A method for detecting an abnormality in soc of a battery pack

Abstract

The application provides an SOC abnormality detection method of a battery pack, comprising the following steps: S1: obtaining a plurality of charging and discharging operation data segments based on historical actual operation data of the battery pack; S2: calculating a sequence corresponding to time and an actual available capacity value under each charging and discharging operation data segment; S3: inputting the plurality of sequences corresponding to time and the actual available capacity value into a mutation detection algorithm and a cumulative change detection algorithm in a time sequence and in a sliding window mode, obtaining an SOC abnormality time point through the mutation detection algorithm, and obtaining an SOC abnormal window segment through the cumulative change detection algorithm.

Description

Technical Field

[0001] This application relates to the field of energy storage, and in particular to a method for detecting SOC anomalies in battery packs. Background Technology

[0002] In the field of power electronics, battery packs are widely used as energy storage components in various scenarios, such as electric vehicles, portable electronic devices, and energy storage applications.

[0003] The battery pack consists of multiple individual battery cells connected in series, parallel, or series-parallel. It also includes a Battery Management System (BMS), which manages and monitors the battery pack, such as acquiring data at regular sampling intervals including the total battery pack voltage, highest individual cell voltage, lowest individual cell voltage, battery pack current, highest temperature, lowest temperature, cumulative charge capacity, and state of charge (SOC).

[0004] The State of Charge (SOC) of a battery refers to the ratio of its remaining capacity to its capacity when fully charged. SOC is crucial for battery energy management, safety monitoring, and health maintenance, so it is essential to estimate SOC.

[0005] Current methods focus on how to accurately estimate SOC, but there are few methods for anomaly detection of the estimated SOC. Even if there are, they have many problems, such as strong parameter dependence, high computational complexity, high resource consumption, and poor performance. Summary of the Invention

[0006] According to one embodiment, this application provides a method for detecting SOC anomalies in a battery pack, comprising: S1: acquiring multiple charging and discharging operation data segments based on historical actual operating data of the battery pack; S2: calculating a sequence corresponding to the time and actual available capacity value under each charging and discharging operation data segment based on the multiple charging and discharging operation data segments; S3: inputting the multiple sequences corresponding to the time and actual available capacity value into a sudden change detection algorithm and a cumulative change detection algorithm in chronological order and using a sliding window method, obtaining the SOC anomaly time point through the sudden change detection algorithm, and obtaining the window segment of SOC anomaly through the cumulative change detection algorithm.

[0007] Furthermore, step S1 also includes: S11: The cloud computing platform obtains the historical actual operation data of the battery pack from the battery management system of the battery pack to form a database to be processed; S12: Obtain the continuous charging and discharging operation data of the same battery pack from the database to be processed; S13: Divide the charging and discharging operation data into multiple charging and discharging operation data segments according to the time column of the charging and discharging operation data.

[0008] Furthermore, the historical actual operating data in step S1 is a database corresponding to the sampling time, battery pack SOC, and battery pack throughput.

[0009] Furthermore, in step S11, the database to be processed in the cloud computing platform is a database corresponding to time and battery pack SOC and battery pack throughput.

[0010] Furthermore, in step S13, the charging and discharging operation data is divided into multiple charging and discharging operation data segments by dividing the charging and discharging operation data into multiple charging and discharging operation data segments by identifying the time positions in the time series of the charging and discharging operation data where the time interval between consecutive points exceeds a threshold time, and using these time positions as cutting points, the charging and discharging operation data is divided into multiple charging and discharging operation data segments.

[0011] Furthermore, step S2 also includes: S21: obtaining each time in the time column of each charging and discharging operation data segment and the corresponding SOC and throughput, calculating the difference in SOC value and the difference in throughput between adjacent times in each charging and discharging operation data segment, forming a sequence corresponding to the time-SOC difference ΔSOC and the time-throughput difference ΔET under each charging and discharging operation data segment; S22: calculating the actual available capacity corresponding to each time in each charging and discharging operation data segment according to the formula Cr=ΔET / ΔSOC, forming a sequence corresponding to the time and actual available capacity value under each charging and discharging operation data segment.

[0012] Furthermore, between steps S21 and S22, the following step is also included: Step S211: Find out whether there is a ΔSOC less than a preset value in the sequence corresponding to the time and SOC difference ΔSOC under each charging and discharging operation data segment. If so, remove the ΔSOC corresponding to that time, and at the same time remove the ΔET corresponding to that time in the sequence corresponding to the time and throughput difference ΔET.

[0013] Furthermore, step S3 also includes: S31: selecting the window length; S32: inputting multiple sequences corresponding to the time and actual available capacity values ​​into the mutation detection algorithm and the cumulative change detection algorithm in chronological order according to the window length; S33: the mutation detection algorithm outputs the SOC abnormal time point, and the cumulative change detection algorithm outputs the window segment of the SOC abnormality.

[0014] Furthermore, step S32 involves: dividing the data from the first time to the nth time in the sequence corresponding to the multiple times and the actual available capacity values ​​into a window segment, where n is the window length; dividing the data from the second time to the (n+1)th time into a window segment; dividing the data from the third time to the (n+2)th time into a window segment; ...; dividing the data from the mth time to the (n+m-1)th time into a window segment, where m is a natural number greater than 1, and then inputting it into the mutation detection algorithm and the cumulative change detection algorithm.

[0015] Furthermore, in step S33, the mutation detection algorithm performs the following: remove the maximum and minimum values ​​within each window segment, then calculate the mean and standard deviation of the window segment, calculate the interval of three times the standard deviation of the window segment based on the mean and standard deviation, determine whether the actual available capacity value corresponding to the next time of the window segment is within the interval of three times the standard deviation, if not, then it is considered that the SOC corresponding to the next time of the window segment has become abnormal, and the mutation detection algorithm outputs the SOC abnormal time point.

[0016] Furthermore, in step S32, the cumulative change detection algorithm performs the following steps: remove the maximum and minimum values ​​within each window segment, then calculate the average value of the window segment, sum the negative differences between all actual available capacity values ​​and the average value within the window segment to obtain the cumulative deviation sum; determine whether the cumulative deviation sum is greater than or equal to a set threshold, and if so, consider that there is an abnormal change in SOC within the window segment, and the cumulative change detection algorithm outputs the window segment with abnormal SOC change.

[0017] The features and technical advantages of this disclosure have been outlined quite extensively above to facilitate a better understanding of the detailed description that follows. Additional features and advantages of this disclosure, which form the subject matter of the claims, will be described below. Those skilled in the art will understand that the disclosed concepts and specific embodiments can be readily used as the basis for modifying or designing other structures or processes for achieving the same purpose as this disclosure. Those skilled in the art will also recognize that such equivalent structures do not depart from the spirit and scope of this disclosure as set forth in the appended claims. Attached Figure Description

[0018] To gain a more complete understanding of this disclosure and its advantages, the following description is provided in conjunction with the accompanying drawings, wherein:

[0019] Figure 1 A schematic flowchart of a battery pack SOC anomaly detection method according to an embodiment of this application is shown;

[0020] Figure 2 This paper illustrates a flowchart of an embodiment of the present application for obtaining multiple charge and discharge operation data segments based on historical actual operation data of a battery pack;

[0021] Figure 3 This illustration shows a flowchart of an embodiment of the present application, which calculates the sequence of time and actual available capacity value corresponding to each charge and discharge operation data segment based on the plurality of charge and discharge operation data segments;

[0022] Figure 4 The illustration shows a flowchart of an embodiment of this application, in which multiple sequences corresponding to the times and actual available capacity values ​​are input into a mutation detection algorithm and a cumulative change detection algorithm in chronological order and using a sliding window method. The mutation detection algorithm is used to obtain the SOC abnormal time point, and the cumulative change detection algorithm is used to obtain the window segment of SOC abnormality.

[0023] Unless otherwise stated, corresponding numbers and symbols in the various figures generally refer to corresponding parts. These figures are drawn to clearly illustrate relevant aspects of the various embodiments and are not necessarily drawn to scale. Detailed Implementation

[0024] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] One embodiment of this application provides a method for detecting SOC anomalies in a battery pack. Please refer to [link / reference]. Figure 1 The schematic diagram shown is a flowchart of a battery pack SOC anomaly detection method according to an embodiment of this application. The battery pack SOC anomaly detection method proposed in this application includes:

[0026] S1: Obtain multiple charging and discharging operation data segments based on the historical actual operation data of the battery pack;

[0027] S2: Based on the multiple charging and discharging operation data segments, calculate the sequence corresponding to the time and actual available capacity value under each charging and discharging operation data segment;

[0028] S3: Input the sequences corresponding to the times and actual available capacity values ​​into the mutation detection algorithm and the cumulative change detection algorithm in chronological order and in a sliding window manner. The mutation detection algorithm obtains the SOC abnormal time point, and the cumulative change detection algorithm obtains the SOC abnormal window segment.

[0029] Please see Figure 2The flowchart shown is a schematic diagram of an embodiment of this application for obtaining multiple charge and discharge operation data segments based on the historical actual operation data of the battery pack. Specifically, step S1 includes: S11: The cloud computing platform obtains the historical actual operation data of the battery pack from the battery management system of the battery pack to form a database to be processed; S12: Obtain the charge and discharge operation data of the same battery pack running continuously from the database to be processed; S13: Divide the charge and discharge operation data into multiple charge and discharge operation data segments by dividing the time sequence of the charge and discharge operation data.

[0030] For step S11, specifically, during the actual operation of the battery pack, the battery management system acquires parameters including but not limited to battery pack voltage, current, temperature, cumulative charge capacity, and SOC at certain sampling intervals (e.g., 10 seconds, but not limited to this, depending on different application scenarios). This data constitutes the battery pack's historical operational data. Therefore, the historical operational data in step S1 represents the actual data generated during the battery pack's historical operation in the real world. In one embodiment of this application, the historical operational data of the battery pack is a database corresponding to sampling time, battery pack SOC, and battery pack throughput.

[0031] In practical applications, the cloud computing platform obtains and stores historical actual operation data from the battery management system. Due to factors such as network transmission and cloud computing platform storage, the time intervals in the time column of the database to be processed in the cloud computing platform may no longer be the sampling interval of the battery management system. Thus, the database to be processed in the cloud computing platform is a database corresponding to time and battery pack SOC and battery pack throughput.

[0032] Specifically, in practical applications, due to factors such as network transmission and cloud computing platform storage, the time intervals between adjacent data in the database to be processed obtained and stored by the cloud computing platform from the battery management system may be relatively long, such as exceeding the threshold time (e.g., ten minutes). If there is no corresponding data to be processed within such a long period of time, it is impossible to know whether the battery pack is in a charging state, a discharging state, or a state of charging and discharging simultaneously. This may result in changes in the battery pack's throughput, but no change in the battery pack's SOC, which will affect the effectiveness of battery pack SOC anomaly detection. Therefore, it is necessary to process the segments with long intervals and exclude them from subsequent calculations, that is, it is necessary to process the charging and discharging operation data obtained in step S12.

[0033] Specifically, in step S13, dividing the charging and discharging operation data into multiple charging and discharging operation data segments based on the time sequence involves identifying time positions in the time sequence where the time interval between consecutive segments exceeds a threshold time, and using these time positions as cutting points to divide the charging and discharging operation data into multiple charging and discharging operation data segments. Thus, each charging and discharging operation data segment corresponds to a time interval with the battery pack's SOC and battery pack throughput, and the time interval between consecutive segments does not exceed the threshold time. In one embodiment, the threshold time is ten minutes, but this application does not limit its specific value and it can be set based on experience.

[0034] In one embodiment of this application, in step S13, Pandas is used to identify the time positions in the time series of the charging and discharging operation data where the time interval between consecutive points exceeds a threshold time. Using these time positions as cutting points, the charging and discharging operation data is divided into multiple charging and discharging operation data segments. However, this application does not limit the specific method used; any applicable method can be applied to this application.

[0035] For example, if the charging and discharging operation data in the cloud computing platform (which only shows the time column and does not show the battery pack SOC and battery pack throughput corresponding to the time column) is as shown in Table 1.

[0036] Table 1

[0037] Time = [

[0038] 2025 / 4 / 8 0:02

[0039] 2025 / 4 / 8 0:07

[0040] 2025 / 4 / 8 0:11

[0041] 2025 / 4 / 8 0:16

[0042] 2025 / 4 / 8 0:20

[0043] 2025 / 4 / 8 0:25

[0044] 2025 / 4 / 8 0:38

[0045] 2025 / 4 / 8 0:47

[0046] 2025 / 4 / 8 0:56

[0047] 2025 / 4 / 8 1:05

[0048] 2025 / 4 / 8 1:10

[0049] 2025 / 4 / 8 1:14

[0050] 2025 / 4 / 8 1:19 ]

[0052] If the threshold time is ten minutes, then the time interval between 0:25 on April 8, 2025 and 0:38 on April 8, 2025 can be identified as thirteen minutes, which is greater than the threshold time (e.g., ten minutes). Therefore, this position can be used as the cutting point to divide the charging and discharging operation data shown in Table 1 (which only shows the time column and does not show the battery pack SOC and battery pack throughput corresponding to the time column) into two charging and discharging operation data segments (which only show the time column and do not show the battery pack SOC and battery pack throughput corresponding to the time column), namely Table 2 and Table 3.

[0053] Table 2

[0054] Time = [

[0055] 2025 / 4 / 8 0:02

[0056] 2025 / 4 / 8 0:07

[0057] 2025 / 4 / 8 0:11

[0058] 2025 / 4 / 8 0:16

[0059] 2025 / 4 / 8 0:20

[0060] 2025 / 4 / 8 0:25 ]

[0062] Table 3

[0063] Time = [

[0064] 2025 / 4 / 8 0:38

[0065] 2025 / 4 / 8 0:47

[0066] 2025 / 4 / 8 0:56

[0067] 2025 / 4 / 8 1:05

[0068] 2025 / 4 / 8 1:10

[0069] 2025 / 4 / 8 1:14

[0070] 2025 / 4 / 8 1:19 ]

[0072] In this way, uncertain data fragments are eliminated, improving the effectiveness of battery pack SOC anomaly detection.

[0073] Please see Figure 3 The flowchart shown is a schematic diagram of an embodiment of this application, illustrating the calculation of the sequence corresponding to the time and actual available capacity value under each charging and discharging operation data segment based on the plurality of charging and discharging operation data segments. Specifically, step S2 includes: S21: obtaining each time in the time column of each charging and discharging operation data segment and the SOC and throughput corresponding to that time, calculating the difference in SOC value and the difference in throughput between adjacent times in each charging and discharging operation data segment, forming the sequence corresponding to the time-SOC difference ΔSOC and the sequence corresponding to the time-throughput difference ΔET under each charging and discharging operation data segment; S22: calculating the actual available capacity (realCapacity) corresponding to each time under each charging and discharging operation data segment according to the formula Cr=ΔET / ΔSOC, forming the sequence corresponding to the time and actual available capacity value under each charging and discharging operation segment.

[0074] For step S21, for example, in one of the charging and discharging operation data segments, the SOC and throughput ET (EnergyThroughput) corresponding to each time in the time column are: SOC = [0,10,20,21,40,50,60,70,80,90,100] (percentage); ET = [0,0.5,1.0,1.5,2.0,2.5,3.0,3.5,4.0,4.5,5.0] (unit: kWh), using the first-order difference formula: Δx t =x t -x {t-1} The sequence corresponding to the time-SOC difference ΔSOC under this charging and discharging operation data segment can be calculated as ΔSOC=[10,10,1,19,10,10,10,10,10,10], and the sequence corresponding to the time-throughput difference ΔET under this charging and discharging operation data segment can be calculated as ΔET=[0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5] (unit: kWh).

[0075] In step S22, the actual available capacity corresponding to each time point in the charging and discharging operation data segment can be calculated according to the formula Cr=ΔET / ΔSOC, forming a sequence Cr=[5,5,50,2.6,5,5,5,5,5,5,5] corresponding to the time and actual available capacity values ​​in the charging and discharging operation data segment.

[0076] This yields a sequence of time and actual available capacity values ​​for each charging and discharging operation data segment, which is then stored on a cloud computing platform.

[0077] In another embodiment of this application, step S211 is further included between step S21 and step S22: find whether there is a ΔSOC less than a preset value in the sequence corresponding to the time and SOC difference ΔSOC under each charging and discharging operation data segment. If so, remove the ΔSOC corresponding to that time and at the same time remove the ΔET corresponding to that time in the sequence corresponding to the time and throughput difference ΔET.

[0078] In one embodiment, a preset value of 2 (percentage) can be taken, but this application does not limit its specific value.

[0079] As described above, in step S22, the actual available capacity at each time point under each charging and discharging operation data segment is calculated according to the formula Cr = ΔET / ΔSOC. Since ΔSOC is used as the denominator in the calculation of the actual available capacity, and both ΔET and ΔSOC are affected by accuracy, when the accuracy is not high enough, the smaller the ΔSOC in the denominator, the greater the impact on the accuracy of the calculated actual available capacity Cr. Furthermore, due to the influence of SOC accuracy, the cloud computing platform only retains integer digits, so the change in SOC is very small when the charging and discharging current is very small, which can easily affect the calculation results. Therefore, to improve the accuracy of the calculated actual available capacity Cr, it is necessary to remove excessively small ΔSOC and simultaneously remove the corresponding ΔET.

[0080] Thus, a sequence of multiple time periods and actual available capacity values ​​corresponding to multiple charging and discharging operation data segments was obtained.

[0081] Please see Figure 4 The flowchart shown is an embodiment of this application illustrating the process of inputting multiple sequences corresponding to the actual available capacity values ​​at different times into a mutation detection algorithm and a cumulative change detection algorithm in chronological order and using a sliding window method. The mutation detection algorithm obtains the SOC abnormal time point, and the cumulative change detection algorithm obtains the SOC abnormal window segment. Specifically, step S3 includes: S31: selecting the window length; S32: inputting multiple sequences corresponding to the actual available capacity values ​​at different times into the mutation detection algorithm and the cumulative change detection algorithm in chronological order and according to the window length; S33: the mutation detection algorithm outputs the SOC abnormal time point, and the cumulative change detection algorithm outputs the SOC abnormal window segment.

[0082] For step S31, the window length can be set according to actual needs. For example, the ten time points in the sequence corresponding to the time and the actual available capacity value and their corresponding actual available capacity values ​​are divided into a window to form data within a window segment.

[0083] For step S32, specifically: the data from the first time to the nth time in the sequence corresponding to the multiple times and the actual available capacity values ​​are divided into a window segment, where n is the window length; the data from the second time to the (n+1)th time are divided into a window segment; the data from the third time to the (n+2)th time are divided into a window segment; ...; the data from the mth time to the (n+m-1)th time are divided into a window segment, where m is a natural number greater than 1, and then input into the mutation detection algorithm and the cumulative change detection algorithm.

[0084] Specifically, for step S33, the mutation detection algorithm performs the following: remove the maximum and minimum values ​​within each window segment, then calculate the mean and standard deviation of the window segment, calculate the interval of three times the standard deviation of the window segment based on the mean and standard deviation, determine whether the actual available capacity value corresponding to the next time of the window segment is within the interval of three times the standard deviation, if not, it is considered that the SOC corresponding to the next time of the window segment is abnormal, and so on, sequentially detect whether the SOC corresponding to each time point after the first window segment is abnormal, and then the mutation detection algorithm outputs the time point of SOC abnormality.

[0085] For example, a window segment of length 10 corresponds to the sequence of actual available capacity values ​​[10, 9, 8, 7, 8, 9, 9, 9, 8, 10] (unit: kWh). The actual available capacity value at the time point after this window segment is 2 kWh. A mutation detection algorithm removes the maximum value of 10 and the minimum value of 7 within this window segment. The average value of the remaining data in this window segment is calculated to be approximately 8.57, and the standard deviation is approximately 0.5. Therefore, the interval three times the standard deviation is [7.07, 10.07]. Thus, the actual available capacity value of 2 kWh at the next time point after this window segment is not within the three-standard-deviation interval.

[0086] Within [7.07, 10.07], the SOC corresponding to the time point after this window segment is considered abnormal, and the abnormal time point is output.

[0087] Specifically, for step S32, the cumulative change detection algorithm performs the following steps: remove the maximum and minimum values ​​within each window segment, then calculate the average value of the window segment, sum the negative differences between all actual available capacity values ​​and the average value within the window segment to obtain the cumulative deviation sum; determine whether the cumulative deviation sum is greater than or equal to a set threshold, if so, then it is considered that there is an abnormal change in SOC within the window segment, and so on, in order to detect whether there is an abnormal change in SOC within each divided window segment, and then the cumulative change detection algorithm outputs the window segment with abnormal SOC change.

[0088] Furthermore, the cumulative change detection algorithm also outputs the time point corresponding to the actual available capacity value that is less than the average value within each window segment, as the SOC abnormal time point.

[0089] For example, a window segment of length 10 corresponds to the actual available capacity value sequence [10,9,8,7,6,5,4,3,2,2] (unit: kWh), with a threshold of 8 kWh. The cumulative change detection algorithm removes the maximum value of 10 and the minimum value of 2 within the window segment, then calculates the average value of the remaining data in the window segment to be approximately 5.5. Then, it accumulates the negative differences between all actual available capacity values ​​within the window segment and the average value of 5.5, resulting in a cumulative deviation sum of 11.5. Since this cumulative deviation sum of 11.5 is greater than the set threshold of 8, it is considered that there is an abnormal change in SOC within the window segment, and the window segment is output. Furthermore, based on the time sequence, it is determined whether the actual available capacity value within the window segment is less than the average value. The time point corresponding to the actual available capacity value within the window segment that is less than the average value is taken as the SOC anomaly start time point. Subsequently, once an actual available capacity value greater than or equal to the average value appears, the time point corresponding to the actual available capacity value greater than or equal to the average value is taken as the SOC anomaly end time point. If no actual available capacity value greater than or equal to the average value appears by the end of the window segment, the end time point of the window segment is taken as the SOC anomaly end time point. The time segment from the SOC anomaly start time point to the SOC anomaly end time point is output as the SOC anomaly time segment. As mentioned above, the time point corresponding to the actual available capacity value 5 is taken as the SOC anomaly start time point. The actual available capacity values ​​after the time point corresponding to the actual available capacity value 5 until the end time point of the window segment are all less than the average value of 5.5. Therefore, the end time point of the window segment is taken as the SOC anomaly end time point. That is, the segment formed by the time point corresponding to the actual available capacity value 5 and the end time point of the window segment is output as the SOC anomaly time segment.

[0090] The aforementioned threshold can be set manually based on experience, and ΔSOC can be dynamically bound to the actual available capacity as described above. The threshold for judging jumps under different cells and different nominal capacities can be corrected based on the accuracy of the SOC anomaly detection method of the battery pack described above. This solves the misjudgment problem caused by the excessive coupling between ΔSOC and rated capacity, improves the accuracy of SOC anomaly detection, and has strong applicability.

[0091] Thus, the SOC anomaly detection method for battery packs provided in this application can locate SOC anomaly points and segments. It is evident that this application does not determine anomalies based on SOC accuracy, but rather dynamically links ΔSOC to the actual usable capacity. This is because the actual usable capacity is dynamically changing, influenced by the actual application scenario, type, and voltage difference of the battery pack, causing SOC to also change. Furthermore, abnormal changes in SOC will be reflected in the calculated actual usable capacity. Therefore, when the actual usable capacity experiences abnormal fluctuations, it can be determined that an abnormal change in SOC has occurred, and thus, the existence of an SOC anomaly can be determined based on the actual usable capacity. It is clear that this application, by processing the SOC calculated using any algorithm and determining whether the SOC accurately follows the dynamic changes in actual usable capacity, makes anomaly detection more accurate.

[0092] In practical applications, the output results of mutation detection algorithms and cumulative change detection algorithms are saved to the cloud computing platform to provide data support for subsequent monitoring, alarms, and in-depth analysis.

[0093] Furthermore, the SOC anomaly detection method for battery packs described in this application is adaptable to application scenarios of different battery products, such as micro-storage, residential energy storage, industrial and commercial energy storage, and power batteries. Regardless of whether the battery cell types and specifications are consistent, it can be seamlessly deployed without configuring additional parameters. It is also applicable to different charging and discharging modes in battery packs, such as shallow charging and shallow discharging, deep charging and deep discharging, and other complex operating conditions, thus exhibiting strong adaptability.

[0094] In practical applications, the SOC anomaly detection method for battery packs provided in this application is executed in real time or offline by a cloud computing platform, which then sends the anomaly analysis results. The cloud computing platform has high memory and computational efficiency, making it suitable for model training with large amounts of data. In practical applications, the battery management system collects battery pack operating data during operation; therefore, the cloud computing platform obtains the aforementioned historical operating data from the battery management system.

[0095] Although embodiments of the present disclosure and their advantages have been described in detail, it should be understood that various changes, substitutions and alterations may be made herein without departing from the spirit and scope of the disclosure as defined by the appended claims.

[0096] Furthermore, the scope of this application is not intended to be limited to the specific embodiments of the processes, machines, manufactures, compositions of matter, apparatuses, methods, and steps described in the specification. As will be readily understood by those skilled in the art from the disclosure of this publication, processes, machines, manufactures, compositions of matter, means, methods, or steps that perform substantially the same function, currently exist or will be developed or implemented thereafter, will yield substantially the same results as the corresponding embodiments described herein that are available according to this disclosure. Therefore, the appended claims are intended to include such processes, machines, manufactures, compositions of matter, apparatuses, methods, or steps within their scope.

Claims

1. A method for detecting SOC anomalies in a battery pack, characterized in that, include: S1: Obtain multiple charging and discharging operation data segments based on the historical actual operation data of the battery pack; S2: Based on the multiple charging and discharging operation data segments, calculate the sequence corresponding to the time and actual available capacity value under each charging and discharging operation data segment; S3: Input the sequences corresponding to the times and actual available capacity values ​​into the mutation detection algorithm and the cumulative change detection algorithm in chronological order and in a sliding window manner. The mutation detection algorithm is used to obtain the SOC abnormal time point, and the cumulative change detection algorithm is used to obtain the SOC abnormal window segment. The mutation detection algorithm performs the following steps: First, it removes the maximum and minimum values ​​within each window segment. Then, it calculates the mean and standard deviation of that window segment. Based on the mean and standard deviation, it calculates a three-standard-deviation interval for that window segment. It then determines whether the actual available capacity value corresponding to the next time point of that window segment is within this three-standard-deviation interval. If not, it considers the SOC corresponding to the next time point of that window segment to be abnormal, and the mutation detection algorithm outputs the SOC abnormality time point. The cumulative change detection algorithm performs the following steps: First, it removes the maximum and minimum values ​​within each window segment. Then, it calculates the mean of that window segment. It accumulates the negative differences between all actual available capacity values ​​and the mean within that window segment to obtain the cumulative deviation sum. It then determines whether the cumulative deviation sum is greater than or equal to a set threshold. If so, it considers there is an abnormal change in SOC within that window segment, and the cumulative change detection algorithm outputs the window segment with the abnormal SOC change.

2. The SOC anomaly detection method for a battery pack according to claim 1, characterized in that, Step S1 also includes: S11: The cloud computing platform obtains the historical actual operating data of the battery pack from the battery management system of the battery pack and forms a database to be processed; S12: Obtain the continuous charging and discharging operation data of the same battery pack from the database to be processed; S13: Divide the charging and discharging operation data into multiple charging and discharging operation data segments by dividing the charging and discharging operation data into time series.

3. The SOC anomaly detection method for a battery pack according to claim 2, characterized in that, The historical actual operating data in step S1 is a database corresponding to the sampling time, battery pack SOC, and battery pack throughput.

4. The SOC anomaly detection method for a battery pack according to claim 2, characterized in that, The database to be processed in the cloud computing platform in step S11 is a database corresponding to time, battery pack SOC, and battery pack throughput.

5. The SOC anomaly detection method for a battery pack according to claim 2, characterized in that, In step S13, the charging and discharging operation data is divided into multiple charging and discharging operation data segments based on the time sequence of the charging and discharging operation data, as follows: Identify the time positions in the time series of the charging and discharging operation data where the time interval between consecutive points exceeds a threshold, and use these time positions as cutting points to divide the charging and discharging operation data into multiple charging and discharging operation data segments.

6. The SOC anomaly detection method for a battery pack according to claim 2, characterized in that, Step S2 also includes: S21: Obtain each time and the corresponding SOC and throughput in the time column of each charging and discharging operation data segment; calculate the difference in SOC value and the difference in throughput between adjacent times in each charging and discharging operation data segment; and form the sequence corresponding to the time-SOC difference ΔSOC and the time-throughput difference ΔET under each charging and discharging operation data segment. S22: Calculate the actual available capacity corresponding to each time point under each charge and discharge operation data segment according to the formula Cr=ΔET / ΔSOC, forming a sequence of time and actual available capacity values ​​under each charge and discharge operation data segment.

7. The SOC anomaly detection method for a battery pack according to claim 6, characterized in that, Between step S21 and step S22, the following is also included: Step S211: Find whether there is a ΔSOC less than a preset value in the sequence corresponding to the time and SOC difference ΔSOC under each charging and discharging operation data segment. If so, remove the ΔSOC corresponding to that time and at the same time remove the ΔET corresponding to that time in the sequence corresponding to the time and throughput difference ΔET.

8. The SOC anomaly detection method for a battery pack according to claim 6, characterized in that, Step S3 also includes: S31: Select window length; S32: Input the multiple sequences corresponding to the time and actual available capacity values ​​into the mutation detection algorithm and the cumulative change detection algorithm in chronological order according to the window length; S33: The mutation detection algorithm outputs the SOC abnormal time point, and the cumulative change detection algorithm outputs the SOC abnormal window segment.

9. The SOC anomaly detection method for a battery pack according to claim 8, characterized in that, Step S32 is as follows: Divide the data from the first time to the nth time in the sequence corresponding to the multiple times and the actual available capacity values ​​into a window segment, where n is the window length; divide the data from the second time to the (n+1)th time into a window segment; divide the data from the third time to the (n+2)th time into a window segment; ...; divide the data from the mth time to the (n+m-1)th time into a window segment, where m is a natural number greater than 1, and then input it into the mutation detection algorithm and the cumulative change detection algorithm.

Images