subsequently, for new battery data, the anomaly diagnostic model is used to detect anomalies and output diagnostic results. Simultaneously, newly input data is also used as subsequent historical battery data. After automatically identifying battery tags using a battery tag recognition method, the data is added to the model training samples of the anomaly diagnostic model for iterative model updates. The anomaly diagnostic model can be a machine learning model or a deep learning model, and it is trained through online incremental training.

[0148] For the diagnostic results of the anomaly diagnosis model, the diagnostic results are verified and reviewed to obtain verification data and labels. The verification results are also added to the training samples, and the previously learned content is verified, so that the anomaly diagnosis model can learn more reasonable information that is closer to the actual operating conditions.

[0149] Compared with traditional machine learning or deep learning models used for diagnosis, the embodiments of this application have lower requirements for computing and storage resources, and the abnormal diagnosis model has high training efficiency. The continuous training of online incremental learning can better adapt to the actual operating state of the battery, effectively solving the problem that offline models have poor diagnostic results due to parameter differences in situations such as battery aging and changes in operating conditions.

[0150] Automatically identifying labels based on battery operating status data can provide continuous model training samples, solving the problems of sample data and labels for model training. By reviewing diagnostic results, the robustness of the model is improved, and the anomaly diagnosis model can be better adapted to learning different anomaly types, thereby further improving the model's generalization ability.

[0151] When faced with situations such as changing battery type or operating conditions, there is no need to adjust the model framework or perform battery calibration and modeling, reducing repetitive work and greatly shortening the development cycle for application deployment.

[0152] This application also provides a battery state detection system.

[0153] like Figure 3 As shown, the battery status detection system includes a data processing module 310 and a diagnostic model module 320.

[0154] The data processing module 310 is used to determine the first battery label corresponding to the first operating state data based on the first operating state data of the battery to be detected. The first operating state data and the first battery label are used to construct incremental training samples.

[0155] The diagnostic model module 320 is connected to the data processing module 310. The diagnostic model module 320 includes an anomaly diagnostic model. The diagnostic model module 320 is used to input the first operating state data into the anomaly diagnostic model to obtain the first battery diagnostic result of the battery to be tested output by the anomaly diagnostic model. The incremental training samples are used to update and train the anomaly diagnostic model online.

[0156] In this embodiment, the data processing module 310 processes the battery data, which can be divided into two parts: data feature extraction and battery tag identification.

[0157] By extracting features from the first operating state data of the battery under test, we can obtain first operating state features that characterize the actual operating state and estimate the information of the battery under test, including but not limited to the battery's voltage, current, temperature, internal resistance, SOC, SOH, etc.

[0158] Battery tags are used to describe whether a battery is abnormal or not. They are used as tags in anomaly diagnosis models and can be automatically identified through inspection methods such as temperature consistency test, voltage consistency test, and internal resistance consistency test, or through learning methods such as clustering or other unsupervised, semi-supervised, or self-supervised learning methods.

[0159] The following describes two specific examples of battery tag identification.

[0160] (1) Voltage consistency test label identification.

[0161] Voltage consistency testing is based on the consistency differences in voltage characteristics of individual cells within the same battery pack under continuous operating conditions. This is used to identify battery tags. The specific implementation process is as follows:

[0162] An average cell is constructed, whose voltage is the average value after removing the maximum and minimum values. The difference between each individual cell and the average cell is then calculated, representing the voltage consistency difference.

[0163]

[0164] ΔU i =U i -U avg

[0165] In the above formula, n is the number of individual battery cells, U i U is the voltage of the i-th battery cell. avg U is the average battery voltage. max U min These represent the maximum and minimum voltage values, respectively.

[0166] To reduce calculation errors caused by factors such as data synchronization and sampling noise, an exponentially weighted moving average is applied to the voltage consistency differences calculated in step a:

[0167]

[0168]

[0169] In the above formula, Let λ be the filtered voltage consistency difference value, and λ be the filtering coefficient. Assuming the sampled data length is m, the final voltage consistency difference value for each battery cell is:

[0170]

[0171] In the above formula, Let be the final voltage consistency difference value of the i-th battery cell. The final voltage consistency difference score can be obtained by Z-score normalization of this score.

[0172] The battery data is labeled based on the final voltage consistency difference score of each individual battery cell. When the final voltage consistency difference score of an individual battery cell exceeds a set threshold, the battery data is labeled as abnormal, and vice versa.

[0173] (2) Label recognition using the Single-pass clustering method.

[0174] Single-pass clustering does not require iterating over the state of each sample like k-means, making it computationally very fast. It also does not require specifying the number of clusters, meaning all samples may potentially cluster into only one class, avoiding the results of forced clustering. Based on the single-pass concept, this method improves battery anomaly detection and label identification. The specific implementation process is as follows:

[0175] The vector characteristics of a single battery cell can be calculated directly using voltage values, or by using multiple characteristic values ​​such as temperature, current, and internal resistance.

[0176] Similarity calculation has been changed to distance calculation because the actual similarity value has very little distinguishing power, and distance measurement is more reasonable. If the distance exceeds the threshold, they are considered to be different classes, and otherwise they are considered to be the same class.

[0177] When calculating distances, the new sample unit does not calculate the distance with every unit in all existing classes. Instead, after each iteration, the centroid of each class is recalculated, and the distance between the new sample unit and the centroid is calculated.

[0178] After all battery data has been categorized, the battery data is labeled and identified. All the final categories are reviewed, and the variance and correlation coefficient of the characteristic values ​​of the battery cells in each category, such as voltage, internal resistance, and temperature, are calculated. If the variance exceeds the threshold and the correlation coefficient is lower than the threshold, it is considered that the fluctuation within the category is large, and the battery data in that category is labeled as abnormal; otherwise, it is labeled as normal.

[0179] The above methods can all be used to automatically label battery operating status data, that is, to give the battery a label category based on the battery's operating status characteristics. The category can be diverse, including but not limited to different categories such as normal, abnormal, internal short circuit, undervoltage, overvoltage, and temperature rise. The label category can be adjusted according to the actual application needs.

[0180] According to the battery state detection system provided in the embodiments of this application, the anomaly diagnosis model is trained through online incremental learning. The anomaly diagnosis model diagnoses the battery state of the battery under test based on the operating state data of the battery under test. It can be adapted to different diagnostic tasks and effectively improve the accuracy and efficiency of battery detection.

[0181] In some embodiments, the battery state detection system may further include a result verification module 330.

[0182] The result verification module 330 is connected to the diagnostic model module 320. The result verification module 330 is used to verify and check the diagnostic results of the first battery to obtain the first battery verification result. The diagnostic model module 320 is used to update and train the abnormal diagnostic model online based on the first battery verification result.

[0183] The result verification module 330 is used to verify and review the battery diagnostic results. For battery information that is incorrectly judged by the abnormal diagnostic model, feedback can be verified manually or by machine and then re-input into the abnormal diagnostic model for training, thereby further improving the robustness of the model, optimizing the iterative model and improving the model's adaptability.

[0184] After obtaining the initial anomaly diagnosis model at the end of the cold start training phase, battery detection is performed using the anomaly diagnosis model. The result verification module 330 verifies and reviews the battery diagnosis results. If the result verification module 330 does not have a message queue for correcting battery diagnosis results, it can use only the data from the data processing module 310 for model training.

[0185] If the result verification module 330 has a message queue for battery diagnostic result correction, the diagnostic model module 320 receives the battery verification result from the result verification module 330 and uses the battery verification result to guide the correction of the data result obtained from the data processing module 310. The data from the two modules are then combined and input into the anomaly diagnostic model for incremental training to obtain the optimized and iterative anomaly diagnostic model.

[0186] In this embodiment, the battery status detection system includes a data processing module 310, a diagnostic model module 320, and a result verification module 330. The data processing module 310 is used to process battery data, extracting features required by the anomaly diagnostic model on the one hand, and automatically identifying battery tags on the other hand. Both are used by the anomaly diagnostic model.

[0187] The diagnostic model module 320 is used to perform model training and detection of the abnormal diagnostic model. It adopts an online incremental learning method to continuously learn the latest state information of the battery. The result verification module 330 is used to verify and check the diagnostic results, and improve the robustness of the abnormal diagnostic model through verification feedback.

[0188] The battery state detection method provided in this application can be executed by a battery state detection device. This application uses an example of a battery state detection device executing the battery state detection method to illustrate the battery state detection device provided in this application.

[0189] This application also provides a battery state detection system.

[0190] like Figure 4 As shown, the battery status detection system includes: a battery management system (BMS) 410, an embedded development board 420, and a cloud platform 430.

[0191] The embedded development board 420 is located in the battery BMS system 410. The embedded development board 420 is used to collect battery operating status data of the battery BMS system 410.

[0192] The cloud 430 is connected to the embedded development board 420. The embedded development board 420 is used to transmit battery operating status data to the cloud 430. The cloud 430 is used to input the battery operating status data into the anomaly diagnosis model, obtain the battery diagnosis result of the battery BMS system 410 output by the anomaly diagnosis model, and send the battery diagnosis result of the battery BMS system 410 back to the embedded development board 420.

[0193] Among them, battery operating status data and the corresponding battery labels are used to construct incremental training samples, which are used to update and train the anomaly diagnosis model online.

[0194] In this embodiment, data information can be transmitted between the embedded development board 420 and the battery BMS system 410, and data information can also be transmitted between the embedded development board 420 and the cloud 430.

[0195] The embedded development board 420 is used to collect battery operating status data from the battery BMS system 410. The battery operating status data includes, but is not limited to, battery current, voltage, temperature, abnormal information, operating status and other status data.

[0196] In actual implementation, the embedded development board 420 can perform preprocessing operations such as data cleaning, data integration, and data transformation on the battery operating status data of the battery BMS system 410, and then transmit the processed battery operating status data to the cloud 430.

[0197] In some embodiments, the architecture of the battery BMS system 410 can support the acquisition and preprocessing of battery operating status data, and the cloud 430 can also communicate directly with the battery BMS system 410, directly obtaining the battery operating status data transmitted by the battery BMS system 410.

[0198] The cloud-based 430 can use the battery operating status data of the battery BMS system 410 to perform online training and updating of the anomaly diagnosis model and diagnostic inference of the battery diagnosis results of the battery BMS system 410, based on the battery status detection method described above.

[0199] Understandably, the cloud 430 can store data such as battery operating status data and battery diagnostic results from the battery BMS system 410.

[0200] In this embodiment, the cloud 430 can send the battery diagnostic results of the battery BMS system 410 back to the embedded development board 420, which then feeds back the results to the battery BMS system 410. The battery BMS system 410 executes corresponding operating strategies based on the feedback results, such as normal operation, prompting, and alarming.

[0201] In some embodiments, the embedded development board 420 can directly process the battery diagnostic results of the battery BMS system 410 into an executable running strategy of the battery BMS system 410 and send it directly to the battery BMS system 410 for execution.

[0202] In this embodiment, the cloud 430 uses an anomaly diagnosis model trained through online incremental learning to detect battery anomalies. This significantly solves the problem of learning new data, enabling the model to better adapt to the current state of the battery and avoid catastrophic forgetting of previously learned knowledge. When the diagnosis type changes, the model can also learn new content and adjust the previously learned content to adapt to different types or different operating conditions of battery anomaly detection. It can also adapt to different detection tasks of the same type of battery, such as detecting abnormal or normal tasks, as well as detecting abnormal voltage, abnormal current, abnormal temperature, etc. The detection target can be changed, that is, it can adapt to different diagnostic tasks, effectively improving the accuracy and efficiency of battery detection.

[0203] According to the battery status detection system provided in this application embodiment, data acquisition and transmission are performed through an embedded development board 420, and model training and diagnostic inference are completed in the cloud 430. The battery status detection of the battery BMS system 410 can be realized simply by deploying the embedded development board 420 on the battery BMS system 410. The main calculation work for diagnosis is performed in the cloud 430, which is more suitable for small-scale battery systems.

[0204] This application also provides a battery state detection system.

[0205] like Figure 5 As shown, the battery status detection system includes a battery management system (BMS) 410 and a server 510.

[0206] The server 510 is connected to the battery BMS system 410. The server 510 is used to obtain the battery operating status data of the battery BMS system 410, input the battery operating status data into the anomaly diagnosis model, obtain the battery diagnosis result of the battery BMS system 410 output by the anomaly diagnosis model, and send the battery diagnosis result of the battery BMS system 410 back to the battery BMS system 410.

[0207] Among them, battery operating status data and the corresponding battery labels are used to construct incremental training samples, which are used to update and train the anomaly diagnosis model online.

[0208] In this embodiment, the server 510 uses an anomaly diagnosis model trained through online incremental learning to detect battery anomalies. This significantly solves the problem of learning new data, enabling the model to better adapt to the current state of the battery and avoid catastrophic forgetting of previously learned knowledge. When the diagnosis type changes, the model can also learn new content and adjust the previously learned content to adapt to different types or different operating conditions of battery anomaly detection. It can also adapt to different detection tasks of the same type of battery, such as detecting abnormal or normal tasks, as well as detecting abnormal voltage, abnormal current, abnormal temperature, etc. The detection target can be changed, that is, it can adapt to different diagnostic tasks, effectively improving the accuracy and efficiency of battery detection.

[0209] In this embodiment, data information can be transmitted between the server 510 and the battery BMS system 410. Data acquisition, data processing, model training, and diagnostic inference are all performed on the server 510. The server 510 has a fast data processing speed and the timeliness of the diagnostic algorithm will not be affected by network communication quality or information lag in the cloud 430.

[0210] According to the battery status detection system provided in this application embodiment, data acquisition, data processing, model training, and diagnostic inference are performed by server 510 to realize battery status detection of battery BMS system 410. The data processing speed of server 510 is fast, which can ensure the timeliness of battery status detection. This application embodiment also provides a battery status detection device.

[0211] like Figure 6 As shown, the battery state detection device includes:

[0212] The acquisition module 610 is used to acquire the first operating status data of the battery to be tested.

[0213] Processing module 620 is used to input the first operating status data into the anomaly diagnosis model and obtain the first battery diagnosis result of the battery to be tested output by the anomaly diagnosis model.

[0214] The first operating state data and the first battery label corresponding to the first operating state data are used to construct incremental training samples, which are used to update and train the anomaly diagnosis model online.

[0215] According to the battery state detection device provided in the embodiments of this application, the anomaly diagnosis model is trained through online incremental learning. The anomaly diagnosis model diagnoses the battery state of the battery under test based on the operating state data of the battery under test. It can be adapted to different diagnostic tasks and effectively improve the accuracy and efficiency of battery detection.

[0216] In some embodiments, the processing module 620 is further configured to verify and review the first battery diagnostic result to obtain the first battery review result;

[0217] The results of the first battery verification are input into the anomaly diagnosis model, and the anomaly diagnosis model is updated and trained online.

[0218] In some embodiments, the processing module 620 is further configured to correct the first battery label based on the first battery verification result;

[0219] The first operating state data and the corrected first battery label are used as incremental training samples to update and train the anomaly diagnosis model online.

[0220] In some embodiments, the processing module 620 is further configured to replace the first battery label with the first correction label when it is determined that the first correction label corresponding to the first battery review result is a historical label, and / or remove sample data with the same label as the first battery label in the incremental training samples.

[0221] In some embodiments, the processing module 620 is further configured to use the second correction label and the first battery label as incremental training samples when it is determined that the second correction label corresponding to the first battery review result is a newly added label.

[0222] Alternatively, if the second corrected label is determined to be a newly added label, the first battery label is replaced with the second corrected label, and the labels of the same type as the first battery label in the incremental training samples are replaced with the second corrected label.

[0223] In some embodiments, the processing module 620 is further configured to train an anomaly diagnosis model through the following steps:

[0224] Historical battery operating status data and historical battery tags are input into the first diagnostic model for model training to obtain an initial anomaly diagnostic model. Historical battery tags are determined based on historical battery operating status data.

[0225] Incremental training samples are input into the anomaly diagnosis model to perform online iterative updates of the anomaly diagnosis model.

[0226] In some embodiments, the processing module 620 is further configured to obtain the first battery tag through the following steps:

[0227] Feature extraction is performed on the first operating state data to obtain the features of the first operating state;

[0228] The first battery tag is determined based on the characteristics of the first operating state.

[0229] The battery status detection device in this application embodiment can be an electronic device or a component within an electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal or other devices besides a terminal. For example, the electronic device can be a mobile phone, tablet computer, laptop computer, PDA, in-vehicle electronic device, mobile internet device (MID), augmented reality (AR) / virtual reality (VR) device, robot, wearable device, ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc. It can also be a server, network attached storage (NAS), personal computer (PC), television (TV), ATM, or self-service machine, etc. This application embodiment does not specifically limit the device.

[0230] The battery status detection device in this application embodiment can be a device with an operating system. This operating system can be Android, iOS, or other possible operating systems; this application embodiment does not specifically limit it.

[0231] The battery state detection device provided in this application embodiment can achieve... Figure 1 and Figure 2 The various processes implemented in the method implementation examples will not be described again here to avoid repetition.

[0232] In some embodiments, such as Figure 7 As shown, this application embodiment also provides an electronic device 700, including a processor 701, a memory 702, and a computer program stored in the memory 702 and executable on the processor 701. When the program is executed by the processor 701, it implements the various processes of the above-described battery state detection method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0233] It should be noted that the electronic devices in the embodiments of this application include the mobile electronic devices and non-mobile electronic devices described above.

[0234] This application also provides a non-transitory computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the above-described battery state detection method embodiments and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0235] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0236] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described battery state detection method.

[0237] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0238] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described battery state detection method embodiments and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0239] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0240] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0241] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0242] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

[0243] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," 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 this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0244] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

Claims

1. A battery state detection method characterized by, The method comprises: obtaining first running state data of a battery to be detected; inputting the first running state data into an abnormality diagnosis model to obtain a first battery diagnosis result of the battery to be detected output by the abnormality diagnosis model; wherein the first running state data and a first battery label corresponding to the first running state data are used to construct an incremental training sample, and the incremental training sample is used to perform online update training on the abnormality diagnosis model; after the first battery diagnosis result output by the abnormality diagnosis model is obtained, the method further comprises: checking and reviewing the first battery diagnosis result to obtain a first battery review result; inputting the first battery review result into the abnormality diagnosis model to perform online update training on the abnormality diagnosis model; the step of inputting the first battery review result into the abnormality diagnosis model to perform online update training on the abnormality diagnosis model comprises: correcting the first battery label based on the first battery review result; using the first running state data and the corrected first battery label as the incremental training sample to perform online update training on the abnormality diagnosis model; wherein correcting the first battery label comprises correcting a historical label and correcting a new label.

2. The battery state detection method according to claim 1, characterized by, the step of correcting the first battery label based on the first battery review result comprises: in a case where a first correction label corresponding to the first battery review result is determined to be a historical label, replacing the first battery label with the first correction label, and / or removing sample data with the same label as the first battery label from the incremental training sample.

3. The battery state detection method according to claim 1, characterized by, the step of correcting the first battery label based on the first battery review result comprises: in a case where a second correction label corresponding to the first battery review result is determined to be a new label, using the second correction label and the first battery label as the incremental training sample; or, in a case where the second correction label is determined to be a new label, replacing the first battery label with the second correction label, and replacing labels of the same type as the first battery label in the incremental training sample with the second correction label.

4. The battery state detection method according to any one of claims 1 to 3, characterized by, The abnormality diagnosis model is trained by the following steps: inputting historical battery running state data and historical battery labels into a first diagnosis model to perform model training, thereby obtaining an initial abnormality diagnosis model, wherein the historical battery labels are determined based on the historical battery running state data; inputting the incremental training sample into the abnormality diagnosis model to perform online iterative update of the abnormality diagnosis model.

5. The battery state detection method according to any one of claims 1 to 3, characterized by, The first battery label is obtained by the following steps: performing feature extraction on the first running state data to obtain first running state features; determining the first battery label based on the first running state features.

6. A battery state detecting device characterized by comprising: The method comprises: an acquisition module configured to acquire first running state data of a battery to be detected; a processing module configured to input the first running state data into an abnormality diagnosis model to obtain a first battery diagnosis result of the battery to be detected output by the abnormality diagnosis model; The first running state data and the first battery label corresponding to the first running state data are used to construct an incremental training sample, and the incremental training sample is used for online update training of the anomaly diagnosis model. The processing module is further configured to verify and review the first battery diagnosis result to obtain a first battery review result. The first battery review result is input into the anomaly diagnosis model to perform online update training on the anomaly diagnosis model. The processing module is further configured to correct the first battery label based on the first battery review result. The first running state data and the corrected first battery label are used as the incremental training sample to perform online update training on the anomaly diagnosis model. The correction of the first battery label includes correction of historical labels and correction of new labels.

7. A battery state detection system characterized by comprising: The data processing module is configured to determine a first battery label corresponding to first running state data of a battery to be detected based on the first running state data, and the first running state data and the first battery label are used to construct an incremental training sample. The diagnosis model module is connected with the data processing module and includes an anomaly diagnosis model, and is configured to input the first running state data into the anomaly diagnosis model, obtain a first battery diagnosis result of the battery to be detected output by the anomaly diagnosis model, and use the incremental training sample to perform online update training on the anomaly diagnosis model. The result review module is connected with the diagnosis model module and is configured to verify and review the first battery diagnosis result to obtain a first battery review result, and the diagnosis model module is configured to perform online update training on the anomaly diagnosis model based on the first battery review result. The result review module is further configured to correct the first battery label based on the first battery review result. The first running state data and the corrected first battery label are used as the incremental training sample to perform online update training on the anomaly diagnosis model. The correction of the first battery label includes correction of historical labels and correction of new labels. The battery BMS system; 8. A battery state detection system characterized by comprising: The embedded development board is configured to collect battery running state data of the battery BMS system; The cloud is in communication connection with the embedded development board, and the embedded development board is configured to transmit the battery running state data to the cloud. The cloud is configured to input the battery running state data into an anomaly diagnosis model, obtain a battery diagnosis result of the battery BMS system output by the anomaly diagnosis model, and transmit the battery diagnosis result of the battery BMS system back to the embedded development board. ​ ​ The battery operation state data and the battery label corresponding to the battery operation state data are used to construct an incremental training sample, and the incremental training sample is used for online update training of the anomaly diagnosis model. After obtaining the battery diagnosis result of the battery to be detected output by the anomaly diagnosis model, the cloud end is further configured to: verify and review the battery diagnosis result to obtain a battery review result; input the battery review result into the anomaly diagnosis model to perform online update training on the anomaly diagnosis model; the inputting of the battery review result into the anomaly diagnosis model to perform online update training on the anomaly diagnosis model comprises: correcting the battery label based on the battery review result; using the battery operation state data and the corrected battery label as the incremental training sample to perform online update training on the anomaly diagnosis model; wherein correcting the battery label comprises correcting historical labels and correcting new labels.

9. A battery state detection system characterized by comprising: comprises: a battery BMS system; a server, which is in communication connection with the battery BMS system, is configured to obtain battery operation state data of the battery BMS system, input the battery operation state data into an anomaly diagnosis model, obtain a battery diagnosis result of the battery BMS system output by the anomaly diagnosis model, and return the battery diagnosis result of the battery BMS system to the battery BMS system; The battery operation state data and the battery label corresponding to the battery operation state data are used to construct an incremental training sample, and the incremental training sample is used for online update training of the anomaly diagnosis model. After obtaining the battery diagnosis result of the battery to be detected output by the anomaly diagnosis model, the cloud end is further configured to: verify and review the battery diagnosis result to obtain a battery review result; input the battery review result into the anomaly diagnosis model to perform online update training on the anomaly diagnosis model; the inputting of the battery review result into the anomaly diagnosis model to perform online update training on the anomaly diagnosis model comprises: correcting the battery label based on the battery review result; using the battery operation state data and the corrected battery label as the incremental training sample to perform online update training on the anomaly diagnosis model; wherein correcting the battery label comprises correcting historical labels and correcting new labels.

10. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to realize the battery state detection method of any one of claims 1-5. 11.A non-transitory computer-readable storage medium having stored thereon a computer program. The computer program is executed by the processor to realize the battery state detection method of any one of claims 1-5.

Images