Anomaly detection method, apparatus, device, and storage medium

Abstract

The application relates to an anomaly detection method, device and equipment and a storage medium. The method comprises the following steps: obtaining relevant index data of a vehicle battery; using a preset first anomaly identification model to identify the relevant index data to obtain an anomaly identification result; in the case that the anomaly identification result indicates that the vehicle battery has an anomaly, determining anomaly index data to be used according to an anomaly index involved in the anomaly identification result, wherein the anomaly index data at least comprises a data segment intercepted from a historical index data sequence corresponding to the anomaly index; using a preset second anomaly identification model to identify the anomaly index data to obtain a target anomaly score of the vehicle battery; and detecting whether the vehicle battery has an anomaly according to the target anomaly score. The application can avoid the interference of invalid data, thereby improving the accuracy of vehicle battery detection.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to an anomaly detection method, apparatus, device, and storage medium. Background Technology

[0002] In today's rapidly developing technological world, vehicles have become an indispensable tool for people's daily lives and the operation of society. As the core source of vehicle power, the performance and health of the battery directly affect the overall performance of the vehicle. Taking electric vehicles as an example, the battery's range, charging and discharging efficiency, and stability determine the user experience and applicability of the vehicle.

[0003] Currently, most common vehicle battery testing methods focus on visual inspection, such as observing for obvious abnormalities like damage or bulges, making it difficult to detect potential internal problems and resulting in low testing accuracy. Summary of the Invention

[0004] This application provides an anomaly detection method, apparatus, device, and storage medium that can avoid interference from invalid data, thereby improving the accuracy of vehicle battery detection.

[0005] Firstly, this application provides an anomaly detection method, the method comprising: Obtain relevant performance data for vehicle batteries; The relevant indicator data are identified using a preset first anomaly identification model to obtain anomaly identification results; When the anomaly identification result indicates that the vehicle battery is abnormal, the abnormal indicator data to be used is determined according to the abnormal indicators involved in the anomaly identification result. The abnormal indicator data includes at least a data segment extracted from the historical indicator data sequence corresponding to the abnormal indicator. Using a preset second anomaly identification model, the abnormal indicator data is identified to obtain the target anomaly score of the vehicle battery. Based on the target anomaly score, detect whether there is an anomaly in the vehicle battery.

[0006] Secondly, this application provides an anomaly detection device, the device comprising: The acquisition unit is used to acquire relevant indicator data of the vehicle battery. The first identification unit is used to identify the relevant indicator data using a preset first anomaly identification model to obtain anomaly identification results; The determining unit is configured to, when the anomaly identification result indicates that the vehicle battery is abnormal, determine the abnormal indicator data to be used based on the abnormal indicators involved in the anomaly identification result, wherein the abnormal indicator data includes at least a data segment extracted from the historical indicator data sequence corresponding to the abnormal indicator. The second identification unit is used to identify the abnormal indicator data using a preset second abnormality identification model to obtain the target abnormality score of the vehicle battery. The detection unit is used to detect whether there is an abnormality in the vehicle battery based on the target abnormality score.

[0007] Thirdly, this application provides an anomaly detection device, comprising: at least one communication interface; at least one bus connected to the at least one communication interface; at least one processor connected to the at least one bus; and at least one memory connected to the at least one bus, wherein the processor is configured to: Obtain relevant performance data for vehicle batteries; The relevant indicator data are identified using a preset first anomaly identification model to obtain anomaly identification results; When the anomaly identification result indicates that the vehicle battery is abnormal, the abnormal indicator data to be used is determined according to the abnormal indicators involved in the anomaly identification result. The abnormal indicator data includes at least a data segment extracted from the historical indicator data sequence corresponding to the abnormal indicator. Using a preset second anomaly identification model, the abnormal indicator data is identified to obtain the target anomaly score of the vehicle battery. Based on the target anomaly score, detect whether there is an anomaly in the vehicle battery.

[0008] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described anomaly detection method.

[0009] Compared with the prior art, the technical solution provided in this application has the following advantages: acquiring relevant indicator data of the vehicle battery; using a preset first anomaly identification model to identify the relevant indicator data and obtain anomaly identification results; when the anomaly identification results indicate that the vehicle battery is abnormal, determining the anomaly indicator data to be used based on the anomaly indicators involved in the anomaly identification results, wherein the anomaly indicator data at least includes data segments extracted from the historical indicator data sequence corresponding to the anomaly indicator; using a preset second anomaly identification model to identify the anomaly indicator data and obtain the target anomaly score of the vehicle battery; and detecting whether the vehicle battery is abnormal based on the target anomaly score. Through the above steps, this application dynamically filters the anomaly indicator data to be used based on the anomaly identification results of the first anomaly identification model. These anomaly indicator data are key data highly correlated with anomalies, thus enabling the second anomaly identification model to focus on key data highly correlated with anomalies, avoiding interference from invalid data, thereby improving the accuracy of vehicle battery detection. Attached Figure Description

[0010] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0013] Figure 1 A flowchart illustrating an anomaly detection method provided in an embodiment of this application; Figure 2 This is a schematic flowchart of an anomaly detection device provided in an embodiment of this application; Figure 3 This is a schematic diagram of an anomaly detection device provided in an embodiment of this application. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] The following disclosure provides numerous different embodiments or examples for implementing various structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the scope of the invention. Furthermore, reference numerals and / or letters may be repeated in different examples. Such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0016] In today's rapidly developing technological world, vehicles have become an indispensable tool for people's daily lives and the operation of society. As the core source of vehicle power, the performance and health of the battery directly affect the overall performance of the vehicle. Taking electric vehicles as an example, the battery's range, charging and discharging efficiency, and stability determine the user experience and applicability of the vehicle. Currently, common vehicle battery testing methods mostly focus on visual inspection, such as observing for obvious abnormalities like damage or bulges, making it difficult to detect potential internal problems and resulting in low testing accuracy.

[0017] To address the aforementioned problems, embodiments of this application provide an anomaly detection method. This method avoids interference from invalid data, thereby improving the accuracy of vehicle battery detection. Figure 1 As shown, the specific steps include: Step 101: Obtain relevant indicator data for the vehicle battery.

[0018] The relevant indicator data are the basic data used to assess the health status of the vehicle battery, including real-time indicator data and / or historical indicator data corresponding to historical periods. This data includes, but is not limited to: battery voltage (V), battery current (I), battery capacity (SOC), battery state of health (SOH), battery temperature, charge / discharge status, historical fault codes, automatic charging command flags and execution status, and vehicle age. The historical period can be flexibly determined according to actual needs, for example, it can be set to the past 1 day, the past 3 days, or the past 7 days; accordingly, the acquired historical indicator data is the data within the corresponding historical period.

[0019] The implementing entity of this application can be a cloud server, acting as a data processing center. This server can acquire relevant indicator data of the vehicle battery from any vehicle and perform corresponding anomaly detection operations based on this data. Specifically, the cloud server can receive relevant indicator data of the vehicle battery uploaded from different vehicles, store, analyze, and identify it uniformly, and then send the detection results to the vehicle or monitoring platform to support battery status assessment and anomaly warning.

[0020] Step 102: Use the preset first anomaly identification model to identify the relevant indicator data and obtain the anomaly identification results.

[0021] Where the anomaly identification result indicates that the vehicle battery is not abnormal, the anomaly identification result may include, but is not limited to, a flag indicating no abnormality, an empty field, or an anomaly level of a specified level (e.g., the lowest anomaly level). Where the anomaly identification result indicates that the vehicle battery is abnormal, the anomaly identification result shall at least include anomaly indicators, and may also include anomaly level, corresponding anomaly time point, etc., which are not specifically limited in this application.

[0022] The first anomaly detection model is used for preliminary anomaly screening of vehicles, aiming to identify vehicles with potential anomalies. This model can be implemented using a threshold-based detection algorithm or a lightweight model. For example, when using a threshold-based detection algorithm, it can determine whether the value of a certain indicator exceeds a preset threshold; if it does, the indicator is identified as an anomaly and used as the anomaly detection result. Furthermore, the model can analyze the anomaly indicators and their degree of anomaly to determine the anomaly level. For example, if the anomaly indicator is voltage, and the voltage value exceeds the corresponding voltage threshold by a large margin, it is classified as a high anomaly level; if the voltage value exceeds the corresponding voltage threshold by a small margin, it is classified as a low anomaly level. The model can also obtain the anomaly time point corresponding to the occurrence of the anomaly indicator; for example, it can obtain the time point corresponding to the voltage value exceeding the corresponding voltage threshold and use it as the anomaly time point. When using a lightweight model, a simpler machine learning model can be used, such as logistic regression, decision trees, or Lightweight Gradient Boosting Machine (LightGBM), to quickly identify relevant indicator data and obtain anomaly detection results. The model was fully trained based on a large amount of historical data before its use. Its training process is similar to the model training methods in existing technologies, and will not be described in detail here.

[0023] Step 103: If the anomaly identification result indicates that there is an anomaly in the vehicle battery, determine the anomaly indicator data to be used based on the anomaly indicators involved in the anomaly identification result.

[0024] The abnormal indicator data includes at least data segments extracted from the historical indicator data sequence corresponding to the abnormal indicator, and may also include data segments extracted from the historical indicator data sequence corresponding to the related indicators of the abnormal indicator. The position and time range covered by these segments in the historical indicator data sequence are determined based on the anomaly identification results. This historical indicator data sequence contains the historical indicator data and its time point for the corresponding indicator, where the time point is the data collection time.

[0025] In this step, when the anomaly identification result indicates an anomaly in the vehicle battery, an anomaly indicator is obtained from the anomaly identification result. A time window of preset duration is then obtained, and the end time of this time window is aligned with the current time point. From the historical indicator data sequence corresponding to the anomaly indicator, historical indicator data corresponding to each time point within this time window is extracted to obtain the anomaly indicator data to be used. Alternatively, the anomaly level is obtained from the anomaly identification result, and the duration of the time window is set according to the anomaly level. Then, the end time of the set time window is aligned with the current time point, and historical indicator data corresponding to each time point within this time window is extracted from the historical indicator data sequence corresponding to the preset indicator to obtain the anomaly indicator data to be used. Other methods can also be used to determine the anomaly indicator data; this is not limited to these methods.

[0026] The preset indicators can be abnormal indicators, or abnormal indicators and related indicators; there are no restrictions here.

[0027] Step 104: Use the preset second anomaly identification model to identify the abnormal indicator data and obtain the target anomaly score of the vehicle battery.

[0028] The second anomaly detection model employs a more complex machine learning model, such as XGBoost, Random Forest, Long Short-Term Memory (LSTM), or Transformer, to identify anomalous indicator data and obtain anomaly scores. This model has been thoroughly trained on a large amount of historical data before use, and its training process is similar to that of existing models, so it will not be elaborated here. The anomaly score is a quantitative value representing the current degree of anomaly in the vehicle battery, ranging from 0 to 1 or 0 to 100; a higher score indicates a higher degree of anomaly.

[0029] In this step, the abnormal indicator data is input into the preset second abnormality identification model to obtain an abnormality score, and this abnormality score is determined as the target abnormality score of the vehicle battery.

[0030] Step 105: Detect whether there is any abnormality in the vehicle battery based on the target anomaly score.

[0031] In this step, a score threshold can be set. When the target abnormal score exceeds the threshold, it is determined that the vehicle battery is abnormal; otherwise, it is determined that the vehicle battery is not abnormal.

[0032] In this embodiment, relevant indicator data of the vehicle battery are acquired; a preset first anomaly identification model is used to identify the relevant indicator data to obtain anomaly identification results; when the anomaly identification results indicate that the vehicle battery is abnormal, abnormal indicator data to be used is determined based on the anomaly identification results, the abnormal indicator data at least includes data segments extracted from the historical indicator data sequence corresponding to the abnormal indicator; a preset second anomaly identification model is used to identify the abnormal indicator data to obtain a target anomaly score for the vehicle battery; and the presence of anomalies in the vehicle battery is detected based on the target anomaly score. Through the above steps, this application dynamically filters the abnormal indicator data to be used based on the anomaly identification results of the first anomaly identification model. These abnormal indicator data are key data highly correlated with anomalies, allowing the second anomaly identification model to focus on key data highly correlated with anomalies, avoiding interference from invalid data, thereby improving the accuracy of vehicle battery detection.

[0033] In this embodiment, considering that the physical state change of a vehicle battery is a dynamic process, the abnormal indicator value at a single moment is difficult to accurately reflect its true abnormality. Therefore, it is necessary to analyze the dynamic change process corresponding to the indicator. Since the dynamic change processes corresponding to different abnormal indicators have different abnormal evolution cycles, the abnormal evolution cycle of the abnormal indicator can be determined first, and then data for the corresponding time period can be extracted from the historical indicator data sequence according to the cycle as abnormal indicator data characterizing the dynamic change process of the abnormal indicator. The specific steps are: determining the first time window corresponding to the abnormal indicator according to the abnormal evolution cycle of the abnormal indicator; and determining the abnormal indicator data to be used from the historical indicator data sequence corresponding to the abnormal indicator according to the first time window.

[0034] Different abnormal indicators exhibit different abnormal evolution processes, leading to variations in their corresponding abnormal evolution cycles. The abnormal evolution process refers to the entire process from the appearance of an abnormal indicator to its complete evolution, while the abnormal evolution cycle is the time span required for this process. It reflects the evolutionary pattern of the abnormality represented by the indicator from its inception to its manifestation, and is set by technicians based on the actual process. For example, the abnormal evolution process of the resting voltage indicator reflects the long-term self-discharge trend or aging process of the battery, which changes relatively slowly, thus corresponding to a longer abnormal evolution cycle, typically around 7 days. Conversely, the abnormal evolution process of the starting voltage indicator reflects the battery's response capability under instantaneous high-current loads, which changes rapidly, thus corresponding to a shorter abnormal evolution cycle, typically several hours or even less.

[0035] The duration of the anomaly evolution cycle is positively correlated with the duration of the first time window. The first time window refers to the time window used to extract historical data; its duration is used to determine which data to extract from the historical indicator data sequence to characterize the dynamic change process of the anomaly indicator. To ensure that the dynamic change trend of the data corresponding to the anomaly indicator can be fully captured, the duration of the first time window is usually determined based on the anomaly evolution cycle. Generally, the duration of the first time window is directly taken as the duration corresponding to the anomaly evolution cycle. Considering the continuity of data sampling and to avoid information loss due to data truncation, in practical applications, a preset buffer duration can be added to the anomaly evolution cycle to ensure that sufficient data is obtained. That is: First time window duration = anomaly evolution cycle duration + buffer duration. The buffer duration can be a fixed duration set by technicians based on experience, or it can be determined based on the anomaly evolution cycle. For example, the buffer duration can be obtained by multiplying the anomaly evolution cycle by a preset ratio. The first time window includes an end time point and an epidemiological start time point, with the end time point being greater than the start time point.

[0036] In this step, abnormal indicators are obtained from the anomaly identification results. Then, based on the pre-set correspondence between indicators and anomaly evolution cycles, the anomaly evolution cycle corresponding to the abnormal indicator is determined. Furthermore, the duration of the first time window is set according to the duration of this anomaly evolution cycle. Finally, based on the first time window, data is extracted from the historical indicator data sequence corresponding to the abnormal indicator to obtain the abnormal indicator data to be used.

[0037] In this application embodiment, there are multiple methods for extracting data from the historical indicator data sequence corresponding to the abnormal indicator according to the first time window to obtain the abnormal indicator data to be used, including: The first method aligns the end time of the first time window with the current time point, and extracts the historical indicator data corresponding to each time point within the first time window from the historical indicator data sequence corresponding to the abnormal indicator, which is then used as the abnormal indicator data to be used.

[0038] In practice, whenever a data point is collected, it is written into the historical data sequence along with its corresponding time point. Therefore, the current time point is the time point corresponding to the last data point in the sequence. The historical data sequence contains at least one historical data point and its corresponding time point.

[0039] In practice, since the more recent the indicator data, the more it reflects the current situation and the higher its reference value, the end time of the first time window can be aligned with the current time point, and the historical indicator data corresponding to each time point within the time window can be extracted as abnormal indicator data to be used.

[0040] The second method involves determining the target abnormal time point corresponding to the abnormal indicator; if the duration from the target abnormal time point to the current time point is less than the target duration, aligning the end time point of the first time window with the current time point, and extracting the historical indicator data within the first time window at the corresponding time point from the historical indicator data sequence corresponding to the abnormal indicator, as the abnormal indicator data to be used; if the duration from the target abnormal time point to the current time point is greater than or equal to the target duration, determining the target time point based on the target abnormal time point and the target duration, aligning the end time point of the first time window with the target time point, and extracting the historical indicator data within the first time window at the corresponding time point from the historical indicator data sequence corresponding to the abnormal indicator, as the abnormal indicator data to be used.

[0041] Here, the abnormal time point is the point in time when the value of the corresponding abnormal indicator becomes abnormal. There may be one or more abnormal time points, depending on the actual situation. When there is only one abnormal time point, it is determined as the target abnormal time point. When there are multiple abnormal time points, one of them can be selected as the target abnormal time point. The specific selection rule can be to use the abnormal time point with the largest rate of change of the corresponding indicator value as the target abnormal time point, or other methods can be used to determine the target abnormal time point. There is no limitation here.

[0042] The target duration is half the duration of the first time window. In other words, if the duration from the target anomalous time point to the current time point is less than the target duration, it means the duration is insufficient for the first time window to extract data from the historical indicator data sequence using the target anomalous time point as an intermediate point. Therefore, the end time point of the first time window is aligned with the current time point to extract data. Conversely, if the duration from the target anomalous time point to the current time point is greater than or equal to the target duration, it means the duration is sufficient for the first time window to extract data from the historical indicator data sequence using the target anomalous time point as an intermediate point. Therefore, the sum of the target anomalous time point and the target duration is used as the target time point aligned with the end time point of the first time window, and the first time window is used to extract data in this way.

[0043] In implementation, the closer the data's corresponding time point is to the current time point, the more accurately it reflects the current situation; and the more concentrated the data's corresponding time points are near the target anomaly time point, the more it reflects the overall anomaly characteristics. Therefore, when selecting data, priority should be given to collecting relevant data before and after the target anomaly time point, and then the latest data should be filtered from it. The specific steps are as follows: when the target anomaly time point is the current time point, or when the target anomaly time point is close to the current time point (i.e., when the time from the target anomaly time point to the current time point is less than the target time point), it indicates that the data before and after the target anomaly time point includes the latest data. Therefore, the end time point of the first time window can be aligned with the current time point, and then the indicator data of the corresponding time point within the first time window can be determined from the historical indicator data sequence corresponding to the anomaly indicator, thus obtaining the anomaly indicator data to be used. When the target abnormal time point is far from the current time point, that is, when the duration from the target abnormal time point to the current time point is greater than or equal to the target duration, it means that the data before and after the target abnormal time point does not include the latest data. It is necessary to prioritize extracting the data before and after the abnormal time point, and take the sum of the target abnormal time point and the target duration as the target time point. Then, align the end time point of the first time window with the target time point, and then determine the indicator data of the corresponding time point in the first time window from the historical indicator data sequence corresponding to the abnormal indicator to obtain the abnormal indicator data to be used.

[0044] The third method involves determining the target abnormal time point corresponding to the abnormal indicator. If the target abnormal time point is the current time point, the end time point of the first time window is aligned with the current time point. Historical indicator data within the first time window for the corresponding time point is extracted from the historical indicator data sequence corresponding to the abnormal indicator and used as the abnormal indicator data to be used. If the target abnormal time point is not the current time point, the target type corresponding to the abnormal indicator is determined. If the target type indicates that the target abnormal time point is the end time point of the first time window, the end time point of the first time window is aligned with the target abnormal time point. Historical indicator data within the first time window for the corresponding time point is extracted from the historical indicator data sequence corresponding to the abnormal indicator and used as the abnormal indicator data to be used. The system identifies the following steps: First, if the target type indicates that the target anomaly time point is the start time point of the first time window, it checks whether the duration from the target anomaly time point to the current time point is greater than the duration of the first time window, obtaining a first detection result. Based on the first detection result and the first time window, it determines the anomaly indicator data to be used from the historical indicator data sequence corresponding to the anomaly indicator. Second, if the target type indicates that the target anomaly time point is the middle time point of the first time window, it checks whether the duration from the target anomaly time point to the current time point is greater than the target duration, obtaining a second detection result. Based on the second detection result and the first time window, it determines the anomaly indicator data to be used from the historical indicator data sequence corresponding to the anomaly indicator.

[0045] In practical applications, early changes in some indicators (such as voltage) are more indicative of whether there are abnormalities in the vehicle battery, while later changes in other indicators (such as temperature) are more significant. Still others (such as internal resistance) can effectively reflect battery abnormalities regardless of changes before or after the abnormal time point. Therefore, a corresponding type can be configured for each indicator. This type includes three categories, used to identify the target abnormal time point as the end, start, and middle time points of the first time window, respectively. The middle time point refers to the abnormal time point located in the middle of the first time window.

[0046] The steps to determine the target type corresponding to the abnormal indicator are as follows: obtain the pre-set correspondence between indicators and types, determine the type corresponding to the abnormal indicator based on the correspondence, and use it as the target type.

[0047] Based on the first detection result and the first time window, the specific steps for determining the abnormal indicator data to be used in the historical indicator data sequence corresponding to the abnormal indicator are as follows: If the first detection result indicates that the duration from the current time point to the target abnormal time point is greater than the duration of the first time window, then the start time point of the first time window is aligned with the target abnormal time point, and the historical indicator data of the corresponding time point within the first time window is extracted as the abnormal indicator data to be used; If the first detection result indicates that the duration from the current time point to the target abnormal time point is less than or equal to the duration of the first time window, then the end time point of the first time window is aligned with the current time point, and the historical indicator data of the corresponding time point within the first time window is extracted as the abnormal indicator data to be used.

[0048] Based on the second detection result and the first time window, the specific steps for determining the anomaly indicator data to be used in the historical indicator data sequence corresponding to the anomaly indicator are as follows: If the second detection result indicates that the duration from the current time point to the target anomaly time point is less than or equal to the target duration, align the end time point of the first time window with the current time point, and extract the historical indicator data within the first time window corresponding to that time point as the anomaly indicator data to be used. If the second detection result indicates that the duration from the current time point to the target anomaly time point is greater than the target duration, determine the target time point by the sum of the target anomaly time point and the target duration, align the end time point of the first time window with the target time point, and extract the historical indicator data within the first time window corresponding to that time point as the anomaly indicator data to be used.

[0049] Fourth, determine the set of abnormal time points corresponding to the abnormal indicators; if the set of abnormal time points includes multiple abnormal time points, determine a first duration based on the minimum and maximum time points in the set of abnormal time points; if the first duration is greater than the duration of the first time window, adjust the duration of the first time window to the first duration; based on the adjusted first time window, determine the abnormal indicator data to be used in the historical indicator data sequence corresponding to the abnormal indicators. To ensure more comprehensive data acquisition, it is necessary to determine whether the first time window can cover these abnormal time points. If not, the duration of the first time window needs to be adjusted to ensure that it covers these abnormal time points.

[0050] Based on the adjusted first time window, the abnormal indicator data to be used is determined from the historical indicator data sequence corresponding to the abnormal indicator. The specific steps include: aligning the end time point of the adjusted first time window with the maximum time point, and extracting the historical indicator data within the adjusted first time window at the corresponding time point from the historical indicator data sequence corresponding to the abnormal indicator, as the abnormal indicator data to be used; or, aligning the start time point of the adjusted first time window with the minimum time point, and extracting the historical indicator data within the adjusted first time window at the corresponding time point from the historical indicator data sequence corresponding to the abnormal indicator, as the abnormal indicator data to be used.

[0051] In this embodiment, when the target analysis result also includes an anomaly level, the second time window used for data extraction can be dynamically adjusted based on the anomaly level, thereby extracting data from the historical indicator data sequence corresponding to each preset indicator based on the second time window. Specific steps include: determining preset indicators based on the anomaly indicators; determining the second time window to be used based on the anomaly level indicated by the anomaly identification result; and determining the anomaly indicator data to be used in the historical indicator data sequence corresponding to the preset indicator based on the current time point and the second time window.

[0052] The anomaly level is positively correlated with the duration of the second time window. A higher anomaly level results in a longer second time window, covering a wider range of historical data and capturing a more complete anomaly evolution process; a lower anomaly level results in a shorter second time window, focusing on recent data and improving the real-time performance of data processing. In practical applications, this application allows for pre-setting the time window duration corresponding to different anomaly levels. Specifically, this application sets three anomaly levels: high, medium, and low. Low anomalies correspond to a 1-day time window, medium anomalies to a 3-day time window, and high anomalies to a 7-day time window.

[0053] The preset indicators can be abnormal indicators or indicators related to abnormal indicators. These include various indicators closely related to battery safety monitoring, such as individual battery cell voltage, battery pack temperature, charging and discharging current, insulation resistance, SOC (state of charge) change rate, and individual cell voltage difference. Each preset indicator corresponds to a historical indicator data sequence, which records the indicator value and the corresponding time point at each time point in chronological order.

[0054] In this step, the anomaly level is obtained from the target analysis results. Based on the correspondence between the anomaly level and the time window duration, the duration of the time window corresponding to that anomaly level is determined, and the duration of the second time window is set accordingly. The end time point of the second time window is aligned with the current time point, and indicator data is extracted from the historical indicator data sequence corresponding to each preset indicator according to the second time window to obtain the anomaly indicator data to be used.

[0055] In this embodiment, based on the abnormal indicators involved in the anomaly identification results, target indicators associated with them are automatically expanded, and the abnormal indicators and target indicators are jointly determined as preset indicators, thereby constructing more targeted and comprehensive battery anomaly monitoring indicators. Specific steps include: determining target indicators related to the abnormal indicators based on the abnormal indicators; and determining preset indicators based on the abnormal indicators and the target indicators.

[0056] Among them, target indicators refer to other battery monitoring indicators that have a physical, electrical, or coupling relationship with the abnormal indicator. For example, when the abnormal indicator is the voltage of a single cell, the related target indicators may include the internal resistance of that cell, the current distribution of the module in which that cell is located, the voltage of adjacent battery cells, and the SOC (state of charge) change rate of that cell. When the abnormal indicator is the local temperature of the battery pack, the related target indicators may include the coolant flow rate near the temperature sensor, the charging and discharging current of the corresponding area, and the temperature values ​​of other temperature sensors in the same area.

[0057] In this step, based on the abnormal indicator, a target indicator related to the abnormal indicator is queried from a preset indicator association table; the abnormal indicator and the target indicator are used together as preset indicators.

[0058] In this embodiment, when the relevant indicator data includes historical indicator data and real-time indicator data, a first anomaly identification model with a dual-branch structure is used for identification and fusion, which can simultaneously capture the long-term degradation trend and short-term abnormal fluctuations of the battery system, thereby improving the accuracy and robustness of anomaly identification. The specific steps include: using the first submodule to identify the historical indicator data and obtain a first identification result; using the second submodule to identify the real-time indicator data and obtain a second identification result; and using the first determining module to determine the anomaly identification result based on the first identification result and the second identification result.

[0059] Historical data reflects the long-term evolution and degradation trends of the battery system, such as changes in the consistency of individual cell voltages, capacity decay curves, internal resistance growth trends, and fluctuations in indicators before and after historical anomalies over the past few days, weeks, or even months. Real-time data reflects the instantaneous state of the battery system at the current moment or within a recent short time window, such as the current battery pack temperature, charging and discharging current, instantaneous values ​​of individual cell voltages, and the rate of change of data over the last few seconds or minutes.

[0060] The first anomaly identification model adopts a dual-branch structure design. The first sub-module is a big data statistical model, which is used to process historical statistical data, extract key statistical features that characterize the vehicle battery status, and compare these key statistical features with thresholds to obtain anomaly indicators. The second sub-module is a behavior identification model, which is used to identify abnormal behavior events of the low-voltage system during operation through real-time statistical data to obtain anomaly indicators. The first determination module fuses the identification results of the two sub-modules to obtain the anomaly identification result.

[0061] Specifically, the first submodule analyzes historical indicator data, extracts abnormal indicators, and outputs the first identification result accordingly. For example, it calculates the percentage of times the voltage is below a set threshold (e.g., 11.8V) during the sampling period. If the percentage exceeds the set threshold (e.g., 30%), the percentage of low voltage values ​​is identified as an abnormal indicator. It also calculates the Pearson correlation coefficient between the voltage value and the SOC value during the sampling period. If the correlation coefficient is below the set threshold (e.g., 0.4), it indicates a weak relationship between the two, and the correlation between voltage and SOC is identified as an abnormal indicator. Finally, it calculates the average value of all SOH values ​​during the sampling period. If the average value is below the set threshold (e.g., 60%), the SOH value is identified as an abnormal indicator.

[0062] After obtaining the aforementioned abnormal indicators, the first submodule can directly identify the abnormal indicator as the first identification result. Alternatively, it can determine the abnormal score corresponding to the abnormal indicator based on the abnormal indicator, its corresponding judgment conditions, and the preset correspondence between abnormal indicators, judgment conditions, and abnormal scores. The abnormal scores of all abnormal indicators are then added together to obtain the total abnormal score. This total abnormal score and the abnormal indicator are then used as the first identification result. The preset correspondence between abnormal indicators, judgment conditions, and abnormal scores is shown in Table 1.

[0063] Table 1

[0064] The second submodule identifies abnormal behavior events based on the judgment conditions corresponding to each abnormal behavior event, determines the corresponding abnormal indicator and abnormal score (e.g., determined according to a preset correspondence between abnormal behavior events, judgment conditions, and abnormal scores), and adds up the abnormal scores corresponding to all abnormal behavior events. The sum and the corresponding abnormal indicator of each abnormal behavior event are then used as the second identification result. Table 2 shows the correspondence between behavior events and judgment conditions, and Table 3 shows the correspondence between abnormal behavior events, judgment conditions, and abnormal scores.

[0065] Table 2

[0066] Table 3

[0067] The first determining module fuses the identification results of the two sub-modules to obtain the anomaly identification result. The specific steps are as follows: If both the first and second identification results include anomaly indicators and anomaly scores, the anomaly indicators in the first and second identification results are merged to obtain the anomaly indicator in the anomaly identification result. The anomaly scores in the first and second identification results are accumulated to obtain the accumulated value. The range of the accumulated value is determined. Based on the range of the accumulated value and the preset correspondence between the range and the anomaly level, the anomaly level corresponding to that range is determined, and this anomaly level is identified as the anomaly level in the anomaly identification result. For example, the preset correspondence between the range and the anomaly level could be: when the accumulated value is greater than or equal to 0 and less than 20, it corresponds to a low-risk level; when the accumulated value is greater than or equal to 20 and less than 50, it corresponds to a medium-risk level; and when the accumulated value is greater than or equal to 50, it corresponds to a high-risk level. It should be noted that the above numerical range is only a specific example, and technicians can adjust the thresholds and level classifications according to actual detection needs.

[0068] In this embodiment, the second anomaly identification model includes multiple third sub-modules and a second determination module. Each third sub-module can determine a reference anomaly score, and the second determination module can determine the final target anomaly score based on these reference anomaly scores. The specific steps are as follows: using each third sub-module, the anomaly index data is identified to obtain a reference anomaly score corresponding to each third sub-module; using the second determination module, the reference anomaly score corresponding to each third sub-module is analyzed to obtain the target anomaly score of the vehicle battery.

[0069] The third submodule is a machine learning model used to identify anomalous indicator data and obtain corresponding reference anomaly scores. Each third submodule has a different type; for example, they may be XGBoost, Random Forest, or Long Short-Term Memory (LSTM) networks. The second determining module, also a machine learning model, is used to synthesize multiple reference anomaly scores to obtain the target anomaly score; specifically, it is a logistic regression module. In practice, the third submodule is first trained based on sample data, and then the second determining module is trained based on the trained third submodule and the sample data to obtain a trained second determining module. Alternatively, the third submodule and the second determining module can be trained simultaneously to obtain both trained third submodules and trained second determining modules. The training method is similar to existing training methods and will not be elaborated further here.

[0070] In this step, abnormal indicator data is input into each third submodule so that each third submodule can identify the abnormal indicator data and obtain a reference abnormal score for each third submodule. Then, the reference abnormal score for each third submodule is input into the second determining module so that the second determining module can analyze the reference abnormal score for each third submodule and obtain the target abnormal score for the vehicle battery.

[0071] Before inputting the outlier data into the third submodule, the outlier data can be preprocessed, and then input into the third submodule to obtain a reference outlier score. For example, preprocessing includes handling missing values, outlier handling, feature construction, and data normalization.

[0072] like Figure 2 As shown, this application embodiment provides an anomaly detection device, which corresponds to the method embodiment, and specifically includes: Acquisition unit 201 is used to acquire relevant indicator data of the vehicle battery; The first identification unit 202 is used to identify the relevant indicator data using a preset first anomaly identification model to obtain anomaly identification results; The determining unit 203 is used to determine the abnormal indicator data to be used based on the abnormal indicators involved in the abnormal identification result when the abnormal identification result indicates that the vehicle battery is abnormal. The abnormal indicator data includes at least a data segment extracted from the historical indicator data sequence corresponding to the abnormal indicator. The second identification unit 204 is used to identify the abnormal indicator data using a preset second abnormality identification model to obtain the target abnormality score of the vehicle battery. The detection unit 205 is used to detect whether there is an abnormality in the vehicle battery based on the target abnormality score.

[0073] Optionally, the determining unit 203 is used for: Based on the abnormal evolution cycle corresponding to the abnormal indicator, a first time window corresponding to the abnormal indicator is determined, wherein the duration of the abnormal evolution cycle is positively correlated with the duration of the first time window. Based on the first time window, determine the abnormal indicator data to be used from the historical indicator data sequence corresponding to the abnormal indicator.

[0074] Optionally, the determining unit 203 is used for: Determine the target abnormal time point corresponding to the abnormal indicator; If the time from the target abnormal time point to the current time point is less than the target time point, the end time point of the first time window is aligned with the current time point, and the historical indicator data of the corresponding time point within the first time window is extracted from the historical indicator data sequence corresponding to the abnormal indicator as the abnormal indicator data to be used. If the duration from the target abnormal time point to the current time point is greater than the target duration, a target time point is determined based on the target abnormal time point and the target duration. The end time point of the first time window is aligned with the target time point. Historical indicator data corresponding to the abnormal indicator within the first time window is extracted from the historical indicator data sequence corresponding to the abnormal indicator and used as the abnormal indicator data to be used.

[0075] Optionally, the determining unit 203 is used for: Determine the target abnormal time point corresponding to the abnormal indicator; If the target anomaly time point is not the current time point, determine the target type corresponding to the anomaly indicator; When the target type indicates that the target abnormal time point is the end time point of the first time window, the end time point of the first time window is aligned with the target abnormal time point, and the historical indicator data of the corresponding time point within the first time window is extracted from the historical indicator data sequence corresponding to the abnormal indicator as the abnormal indicator data to be used. When the target type indicates that the target abnormal time point is the start time point of the first time window, it is detected whether the duration from the target abnormal time point to the current time point is greater than the duration of the first time window, and a first detection result is obtained. Based on the first detection result and the first time window, the abnormal indicator data to be used is determined in the historical indicator data sequence corresponding to the abnormal indicator. When the target type indicates that the target abnormal time point is the middle time point of the first time window, it is detected whether the duration from the target abnormal time point to the current time point is greater than the target duration, and a second detection result is obtained. Based on the second detection result and the first time window, the abnormal indicator data to be used is determined in the historical indicator data sequence corresponding to the abnormal indicator.

[0076] Optionally, the determining unit 203 is used for: Determine the set of abnormal time points corresponding to the abnormal indicators; When the set of abnormal time points includes multiple abnormal time points, a first duration is determined based on the minimum and maximum time points in the set of abnormal time points. If the first duration is longer than the duration of the first time window, the duration of the first time window is adjusted to the first duration. Based on the adjusted first time window, the abnormal indicator data to be used is determined from the historical indicator data sequence corresponding to the abnormal indicator.

[0077] Optionally, the determining unit 203 is used for: Align the end time point of the adjusted first time window with the maximum time point, and extract the historical indicator data of the corresponding time point within the adjusted first time window from the historical indicator data sequence corresponding to the abnormal indicator, as the abnormal indicator data to be used. Alternatively, align the starting time of the adjusted first time window with the minimum time point, and extract the historical indicator data within the adjusted first time window at the corresponding time point from the historical indicator data sequence corresponding to the abnormal indicator, as the abnormal indicator data to be used.

[0078] Optionally, the determining unit 203 is used for: Based on the aforementioned abnormal indicators, determine the preset indicators; Based on the anomaly level indicated by the anomaly identification result, a second time window to be used is determined, wherein the anomaly level is positively correlated with the duration of the second time window; Based on the current time point and the second time window, determine the abnormal indicator data to be used from the historical indicator data sequence corresponding to the preset indicator.

[0079] Optionally, the determining unit 203 is used for: Based on the abnormal indicators, determine the target indicators related to the abnormal indicators; Based on the abnormal indicators and the target indicators, preset indicators are determined.

[0080] Optionally, the relevant indicator data includes historical indicator data and real-time indicator data. The first anomaly identification model includes a first sub-module, a second sub-module, and a first determination module. The first identification unit 202 is used for: Using the first submodule, the historical indicator data is identified to obtain a first identification result; The second submodule is used to identify the real-time indicator data to obtain a second identification result; Using the first determining module, an anomaly identification result is determined based on the first identification result and the second identification result.

[0081] Optionally, the second anomaly recognition model includes multiple different types of third sub-modules and second determination modules, and the second recognition unit 204 is used for: Using each third submodule, the abnormal indicator data is identified to obtain a reference abnormal score corresponding to each third submodule; Using the second determining module, the reference anomaly score corresponding to each third sub-module is analyzed to obtain the target anomaly score of the vehicle battery.

[0082] like Figure 3 As shown in the figure, this application embodiment provides an anomaly detection device, including a processor 301, a communication interface 302, a memory 303, and a communication bus 304, wherein the processor 301, the communication interface 302, and the memory 303 communicate with each other through the communication bus 304. Memory 303 is used to store computer programs; In one embodiment of this application, when the processor 301 executes a program stored in the memory 303, it implements the anomaly detection method provided in any of the foregoing method embodiments, including: Obtain relevant performance data for vehicle batteries; The relevant indicator data are identified using a preset first anomaly identification model to obtain anomaly identification results; When the anomaly identification result indicates that the vehicle battery is abnormal, the abnormal indicator data to be used is determined according to the abnormal indicators involved in the anomaly identification result. The abnormal indicator data includes at least a data segment extracted from the historical indicator data sequence corresponding to the abnormal indicator. Using a preset second anomaly identification model, the abnormal indicator data is identified to obtain the target anomaly score of the vehicle battery. Based on the target anomaly score, detect whether there is an anomaly in the vehicle battery.

[0083] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps performed by the anomaly detection method provided in any of the foregoing method embodiments.

[0084] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0085] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0086] It should be understood that the terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “described” as used herein may also include the plural forms. The terms “comprising,” “including,” “containing,” and “having” are inclusive and therefore indicate the presence of the stated features, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or combinations thereof. The method steps, processes, and operations described herein are not construed as requiring them to be performed in a particular order described or illustrated unless the order of performance is explicitly indicated. It should also be understood that additional or alternative steps may be used.

[0087] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. An anomaly detection method, characterized in that, The method includes: Obtain relevant performance data for vehicle batteries; The relevant indicator data are identified using a preset first anomaly identification model to obtain anomaly identification results; When the anomaly identification result indicates that the vehicle battery is abnormal, the abnormal indicator data to be used is determined according to the abnormal indicators involved in the anomaly identification result. The abnormal indicator data includes at least a data segment extracted from the historical indicator data sequence corresponding to the abnormal indicator. Using a preset second anomaly identification model, the abnormal indicator data is identified to obtain the target anomaly score of the vehicle battery. Based on the target anomaly score, detect whether there is an anomaly in the vehicle battery; Based on the anomaly indicators involved in the anomaly identification results, the anomaly indicator data to be used is determined, including: Based on the abnormal evolution cycle corresponding to the abnormal indicator, a first time window corresponding to the abnormal indicator is determined, wherein the duration of the abnormal evolution cycle is positively correlated with the duration of the first time window, and the abnormal evolution cycle is the time span required for the abnormal indicator to complete its full evolution process from its appearance. Determine the target abnormal time point corresponding to the abnormal indicator; If the time from the target abnormal time point to the current time point is less than the target time point, the end time point of the first time window is aligned with the current time point, and the historical indicator data of the corresponding time point within the first time window is extracted from the historical indicator data sequence corresponding to the abnormal indicator as the abnormal indicator data to be used. If the duration from the target abnormal time point to the current time point is greater than or equal to the target duration, a target time point is determined based on the target abnormal time point and the target duration. The end time point of the first time window is aligned with the target time point. Historical indicator data corresponding to the abnormal indicator within the first time window is extracted from the historical indicator data sequence corresponding to the abnormal indicator and used as the abnormal indicator data to be used.

2. An anomaly detection method, characterized in that, The method includes: Obtain relevant performance data for vehicle batteries; The relevant indicator data are identified using a preset first anomaly identification model to obtain anomaly identification results; When the anomaly identification result indicates that the vehicle battery is abnormal, the abnormal indicator data to be used is determined according to the abnormal indicators involved in the anomaly identification result. The abnormal indicator data includes at least a data segment extracted from the historical indicator data sequence corresponding to the abnormal indicator. Using a preset second anomaly identification model, the abnormal indicator data is identified to obtain the target anomaly score of the vehicle battery. Based on the target anomaly score, detect whether there is an anomaly in the vehicle battery; Based on the anomaly indicators involved in the anomaly identification results, the anomaly indicator data to be used is determined, including: Based on the abnormal evolution cycle corresponding to the abnormal indicator, a first time window corresponding to the abnormal indicator is determined, wherein the duration of the abnormal evolution cycle is positively correlated with the duration of the first time window, and the abnormal evolution cycle is the time span required for the abnormal indicator to complete its full evolution process from its appearance. Determine the target abnormal time point corresponding to the abnormal indicator; If the target anomaly time point is not the current time point, determine the target type corresponding to the anomaly indicator; When the target type indicates that the target abnormal time point is the end time point of the first time window, the end time point of the first time window is aligned with the target abnormal time point, and the historical indicator data of the corresponding time point within the first time window is extracted from the historical indicator data sequence corresponding to the abnormal indicator as the abnormal indicator data to be used. When the target type indicates that the target abnormal time point is the start time point of the first time window, it is detected whether the duration from the target abnormal time point to the current time point is greater than the duration of the first time window, and a first detection result is obtained. Based on the first detection result and the first time window, the abnormal indicator data to be used is determined in the historical indicator data sequence corresponding to the abnormal indicator. When the target type indicates that the target abnormal time point is the middle time point of the first time window, it is detected whether the duration from the target abnormal time point to the current time point is greater than the target duration, and a second detection result is obtained. Based on the second detection result and the first time window, the abnormal indicator data to be used is determined in the historical indicator data sequence corresponding to the abnormal indicator.

3. An anomaly detection method, characterized in that, The method includes: Obtain relevant performance data for vehicle batteries; The relevant indicator data are identified using a preset first anomaly identification model to obtain anomaly identification results; When the anomaly identification result indicates that the vehicle battery is abnormal, the abnormal indicator data to be used is determined according to the abnormal indicators involved in the anomaly identification result. The abnormal indicator data includes at least a data segment extracted from the historical indicator data sequence corresponding to the abnormal indicator. Using a preset second anomaly identification model, the abnormal indicator data is identified to obtain the target anomaly score of the vehicle battery. Based on the target anomaly score, detect whether there is an anomaly in the vehicle battery; Based on the anomaly indicators involved in the anomaly identification results, the anomaly indicator data to be used is determined, including: Based on the abnormal evolution cycle corresponding to the abnormal indicator, a first time window corresponding to the abnormal indicator is determined, wherein the duration of the abnormal evolution cycle is positively correlated with the duration of the first time window, and the abnormal evolution cycle is the time span required for the abnormal indicator to complete its full evolution process from its appearance. Determine the set of abnormal time points corresponding to the abnormal indicators; When the set of abnormal time points includes multiple abnormal time points, a first duration is determined based on the minimum and maximum time points in the set of abnormal time points. If the first duration is longer than the duration of the first time window, the duration of the first time window is adjusted to the first duration. Based on the adjusted first time window, the abnormal indicator data to be used is determined from the historical indicator data sequence corresponding to the abnormal indicator.

4. The method according to claim 3, characterized in that, Based on the adjusted first time window, the abnormal indicator data to be used is determined from the historical indicator data sequence corresponding to the abnormal indicator, including: Align the end time point of the adjusted first time window with the maximum time point, and extract the historical indicator data of the corresponding time point within the adjusted first time window from the historical indicator data sequence corresponding to the abnormal indicator, as the abnormal indicator data to be used. Alternatively, align the starting time of the adjusted first time window with the minimum time point, and extract the historical indicator data within the adjusted first time window at the corresponding time point from the historical indicator data sequence corresponding to the abnormal indicator, as the abnormal indicator data to be used.

5. An anomaly detection device, characterized in that, The apparatus is applied to the method according to any one of claims 1-4, comprising: The acquisition unit is used to acquire relevant indicator data of the vehicle battery. The first identification unit is used to identify the relevant indicator data using a preset first anomaly identification model to obtain anomaly identification results; The determining unit is configured to, when the anomaly identification result indicates that the vehicle battery is abnormal, determine the abnormal indicator data to be used based on the abnormal indicators involved in the anomaly identification result, including: determining a first time window corresponding to the abnormal indicator based on the abnormal evolution cycle corresponding to the abnormal indicator, wherein the duration of the abnormal evolution cycle is positively correlated with the duration of the first time window, and the abnormal evolution cycle is the time span required for the abnormal indicator to complete its full evolution process from its appearance; and determining the abnormal indicator data to be used in the historical indicator data sequence corresponding to the abnormal indicator based on the first time window, wherein the abnormal indicator data includes at least a data segment extracted from the historical indicator data sequence corresponding to the abnormal indicator. The second identification unit is used to identify the abnormal indicator data using a preset second abnormality identification model to obtain the target abnormality score of the vehicle battery. The detection unit is used to detect whether there is an abnormality in the vehicle battery based on the target abnormality score.

6. An anomaly detection device, characterized in that, The device is applied to the method of any one of claims 1-4, comprising: at least one communication interface; at least one bus connected to the at least one communication interface; at least one processor connected to the at least one bus; and at least one memory connected to the at least one bus, wherein the processor is configured to: Obtain relevant performance data for vehicle batteries; The relevant indicator data are identified using a preset first anomaly identification model to obtain anomaly identification results; When the anomaly identification result indicates that the vehicle battery is abnormal, the abnormal indicator data to be used is determined according to the abnormal indicators involved in the anomaly identification result, including: determining a first time window corresponding to the abnormal indicator according to the abnormal evolution cycle corresponding to the abnormal indicator, wherein the duration of the abnormal evolution cycle is positively correlated with the duration of the first time window, and the abnormal evolution cycle is the time span required for the abnormal indicator to complete its full evolution process from its appearance; determining the abnormal indicator data to be used in the historical indicator data sequence corresponding to the abnormal indicator according to the first time window; the abnormal indicator data includes at least a data segment extracted from the historical indicator data sequence corresponding to the abnormal indicator. Using a preset second anomaly identification model, the abnormal indicator data is identified to obtain the target anomaly score of the vehicle battery. Based on the target anomaly score, detect whether there is an anomaly in the vehicle battery.

7. 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 method described in any one of claims 1-4.

Images