A data mining-based electric vehicle battery system defect identification method

Abstract

The application provides a data mining-based electric vehicle battery system defect identification method, which firstly prepares different vehicle data and uploads the data to a big data platform, and then sequentially performs data preprocessing, defect feature parameter extraction and defect feature database construction by the big data platform to realize construction of a battery monomer inconsistency defect feature database, and whether the battery system defect exists can be determined by comparing the charging or driving segments of each vehicle in different regions, seasons and mileage intervals with corresponding thresholds, and the accuracy and real-time performance of the battery defect identification are significantly improved compared with the prior art.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of electric vehicle battery performance and defect detection, and particularly relates to a defect identification method for an electric vehicle battery system based on data mining. BACKGROUND

[0002] With the development and popularization of new energy vehicles, the defect problems existing in the battery system gradually appear. From the statistical data of the vehicle battery pack defect caused by the manufacturer recall event in recent years, it can be seen that the number of serious safety hazards such as thermal runaway caused by the difference between battery monomers accounts for more than 80%, so it is necessary to improve the importance of battery system defects. At present, most of the defects of the electric vehicle battery system are based on the risk assessment system established under specific conditions. The competent department and relevant experts combine the actual situation to judge whether there is a defect, which is mostly limited to the defect information complained by consumers. However, after the actual failure occurs, the defect judgment is not only lagging in time, but also the number of real vehicle data samples as the basis for judgment is often not enough, so the general applicability of the related evaluation system obviously cannot meet the requirements of early detection and timely treatment of battery system defects. SUMMARY

[0003] Therefore, in view of the above technical problems existing in the prior art, the present application provides a defect identification method for an electric vehicle battery system based on data mining, which specifically comprises the following steps:

[0004] Step 1: For all vehicles of a vehicle model using the same specification battery system, real-time acquisition of original running data through a vehicle-mounted terminal, and uploading to a big data platform in the form of fields; different fields correspond to data acquisition time, vehicle state, charging state, speed, mileage, total voltage, total current, SOC, longitude, latitude, and battery monomer voltage value;

[0005] Step 2: After the big data platform receives the original running data uploaded by each vehicle, the following data preprocessing is performed in sequence: eliminating data frames with abnormal values; judging which state the data frame corresponds to among driving, charging and full-power standing based on the corresponding fields of each data frame, and combining the time interval between data frames to divide the running data into driving segments, charging segments and full-power standing segments; determining the extended features of each frame based on the corresponding fields of each data frame, including: month, season, region, frame-by-frame integral calculation result of charging working condition frame, and charging segment label item and other extended features;

[0006] Step 3: For each charging segment obtained by division, calculate the average of the battery monomer voltage variation coefficient of all vehicles corresponding to a specific region and / or season and / or mileage interval and the total capacity degradation rate k of the battery pack C, as the charging segment defect feature; for each driving segment obtained by segmentation, all vehicle battery cell discharge voltage difference factors corresponding to a specific region and / or season and / or mileage interval are calculated with the battery pack internal resistance rising rate factor k R , as the driving segment defect feature;

[0007] Step four, using the charging segment defect features and driving segment defect features corresponding to different specific regions and / or seasons and / or mileage intervals obtained from step three, a monomer inconsistency defect feature database of the battery system of the vehicle model throughout its life cycle is constructed, and the corresponding thresholds of each charging segment defect feature and driving segment defect feature are set;

[0008] Step five, by comparing the charging segment defect features and driving segment defect features of each vehicle itself with the corresponding thresholds of specific regions and / or seasons and / or mileage intervals, it is determined whether the vehicle battery system has monomer inconsistency defects.

[0009] Further, in the preprocessing in step two, the removing operation specifically includes deleting duplicate frames; performing sliding window filtering average or difference method for frames with sudden changes in current, voltage, SOC or 0 value; after replacing the “\r” character in the data with a null value, performing time sequence arrangement and data cleaning, and deleting frames with abnormal time fields;

[0010] The specific process of segmenting different segments includes: first, the current, SOC and vehicle speed fields are labeled with driving, charging or full power standing state labels for each frame; then, the state label is compared with the vehicle state information contained in the original data, if the same rate meets the predetermined value (such as greater than or equal to 95%), the state label is reset according to the original data and the vehicle state is determined, otherwise the state label remains unchanged; finally, it is determined whether the data frames belong to the same segment according to the time interval between the data frames with the same state label;

[0011] Determining the extended feature items of each frame specifically includes generating corresponding month and season labels according to the time fields of each frame of preprocessed data; generating regional information according to the latitude and longitude fields; generating intermediate variable labels corresponding to the integral calculation results of each frame; and distinguishing segment number labels of different segments.

[0012] Further, the process of obtaining the charging segment defect feature in step three specifically includes:

[0013] First, for all charging segments under a specific season and / or region and / or mileage interval, the monomer voltage list of the last 5% SOC (if less than 5%, all are extracted) data segment of each segment is extracted, and for a segment M, the monomer voltage list data is split to obtain the following monomer voltage matrix V M:

[0014]

[0015] where m represents the total number of voltage data frames, n is the number of single bodies, v ij represents the voltage of single body i in the jth frame;

[0016] The mean voltage vector is obtained by calculating the mean value of the voltage of all frames of different single bodies:

[0017]

[0018] where the mean voltage of single body i is The following formula is used for calculation:

[0019]

[0020] The single body voltage variation coefficient ε of segment M is calculated by the following formula M :

[0021]

[0022] where μ M represents the average value of each element of the mean voltage vector of segment M;

[0023] The mean value of the variation coefficient of all charging segments of a certain vehicle model in a specific season and / or region and / or mileage interval is calculated as one of the charging segment defect features;

[0024] The capacity degradation rate is calculated based on the relative capacity at a unit SOC, and for segment M, the relative capacity C M can be represented as:

[0025]

[0026] where t s represents the start time of the segment, t e represents the end time of the segment, I t represents the total current at time t, and ΔSOC represents the segment SOC change;

[0027] The relative capacity of all charging segments of a certain vehicle model in a specific region and / or season and / or mileage interval is calculated to obtain the relative capacity vector [C1 C2…C N ], N represents the number of segments, the capacity vector is min-max normalized, and the least squares method is used to perform linear fitting on the data points constructed by the normalized vector and the mileage, and the slope k C is taken as another charging segment defect feature;

[0028] The process of obtaining the discharge segment defect feature specifically comprises:

[0029] a single-cell voltage matrix V of the driving segment M in a specific season and / or region and / or mileage interval M The median voltage of each frame of different single cells is taken as a reference voltage to construct a reference voltage vector:

[0030]

[0031] The correlation coefficient between any single cell i and the reference voltage is calculated by the following formula

[0032]

[0033] The correlation coefficient vector of the driving segment M is obtained The minimum value of the correlation coefficient vector is taken The minimum value of the correlation coefficient of all driving segments of a certain vehicle model in a specific region and / or season and / or mileage interval is averaged to obtain a discharge voltage difference factor as one of the discharge segment defect features;

[0034] The internal resistance increase rate is specifically obtained based on an RC equivalent circuit model, comprising:

[0035] First, the following first-order RC equivalent circuit model is established for the battery pack:

[0036]

[0037] In the formula, Uoc represents the open circuit voltage, U represents the terminal voltage, R1 represents the ohmic resistance, which can reflect the battery aging state, R2 and C represent the polarization resistance and polarization capacitance of the battery pack respectively, which can reflect the dynamic characteristics of the battery, I represents the total current, j represents the imaginary unit, and w represents the frequency;

[0038] The above formula is converted to the time domain to obtain the following relationship:

[0039]

[0040] In the formula, ΔT represents the time interval of uploading data to the platform, and k represents a certain time;

[0041] Let a, b, c, and a, b, and c are obtained by parameter identification through the least square method combined with the forgetting factor, and the values of R1, R2, and C are obtained by solving the equation;

[0042] Calculate the battery pack internal resistance for all driving segments of a certain vehicle model under a specific region and / or season and / or mileage range, and obtain the internal resistance vector [R1 R2…R N [N] represents the number of segments. The internal resistance vector is normalized using a min-max method, and the normalized vector is fitted to the data points constructed from the mileage using the least squares method. The slope k of the fitted vector is then used. R This serves as another characteristic of discharge segment defects.

[0043] Furthermore, the thresholds set for each defect feature in step four are specifically based on the following rules:

[0044] (1) Defect threshold calculation: All vehicle models in each subset of the dataset Calculate the mean and standard deviation use Vehicles exceeding a defect threshold are considered to have defects.

[0045] (2)k C Defect threshold calculation: for all car models k in each subset of the dataset C Calculate the mean and standard deviation use Vehicles with defects below a threshold are considered to have defects.

[0046] (3) Defect threshold calculation: All vehicle models in each subset of the dataset Calculate the mean and standard deviation use Vehicles exceeding a defect threshold are considered to have defects.

[0047] (4)k R Defect threshold calculation: for all car models k in each subset of the dataset R Calculate the mean and standard deviation use Vehicles exceeding a defect threshold are considered to have defects.

[0048] The electric vehicle battery system defect identification method based on data mining provided by the present invention first prepares data of different vehicle models and uploads it to a big data platform. Then, the big data platform sequentially performs several steps, including data preprocessing, defect feature parameter extraction, and defect feature database construction, to build a database of inconsistent defect features of individual battery cells. It can compare the charging or driving segments of each vehicle in different regions, seasons, and mileage ranges with corresponding thresholds to determine whether there are battery system defects. Compared with the prior art, it significantly improves the accuracy and real-time performance of battery defect identification. Attached Figure Description

[0049] Figure 1 This is a general flowchart of the method provided by the present invention;

[0050] Figure 2 This is an optional extraction format for the original data;

[0051] Figure 3 A flowchart illustrating the specific process of segmenting a data frame into fragments;

[0052] Figure 4 This is a schematic diagram illustrating the process of constructing a database of defect features for battery systems. Detailed Implementation

[0053] The technical solution of the present invention 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 the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] The data mining-based defect identification method for electric vehicle battery systems provided by this invention, such as... Figure 1 As shown, the specific steps include:

[0055] Step 1: For all vehicles of a certain model using the same battery system, collect raw operating data in real time through the onboard terminal and upload it to the big data platform in the form of fields. Different fields correspond to data collection time, vehicle status, charging status, vehicle speed, mileage, total voltage, total current, SOC, longitude, latitude, and individual battery cell voltage value, respectively. Specific data and field formats that can be used are as follows: Figure 2 As shown.

[0056] Step two, after receiving the raw operation data uploaded by each vehicle, the big data platform sequentially performs the following data preprocessing: eliminating data frames with abnormal values; judging which state among driving, charging and full-charge standing each data frame corresponds to based on the corresponding fields of each data frame, and combining the time intervals between data frames, the operation data is divided into driving segments, charging segments and full-charge standing segments; determining the extended features of each frame based on the corresponding fields of each data frame, including: month, season, region, frame-by-frame integral calculation results of charging working conditions and charging segment label items and other extended features;

[0057] Step three, for each charging segment obtained by division, the average of the battery cell voltage variation coefficients corresponding to a specific region and / or season and / or mileage interval of all vehicles and the total capacity degradation rate k of the battery pack C are calculated as charging segment defect features; for each driving segment obtained by division, the battery cell discharge voltage difference factor and the internal resistance rise rate factor k of the battery pack corresponding to a specific region and / or season and / or mileage interval of all vehicles are calculated as driving segment defect features; R

[0058] Step four, using the charging segment defect features and driving segment defect features corresponding to different specific regions and / or seasons and / or mileage intervals obtained by step three, a monomer inconsistency defect feature database of the battery system full life cycle of the vehicle model is constructed, and the corresponding thresholds of each charging segment defect feature and driving segment defect feature are set respectively;

[0059] Step five, by comparing the charging segment defect features and driving segment defect features of each vehicle itself with the corresponding thresholds of a specific region and / or season and / or mileage interval, it is judged whether the vehicle battery system has monomer inconsistency defect.

[0060] In a preferred embodiment of the present application, in the preprocessing in step two, the elimination can be processed as shown in Table 1:

[0061] Table 1 Abnormal value problem and processing method

[0062]

[0063]

[0064] The elimination operation specifically includes deleting duplicate frames; performing sliding window filtering average or difference method for processing on frames with sudden changes in current, voltage and SOC or 0 values; after replacing the “\r” character in the data with a null value, performing time sequence arrangement and data cleaning, and deleting frames with abnormal time fields;

[0065] ​The specific process of segmenting different segments includes: first, the current, SOC and vehicle speed fields are respectively marked with driving, charging or full power standing state labels; then, the state labels are compared with the vehicle state information contained in the original data, if the same rate meets the predetermined value (such as greater than or equal to 95%), the state label is reset according to the original data and the vehicle state is determined, otherwise the state label remains unchanged; finally, whether the data frames belong to the same segment is determined according to the time interval between the data frames with the same state label (such as less than or equal to 5 minutes), and the specific process is as shown in Figure 3 .

[0066] The specific process of determining the extension feature of each frame includes generating the corresponding month and season labels according to the time field of each frame of preprocessed data; generating regional information according to the latitude and longitude field; generating the intermediate variable label corresponding to the integral calculation result of each frame; and distinguishing the segment number label of different segments.

[0067] Since the poor consistency of single cells directly manifests as obvious differences between the voltages of single cells, at the same time, with the use of the battery, the formation and growth of SEI film, the decomposition reaction of electrolyte, the internal mechanism changes such as lithium dendrite will cause the battery to age, leading to the reduction of available lithium ions and the poor traffic, which specifically manifests as the reduction of available capacity and the increase of battery internal resistance. The deterioration of the consistency of single cells will accelerate the aging of the battery pack, which manifests as the increase of internal resistance rate and the acceleration of capacity decay rate. In the segmented data segments, the data segments with charging segments are screened, and the explicit representation of the voltage difference of single cells and the implicit representation of the capacity decay rate are constructed as charging segment defect features. Therefore, in a preferred embodiment of the present application, the process of obtaining the charging segment defect features in step three specifically includes:

[0068] First, for all charging segments under a specific season and / or region and / or mileage interval, the single cell voltage list of the last 5% SOC (if less than 5%, all are extracted) data segment of each segment is extracted, and for a segment M, the single cell voltage list data is split to obtain the following single cell voltage matrix V M :

[0069]

[0070] In the formula, m represents the total number of voltage data frames, n is the number of single cells, v ij represents the voltage of single cell i in the jth frame;

[0071] The average voltage vector is obtained by calculating the average value of the voltage of all frames of different single cells:

[0072]

[0073] The average voltage of single cell i is The average voltage of single cell i is calculated by the following formula:

[0074]

[0075] The monomer voltage variation coefficient ε of the segment M is calculated by the following formula M :

[0076]

[0077] In the formula, μ M represents the average value of each element of the mean voltage vector of the segment M;

[0078] The average value of the variation coefficients of all charging segments of a certain vehicle model in a certain season and / or region and / or mileage interval is calculated as one of the charging segment defect features;

[0079] The capacity degradation rate is specifically calculated based on the relative capacity at a unit SOC, and for the segment M, the relative capacity C M may be represented as:

[0080]

[0081] In the formula, t s represents the start time of the segment, t e represents the end time of the segment, I t represents the total current at time t, and ΔSOC represents the segment SOC change;

[0082] The relative capacity of all charging segments of a certain vehicle model in a certain region and / or season and / or mileage interval is calculated to obtain a relative capacity vector [C1 C2…C N ], N represents the number of segments, the capacity vector is min-max normalized, and a least squares method is used to perform linear fitting on the data points constructed by the normalized vector and the mileage, and the slope k C of the fitting is taken as another charging segment defect feature;

[0083] The process of obtaining the discharge segment defect feature specifically includes:

[0084] The monomer voltage matrix V M of the driving segment M in a certain season and / or region and / or mileage interval is taken, the median voltage of each frame of different monomers is taken as a reference voltage, and a reference voltage vector is constructed:

[0085]

[0086] The correlation coefficient between any monomer i and the reference voltage is calculated by the following formula

[0087]

[0088] Obtain the correlation coefficient vector of the driving segment M Take the minimum value among them Further, obtain the discharge voltage difference factor by averaging the minimum values of the correlation coefficients of all driving segments of a certain vehicle model in a certain region and / or season and / or mileage interval As one of the discharge segment defect features;

[0089] The internal resistance increase rate is specifically obtained based on an RC equivalent circuit model, including:

[0090] First, establish the following first-order RC equivalent circuit model for the battery pack:

[0091]

[0092] In the formula, Uoc represents the open circuit voltage, U represents the terminal voltage, R1 represents the ohmic internal resistance, which can reflect the battery aging state, R2 and C represent the polarization internal resistance and polarization capacitance of the battery pack, respectively, which can reflect the dynamic characteristics of the battery, I represents the total current, j represents the imaginary unit, and w represents the frequency;

[0093] After converting the above formula to the time domain, the following relationship is obtained:

[0094]

[0095] In the formula, ΔT represents the time interval of uploading data to the platform, and k represents a certain time;

[0096] Let a, b, c, and obtain the values of R1, R2, and C by parameter identification through the least squares method combined with the forgetting factor;

[0097] Calculate the battery pack internal resistance of all driving segments of a certain vehicle model in a certain region and / or season and / or mileage interval to obtain the internal resistance vector [R1 R2…R N ], N represents the number of segments, perform min-max normalization on the internal resistance vector, and use the least squares method to perform linear fitting on the data points constructed by the normalized vector and the mileage, and the slope k R As another discharge segment defect feature.

[0098] Further, the threshold values set for each defect feature in step four are obtained based on the following rules:

[0099] (1) Defect threshold calculation of : in each sub-data set, all vehicle models Calculate the mean and the standard deviation Use As the defect threshold, the vehicle model greater than the threshold is considered to have a defect;

[0100] (2) k C Defect threshold calculation: calculate the mean C and standard deviation of all vehicle models k in each sub-data set. As the defect threshold, the vehicle model less than the threshold is considered to have a defect;

[0101] (3) Defect threshold calculation: calculate the mean and standard deviation of all vehicle models k in each sub-data set. As the defect threshold, the vehicle model greater than the threshold is considered to have a defect;

[0102] (4) k R Defect threshold calculation: calculate the mean R and standard deviation of all vehicle models k in each sub-data set. As the defect threshold, the vehicle model greater than the threshold is considered to have a defect.

[0103] Thus, by the process shown in Figure 4 , the battery system defect feature database of each vehicle model using different specifications of battery system can be constructed.

[0104] It should be understood that the size of the serial number of each step in the embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0105] Although embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for defect identification in electric vehicle battery systems based on data mining, characterized in that: Specifically, the following steps are included: Step 1: For all vehicles of a certain model that use the same battery system, collect raw operating data in real time through the vehicle terminal and upload it to the big data platform in the form of fields; different fields correspond to data collection time, vehicle status, charging status, vehicle speed, mileage, total voltage, total current, SOC, longitude, latitude, and battery cell voltage value, respectively. Step 2: After receiving the raw operating data uploaded by each vehicle, the big data platform performs the following data preprocessing in sequence: remove data frames with outliers; determine which state (driving, charging, or fully charged idle) each data frame corresponds to based on its corresponding fields, and divide the operating data into driving segments, charging segments, and fully charged idle segments based on the time interval between data frames; determine the extended features of each frame based on its corresponding fields, including: the frame-by-frame integral calculation results of month, season, region, charging condition frames, and extended features of charging segment tag items; Step 3: For each segmented charging unit, calculate the mean coefficient of variation of the voltage of all individual battery cells in the vehicles corresponding to a specific region and / or season and / or mileage range. With the rate of decay of the total battery pack capacity k C As a characteristic of charging segment defects; for each segmented driving route, calculate the discharge voltage difference factor of all vehicle battery cells corresponding to a specific region and / or season and / or mileage range. With the rate of increase of battery pack internal resistance k R , as a defect feature of the driving segment; Step 4: Using the charging segment defect features and driving segment defect features obtained in Step 3 that correspond to different specific regions and / or seasons and / or mileage ranges, construct a database of individual cell inconsistency defect features for the entire life cycle of the vehicle battery system, and set corresponding thresholds for each charging segment defect feature and driving segment defect feature respectively. Step 5: By comparing the charging segment defect characteristics and driving segment defect characteristics of each vehicle with the corresponding thresholds for specific regions and / or seasons and / or mileage ranges, determine whether there are individual cell inconsistencies in the vehicle battery system.

2. The method as described in claim 1, characterized in that: In the preprocessing step two, the operation of removing data frames with outliers specifically includes deleting duplicate frames; performing sliding window filtering and averaging or difference methods on frames with sudden changes in current, voltage, or SOC or a value of 0; replacing the "\r" character in the data with null values, performing time-series arrangement and data cleaning, and deleting frames with outliers in the time field. The specific process of segmenting different segments includes: First, the current, SOC and vehicle speed fields are used to label each frame with a status tag indicating whether it is driving, charging or fully charged and stationary; then, the status tag is compared with the vehicle status information contained in the original data. If the sameness rate meets a predetermined value, the status tag is reset according to the original data and the vehicle status is determined. Otherwise, the status tag remains unchanged; finally, the time interval between data frames with the same status tag is used to determine whether the data frames belong to the same segment. The determination of extended feature terms for each frame specifically includes generating corresponding month and season labels based on the time field corresponding to each frame of data after preprocessing; generating regional information based on the latitude and longitude fields; generating intermediate variable labels corresponding to the frame-by-frame integration calculation results; and segment number labels to distinguish different segments.

3. The method as described in claim 1, characterized in that: Step three, the process of obtaining the defect features of the charging segment, specifically includes: First, for all charging segments under a specific season and / or region and / or mileage range, extract the individual cell voltage list of the last 5% SOC data segment of each segment. For a certain segment M, after splitting its individual cell voltage list data, the following individual cell voltage matrix V is obtained. M : In the formula, m represents the total number of voltage data frames, n is the number of individual units, and v ij This represents the voltage of unit i in frame j; The mean voltage vector is obtained by calculating the mean voltage for all frames of different units: Among them, the average voltage of cell i Calculate using the following formula: The single-unit voltage variation coefficient ε of segment M is calculated using the following formula. M : In the formula, μ M This represents the average value of each element in the mean voltage vector of segment M; Calculate the mean of the coefficient of variation of all charging segments for a specific vehicle model under a specific season and / or region and / or mileage range. As one of the defect characteristics of a charging segment; The capacity decay rate is specifically calculated based on the relative capacity per unit SOC, where the relative capacity C for segment M is... M It can be represented as: In the formula, t s Indicates the start time of the segment, t e Indicates the end time of the segment, I t ΔSOC represents the total current at time t, and ΔSOC represents the change in SOC during the segment. Calculate the relative capacity of all charging segments for a specific vehicle model within a specific region and / or season and / or mileage range to obtain the relative capacity vector [C1 C2 … C N [], N represents the number of segments. The capacity vector is normalized using min-max normalization, and the normalized vector is fitted to the data points constructed by the mileage using the least squares method. The slope k of the fitted vector is then used. C As another characteristic of charging segment defects; The process of obtaining the defect features of the driving segment specifically includes: The individual voltage matrix V for a driving segment M under a specific season and / or region and / or mileage range M The median voltage of each individual unit in each frame is taken as the reference voltage, and a reference voltage vector is constructed: The correlation coefficient between any single cell i and the reference voltage can be calculated using the following formula. Obtain the correlation coefficient vector of driving segment M Then take the minimum value. Then, the mean value of the minimum correlation coefficients for all driving segments of a certain vehicle model under a specific region and / or season and / or mileage range is calculated to obtain the discharge voltage difference factor. As one of the defective features of the driving segment; The rate of increase in internal resistance is specifically obtained based on the RC equivalent circuit model, including: First, the following first-order RC equivalent circuit model is established for the battery pack: In the formula, Uoc represents the open-circuit voltage, U represents the terminal voltage, R1 represents the ohmic internal resistance, which can reflect the aging state of the battery, R2 and C represent the polarization internal resistance and polarization capacitance of the battery pack, respectively, which can reflect the dynamic characteristics of the battery, I represents the total current, j represents the imaginary part unit, and w represents the frequency. Transforming the above equation to the time domain yields the following relationship: In the formula, ΔT represents the time interval for data to be uploaded to the platform, and k represents a certain moment; Will Let it be a, Let it be b. Let c be the parameter. By combining the least squares method with the forgetting factor, we can identify the parameters a, b, and c. Solving the equation will give us the values ​​of R1, R2, and C. Calculate the battery pack internal resistance for all driving segments of a certain vehicle model under a specific region and / or season and / or mileage range, and obtain the internal resistance vector [R1 R2 … R N [N] represents the number of segments. The internal resistance vector is normalized using a min-max method, and the normalized vector is fitted to the data points constructed from the mileage using the least squares method. The slope k of the fitted vector is then used. R This is another characteristic of defects in the driving segment.

4. The method as described in claim 1, characterized in that: The thresholds for each defect feature in step four are specifically set based on the following rules: (1) Defect threshold calculation: All vehicle models in each subset of the dataset Calculate the mean and standard deviation use Vehicles exceeding a defect threshold are considered to have defects. (2)k C Defect threshold calculation: for all car models k in each subset of the dataset C Calculate the mean and standard deviation use Vehicles with defects below a threshold are considered to have defects. (3) Defect threshold calculation: All vehicle models in each subset of the dataset Calculate the mean and standard deviation use Vehicles exceeding a defect threshold are considered to have defects. (4)k R Defect threshold calculation: for all car models k in each subset of the dataset R Calculate the mean and standard deviation use Vehicles exceeding a defect threshold are considered to have defects.

Images