A method and device for identifying abnormal battery cells of a new energy bus power battery

Abstract

The application provides a new energy bus power battery abnormal cell identification method and device, the method comprises: obtaining the voltage time series data set of each target cell in the power battery of the target new energy bus under multiple vehicle working conditions; entropy anomaly detection, fluctuation anomaly detection, voltage outlier anomaly detection, correlation anomaly detection and step anomaly detection are carried out on each target cell respectively, and entropy anomaly detection result, fluctuation anomaly detection result, voltage outlier anomaly detection result, correlation anomaly detection result and step anomaly detection result are determined correspondingly; the discrete score value corresponding to each target cell is determined by using the intersection logic of abnormal characteristics, and the target score value is determined, and the abnormal cell is identified according to the target score value. Through the above method, the new energy bus power battery cell full working condition, non-intrusive and high accuracy abnormal cell identification is realized, and the efficiency of fault detection and operation and maintenance disposal is improved.

Description

Technical Field

[0001] This application relates to the field of abnormal detection technology for power battery cells in new energy buses, and in particular to a method and device for identifying abnormal power battery cells in new energy buses. Background Technology

[0002] Current technologies for identifying abnormal cells in the power batteries of new energy buses mostly use a single core algorithm for anomaly identification. They can only set fixed rules in a single dimension. When the monitoring data of a certain cell exceeds the range of the rule, it is judged as abnormal. At the same time, existing technologies generally require disassembling the battery pack to carry out intrusive operation and maintenance inspections.

[0003] The aforementioned technical solutions have several drawbacks, including: a high false alarm rate for anomaly identification; the working states of power batteries are complex under different conditions such as charging and discharging, driving, and resting, with significant differences in voltage fluctuations and internal resistance changes; a single algorithm is easily affected by operating conditions and may produce misjudgments, failing to accurately pinpoint truly faulty cells, frequently generating invalid alarms, and increasing the workload of maintenance and identification; a single judgment dimension; existing algorithms mostly focus only on a single battery working state, resulting in a one-sided detection dimension that cannot comprehensively and reasonably reflect the true health status of the cells, easily leading to missed detection of early hidden faults; and high maintenance costs and low detection efficiency; fault detection requires forcibly stopping the vehicle and disassembling the battery pack for physical invasive testing, which not only consumes a lot of manpower, material resources, and financial resources, but also significantly increases battery maintenance costs and affects the stability and continuity of public transportation services. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a method and device for identifying abnormal cells in the power battery of new energy buses. By collecting voltage time-series data of each cell of the power battery under all operating conditions, parallel multi-dimensional anomaly detection is performed on the cells. The discrete score value of the cell is determined by the intersection logic of the anomaly features. Based on the target score value corresponding to the discrete score value, the abnormal cell is identified among multiple cells. This solves the problems of poor reliability of single-dimensional detection, incomplete fault identification, intrusive detection method, and low cost-effectiveness of operation and maintenance in the prior art. It realizes the identification of abnormal cells in the power battery of new energy buses under all operating conditions, non-intrusive, and with high accuracy. It effectively avoids the problems of false alarms and missed detection of hidden faults. It does not require vehicle downtime and battery pack disassembly, which greatly reduces operation and maintenance costs and improves the efficiency of fault detection and operation and maintenance.

[0005] This application provides a method for identifying abnormal battery cells in the power batteries of new energy buses. The identification method includes: For multiple target cells in the power battery of a target new energy bus, voltage time series datasets corresponding to each target cell under various vehicle operating conditions are obtained by the on-board terminal equipment in the target new energy bus; wherein, the vehicle operating conditions include charging condition, driving condition and stationary condition; Based on the voltage time series dataset, entropy anomaly detection, fluctuation anomaly detection, voltage outlier anomaly detection, correlation anomaly detection, and step size anomaly detection are performed on each target cell, and the corresponding entropy anomaly detection results, fluctuation anomaly detection results, voltage outlier anomaly detection results, correlation anomaly detection results, and step size anomaly detection results for each target cell are determined. Based on the entropy anomaly detection results, the fluctuation anomaly detection results, the voltage outlier anomaly detection results, the correlation anomaly detection results, and the step size anomaly detection results, the discrete score value corresponding to each target cell is determined using a preset anomaly feature intersection logic. Based on the discrete score values, a target score value corresponding to each target cell is determined, and abnormal cells are identified among the multiple target cells according to the target score values.

[0006] Furthermore, the entropy anomaly detection result for each target cell is determined through the following steps: Determine the voltage timing data corresponding to each target cell in the voltage timing dataset; For each target cell, the voltage timing data corresponding to the target cell is divided into window voltage data under multiple sliding windows according to a preset sliding window size, and the window voltage data under each sliding window is divided into multiple interval voltage data according to a preset number of intervals. Determine the voltage occurrence frequency value corresponding to each of the voltage intervals, and based on each voltage occurrence frequency value, determine the information entropy corresponding to each of the sliding windows, so as to determine the information entropy sequence data corresponding to the target cell; Determine the expected value and standard deviation of the information entropy sequence data, and based on the information entropy sequence data, the expected value and the standard deviation, determine the standardized outlier score of the target battery cell in each sliding window; Based on the standardized anomaly score and the preset entropy anomaly detection threshold corresponding to each sliding window, the entropy anomaly detection result corresponding to the target cell is determined; wherein, the entropy anomaly detection result includes one of the following: first entropy anomaly, second entropy anomaly, and no entropy anomaly, and the entropy anomaly detection result also includes an entropy anomaly time set.

[0007] Furthermore, the fluctuation anomaly detection result for each target cell is determined through the following steps: In the voltage time series dataset, the cell voltage dataset corresponding to each time moment and the voltage time series data corresponding to each target cell are determined respectively; The median of the cell voltage dataset is determined as the voltage reference value at each time point, and for each target cell, the voltage deviation at each time point is determined based on the voltage time series data and the voltage reference value. Based on the voltage deviation at each time point, determine the standard deviation of the deviation for the target cell. Based on the voltage time series data corresponding to each target cell, the average voltage sequence data corresponding to all target cells is determined, and based on the voltage time series data and the average voltage sequence data corresponding to the target cell, the mean square error value corresponding to the target cell is determined. Determine whether the standard deviation of the deviation is greater than a preset standard deviation threshold, and determine whether the mean square error is greater than a preset mean square error threshold; If the deviation standard deviation value is greater than the preset deviation standard deviation threshold, or the mean square error value is greater than the preset mean square error threshold, then the fluctuation anomaly detection result corresponding to the target cell is determined to be that there is a fluctuation anomaly. If the deviation standard deviation is less than or equal to a preset deviation standard deviation threshold, and / or the mean square error is less than or equal to a preset mean square error threshold, then the fluctuation anomaly detection result corresponding to the target cell is determined to be that there is no fluctuation anomaly.

[0008] Furthermore, the voltage outlier detection result for each target cell is determined through the following steps: In the voltage time series dataset, the cell voltage dataset corresponding to each time moment and the voltage time series data corresponding to each target cell are determined respectively; In the cell voltage dataset corresponding to each time moment, the upper quartile and lower quartile of the voltage are determined respectively, and the difference between the upper quartile and the lower quartile is determined as the voltage interquartile range corresponding to each time moment; Based on the lower voltage quartile and the voltage interquartile range, the lower outlier limit value corresponding to each time moment is determined, and based on the upper voltage quartile and the voltage interquartile range, the upper outlier limit value corresponding to each time moment is determined. For each target cell, the median of the cell voltage dataset is determined as the voltage reference value at each time point. Based on the outlier lower limit, the outlier upper limit, the voltage interquartile range, the voltage reference value, and the voltage time series data corresponding to the target cell, the single-point offset score value of the target cell at each time point is determined. The average value of the single-point offset score corresponding to the target cell at each time moment is determined as the average offset score value corresponding to the target cell, and it is determined whether the average offset score value is greater than or equal to a preset offset score threshold. If the average offset score is greater than or equal to the preset offset score threshold, then the voltage outlier detection result corresponding to the target cell is determined to be a voltage outlier. If the average offset score is less than the preset offset score threshold, then the voltage outlier detection result for the target cell is determined to be that there is no voltage outlier.

[0009] Furthermore, the correlation anomaly detection result for each target cell is determined through the following steps: Determine the voltage timing data corresponding to each target cell in the voltage timing dataset; For each target cell, the average voltage value corresponding to the target cell is determined based on the voltage timing data corresponding to the target cell. Based on the average voltage value and voltage time series data corresponding to the target cell, and the average voltage value and voltage time series data corresponding to other target cells, the Pearson correlation coefficient between the target cell and each of the other target cells is determined; The average value of the Pearson correlation coefficients between the target cell and each of the other target cells is determined as the correlation score value corresponding to the target cell. Determine whether the relevance score is less than a preset relevance score threshold; If the correlation score is less than the preset correlation score threshold, then the correlation anomaly detection result corresponding to the target battery cell is determined to be an anomaly. If the correlation score is greater than or equal to the preset correlation score threshold, then the correlation anomaly detection result corresponding to the target battery cell is determined to be that there is no correlation anomaly.

[0010] Furthermore, the step size anomaly detection result for each target cell is determined through the following steps: In the voltage timing dataset, determine the voltage timing data corresponding to each target cell under the charging condition and the driving condition, respectively; For each target cell, based on the voltage value corresponding to each moment in the voltage timing data, the voltage difference between the voltage value corresponding to each moment and the voltage value corresponding to the adjacent moment before that moment is determined, and the voltage difference is determined as the voltage step size of the target cell at each moment; The median of the voltage steps corresponding to all the target cells at each time point is determined as the reference step size for each time point. Based on the reference step size for each time point and the voltage step size for the target cell at each time point, the extreme value difference multiple for the target cell at each time point is determined. The average value of the extreme difference multiple corresponding to the target cell at each time moment is determined as the average difference multiple corresponding to the target cell, and it is determined whether the average difference multiple is greater than or equal to a preset difference multiple threshold. If the average difference multiple is greater than or equal to the preset difference multiple threshold, then the step size anomaly detection result corresponding to the target cell is determined to be that there is a step size anomaly. If the average difference multiple is less than the preset difference multiple threshold, then the step size anomaly detection result corresponding to the target cell is determined to be that there is no step size anomaly.

[0011] Furthermore, the step of determining the discrete score value corresponding to each target cell based on the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, the correlation anomaly detection result, and the step size anomaly detection result, using a preset anomaly feature intersection logic, includes: For each target cell, the entropy anomaly detection result, fluctuation anomaly detection result, voltage outlier anomaly detection result, correlation anomaly detection result and step size anomaly detection result corresponding to the target cell are traversed to construct an anomaly feature vector corresponding to the target cell, including each type of detection result; When the abnormal feature vector indicates that the correlation anomaly detection result is abnormal and the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, and the step size anomaly detection result are all normal, the discrete score value corresponding to the target cell is determined to be a first preset score value; wherein, the first preset score value indicates that the anomaly type of the target cell is a simple correlation anomaly. When the abnormal feature vector indicates that there are abnormalities in the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, and the step size anomaly detection result, the discrete score value corresponding to the target cell is determined to be a second preset score value; wherein, the second preset score value indicates that the anomaly type of the target cell is severe consistency collapse; When the abnormal feature vector indicates that there are abnormalities in both the entropy anomaly detection result and the step size anomaly detection result, or when the abnormal feature vector indicates that there are abnormalities in both the entropy anomaly detection result and the fluctuation anomaly detection result, the discrete score value corresponding to the target cell is determined to be a third preset score value; wherein, the third preset score value indicates that the anomaly type of the target cell is entropy step size joint anomaly or entropy fluctuation joint anomaly. When the abnormal feature vector indicates that the correlation anomaly detection result is abnormal and at least one of the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, and the step size anomaly detection result is abnormal, the discrete score value corresponding to the target cell is determined to be a fourth preset score value; wherein, the fourth preset score value indicates that the anomaly type of the target cell is a correlation joint anomaly.

[0012] Furthermore, the step of determining the target score value corresponding to each target cell based on the discrete score value, and identifying abnormal cells among multiple target cells according to the target score value, includes: For each target cell, the target score value corresponding to the target cell is determined by using a preset piecewise quadratic function based on the discrete score value corresponding to the target cell. Determine whether the target score is greater than or equal to a preset scoring threshold; If the target score is greater than or equal to a preset scoring threshold, the target cell is determined to be an abnormal cell. If the target score is less than the preset scoring threshold, then the target cell is determined not to be an abnormal cell.

[0013] This application embodiment also provides a device for identifying abnormal battery cells in the power battery of a new energy bus, the identification device comprising: The data acquisition module is used to acquire voltage time series datasets for each target cell in the power battery of the target new energy bus under various vehicle operating conditions, collected by the on-board terminal equipment of the target new energy bus; wherein, the vehicle operating conditions include charging condition, driving condition and stationary condition. An anomaly detection module is used to perform entropy anomaly detection, fluctuation anomaly detection, voltage outlier anomaly detection, correlation anomaly detection, and step size anomaly detection on each target cell based on the voltage time series dataset, and to determine the corresponding entropy anomaly detection result, fluctuation anomaly detection result, voltage outlier anomaly detection result, correlation anomaly detection result, and step size anomaly detection result for each target cell. The result quantization module is used to determine the discrete score value corresponding to each target cell based on the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, the correlation anomaly detection result, and the step size anomaly detection result, using a preset anomaly feature intersection logic. An anomaly identification module is used to determine the target score value corresponding to each target cell based on the discrete score value, and to identify the abnormal cell among the multiple target cells according to the target score value.

[0014] This application also provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, the steps of the above-described method for identifying abnormal cells in the power battery of a new energy bus are performed.

[0015] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the above-described method for identifying abnormal cells in the power battery of a new energy bus.

[0016] This application provides a method and apparatus for identifying abnormal battery cells in the power battery of a new energy bus. The identification method includes: for multiple target battery cells in the power battery of a target new energy bus, acquiring voltage time-series datasets for each target battery cell under various vehicle operating conditions, collected by an on-board terminal device in the target new energy bus; wherein the vehicle operating conditions include charging conditions, driving conditions, and stationary conditions; based on the voltage time-series datasets, performing entropy anomaly detection, fluctuation anomaly detection, voltage outlier anomaly detection, correlation anomaly detection, and step size anomaly detection on each target battery cell, respectively. The following steps are taken to determine the entropy anomaly detection result, fluctuation anomaly detection result, voltage outlier anomaly detection result, correlation anomaly detection result, and step size anomaly detection result for each target cell. Based on the entropy anomaly detection result, fluctuation anomaly detection result, voltage outlier anomaly detection result, correlation anomaly detection result, and step size anomaly detection result, a preset anomaly feature intersection logic is used to determine the discrete score value for each target cell. Based on the discrete score value, a target score value for each target cell is determined, and based on the target score value, an abnormal cell is identified among multiple target cells.

[0017] Compared with existing technologies that use a single core algorithm for anomaly identification, this method collects voltage and timing data of each cell in the power battery under all operating conditions, performs parallel multi-dimensional anomaly detection on the cells, and uses anomaly feature intersection logic to determine the discrete score value of the cell. Based on the target score value corresponding to the discrete score value, the abnormal cell is identified among multiple cells. This solves the problems of poor reliability of single-dimensional detection, incomplete fault identification, intrusive detection method, and low operation and maintenance cost of existing technologies. It realizes full-condition, non-intrusive, and highly accurate anomaly cell identification of power battery cells for new energy buses, effectively avoids false alarms and missed detection of hidden faults, eliminates the need for vehicle downtime and battery pack disassembly, significantly reduces operation and maintenance costs, and improves the efficiency of fault detection and operation and maintenance.

[0018] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart illustrating a method for identifying abnormal battery cells in a new energy bus power battery, provided in an embodiment of this application; Figure 2 A schematic diagram of an overall data flow architecture provided in an embodiment of this application; Figure 3 A schematic diagram of the structure of a device for identifying abnormal cells in the power battery of a new energy bus provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0021] 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, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. Based on the embodiments of this application, every other embodiment obtained by those skilled in the art without inventive effort falls within the scope of protection of this application.

[0022] Research has found that current technologies for identifying abnormal battery cells in new energy buses mostly rely on a single core algorithm (e.g., a single voltage threshold judgment or simple statistical distribution detection) for anomaly identification.

[0023] Existing conventional technologies can only set fixed rules in a single dimension (e.g., only for voltage fluctuations or only for voltage deviations). When the monitoring data of a certain cell exceeds the range of this rule, it is judged as abnormal. At the same time, existing technologies lack non-invasive identification solutions based on historical data during normal vehicle operation, and generally require disassembling the battery pack for invasive maintenance inspections. The overall technical solution has many defects, including: First, the false alarm rate is high. A single type of algorithm can only reflect the cell status from a single dimension. The working state of power batteries is complex under different operating conditions such as charging and discharging, driving, and sitting. The voltage fluctuation and internal resistance change characteristics are very different. A single algorithm is very susceptible to interference from operating conditions and will make false judgments. It cannot accurately locate the abnormal cells that are actually faulty, and frequently generates invalid alarms, increasing the workload of operation and maintenance identification.

[0024] Second, the judgment dimensions are too narrow. Most existing algorithms only focus on a single working state of the battery, or only detect the charging state or only detect the discharging driving state, without comprehensively considering the abnormal performance of the battery cell in all scenarios of charging, driving, and sitting. The detection dimensions are one-sided and cannot fully and reasonably reflect the true health status of the battery cell, which easily leads to the problem of missing early hidden faults.

[0025] Third, the operation and maintenance costs are high and the detection efficiency is low. Existing technologies do not have the ability to identify faults online in real time based on normal vehicle operation data. Fault detection requires forcibly stopping the vehicle and disassembling the battery pack for physical intrusive testing. This not only consumes a lot of manpower, material resources and financial resources, significantly increasing the cost of battery operation and maintenance, but also interrupts the normal operation of new energy buses, affecting the stability and continuity of public transportation services.

[0026] Based on this, this application provides a method for identifying abnormal cells in the power battery of new energy buses. By collecting voltage time-series data of each cell of the power battery under all operating conditions, parallel multi-dimensional anomaly detection is performed on the cells. The discrete score value of the cell is determined by the intersection logic of the anomaly features. Based on the target score value corresponding to the discrete score value, the abnormal cell is identified among multiple cells. This method solves the problems of poor reliability of single-dimensional detection, incomplete fault identification, intrusive detection method, and low cost-effectiveness of operation and maintenance in the prior art. It realizes the identification of abnormal cells in the power battery of new energy buses under all operating conditions, non-intrusive, and with high accuracy. It effectively avoids the problems of false alarms and missed detection of hidden faults. It does not require vehicle downtime and battery pack disassembly, which greatly reduces operation and maintenance costs and improves the efficiency of fault detection and operation and maintenance.

[0027] Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for identifying abnormal battery cells in a new energy bus power battery, as provided in an embodiment of this application. Figure 1 As shown in the embodiments of this application, the method for identifying abnormal cells in the power battery of a new energy bus includes: S100. For multiple target cells in the power battery of the target new energy bus, obtain the voltage time series dataset corresponding to each target cell under various vehicle operating conditions, collected by the on-board terminal equipment in the target new energy bus.

[0028] Here, new energy buses are urban public transport vehicles equipped with large-capacity power batteries as the sole or primary power source. They rely on the charging and discharging cycles of the power batteries to drive the vehicle and power all onboard auxiliary equipment, and have the characteristics of zero emissions and low noise operation. The power battery is an energy storage battery pack adapted to the high-power and long-range operation requirements of new energy buses. It is composed of multiple individual cells integrated in series and parallel. It is the core power storage and power supply component of the vehicle, and its working state directly determines the vehicle's driving safety and range.

[0029] Among them, the target cell is an independent single energy storage cell in the power battery pack. It is the smallest energy storage and electrical detection unit of the power battery and the core detection object for abnormal state identification and health detection in the method shown in the embodiments of this application. All detection algorithms take a single target cell as the smallest execution unit.

[0030] In this embodiment, the vehicle-mounted terminal device (e.g., T-box) is exclusively deployed on the remote intelligent data acquisition and transmission terminal of the new energy bus. It can collect core operating data such as vehicle operating conditions, power battery cell voltage, current, and SOC in real time around the clock, and stably upload them to the background data analysis system through TCP encryption protocol.

[0031] Here, the voltage time series dataset is a set of voltage data of the target battery cell that is continuously collected by the vehicle terminal at a fixed collection frequency and has a unique timestamp. It has strong time continuity and time sequence correlation, and can completely restore the voltage change pattern of the battery cell under different time periods and different operating conditions. The vehicle operating condition refers to the real-time operating status of the new energy bus during actual operation. Combining the characteristics of bus operation and the working principle of the battery, the vehicle operating condition includes, but is not limited to, charging condition, driving condition and stationary condition.

[0032] In this step, for multiple target cells configured in series or parallel within the power battery pack of the target new energy bus, the raw cell voltage data is collected in real time around the clock by the on-board terminal equipment (T-box) under all vehicle operating conditions. Combined with core operating parameters such as vehicle speed, battery current, and resting time, the three core operating conditions of the vehicle are accurately distinguished, and the operating condition classification and filtering of the raw data are completed. The continuous voltage time series data of each target cell under charging, driving, and resting conditions are collected and organized respectively. Finally, the data are integrated to form a standardized, operating condition-specific voltage time series dataset that can be directly used for algorithm detection, so as to completely avoid the problem of mixed multi-operating condition data.

[0033] In the embodiments of this application, for example, the definition criteria for vehicle operating conditions may include: charging condition is a vehicle standby charging state where the vehicle speed is 0, the charging current is continuously greater than 0 and the duration is ≥5min; driving condition is a vehicle driving state where the vehicle speed is greater than 0 and the battery is in the process of discharging or regenerative braking charging; stationary condition is a vehicle stationary standby state where the vehicle speed is 0, the absolute value of the current is <1A and the duration is ≥10min.

[0034] This approach distinguishes the differences in cell voltage variation patterns under various operating conditions, avoiding misjudgments in subsequent algorithms due to mixed data from multiple operating conditions. It provides a scenario-based and standardized data foundation for accurate multi-dimensional anomaly detection. By achieving precise isolation of operating condition data, it avoids detection interference caused by complex changes in operating conditions from the data source, ensuring the accurate operation of the five subsequent detection algorithms in their respective scenarios and significantly reducing the probability of basic misjudgments.

[0035] S200. Based on the voltage time series dataset, perform entropy anomaly detection, fluctuation anomaly detection, voltage outlier anomaly detection, correlation anomaly detection, and step size anomaly detection on each target cell, and determine the corresponding entropy anomaly detection result, fluctuation anomaly detection result, voltage outlier anomaly detection result, correlation anomaly detection result, and step size anomaly detection result for each target cell.

[0036] In this step, based on the standardized voltage time series dataset of different working conditions after data cleaning and working condition decomposition preprocessing, five types of differentiated and complementary cell anomaly detection algorithms are launched synchronously and in parallel, with each target cell as an independent detection unit. These algorithms are: information entropy constant detection, voltage time series fluctuation anomaly detection, statistical distribution voltage outlier anomaly detection, cell correlation consistency anomaly detection, and voltage change step size anomaly detection.

[0037] Among them, entropy anomaly detection, fluctuation anomaly detection, voltage outlier anomaly detection, correlation anomaly detection, and step size anomaly detection capture the explicit faults and early latent anomaly characteristics of the battery cell from five independent core dimensions: voltage distribution disorder, voltage time-series fluctuation stability, voltage cluster deviation, cell coordinated change consistency, and voltage dynamic change rate. Finally, they output five independent anomaly detection results for each target battery cell without interference.

[0038] This approach overcomes the limitations of existing single-dimensional detection technologies, comprehensively covering both overt and early-stage latent cell faults, and fully restoring the cell's health status under all operating conditions. Multi-dimensional complementary detection avoids the problems of missed or false detections associated with single algorithms, achieving full-coverage identification of cell faults and improving the comprehensiveness and reliability of anomaly detection.

[0039] In one possible implementation of this application, in specific implementation, the step of determining the entropy anomaly detection result corresponding to each target cell in step S200 may include: S211. Determine the voltage timing data corresponding to each target cell in the voltage timing dataset.

[0040] In this step, the complete continuous time voltage data corresponding to a single target cell is accurately screened and extracted from the preprocessed full-condition voltage time series dataset of charging, driving, and stationary operation. Other cell data and invalid operation condition data are removed to form exclusive original detection data corresponding only to the currently detected cell, providing a clean and accurate basic data source for subsequent entropy calculation and abnormal feature extraction.

[0041] S212. For each target cell, the voltage timing data corresponding to the target cell is divided into window voltage data under multiple sliding windows according to a preset sliding window size, and the window voltage data under each sliding window is divided into multiple interval voltage data according to a preset number of intervals.

[0042] Here, the sliding window is a general segmented processing unit for time series data analysis. It extracts and slides through continuous voltage time series data segment by segment using a fixed-length window, thereby realizing dynamic time series feature extraction and adapting to the detection scenario of real-time changes in battery cell voltage.

[0043] In this embodiment of the application, the preset sliding window size can be set to 60s; the preset number of intervals can be set to 10.

[0044] In this step, a preset sliding window size is used to slide and extract the continuous voltage time series data of a single cell segment by segment, and the resulting data is split into multiple continuous window voltage data covering the entire time period.

[0045] Furthermore, based on the standard of dividing intervals into equal-width intervals, all voltage sample values ​​within each sliding window are evenly divided into multiple independent interval voltage data, thus completing the refined binning preprocessing of time series data.

[0046] S213. Determine the voltage occurrence frequency value corresponding to each of the voltage intervals, and based on each of the voltage occurrence frequency values, determine the information entropy corresponding to each of the sliding windows, so as to determine the information entropy sequence data corresponding to the target cell.

[0047] Here, information entropy is a statistical indicator used to quantify the degree of disorder in the distribution of time-series data. When applied to cell voltage detection, normal cell voltage data is concentrated and the entropy value fluctuates stably, while abnormal cell voltage data is scattered and irregular and the entropy value will suddenly shift, which can identify hidden distribution anomalies.

[0048] In this step, the frequency of occurrence of cell voltage sampling data in each voltage range is counted one by one, the voltage occurrence frequency of the corresponding range is calculated, and the voltage distribution information entropy corresponding to each sliding window is calculated according to the information entropy calculation formula.

[0049] In this embodiment of the application, the expression for the information entropy calculation formula is as follows.

[0050] .

[0051] in, This represents the information entropy corresponding to each sliding window. Indicates the first The frequency of voltage occurrence in each interval, The number of intervals.

[0052] Furthermore, by traversing and calculating window by window, all sliding window data are continuously processed and finally spliced ​​together to form a complete information entropy sequence data covering the entire detection period of the target battery cell, which intuitively reflects the dynamic and disordered change characteristics of the battery cell voltage distribution.

[0053] S214. Determine the expected value and standard deviation of the information entropy sequence data, and based on the information entropy sequence data, the expected value and the standard deviation, determine the standardized anomaly score of the target battery cell in each sliding window.

[0054] Here, the standardized outlier score is determined by a data normalization algorithm (Z-Score normalization), which can eliminate the influence of differences in different data units and numerical ranges, convert the original entropy data into a uniform standard outlier score, and realize the quantitative rating of the degree of outlier.

[0055] In this step, the information entropy sequence data of the generated target battery cell is subjected to overall statistical analysis, the expected value and standard deviation of the overall sequence are calculated, and the entropy value fluctuation benchmark of the battery cell under normal conditions is obtained.

[0056] Furthermore, the information entropy corresponding to each sliding window is standardized dimensionlessly using the Z-Score standardization formula to unify the data evaluation criteria, and finally obtains the standardized anomaly score corresponding to each sliding window, thereby realizing the quantitative expression of the degree of entropy anomaly.

[0057] In this embodiment of the application, the expression of the Z-Score normalization formula is as follows.

[0058] .

[0059] in, This represents the standardized outlier score for each sliding window. This represents the information entropy corresponding to each sliding window. and These represent the expected value and standard deviation of the information entropy sequence data, respectively.

[0060] S215. Based on the standardized anomaly score value corresponding to each sliding window and the preset entropy anomaly detection threshold, determine the entropy anomaly detection result corresponding to the target cell.

[0061] The entropy anomaly detection results include one of the following: first entropy anomaly, second entropy anomaly, and entropy-free anomaly. The entropy anomaly detection results also include an entropy anomaly time set.

[0062] Here, the first entropy anomaly is characterized as a slight anomaly, and the second entropy anomaly is characterized as a severe anomaly.

[0063] In this embodiment, the preset entropy anomaly detection threshold includes a first entropy anomaly detection threshold and a second entropy anomaly detection threshold, wherein the second entropy anomaly detection threshold is greater than the first entropy anomaly detection threshold.

[0064] In this step, the standardized anomaly score corresponding to each sliding window is compared with the first entropy anomaly detection threshold and the second entropy anomaly detection threshold, respectively. When the standardized anomaly score is greater than or equal to the first entropy anomaly detection threshold and less than the second entropy anomaly detection threshold, the entropy anomaly detection result corresponding to the target cell is determined to be a first entropy anomaly. When the standardized anomaly score is greater than or equal to the second entropy anomaly detection threshold, the entropy anomaly detection result corresponding to the target cell is determined to be a second entropy anomaly. When the standardized anomaly score is less than the first entropy anomaly detection threshold, the entropy anomaly detection result corresponding to the target cell is determined to be an entropy-free anomaly.

[0065] Furthermore, the time nodes corresponding to all sliding windows that generate anomalies are accurately counted and summarized to form a complete entropy anomaly time set, accurately recording the complete time period during which the anomalies occur and continue.

[0066] For example, the preset first entropy anomaly detection threshold Z1 is 2, that is, the slight entropy anomaly threshold; the preset second entropy anomaly detection threshold Z2 is 3, that is, the severe entropy anomaly threshold; according to the magnitude of the standardized anomaly score of each window, the anomaly level is classified: the standardized anomaly score between 2 and 3 is judged as the first entropy anomaly (slight anomaly), the standardized anomaly score ≥ 3 is judged as the second entropy anomaly (severe anomaly), and the standardized anomaly score < 2 is judged as the entropy-free anomaly.

[0067] In one possible implementation of this application, in specific implementation, the step of determining the fluctuation anomaly detection result corresponding to each target cell in step S200 may include: S221. Determine the cell voltage dataset corresponding to each time moment and the voltage time series data corresponding to each target cell in the voltage time series dataset.

[0068] In this step, the complete voltage set of all series cells in the power battery pack at each acquisition time is extracted from the preprocessed voltage time series data under different working conditions according to the acquisition timestamp. At the same time, the continuous voltage time series data of a single target cell covering the entire testing period is extracted separately. The cell cluster comparison dataset and the single cell test dataset are constructed respectively, providing dual data support for the accurate calculation of voltage deviation and fluctuation error in the future.

[0069] S222. Determine the median of the cell voltage dataset as the voltage reference value for each time moment, and for each target cell, determine the voltage deviation of the target cell at each time moment based on the voltage time series data and the voltage reference value.

[0070] Here, voltage deviation is the difference between the real-time detected voltage of a single cell and the median voltage of the cells in the power battery cluster. It is used to intuitively quantify the instantaneous deviation of the voltage of a single cell relative to the overall normal operating level of the battery pack.

[0071] In this step, to avoid interference from individual extreme abnormal cell voltages on the overall benchmark, the median of all cell voltages at each moment is selected as the voltage benchmark value at that moment, accurately reflecting the normal voltage level of the battery pack cluster.

[0072] Furthermore, the difference between the real-time voltage of a single target cell and the cluster reference voltage is calculated at each time step using the deviation calculation formula, so as to obtain the voltage deviation corresponding to each time step and quantify the instantaneous voltage deviation of the cell.

[0073] In this embodiment of the application, the expression for the deviation calculation formula is as follows.

[0074] .

[0075] in, For the target cell at every moment The corresponding voltage deviation, This is voltage timing data. This is the voltage reference value.

[0076] S223. Based on the voltage deviation at each time point, determine the standard deviation of the deviation corresponding to the target cell.

[0077] In this step, voltage deviation data of the target cell at all times throughout the entire period are collected to form a complete deviation time series. The series is then processed using the statistical standard deviation calculation formula to obtain the deviation standard deviation value corresponding to the target cell, which accurately characterizes the overall dispersion and stability of voltage fluctuations of the cell during full-operation process.

[0078] S224. Based on the voltage time series data corresponding to each target cell, determine the average voltage sequence data corresponding to all the target cells, and based on the voltage time series data and the average voltage sequence data corresponding to the target cell, determine the mean square error value corresponding to the target cell.

[0079] Here, mean square error (MSE) is a statistical error analysis index used to measure the overall deviation between the voltage time-series fluctuation trajectory of a single cell and the average fluctuation trajectory of all cells in the battery pack, accurately reflecting the long-term fluctuation consistency of the cells.

[0080] In this step, the voltage timing data of all cells are first integrated to calculate the average voltage sequence data of the entire battery pack, which represents the standard fluctuation trajectory of the cell cluster.

[0081] Furthermore, by comparing the difference between the voltage trajectory of a single cell and the average trajectory of the cluster using the mean square error calculation formula, the mean square error value corresponding to a single target cell is accurately calculated, thereby quantifying the degree of deviation of the long-term fluctuation trajectory of the cell.

[0082] In this embodiment of the application, the expression for the mean squared error calculation formula is as follows.

[0083] .

[0084] in, For the first The mean square error value corresponding to each target cell For the first Voltage timing data corresponding to each target cell For average voltage sequence data, This represents the total number of moments.

[0085] S225. Determine whether the standard deviation of the deviation is greater than a preset standard deviation threshold, and determine whether the mean square error is greater than a preset mean square error threshold.

[0086] S226. If the deviation standard deviation value is greater than the preset deviation standard deviation threshold, or the mean square error value is greater than the preset mean square error threshold, then the fluctuation anomaly detection result corresponding to the target cell is determined to be a fluctuation anomaly.

[0087] S227. If the deviation standard deviation value is less than or equal to the preset deviation standard deviation threshold, and / or the mean square error value is less than or equal to the preset mean square error threshold, then the fluctuation anomaly detection result corresponding to the target cell is determined to be that there is no fluctuation anomaly.

[0088] Here, based on a large amount of normal bus battery operation data, a preset deviation standard deviation threshold and a preset mean square error threshold are calibrated in advance as the criteria for judging normal battery cells.

[0089] In this embodiment, the preset deviation standard deviation threshold can be set as a preset multiple of the average deviation standard deviation of all cells; the preset mean square error threshold can be obtained from historical normal data statistics.

[0090] In this step, when the standard deviation of the target cell exceeds the corresponding threshold or the mean square error exceeds the corresponding threshold, the cell can be determined to have abnormal voltage fluctuations. Only when both indicators are lower than or equal to the corresponding thresholds can the cell be determined to have no abnormal fluctuations. The dual-condition complementary judgment improves the reliability of the detection.

[0091] In one possible implementation of this application, in specific implementation, the step of determining the voltage outlier detection result corresponding to each target cell in step S200 may include: S231. Determine the cell voltage dataset corresponding to each time moment and the voltage time series data corresponding to each target cell in the voltage time series dataset.

[0092] In this step, the full-vehicle cell voltage set of the power battery pack is extracted one by one according to the time dimension from the preprocessed data of charging, driving and stationary conditions. At the same time, the full-time time-series voltage data of a single target cell is extracted separately, and cluster statistical data set and single cell detection data set are constructed respectively, providing complete data support for the subsequent adaptive calculation of voltage distribution characteristics and outlier degree.

[0093] S232. Determine the upper quartile and lower quartile of the voltage at each time point from the cell voltage dataset at each time point, and determine the difference between the upper quartile and the lower quartile as the interquartile range of the voltage at each time point.

[0094] Here, the quartiles divide the ordered cell voltage data into four equal numerical nodes, including the lower voltage quartile (Q1, 25th percentile) and the upper voltage quartile (Q3, 75th percentile), which are used to characterize the centralized distribution of the data. The voltage interquartile range (IQR) is the difference between the upper voltage quartile and the lower voltage quartile. It can stably reflect the concentration range of the cell voltage data, is not affected by extreme outliers, and is the core parameter for adaptive anomaly interval determination.

[0095] In this step, the full vehicle cell voltage data at each collection time are sorted in ascending order. The lower quartile Q1 at the 25th percentile and the upper quartile Q3 at the 75th percentile are accurately calculated using the statistical quartile calculation method. Then, the interquartile range of the cell voltage at each time is calculated using the difference calculation formula IQR=Q3-Q1, which accurately reflects the concentrated distribution range of the battery voltage data at the current time.

[0096] S233. Based on the lower voltage quartile and the voltage interquartile range, determine the lower outlier limit value corresponding to each time moment, and based on the upper voltage quartile and the voltage interquartile range, determine the upper outlier limit value corresponding to each time moment.

[0097] In this step, based on the lower quartile, upper quartile, and interquartile range obtained at each time point, the outlier lower limit and outlier upper limit of the cell voltage data at each time point are calculated using the adaptive interval calculation formula, and the normal fluctuation range of cell voltage that is adapted to the current operating conditions and SOC state is dynamically defined.

[0098] In this embodiment of the application, the expression of the adaptive interval calculation formula is as follows.

[0099] lower(t) = Q1(t) 1.5×IQR(t); upper(t)=Q3(t)+1.5×IQR(t).

[0100] Where lower(t) represents each time step. The corresponding outlier lower bound, upper(t), is the value at each time step. The corresponding outlier upper limit values ​​are Q1(t), which is the lower quartile of the voltage, Q3(t), which is the upper quartile of the voltage, and IQR(t), which is the voltage interquartile range.

[0101] S234. For each target cell, the median of the cell voltage dataset is determined as the voltage reference value at each time moment. Based on the outlier lower limit, the outlier upper limit, the voltage interquartile range, the voltage reference value, and the voltage time series data corresponding to the target cell, the single-point offset score value of the target cell at each time moment is determined.

[0102] Here, the single-point offset score is an outlier quantification index for the cell voltage at a single acquisition moment, used to accurately distinguish the degree of anomaly of slight, moderate and severe outlier cell voltage offset.

[0103] In this embodiment of the application, if the first Each target cell at every moment If the corresponding voltage value exceeds the numerical range corresponding to the lower and upper outlier limits, the single-point offset score value of the target cell at each moment is determined by the following formula.

[0104] .

[0105] in, For the first Each target cell at every moment The corresponding single-point offset score, For the first Voltage timing data corresponding to each target cell For each moment The corresponding voltage reference value, The voltage interquartile range, It is a minimum value to prevent the denominator from being 0.

[0106] Furthermore, if the first Each target cell at every moment If the corresponding voltage value falls within the numerical range corresponding to the lower and upper outlier limits, then the single-point offset score for that moment is 0 by default, thus accurately quantifying the degree of voltage outlier at each moment.

[0107] S235. Determine the average value of the single-point offset score corresponding to the target cell at each time moment as the average offset score value corresponding to the target cell, and determine whether the average offset score value is greater than or equal to the preset offset score threshold.

[0108] In this step, the single-point offset scores of the target cell at all collection times throughout the entire period are summarized, and the arithmetic mean is calculated to obtain the average offset score of the cell, which represents the overall persistent outlier level of the cell.

[0109] Furthermore, the average offset score is compared with a preset offset score threshold to filter out invalid offset data caused by instantaneous voltage fluctuations and determine whether the cell has a persistent outlier anomaly.

[0110] S236. If the average offset score is greater than or equal to the preset offset score threshold, then the voltage outlier detection result corresponding to the target cell is determined to be a voltage outlier.

[0111] S237. If the average offset score is less than the preset offset score threshold, then the voltage outlier detection result corresponding to the target cell is determined to be that there is no voltage outlier.

[0112] In this embodiment of the application, when the average offset score of the target cell is greater than or equal to the preset offset score threshold, it proves that the cell has a long-term voltage cluster outlier phenomenon and is determined to have a voltage outlier anomaly; when the average offset score is less than the preset offset score threshold, it indicates that the cell voltage always fits the normal distribution range of the cluster and is determined to have no voltage outlier anomaly.

[0113] In one possible implementation of this application, in specific implementation, the step of determining the correlation anomaly detection result corresponding to each target cell in step S200 may include: S241. Determine the voltage timing data corresponding to each target cell in the voltage timing dataset.

[0114] In this step, complete and continuous voltage time series data corresponding to a single target cell are accurately extracted from the preprocessed working condition voltage time series dataset. Redundant data and data from other cells are removed, and the time series change characteristics of the currently detected cell are independently retained, providing pure single-cell data for subsequent inter-cell collaborative consistency comparison and correlation coefficient calculation.

[0115] S242. For each target cell, determine the average voltage value corresponding to the target cell based on the voltage timing data corresponding to the target cell.

[0116] In this step, the average voltage value of the extracted single-target cell under the current operating conditions is calculated by the arithmetic mean algorithm. This average value serves as the overall voltage reference parameter for the cell and provides the basic mean data for the accurate calculation of the Pearson correlation coefficient between cells.

[0117] S243. Based on the average voltage value and voltage timing data corresponding to the target cell, and the average voltage value and voltage timing data corresponding to other target cells, determine the Pearson correlation coefficient between the target cell and each of the other target cells.

[0118] Here, the Pearson correlation coefficient is a quantitative indicator that measures the degree of linear synchronization between two sets of time series data. Its value range is [-1, 1]. Normally, the voltage changes between cells are highly synchronized, and the correlation coefficient approaches 1 infinitely.

[0119] In this step, based on the average voltage value and timing voltage data of the target cell, as well as the corresponding data of all other cells in the battery pack, the timing voltage correlation coefficient between the current target cell and each of the other cells is calculated one by one using the standard Pearson correlation coefficient calculation formula, thereby quantifying the degree of synchronous consistency of voltage changes between pairs of cells.

[0120] In this embodiment of the application, the expression for the standard Pearson correlation coefficient calculation formula is as follows.

[0121] .

[0122] in, For the first The target cell and the first Pearson correlation coefficient between target cells and For the first Average voltage value and voltage timing data for each target cell and For the first Average voltage value and voltage timing data for each target cell This represents the time window corresponding to the voltage timing data.

[0123] S244. The average value of the Pearson correlation coefficients between the target cell and each of the other target cells is determined as the correlation score value corresponding to the target cell.

[0124] In this step, the corresponding Pearson correlation coefficients of the target cell and all other cells in the battery pack are summarized, and the arithmetic mean of all correlation coefficients is calculated to obtain the correlation score value corresponding to the target cell, which intuitively reflects the overall level of coordinated change between the target cell and the cells in the entire battery pack cluster.

[0125] S245. Determine whether the relevance score is less than the preset relevance score threshold.

[0126] S246. If the correlation score is less than the preset correlation score threshold, then the correlation anomaly detection result corresponding to the target cell is determined to be a correlation anomaly.

[0127] S247. If the correlation score is greater than or equal to the preset correlation score threshold, then the correlation anomaly detection result corresponding to the target cell is determined to be that there is no correlation anomaly.

[0128] For example, a preset correlation score threshold of 0.9 is set, and the total correlation score of the target cell is compared with this threshold. If the correlation score is lower than 0.9, it indicates that the cell voltage change is not synchronized with the cluster, and it is determined that there is an abnormal correlation. If the correlation score is greater than or equal to 0.9, it indicates that the cell has good coordination, and it is determined that there is no abnormal correlation.

[0129] In one possible implementation of this application, in specific implementation, the step S200 of determining the step size anomaly detection result corresponding to each target cell may include: S251. Determine the voltage timing data corresponding to each target cell under the charging condition and the driving condition in the voltage timing dataset.

[0130] In this step, since the voltage change under static conditions is minimal and the step size has no statistical significance, based on the principle of operating condition adaptability, the effective voltage time series data of the target cell under charging and driving conditions are screened. Invalid data with no obvious voltage fluctuations under static conditions are completely eliminated to avoid invalid data interfering with the accuracy of step size calculation, thus providing a highly adaptable and effective dataset for detecting abnormal voltage change rates.

[0131] S252. For each target cell, based on the voltage value corresponding to each moment in the voltage timing data, determine the voltage difference between the voltage value corresponding to each moment and the voltage value corresponding to the adjacent moment before that moment, and determine the voltage difference as the voltage step size corresponding to the target cell at each moment.

[0132] Here, the voltage step size is the difference in cell voltage between two adjacent acquisition times of the power battery. It directly represents the real-time rate of change of cell voltage and can reflect the dynamic response characteristics during battery charging and discharging and operating condition switching.

[0133] In this step, the voltage time series data of a single cell under effective operating conditions is traversed, and the voltage values ​​of the current time and the adjacent previous time are extracted one by one. The voltage change difference is calculated by the difference formula, and the difference is defined as the voltage step size at the current time, so as to obtain the complete voltage change rate sequence of the cell for the whole time period.

[0134] In this embodiment of the application, the expression for the difference formula is as follows.

[0135] .

[0136] in, Indicates time The corresponding voltage step size, Represents the time in voltage timing data The corresponding voltage value, Represents the time in voltage timing data The corresponding voltage value.

[0137] S253. Determine the median of the voltage step lengths corresponding to all the target cells at each time step as the reference step length for each time step, and determine the extreme value difference multiple for the target cell at each time step based on the reference step length for each time step and the voltage step length for the target cell at each time step.

[0138] Here, the extreme value difference multiple is the deviation multiple between the voltage step size of a single cell and the reference step size of the battery pack cluster, which is used to quantify the degree of deviation between the voltage change rate of a single cell and the overall cluster rate.

[0139] In this step, the voltage step size of all cells at each acquisition time is sorted, and the median is selected as the cluster reference step size at the current time to avoid interference from extreme step size data; the extreme value difference multiple between the single cell step size and the reference step size is calculated at each time using the extreme value difference multiple calculation formula to quantify the rate deviation.

[0140] In the embodiments of this application, the expression of the formula for calculating the extreme value difference multiple is as follows.

[0141] .

[0142] in, For the first Each target cell at time The corresponding extreme value difference multiple, Indicates time The corresponding voltage step size, For a moment The corresponding baseline step size, It is a minimum value to prevent the denominator from being 0.

[0143] S254. Determine the average value of the extreme difference multiple corresponding to the target cell at each time moment as the average difference multiple corresponding to the target cell, and determine whether the average difference multiple is greater than or equal to the preset difference multiple threshold.

[0144] In this step, the extreme difference multiples of the target cell under all time conditions of charging and driving are summarized, and the average difference multiple is calculated by arithmetic average to eliminate the single data deviation caused by instantaneous operating condition fluctuations and accurately reflect the abnormal level of the long-term voltage change rate of the cell.

[0145] Furthermore, the average difference multiple is compared with a preset difference multiple threshold to determine whether the average difference multiple is greater than or equal to the preset difference multiple threshold.

[0146] S255. If the average difference multiple is greater than or equal to the preset difference multiple threshold, then the step size anomaly detection result corresponding to the target cell is determined to be an anomaly in step size.

[0147] S256. If the average difference multiple is less than the preset difference multiple threshold, then the step size anomaly detection result corresponding to the target cell is determined to be that there is no step size anomaly.

[0148] For example, 2 is set as the preset difference multiple threshold, and the average difference multiple of the target cell is compared with the threshold. If the average difference multiple is greater than or equal to the threshold, it indicates that the cell voltage change rate deviates too much from the cluster, and it is determined that there is an abnormal step size. If the average difference multiple is less than the threshold, it is determined that the cell voltage change rate is normal and there is no abnormal step size.

[0149] S300. Based on the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, the correlation anomaly detection result, and the step size anomaly detection result, the discrete score value corresponding to each target cell is determined using a preset anomaly feature intersection logic.

[0150] Here, the discrete score is a discrete quantized score that is based on the intersection logic of abnormal features and is used to classify and assign values ​​to the cell abnormality type and abnormality level. It is used to initially distinguish cells with different fault types.

[0151] In this step, the entropy anomaly detection results, fluctuation anomaly detection results, voltage outlier anomaly detection results, correlation anomaly detection results, and step size anomaly detection results corresponding to each target cell in the power battery pack are integrated one by one. Each type of anomaly result is quantified and labeled to construct a dedicated structured anomaly feature vector adapted to the fault characteristics of a single cell. Furthermore, by utilizing the pre-set multi-algorithm abnormal feature intersection matching logic, different abnormal feature combination scenarios are accurately classified and judged. Combining the fault severity level corresponding to various abnormal combinations, the abnormal state of the battery cell is quantitatively graded and scored. Finally, a standardized discrete score value corresponding to each target battery cell is stably generated, completing the preliminary classification of abnormalities and the screening of false abnormalities.

[0152] In this way, through cross-validation of multiple detection results, genuine and valid anomalies are screened out, while false anomalies caused by interference from operating conditions and data due to a single algorithm are eliminated, achieving preliminary anomaly classification. This addresses the industry pain point of high false alarm rates with single algorithms, accurately distinguishes between false anomalies and real faults, and provides a reliable basis for subsequent accurate scoring and judgment.

[0153] In one possible implementation of this application, step S300 may include: S301. For each target cell, traverse the entropy anomaly detection result, fluctuation anomaly detection result, voltage outlier anomaly detection result, correlation anomaly detection result, and step size anomaly detection result corresponding to the target cell, and construct an anomaly feature vector corresponding to the target cell that includes each type of detection result.

[0154] Here, the abnormal feature vector is a set of features that are structured and quantified from 0 to 1 for the five types of abnormality detection results of each target cell. 1 represents the existence of the corresponding abnormality type, and 0 represents no abnormality. It is the basic data structure for intersection logic operation.

[0155] For example, the expression form of anomaly feature vector is: F i = [fau4, flu, fau3_5, pe, step], where F i represents the abnormal feature vector corresponding to the i-th target cell; fau4, flu, fau3_5, pe, and step respectively represent the entropy anomaly detection result, fluctuation anomaly detection result, voltage outlier anomaly detection result, correlation anomaly detection result, and step size anomaly detection result. Each element in the abnormal feature vector is 0 or 1, where 1 indicates the existence of this type of anomaly and 0 indicates the absence of this type of anomaly.

[0156] S302. When the abnormal feature vector indicates that the correlation anomaly detection result is abnormal and the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, and the step size anomaly detection result are all normal, the discrete score value corresponding to the target cell is determined to be the first preset score value.

[0157] The first preset score indicates that the anomaly type of the target battery cell is a simple correlation anomaly.

[0158] In this step, when the abnormal feature vector of the target cell only marks the existence of correlation anomaly, and the four types of anomalies such as entropy anomaly, fluctuation anomaly, voltage outlier anomaly, and step size anomaly are not marked, it is determined that the cell only has inter-cell coordination deviation and no substantial voltage fault. It is defined as simple correlation anomaly and uniformly assigned the first preset score value, corresponding to the minor anomaly level.

[0159] In this embodiment of the application, the first preset score can be set to 20 points.

[0160] S303. When the abnormal feature vector indicates that there are abnormalities in the entropy abnormality detection result, the fluctuation abnormality detection result, the voltage outlier abnormality detection result, and the step size abnormality detection result, the discrete score value corresponding to the target cell is determined to be the second preset score value.

[0161] The second preset score indicates that the abnormality type of the target cell is severe consistency collapse.

[0162] In this step, when the abnormal feature vector of the target cell simultaneously marks the existence of all four types of abnormalities: entropy abnormality, fluctuation abnormality, voltage outlier abnormality, and step size abnormality, it indicates that the cell voltage distribution, fluctuation, offset, and rate of change have all failed. It is judged as a severe consistency collapse fault and is uniformly assigned a second preset score value, corresponding to the severe fault level.

[0163] In this embodiment of the application, the second preset score can be set to 25 points.

[0164] S304. When the abnormal feature vector indicates that there are abnormalities in both the entropy abnormality detection result and the step size abnormality detection result, or when the abnormal feature vector indicates that there are abnormalities in both the entropy abnormality detection result and the fluctuation abnormality detection result, the discrete score value corresponding to the target cell is determined to be a third preset score value.

[0165] The third preset score value indicates that the abnormality type of the target battery cell is either an entropy step size joint abnormality or an entropy fluctuation joint abnormality.

[0166] In this step, when the abnormal feature vector of the target cell shows a combination of entropy abnormality and step size abnormality, or entropy abnormality and fluctuation abnormality, it is determined that the cell voltage distribution abnormality and voltage change rate and fluctuation state are linked and failed. This is defined as entropy step size joint abnormality or entropy fluctuation joint abnormality, and a third preset score value is uniformly assigned, corresponding to a medium fault level.

[0167] In this embodiment of the application, the third preset score can be set to 30 points.

[0168] S305. When the abnormal feature vector indicates that the correlation anomaly detection result is abnormal and at least one of the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result and the step size anomaly detection result is abnormal, the discrete score value corresponding to the target cell is determined to be the fourth preset score value.

[0169] The fourth preset score value indicates that the abnormality type of the target battery cell is a correlated joint abnormality.

[0170] In this step, when the abnormal feature vector of the target cell has a correlation abnormality, and is accompanied by at least one type of substantial voltage fault abnormality among entropy abnormality, fluctuation abnormality, voltage outlier abnormality, and step size abnormality, it is determined to be a composite abnormality of coordinated abnormality superimposed with substantial fault, defined as a correlation joint abnormality, and uniformly assigned a fourth preset score value, corresponding to a more severe fault level.

[0171] In this embodiment of the application, the fourth preset score can be set to 40 points.

[0172] S400. Based on the discrete score values, determine the target score value corresponding to each target cell, and identify abnormal cells among the multiple target cells according to the target score values.

[0173] Here, the target score is a continuous quantization score from 0 to 100 obtained by finely transforming the discrete score through a preset piecewise quadratic function, which accurately represents the severity of cell abnormalities.

[0174] In this step, for each target cell, the discrete score value is calculated, and a preset piecewise quadratic fitting function is called to perform a refined numerical conversion, converting the discrete graded scores into standardized target scores that are continuously distributed in the range of 0 to 100, thereby achieving a refined quantification of the severity of cell anomalies.

[0175] Furthermore, the quantified target score is accurately compared with the preset official anomaly scoring threshold. Based on the comparison results, normal and abnormal cells in the power battery pack are quickly screened and determined to complete the final identification of faulty cells.

[0176] This transforms discrete grading scores into continuous, refined ratings, quantifying the severity of cell anomalies and enabling precise identification and grading of abnormal cells. The more refined criteria better align with actual operating conditions, accurately identifying abnormal cells and providing maintenance personnel with a reference for anomaly severity, thus supporting tiered maintenance operations.

[0177] In one possible implementation of this application, step S400 may include: S401. For each target cell, based on the discrete score value corresponding to the target cell, the target score value corresponding to the target cell is determined using a preset piecewise quadratic function.

[0178] Here, the piecewise quadratic function is a preset fraction conversion fitting function in the embodiments of this application, which adapts to the conversion requirements of discrete fractions of different levels and realizes the refined and continuous classification of abnormal levels.

[0179] In this step, a differentiated segmented secondary transformation rule is adopted to achieve refined scoring fitting for the discrete scores of different grades of battery cells. Through segmented adaptation, the grade-based discrete scores are uniformly converted into continuous target scores in the range of 0 to 100.

[0180] S402. Determine whether the target score is greater than or equal to a preset scoring threshold.

[0181] S403. If the target score is greater than or equal to the preset scoring threshold, then the target cell is determined to be an abnormal cell.

[0182] S404. If the target score is less than the preset scoring threshold, then the target cell is determined not to be an abnormal cell.

[0183] For example, a preset scoring threshold of 50 points is set for determining cell abnormality. The converted target score is then precisely compared with the preset scoring threshold. When the target score is greater than or equal to 50 points, it indicates a high degree of cell abnormality, and the cell is determined to be abnormal. When the target score is less than 50 points, it indicates that the cell is in a stable state or has only a slight, negligible deviation, and the cell is determined to be normal.

[0184] For example, please refer to Figure 2 , Figure 2 This is a schematic diagram of an overall data flow architecture provided in an embodiment of this application. Figure 2 As shown in the figure, the overall data transmission and processing link of the method described in this application embodiment fully covers the specific flow process from vehicle data collection to result display, including: the new energy bus on-board T-box terminal cluster encrypts and uploads the original data through the TCP protocol, which is received by the background Gateway gateway, and then the data is smoothed and distributed through the Kafka message queue and stored in the ODS source data layer; then the data is synchronously transmitted to the DWD data detail layer to complete data cleaning and working condition splitting preprocessing; then it enters the DWS data aggregation layer to perform parallel calculation of five types of algorithms and anomaly scoring; finally, the processing result is pushed to the ADS data display layer to realize the visualization of abnormal results, alarm push and data exposure, without manual intervention, vehicle shutdown and battery disassembly operation.

[0185] In summary, the method described in this application, compared with the prior art, has the following technical advantages and effects: It boasts high accuracy and an extremely low false alarm rate. Abandoning the traditional single-algorithm, single-dimensional detection model, it employs five independent algorithms for multi-dimensional parallel detection, combined with cross-validation logic based on the intersection of abnormal features. Multiple filters address false anomalies caused by operational interference and data fluctuations, completely resolving the high false alarm rate issue of existing technologies and significantly improving anomaly identification accuracy.

[0186] Comprehensive testing with highly reliable results. Covering all vehicle operating conditions—charging, driving, and stationary—it comprehensively tests the health status of battery cells from five core dimensions: voltage distribution, fluctuation stability, cluster deviation, cell synergy, and voltage change rate. It fully covers early latent and severe overt cell faults, completely reproducing the real working state of battery cells in all scenarios, and the test results are comprehensive, reasonable, and reliable.

[0187] Non-invasive testing, high efficiency and low cost of operation and maintenance. Utilizing historical data collected during the normal operation of new energy buses, online real-time testing is achieved without stopping the vehicle or disassembling the battery pack for invasive physical testing. This does not affect normal public transportation operations, significantly saving manpower and material resources for operation and maintenance, and significantly improving the efficiency of battery fault detection and maintenance.

[0188] The method for identifying abnormal cells in the power batteries of new energy buses provided in this application collects voltage time-series data of each cell under all operating conditions, performs parallel multi-dimensional anomaly detection on the cells, determines the discrete score value of the cells using anomaly feature intersection logic, and identifies the abnormal cells among multiple cells based on the target score value corresponding to the discrete score value. This solves the problems of poor reliability of single-dimensional detection, incomplete fault identification, intrusive detection method, and low cost-effectiveness of operation and maintenance in the prior art. It realizes the full-condition, non-intrusive, and highly accurate identification of abnormal cells in the power batteries of new energy buses, effectively avoids false alarms and missed detection of hidden faults, eliminates the need for vehicle downtime and battery pack disassembly, significantly reduces operation and maintenance costs, and improves the efficiency of fault detection and operation and maintenance.

[0189] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a device for identifying abnormal battery cells in a new energy bus power battery, provided in an embodiment of this application. Figure 3 As shown, the identification device 300 includes: The data acquisition module 310 is used to acquire, for multiple target cells in the power battery of the target new energy bus, the voltage time series datasets corresponding to each target cell under various vehicle operating conditions collected by the on-board terminal equipment of the target new energy bus; wherein, the vehicle operating conditions include charging condition, driving condition and stationary condition. The anomaly detection module 320 is used to perform entropy anomaly detection, fluctuation anomaly detection, voltage outlier anomaly detection, correlation anomaly detection and step size anomaly detection on each target cell based on the voltage time series dataset, and to determine the entropy anomaly detection result, fluctuation anomaly detection result, voltage outlier anomaly detection result, correlation anomaly detection result and step size anomaly detection result corresponding to each target cell. The result quantization module 330 is used to determine the discrete score value corresponding to each target cell based on the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, the correlation anomaly detection result, and the step size anomaly detection result, using a preset anomaly feature intersection logic. The anomaly identification module 340 is used to determine the target score value corresponding to each target cell based on the discrete score value, and to identify the abnormal cell among the multiple target cells according to the target score value.

[0190] Furthermore, when determining the entropy anomaly detection result for each target cell, the anomaly detection module 320 is used to: Determine the voltage timing data corresponding to each target cell in the voltage timing dataset; For each target cell, the voltage timing data corresponding to the target cell is divided into window voltage data under multiple sliding windows according to a preset sliding window size, and the window voltage data under each sliding window is divided into multiple interval voltage data according to a preset number of intervals. Determine the voltage occurrence frequency value corresponding to each of the voltage intervals, and based on each voltage occurrence frequency value, determine the information entropy corresponding to each of the sliding windows, so as to determine the information entropy sequence data corresponding to the target cell; Determine the expected value and standard deviation of the information entropy sequence data, and based on the information entropy sequence data, the expected value and the standard deviation, determine the standardized outlier score of the target battery cell in each sliding window; Based on the standardized anomaly score and the preset entropy anomaly detection threshold corresponding to each sliding window, the entropy anomaly detection result corresponding to the target cell is determined; wherein, the entropy anomaly detection result includes one of the following: first entropy anomaly, second entropy anomaly, and no entropy anomaly, and the entropy anomaly detection result also includes an entropy anomaly time set.

[0191] Furthermore, when determining the fluctuation anomaly detection result for each target cell, the anomaly detection module 320 is used to: In the voltage time series dataset, the cell voltage dataset corresponding to each time moment and the voltage time series data corresponding to each target cell are determined respectively; The median of the cell voltage dataset is determined as the voltage reference value at each time point, and for each target cell, the voltage deviation at each time point is determined based on the voltage time series data and the voltage reference value. Based on the voltage deviation at each time point, determine the standard deviation of the deviation for the target cell. Based on the voltage time series data corresponding to each target cell, the average voltage sequence data corresponding to all target cells is determined, and based on the voltage time series data and the average voltage sequence data corresponding to the target cell, the mean square error value corresponding to the target cell is determined. Determine whether the standard deviation of the deviation is greater than a preset standard deviation threshold, and determine whether the mean square error is greater than a preset mean square error threshold; If the deviation standard deviation value is greater than the preset deviation standard deviation threshold, or the mean square error value is greater than the preset mean square error threshold, then the fluctuation anomaly detection result corresponding to the target cell is determined to be that there is a fluctuation anomaly. If the deviation standard deviation is less than or equal to a preset deviation standard deviation threshold, and / or the mean square error is less than or equal to a preset mean square error threshold, then the fluctuation anomaly detection result corresponding to the target cell is determined to be that there is no fluctuation anomaly.

[0192] Furthermore, when determining the voltage outlier detection result for each target cell, the anomaly detection module 320 is used to: In the voltage time series dataset, the cell voltage dataset corresponding to each time moment and the voltage time series data corresponding to each target cell are determined respectively; In the cell voltage dataset corresponding to each time moment, the upper quartile and lower quartile of the voltage are determined respectively, and the difference between the upper quartile and the lower quartile is determined as the voltage interquartile range corresponding to each time moment; Based on the lower voltage quartile and the voltage interquartile range, the lower outlier limit value corresponding to each time moment is determined, and based on the upper voltage quartile and the voltage interquartile range, the upper outlier limit value corresponding to each time moment is determined. For each target cell, the median of the cell voltage dataset is determined as the voltage reference value at each time point. Based on the outlier lower limit, the outlier upper limit, the voltage interquartile range, the voltage reference value, and the voltage time series data corresponding to the target cell, the single-point offset score value of the target cell at each time point is determined. The average value of the single-point offset score corresponding to the target cell at each time moment is determined as the average offset score value corresponding to the target cell, and it is determined whether the average offset score value is greater than or equal to a preset offset score threshold. If the average offset score is greater than or equal to the preset offset score threshold, then the voltage outlier detection result corresponding to the target cell is determined to be a voltage outlier. If the average offset score is less than the preset offset score threshold, then the voltage outlier detection result for the target cell is determined to be that there is no voltage outlier.

[0193] Furthermore, when determining the correlation anomaly detection result corresponding to each of the target cells, the anomaly detection module 320 is used to: Determine the voltage timing data corresponding to each target cell in the voltage timing dataset; For each target cell, the average voltage value corresponding to the target cell is determined based on the voltage timing data corresponding to the target cell. Based on the average voltage value and voltage time series data corresponding to the target cell, and the average voltage value and voltage time series data corresponding to other target cells, the Pearson correlation coefficient between the target cell and each of the other target cells is determined; The average value of the Pearson correlation coefficients between the target cell and each of the other target cells is determined as the correlation score value corresponding to the target cell. Determine whether the relevance score is less than a preset relevance score threshold; If the correlation score is less than the preset correlation score threshold, then the correlation anomaly detection result corresponding to the target battery cell is determined to be an anomaly. If the correlation score is greater than or equal to the preset correlation score threshold, then the correlation anomaly detection result corresponding to the target battery cell is determined to be that there is no correlation anomaly.

[0194] Furthermore, when determining the step-size anomaly detection result for each target cell, the anomaly detection module 320 is used to: In the voltage timing dataset, determine the voltage timing data corresponding to each target cell under the charging condition and the driving condition, respectively; For each target cell, based on the voltage value corresponding to each moment in the voltage timing data, the voltage difference between the voltage value corresponding to each moment and the voltage value corresponding to the adjacent moment before that moment is determined, and the voltage difference is determined as the voltage step size of the target cell at each moment; The median of the voltage steps corresponding to all the target cells at each time point is determined as the reference step size for each time point. Based on the reference step size for each time point and the voltage step size for the target cell at each time point, the extreme value difference multiple for the target cell at each time point is determined. The average value of the extreme difference multiple corresponding to the target cell at each time moment is determined as the average difference multiple corresponding to the target cell, and it is determined whether the average difference multiple is greater than or equal to a preset difference multiple threshold. If the average difference multiple is greater than or equal to the preset difference multiple threshold, then the step size anomaly detection result corresponding to the target cell is determined to be that there is a step size anomaly. If the average difference multiple is less than the preset difference multiple threshold, then the step size anomaly detection result corresponding to the target cell is determined to be that there is no step size anomaly.

[0195] Furthermore, when the result quantization module 330 is used to determine the discrete value corresponding to each target cell based on the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, the correlation anomaly detection result, and the step size anomaly detection result, using a preset anomaly feature intersection logic, the result quantization module 330 is used to: For each target cell, the entropy anomaly detection result, fluctuation anomaly detection result, voltage outlier anomaly detection result, correlation anomaly detection result and step size anomaly detection result corresponding to the target cell are traversed to construct an anomaly feature vector corresponding to the target cell, including each type of detection result; When the abnormal feature vector indicates that the correlation anomaly detection result is abnormal and the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, and the step size anomaly detection result are all normal, the discrete score value corresponding to the target cell is determined to be a first preset score value; wherein, the first preset score value indicates that the anomaly type of the target cell is a simple correlation anomaly. When the abnormal feature vector indicates that there are abnormalities in the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, and the step size anomaly detection result, the discrete score value corresponding to the target cell is determined to be a second preset score value; wherein, the second preset score value indicates that the anomaly type of the target cell is severe consistency collapse; When the abnormal feature vector indicates that there are abnormalities in both the entropy anomaly detection result and the step size anomaly detection result, or when the abnormal feature vector indicates that there are abnormalities in both the entropy anomaly detection result and the fluctuation anomaly detection result, the discrete score value corresponding to the target cell is determined to be a third preset score value; wherein, the third preset score value indicates that the anomaly type of the target cell is entropy step size joint anomaly or entropy fluctuation joint anomaly. When the abnormal feature vector indicates that the correlation anomaly detection result is abnormal and at least one of the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, and the step size anomaly detection result is abnormal, the discrete score value corresponding to the target cell is determined to be a fourth preset score value; wherein, the fourth preset score value indicates that the anomaly type of the target cell is a correlation joint anomaly.

[0196] Furthermore, when the anomaly identification module 340 is used to determine the target score value corresponding to each target cell based on the discrete score value, and to identify the abnormal cell among the multiple target cells according to the target score value, the anomaly identification module 340 is used to: For each target cell, the target score value corresponding to the target cell is determined by using a preset piecewise quadratic function based on the discrete score value corresponding to the target cell. Determine whether the target score is greater than or equal to a preset scoring threshold; If the target score is greater than or equal to a preset scoring threshold, the target cell is determined to be an abnormal cell. If the target score is less than the preset scoring threshold, then the target cell is determined not to be an abnormal cell.

[0197] The device for identifying abnormal cells in the power battery of new energy buses provided in this application collects voltage time-series data of each cell under all operating conditions, performs parallel multi-dimensional anomaly detection on the cells, determines the discrete score value of the cells using the intersection logic of abnormal features, and identifies the abnormal cells among multiple cells based on the target score value corresponding to the discrete score value. This solves the problems of poor reliability of single-dimensional detection, incomplete fault identification, intrusive detection method, and low cost-effectiveness of operation and maintenance in the prior art. It realizes the identification of abnormal cells in the power battery of new energy buses under all operating conditions, non-intrusive, and with high accuracy, effectively avoiding false alarms and missed detection of hidden faults. It does not require vehicle downtime or battery pack disassembly, significantly reduces operation and maintenance costs, and improves the efficiency of fault detection and operation and maintenance.

[0198] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 4 As shown, the electronic device 400 includes a processor 410, a memory 420, and a bus 430.

[0199] The memory 420 stores machine-readable instructions executable by the processor 410. When the electronic device 400 is running, the processor 410 communicates with the memory 420 via the bus 430. When the machine-readable instructions are executed by the processor 410, they can perform the operations described above. Figure 1 The steps of the method for identifying abnormal cells in the power battery of new energy buses in the method embodiment shown are described in detail in the method embodiment, and will not be repeated here.

[0200] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described actions. Figure 1 The steps of the method for identifying abnormal cells in the power battery of new energy buses in the method embodiment shown are described in detail in the method embodiment, and will not be repeated here.

[0201] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0202] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0203] 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 units can be selected to achieve the purpose of this embodiment according to actual needs.

[0204] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0205] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0206] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The scope of protection of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for identifying abnormal battery cells in the power battery of a new energy bus, characterized in that, The identification method includes: For multiple target cells in the power battery of a target new energy bus, voltage time series datasets corresponding to each target cell under various vehicle operating conditions are obtained by the on-board terminal equipment in the target new energy bus; wherein, the vehicle operating conditions include charging condition, driving condition and stationary condition; Based on the voltage time series dataset, entropy anomaly detection, fluctuation anomaly detection, voltage outlier anomaly detection, correlation anomaly detection, and step size anomaly detection are performed on each target cell, and the corresponding entropy anomaly detection results, fluctuation anomaly detection results, voltage outlier anomaly detection results, correlation anomaly detection results, and step size anomaly detection results for each target cell are determined. Based on the entropy anomaly detection results, the fluctuation anomaly detection results, the voltage outlier anomaly detection results, the correlation anomaly detection results, and the step size anomaly detection results, the discrete score value corresponding to each target cell is determined using a preset anomaly feature intersection logic. The step of determining the discrete score value corresponding to each target cell based on the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, the correlation anomaly detection result, and the step size anomaly detection result, using a preset anomaly feature intersection logic, includes: For each target cell, the entropy anomaly detection result, fluctuation anomaly detection result, voltage outlier anomaly detection result, correlation anomaly detection result and step size anomaly detection result corresponding to the target cell are traversed to construct an anomaly feature vector corresponding to the target cell, including each type of detection result; When the abnormal feature vector indicates that the correlation anomaly detection result is abnormal and the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, and the step size anomaly detection result are all normal, the discrete score value corresponding to the target cell is determined to be a first preset score value; wherein, the first preset score value indicates that the anomaly type of the target cell is a simple correlation anomaly. When the abnormal feature vector indicates that there are abnormalities in the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, and the step size anomaly detection result, the discrete score value corresponding to the target cell is determined to be a second preset score value; wherein, the second preset score value indicates that the anomaly type of the target cell is severe consistency collapse; When the abnormal feature vector indicates that there are abnormalities in both the entropy anomaly detection result and the step size anomaly detection result, or when the abnormal feature vector indicates that there are abnormalities in both the entropy anomaly detection result and the fluctuation anomaly detection result, the discrete score value corresponding to the target cell is determined to be a third preset score value; wherein, the third preset score value indicates that the anomaly type of the target cell is entropy step size joint anomaly or entropy fluctuation joint anomaly. When the abnormal feature vector indicates that the correlation anomaly detection result is abnormal and at least one of the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, and the step size anomaly detection result is abnormal, the discrete score value corresponding to the target cell is determined to be a fourth preset score value; wherein, the fourth preset score value indicates that the anomaly type of the target cell is a correlation joint anomaly; Based on the discrete score values, a target score value corresponding to each target cell is determined, and abnormal cells are identified among the multiple target cells according to the target score values.

2. The method according to claim 1, characterized in that, The entropy anomaly detection result for each target battery cell is determined through the following steps: Determine the voltage timing data corresponding to each target cell in the voltage timing dataset; For each target cell, the voltage timing data corresponding to the target cell is divided into window voltage data under multiple sliding windows according to a preset sliding window size, and the window voltage data under each sliding window is divided into multiple interval voltage data according to a preset number of intervals. Determine the voltage occurrence frequency value corresponding to each of the voltage intervals, and based on each voltage occurrence frequency value, determine the information entropy corresponding to each of the sliding windows, so as to determine the information entropy sequence data corresponding to the target cell; Determine the expected value and standard deviation of the information entropy sequence data, and based on the information entropy sequence data, the expected value and the standard deviation, determine the standardized outlier score of the target battery cell in each sliding window; Based on the standardized anomaly score and the preset entropy anomaly detection threshold corresponding to each sliding window, the entropy anomaly detection result corresponding to the target cell is determined; wherein, the entropy anomaly detection result includes one of the following: first entropy anomaly, second entropy anomaly, and no entropy anomaly, and the entropy anomaly detection result also includes an entropy anomaly time set.

3. The method according to claim 1, characterized in that, The fluctuation anomaly detection result for each target battery cell is determined through the following steps: In the voltage time series dataset, the cell voltage dataset corresponding to each time moment and the voltage time series data corresponding to each target cell are determined respectively; The median of the cell voltage dataset is determined as the voltage reference value at each time point, and for each target cell, the voltage deviation at each time point is determined based on the voltage time series data and the voltage reference value. Based on the voltage deviation at each time point, determine the standard deviation of the deviation for the target cell. Based on the voltage time series data corresponding to each target cell, the average voltage sequence data corresponding to all target cells is determined, and based on the voltage time series data and the average voltage sequence data corresponding to the target cell, the mean square error value corresponding to the target cell is determined. Determine whether the standard deviation of the deviation is greater than a preset standard deviation threshold, and determine whether the mean square error is greater than a preset mean square error threshold; If the deviation standard deviation value is greater than the preset deviation standard deviation threshold, or the mean square error value is greater than the preset mean square error threshold, then the fluctuation anomaly detection result corresponding to the target cell is determined to be that there is a fluctuation anomaly. If the deviation standard deviation is less than or equal to a preset deviation standard deviation threshold, and / or the mean square error is less than or equal to a preset mean square error threshold, then the fluctuation anomaly detection result corresponding to the target cell is determined to be that there is no fluctuation anomaly.

4. The method according to claim 1, characterized in that, The voltage outlier detection result for each target cell is determined through the following steps: In the voltage time series dataset, the cell voltage dataset corresponding to each time moment and the voltage time series data corresponding to each target cell are determined respectively; In the cell voltage dataset corresponding to each time moment, the upper quartile and lower quartile of the voltage are determined respectively, and the difference between the upper quartile and the lower quartile is determined as the voltage interquartile range corresponding to each time moment; Based on the lower voltage quartile and the voltage interquartile range, the lower outlier limit value corresponding to each time moment is determined, and based on the upper voltage quartile and the voltage interquartile range, the upper outlier limit value corresponding to each time moment is determined. For each target cell, the median of the cell voltage dataset is determined as the voltage reference value at each time point. Based on the outlier lower limit, the outlier upper limit, the voltage interquartile range, the voltage reference value, and the voltage time series data corresponding to the target cell, the single-point offset score value of the target cell at each time point is determined. The average value of the single-point offset score corresponding to the target cell at each time moment is determined as the average offset score value corresponding to the target cell, and it is determined whether the average offset score value is greater than or equal to a preset offset score threshold. If the average offset score is greater than or equal to the preset offset score threshold, then the voltage outlier detection result corresponding to the target cell is determined to be a voltage outlier. If the average offset score is less than the preset offset score threshold, then the voltage outlier detection result for the target cell is determined to be that there is no voltage outlier.

5. The method according to claim 1, characterized in that, The correlation anomaly detection result for each target battery cell is determined through the following steps: Determine the voltage timing data corresponding to each target cell in the voltage timing dataset; For each target cell, the average voltage value corresponding to the target cell is determined based on the voltage timing data corresponding to the target cell. Based on the average voltage value and voltage time series data corresponding to the target cell, and the average voltage value and voltage time series data corresponding to other target cells, the Pearson correlation coefficient between the target cell and each of the other target cells is determined; The average value of the Pearson correlation coefficients between the target cell and each of the other target cells is determined as the correlation score value corresponding to the target cell. Determine whether the relevance score is less than a preset relevance score threshold; If the correlation score is less than the preset correlation score threshold, then the correlation anomaly detection result corresponding to the target battery cell is determined to be an anomaly. If the correlation score is greater than or equal to the preset correlation score threshold, then the correlation anomaly detection result corresponding to the target battery cell is determined to be that there is no correlation anomaly.

6. The method according to claim 1, characterized in that, The step size anomaly detection result for each target cell is determined through the following steps: In the voltage timing dataset, determine the voltage timing data corresponding to each target cell under the charging condition and the driving condition, respectively; For each target cell, based on the voltage value corresponding to each moment in the voltage timing data, the voltage difference between the voltage value corresponding to each moment and the voltage value corresponding to the adjacent moment before that moment is determined, and the voltage difference is determined as the voltage step size of the target cell at each moment; The median of the voltage steps corresponding to all the target cells at each time point is determined as the reference step size for each time point. Based on the reference step size for each time point and the voltage step size for the target cell at each time point, the extreme value difference multiple for the target cell at each time point is determined. The average value of the extreme difference multiple corresponding to the target cell at each time moment is determined as the average difference multiple corresponding to the target cell, and it is determined whether the average difference multiple is greater than or equal to a preset difference multiple threshold. If the average difference multiple is greater than or equal to the preset difference multiple threshold, then the step size anomaly detection result corresponding to the target cell is determined to be that there is a step size anomaly. If the average difference multiple is less than the preset difference multiple threshold, then the step size anomaly detection result corresponding to the target cell is determined to be that there is no step size anomaly.

7. The method according to claim 1, characterized in that, The step of determining the target score value corresponding to each target cell based on the discrete score value, and identifying abnormal cells among multiple target cells according to the target score value, includes: For each target cell, the target score value corresponding to the target cell is determined by using a preset piecewise quadratic function based on the discrete score value corresponding to the target cell. Determine whether the target score is greater than or equal to a preset scoring threshold; If the target score is greater than or equal to a preset scoring threshold, the target cell is determined to be an abnormal cell. If the target score is less than the preset scoring threshold, then the target cell is determined not to be an abnormal cell.

8. A device for identifying abnormal cells in the power battery of a new energy bus, characterized in that, The identification device includes: The data acquisition module is used to acquire voltage time series datasets for each target cell in the power battery of the target new energy bus under various vehicle operating conditions, collected by the on-board terminal equipment of the target new energy bus; wherein, the vehicle operating conditions include charging condition, driving condition and stationary condition. An anomaly detection module is used to perform entropy anomaly detection, fluctuation anomaly detection, voltage outlier anomaly detection, correlation anomaly detection, and step size anomaly detection on each target cell based on the voltage time series dataset, and to determine the corresponding entropy anomaly detection result, fluctuation anomaly detection result, voltage outlier anomaly detection result, correlation anomaly detection result, and step size anomaly detection result for each target cell. The result quantization module is used to determine the discrete score value corresponding to each target cell based on the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, the correlation anomaly detection result, and the step size anomaly detection result, using a preset anomaly feature intersection logic. An anomaly identification module is used to determine the target score value corresponding to each target cell based on the discrete score value, and to identify the abnormal cell among the multiple target cells according to the target score value. When the result quantization module is used to determine the discrete score value corresponding to each target cell based on the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, the correlation anomaly detection result, and the step size anomaly detection result, using a preset anomaly feature intersection logic, the result quantization module is used to: For each target cell, the entropy anomaly detection result, fluctuation anomaly detection result, voltage outlier anomaly detection result, correlation anomaly detection result and step size anomaly detection result corresponding to the target cell are traversed to construct an anomaly feature vector corresponding to the target cell, including each type of detection result; When the abnormal feature vector indicates that the correlation anomaly detection result is abnormal and the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, and the step size anomaly detection result are all normal, the discrete score value corresponding to the target cell is determined to be a first preset score value; wherein, the first preset score value indicates that the anomaly type of the target cell is a simple correlation anomaly. When the abnormal feature vector indicates that there are abnormalities in the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, and the step size anomaly detection result, the discrete score value corresponding to the target cell is determined to be a second preset score value; wherein, the second preset score value indicates that the anomaly type of the target cell is severe consistency collapse; When the abnormal feature vector indicates that there are abnormalities in both the entropy anomaly detection result and the step size anomaly detection result, or when the abnormal feature vector indicates that there are abnormalities in both the entropy anomaly detection result and the fluctuation anomaly detection result, the discrete score value corresponding to the target cell is determined to be a third preset score value; wherein, the third preset score value indicates that the anomaly type of the target cell is entropy step size joint anomaly or entropy fluctuation joint anomaly. When the abnormal feature vector indicates that the correlation anomaly detection result is abnormal and at least one of the entropy anomaly detection result, the fluctuation anomaly detection result, the voltage outlier anomaly detection result, and the step size anomaly detection result is abnormal, the discrete score value corresponding to the target cell is determined to be a fourth preset score value; wherein, the fourth preset score value indicates that the anomaly type of the target cell is a correlation joint anomaly.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. The machine-readable instructions are executed by the processor to perform the steps of the method for identifying abnormal cells in the power battery of a new energy bus as described in any one of claims 1 to 7.

Images