Method and system for detecting short-circuit fault in battery pack based on multi-feature fusion

Abstract

The invention discloses a method and a system for detecting a short-circuit fault in a battery pack based on multi-feature fusion. The method comprises the following steps: sampling voltage data, current data and probe temperature data of each monomer in a battery pack at intervals in real time; calculating an average maximum voltage difference F1 according to the voltage data; performing decentration processing according to the voltage data, and calculating to obtain a maximum decentration voltage deviation F2; calculating a weighted absolute current-temperature rise ratio F3 according to the current data and the probe temperature data; integrating the average maximum voltage difference F1, the maximum decentralized voltage deviation F2 and the weighted absolute current-temperature rise ratio F3 into feature points, and integrating and mapping the feature points into data points; and clustering the data points by using a DBSCAN clustering algorithm to obtain abnormal data points, and obtaining short-circuited single batteries in the battery pack according to the abnormal data points. According to the method, voltage, current and temperature data in the battery pack are collected, three characteristics are extracted to represent the state of the battery pack, and the detection accuracy and the detection timeliness can be obviously improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of lithium-ion battery early warning, and in particular relates to a method for detecting short-circuit faults in a battery pack based on multi-feature fusion. Background Art

[0002] Lithium-ion batteries are currently widely used in electric vehicles and energy storage systems, but short circuits in battery packs occur from time to time. If the short circuit in the battery pack cannot be detected in a timely and effective manner, and the battery management system (BMS) does not respond in time, extreme situations such as thermal runaway or even fire of the battery pack are very likely to occur. Therefore, it is very necessary to perform internal short circuit detection on the battery pack.

[0003] The internal short circuit detection algorithms of lithium-ion batteries mainly include gas detection, temperature rise detection, data-driven methods, etc. In recent years, domestic and foreign scholars have designed corresponding monitoring systems and mathematical models for the internal short circuit problem of lithium-ion batteries. There are usually no gas sensors in the actual deployed lithium-ion battery packs, so the methods using gas sensors are often targeted at laboratories and other scenarios. For the temperature rise detection method, some scholars have designed a three-level warning system for a certain lithium-ion battery, and used 50℃, 70℃, and 80℃ as the three-level warning temperatures. This method of monitoring using only a certain feature will inevitably encounter problems such as sensor data distortion and even failure in the actual operation of lithium-ion batteries. Moreover, it is difficult to guarantee the warning speed by only using temperature data to predict internal short circuits. Some scholars also predict battery internal short circuits based on charging data, but in reality, many battery packs have internal short circuits, not only in charging conditions, but also in discharge conditions and frequent charging and discharging conditions. Although many internal short circuit detection algorithms for lithium-ion batteries have been proposed, they are usually only applicable to a single condition or laboratory conditions. In actual energy storage systems and electric vehicle battery packs, the battery packs are usually frequently charged and discharged, so it is difficult to cover most operating conditions by only predicting internal short circuits during the charging phase. Moreover, the working environment of lithium-ion batteries is not as good as that of laboratories. There are often a large number of lithium-ion batteries in the battery pack, and data is inevitably lost and distorted. In addition, as the battery pack ages and its health status decreases, the electrical performance parameters of the battery and its various sensors decrease, resulting in low accuracy in detecting short circuits in the battery pack, and extended detection time, which leads to short warning time. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a method and system for detecting short circuit faults in a battery pack based on multi-feature fusion.

[0005] In a first aspect, the present invention proposes a method for detecting short circuit faults in a battery pack based on multi-feature fusion, the method comprising:

[0006] S1: real-time interval sampling of voltage data, current data and probe temperature data of each cell in the battery pack;

[0007] S2: Calculate the average maximum voltage difference F1 based on the voltage data of each cell in the battery pack;

[0008] S3: According to the voltage data of each cell in the battery pack, after decentralization processing, the maximum decentralized voltage deviation F2 is calculated;

[0009] S4: Calculate the weighted absolute current-temperature rise ratio F3 according to the current data of each cell in the battery pack and the probe temperature data;

[0010] S5: integrating the average maximum voltage difference F1, the maximum decentralized voltage deviation F2 and the weighted absolute current-temperature rise ratio F3 into feature points, and mapping the feature points into data points;

[0011] S6: clustering the mapped data points using the DBSCAN clustering algorithm to obtain abnormal data points, and obtaining short-circuited battery cells in the battery pack based on the abnormal data points.

[0012] In a second aspect, the present invention proposes a battery pack internal short circuit fault detection system based on multi-feature fusion, the system comprising:

[0013] A data acquisition module, used to respond to a fault detection request and collect voltage data, current data and probe temperature data of each cell in the battery pack by sampling in real time and at intervals;

[0014] A data processing module is used to process the voltage data, current data and probe temperature data of each monomer to obtain the average maximum voltage difference F1, the maximum decentralized voltage deviation F2 and the weighted absolute current-temperature rise ratio F3, and normalize them to obtain the corresponding characteristics after normalization;

[0015] A data integration module is used to integrate and map the corresponding features after normalization into data points;

[0016] The clustering module is used to cluster the mapped data points using the DBSCAN clustering algorithm to obtain the abnormal data points;

[0017] The output module is used to output the short circuit fault detection result in the battery pack according to each abnormal data point.

[0018] In a third aspect, the present invention provides an electronic device, comprising:

[0019] one or more processors;

[0020] A memory for storing one or more instructions, wherein when the one or more instructions are executed by the one or more processors, the one or more processors implement the method proposed in the first aspect of the present invention.

[0021] In a fourth aspect, the present invention proposes a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enables the processor to implement the method proposed in the first aspect of the present invention.

[0022] Beneficial effects of the present invention: The present invention uses a sliding window method, and adopts the voltage, current, and temperature data in the battery pack that can be extracted from the BMS system in actual work, and extracts three features of average maximum voltage difference, maximum decentralized deviation, and weighted absolute current-temperature rise ratio to characterize the battery pack state. It can score whether the battery pack has an internal short circuit, and then use the DBSCAN unsupervised clustering method to more accurately find the abnormal time point and locate the internal short-circuited battery cell, which not only overcomes the problem of false alarm caused by partial sensor errors when using methods such as only a single feature, but also improves the fault detection speed and detection accuracy, provides valuable decision-making time for the BMS system, and greatly improves the safety performance of the BMS system and the thermal management system. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A flowchart of a method for detecting a short circuit fault in a battery pack according to an embodiment of the present invention;

[0024] Figure 2 Schematic diagram of a flow chart of a method for detecting a short circuit fault in a battery pack according to an embodiment of the present invention;

[0025] Figure 3 Schematic diagram of the structure of a short circuit fault detection system in a battery pack according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] The present invention has conducted a long-term investigation on the existing BMS data and found that during the operation of lithium-ion battery packs, there will be occasional frame loss, distortion, etc., and as the battery pack and BMS age, the electrical performance parameters of various sensors in the battery will also decrease, resulting in higher system errors. If they cannot be effectively identified, the false alarm rate and missed alarm rate of the detection results will be very high, and it will take a longer time to obtain the detection results (that is, the warning time left for the BMS is shortened).

[0028] The embodiment of the present invention proposes a method for detecting short circuit faults in a battery pack based on multi-feature fusion. Figure 2 , Figure 3 As shown, the method includes:

[0029] S1: Real-time interval sampling of the voltage data, current data and probe temperature data of each battery cell in the battery pack.

[0030] Specifically, the battery pack is composed of an array of N battery cells (referred to as cells). During operation, the battery pack generates various operating data in real time, and the collected data is continuously fed back to the battery management system BMS through sensors. For example, the operating parameters of the battery pack in an electric vehicle include: voltage, current, battery temperature, battery capacity, battery energy, battery health state (State of Health, SOH), etc. The real-time voltage, current and battery temperature of each battery cell in the battery pack are obtained through voltage sensors, current sensors and temperature sensors, and fed back to the battery management system BMS.

[0031] In response to a fault detection request, the BMS uses a sliding window method to sample the voltage data, current data and probe temperature data of each battery cell in the battery pack in real time, that is, M frames of voltage data, current data and probe temperature data are selected by sliding, and at least S data window fragments are selected.

[0032] S2: Calculate the average maximum voltage difference based on the voltage data of each cell in the battery pack.

[0033] Specifically, in the sampling segment, the maximum voltage difference between each cell is solved according to the voltage data of each cell in the battery pack, and the maximum voltage difference between each battery cell is averaged to obtain the average maximum voltage difference.

[0034] The average maximum voltage difference, that is, the average value of the maximum voltage difference between monomers in the sampling segment, is calculated as follows:

[0035] S201: In the sampling segment, according to the voltage data of each cell in the battery pack, the voltage matrix of each cell in the time window (that is, the voltage matrix of the battery pack) is obtained, which is specifically expressed as:

[0036]

[0037] and then,

[0038] Where V represents the voltage matrix in the time window, V i, Represents the row vector of the cell voltage of all batteries at the i-th moment, V ,j represents the voltage column vector of the jth cell in this time window, u i,j represents the voltage of the jth cell at the i-th moment, It means to find the matrix transpose, M represents the length of the sliding window, and N represents the number of battery cells.

[0039] S202: Find the voltage value of each row from the voltage matrix V in the time window, and calculate the maximum battery cell voltage difference between the battery cells in each row:

[0040] V i_max_diff =max(u i,j -u i,k ),i∈[1,M];j,k∈[1,N]

[0041] Among them, V i_max_diff represents the maximum voltage difference between battery cells at time i, max(.) represents the maximum value, u i,j represents the voltage of the jth monomer at the i-th moment, i represents the i-th moment, j represents the j-th monomer, and k represents the k-th monomer.

[0042] S203: According to the V of each row in the voltage matrix within the time window i_max_diff , calculate the average value of the maximum voltage difference:

[0043]

[0044] Among them, V max_diff_mean Indicates the maximum voltage difference within the time window. V max_diff_mean Expressed by F1, that is, F1 = V max_diff_mean .

[0045] In a battery pack, due to various factors such as manufacturing process, materials, and usage conditions, there will be certain performance differences between each battery cell. These differences will cause inconsistent cell voltages during the operation of the battery pack (especially charging and discharging). If these differences are not managed and adjusted, as the charge and discharge cycles increase, the voltages of each cell will gradually diverge, and the service life will be greatly reduced. In consideration of battery balancing management, the present invention decentralizes the voltage value of each battery cell. Compared with the existing decentralized processing of voltage values, the embodiment of the present invention does not simply subtract the matrix mean, but first subtracts the row mean and then subtracts the column mean. Subtracting the row mean is to filter out the voltage changes caused by different load conditions at different time points, and subtracting the column mean is to filter out the physical differences between different cells, so as to better capture the voltage drop caused by external factors such as non-load at different times of the same cell, and thus be able to extract more accurate voltage characteristic values.

[0046] S3: Based on the voltage data of each cell in the battery pack, after decentralized processing, the maximum decentralized voltage deviation is calculated.

[0047] Specifically: within the sampling segment, according to the voltage data of each cell in the battery pack, the voltage value of each battery cell is decentralized, and the maximum decentralized voltage deviation is calculated.

[0048] The specific process of calculating the maximum decentralized pressure deviation is:

[0049] S301: Obtain the voltage matrix V of each battery cell in the time window, calculate the relative difference between the voltage matrix V and its row mean, and obtain the row mean difference matrix V diff .

[0050] Specifically, in the sampling segment, according to the voltage data of each cell in the battery pack, the relative difference between the voltage matrix window and the row mean is calculated, and the row mean difference matrix V between the voltage matrix and the row mean can be obtained. diff , which are specifically:

[0051]

[0052] Where V represents the voltage matrix in the time window, V row_mean It means that the voltage matrix V takes the row mean, that is, all elements in it are equal to the mean of their row; V diff Represents the voltage matrix V minus the row mean V rowmean The row mean difference matrix, V i, Represents the row vector of the cell voltage of all batteries at the i-th moment, V i,_mean represents the row vector representing the average value of the single cell voltage of all batteries at the i-th moment, with a length of N.

[0053] S302: Calculate the voltage matrix V and find the column mean V according to the voltage data of each cell in the battery pack col_mean , calculate the row mean difference matrix V diff and the voltage matrix V column mean V col_mean The relative difference of is used to obtain the monomer voltage decentralized matrix.

[0054] The specific calculation process of S302 is:

[0055] V dece_volt =V diff -V col_mean ,

[0056]

[0057] and then,

[0058] Among them, V dece_volt represents the cell voltage decentralization matrix, V diff represents the matrix V minus the row mean, V col_mean It means that the voltage matrix V takes the column mean, that is, all elements in it are equal to the mean of their columns, V ,j_diff Indicates V diff The jth column of V ,j_mean Indicates V diff The jth monomer in this window has V diff A column vector of length M containing the means of .

[0059] The monomer voltage decentralized matrix reflects the voltage change value within the window of each monomer after excluding the influence of its own constitution.

[0060] S303: According to the cell voltage decentralization matrix, the maximum decentralized voltage deviation is calculated by subtracting the decentralized voltage average value of the cell voltages of the last 10 frames from the decentralized voltage average value of the cell voltages of the first 10 frames in the time window.

[0061] Specifically, the decentralized voltage average mean (V dece_volt (j)[1:10]), minus the decentralized voltage mean (V dece_volt (j)[M-9:M]), obtain the decentralized voltage deviation of the monomer in this window, and find its maximum decentralized voltage deviation V max_dece_diff :

[0062] V dece_diff (j) = mean(V dece_volt (j)[1:10])-mean(V dece_volt (j)[M-9:M])

[0063] V max_dece_diff =max(V dece_diff )

[0064] Where j represents the jth battery cell in the battery pack, and [x:y] represents the data from the xth to yth frames in the selection. Maximum decentralized voltage deviation V max_dece_diff Expressed by F2, that is, F2 = V max_dece_diff .

[0065] Temperature-related detection in lithium-ion batteries is currently usually performed by directly setting a threshold value for judgment. If the threshold value is exceeded, it is considered abnormal. However, in a cold environment, at the initial stage of a short circuit in the battery, although the temperature in the battery also rises rapidly, it cannot reach the fault threshold point in time. At the same time, if the battery itself is undergoing high-current discharge and charging, the battery temperature will also rise rapidly. Therefore, the present invention considers this weighted absolute current-temperature rise ratio to determine whether there is abnormal temperature rise in the battery pack.

[0066] S4: Calculate the weighted absolute current-temperature rise ratio according to the current data of each cell in the battery pack and the probe temperature data.

[0067] Specifically, in the sampling segment, the current value of each frame in each battery cell and the probe temperature difference between every two frames are obtained according to the voltage data and probe temperature data of each cell in the battery pack, and the weighted absolute current-temperature rise ratio is calculated.

[0068] The weighted absolute current-temperature rise ratio can reflect the ratio of the current battery temperature rise to the work done. Normally, the battery will only have a temperature rise when it is working and charging. If there is an abnormal temperature rise, this feature can be well captured.

[0069] The weighted absolute current-temperature rise ratio is calculated as follows:

[0070] S401: Calculate the average value of the absolute value between the currents of two frames in the window according to the current data of each cell in the battery pack to obtain the absolute current average value, which is specifically:

[0071] First, according to the current data of each cell in the battery pack, the total current vector of the battery pack in the sliding window is calculated, which is expressed as:

[0072]

[0073] Where C represents the total current matrix of the battery pack in the time segment, C i Indicates the total current of the battery pack at time i.

[0074] Then, the absolute current average value between the two frames of current in the time window is calculated, which is specifically:

[0075]

[0076] Where C represents the current vector in the time segment, with a length of M, C avg Represents the absolute current average value between two frames of current in the window, with a length of M-1. C[x:y] represents the vector composed of the xth frame to the yth frame in C.

[0077] S402: Calculate the probe temperature difference between every two frames in the window according to the probe temperature data of each cell in the battery pack, which is specifically:

[0078] First, according to the probe temperature data of each cell in the battery pack, the temperature matrix in the time window is obtained as follows:

[0079]

[0080] and then,

[0081] Where T represents the temperature matrix in the time window, T i, represents the temperature row vector of all temperature probes at the i-th moment, T ,j represents the column vector of all temperatures of the jth temperature probe in this time window, t i,j represents the temperature value of the jth temperature probe at the i-th time, (.) T It means to find the matrix transpose, M represents the length of the sliding window, and N represents the number of temperature probes.

[0082] Then, the probe temperature difference between every two frames in the window is calculated, which is specifically:

[0083] T ,j_diff =T ,j [2:M]-T ,j [1:M-1]

[0084]

[0085] Among them, T ,j_diff represents the temperature difference vector of the jth temperature probe in the time window, T ,j [x:y] represents T ,j A vector consisting of the xth to yth values ​​of diff Represents the matrix formed by each temperature probe calculating the temperature difference vector, with a size of (M-1)*S, T i,_diff Indicates T diff The row vector composed of the i-th row elements of is the row vector composed of the temperature rise values ​​of all temperature probes at the i-th to (i+1)-th time. j represents the temperature probe in the j-th battery pack, and S represents the number of temperature probes in the battery pack.

[0086] S403: Calculate the maximum probe temperature difference between every two frames according to the probe temperature difference between every two frames in the window.

[0087] The embodiment of the present invention uses the maximum temperature rise sequence to represent the maximum temperature difference of the probe, which is specifically:

[0088]

[0089] Among them, T max_diff Indicates the maximum temperature rise sequence, max(.) means taking the maximum value, max(T i,_diff ) represents the maximum value of the temperature rise of all temperature probes at time i to (i+1).

[0090] S404: Calculate a weighted absolute current-temperature rise ratio and WS according to the absolute current average value and the probe maximum temperature difference.

[0091] Specifically, the element-wise product of the absolute current average value and the maximum temperature difference of the probe is calculated, and further divided by the time interval, and then weighted summed to obtain the weighted absolute current-temperature rise ratio sum.

[0092] In the embodiment of the present invention, the weight is set to the number of rows of the calculated ratio, which is specifically:

[0093]

[0094] Where WS represents the weighted absolute current-temperature rise ratio and i is the iterative variable and the set weight. The closer to the time point where the sliding window is located, the higher the weight value. max_diff (i) indicates the Tth max_diff The i-th element in, that is, max(T i,_diff ), |.| is the absolute value, C avg (i) indicates C avg In order to avoid calculation errors or abnormalities in the weighted absolute current-temperature rise ratio caused by too small current values, when |C avg (i) When |<1, let |C avg (i)|=1, Δt(i) represents the time interval from the i-th frame to the i+1-th frame in the sliding window.

[0095] S405: Calculate the weighted absolute current-temperature rise ratio according to the weighted absolute current-temperature rise ratio and WS and the sliding segment length, and the calculation formula is:

[0096] F3=WAER,

[0097]

[0098] and then,

[0099] Wherein, WAER represents the weighted absolute current-temperature rise ratio, WAER is represented by F3, i represents the weight of each iteration, and M represents the sliding window length.

[0100] S5: Integrate the average maximum voltage difference F1, the maximum decentralized voltage deviation F2 and the weighted absolute current-temperature rise ratio F3 into feature points (F1, F2, F3), and map the integration into data points.

[0101] Specifically, in three-dimensional Euclidean coordinates, the average maximum voltage difference F1, the maximum decentralized voltage deviation F2 and the weighted absolute current-temperature rise ratio F3 are integrated into a feature point (F1, F2, F3), and the feature point (F1, F2, F3) is integrated and mapped into a data point.

[0102] After repeated sampling and processing of steps S1-S5, a data point set after feature integration is obtained. In order to study the regularity of the data point set, the embodiment of the present invention adopts the DBSCAN algorithm.

[0103] Density-Based Spatial Clustering of Applications with Noise (DBSCAN) is a density-based spatial clustering algorithm that divides areas with sufficient density into clusters and finds clusters of arbitrary shapes in a spatial database with noise. It defines a cluster as the largest set of density-connected points.

[0104] DBSCAN is a scale-sensitive clustering algorithm, so the three features (F1, F2, and F3) need to be normalized.

[0105] Furthermore, the average maximum voltage difference F1, the maximum decentralized voltage deviation F2 and the weighted absolute current-temperature rise ratio F3 are normalized, and the normalized features are integrated into feature points.

[0106] The calculation formula for the normalization process is:

[0107]

[0108] Among them, F scaled represents the normalized feature, F represents the original value of the feature, and F min represents the minimum characteristic value, F max Indicates the maximum value of a feature.

[0109] After step S5, the data points mapped by feature integration include the characteristics of voltage, current and temperature rise in the battery pack. The diagnosis basis of the short circuit in the battery in the present invention no longer relies on a single feature, which reduces the misjudgment rate and makes full use of the correlation between the three features, so that the short circuit in the battery pack can be detected as early as possible.

[0110] S6: clustering the mapped data points using the DBSCAN clustering algorithm to obtain abnormal data points, and obtaining short-circuited battery cells in the battery pack based on the abnormal data points.

[0111] Select an appropriate DBSCAN hyperparameter ε and define the neighborhood radius MinPts: the minimum number of neighborhood points required to be considered a core point.

[0112] Perform DBSCAN clustering on each data point, and the specific clustering steps are as follows:

[0113] S601: Mark all data points as unprocessed;

[0114] S602: For each unprocessed point P, obtain all points in the ε neighborhood of point P, record as set N, then check the core points, if the number of points in set N is greater than or equal to MinPts, mark point P as a core point, otherwise temporarily mark the data point as a boundary point;

[0115] S603: For each point Q in the set N, if point Q is an unprocessed point, obtain the ε neighborhood N(Q) of point Q. If the number of points in the neighborhood N(Q) is greater than or equal to MinPts, mark point Q as a core point, and add all points in the neighborhood N(Q) to the current cluster; if point Q does not belong to any cluster, temporarily mark it as a boundary point;

[0116] S604: Repeat S602 and S603 until all data points are processed, and the data points that are not added to any cluster are determined as boundary points, which are abnormal data points.

[0117] After finding the abnormal data point, enter the data corresponding to the segment and find the cell with the largest voltage drop in the battery module corresponding to the temperature probe with the largest temperature rise, that is, the battery cell detected as having an internal short circuit.

[0118] Based on the same inventive concept, the embodiment of the present invention proposes a battery pack internal short circuit fault detection system based on multi-feature fusion. In this system, the same or similar technical features as the above-mentioned battery pack internal short circuit fault detection method are provided, which will not be described in detail below.

[0119] Reference Figure 3 As shown, the battery pack short circuit fault detection system based on multi-feature fusion includes:

[0120] A data acquisition module, used to respond to a fault detection request and collect voltage data, current data and probe temperature data of each cell in the battery pack by sampling in real time and at intervals;

[0121] A data processing module is used to process the voltage data, current data and probe temperature data of each monomer to obtain the average maximum voltage difference F1, the maximum decentralized voltage deviation F2 and the weighted absolute current-temperature rise ratio F3, and normalize them to obtain the corresponding characteristics after normalization;

[0122] A data integration module is used to integrate and map the corresponding features after normalization into data points;

[0123] The clustering module is used to cluster the mapped data points using the DBSCAN clustering algorithm to obtain the abnormal data points;

[0124] The output module is used to output the short circuit fault detection result in the battery pack according to each abnormal data point.

[0125] An embodiment of the present invention provides an electronic device, including:

[0126] one or more processors;

[0127] A memory for storing one or more instructions, wherein when the one or more instructions are executed by the one or more processors, the one or more processors implement the method proposed in the first aspect of the present invention.

[0128] An embodiment of the present invention provides a computer-readable storage medium having executable instructions stored thereon. When the executable instructions are executed by a processor, the processor implements the method provided in the first aspect of the present invention.

[0129] Beneficial effects of the present invention: The present invention uses a sliding window method, adopts the voltage, current, and temperature data that can be extracted from the BMS system in actual work, extracts three features (F1, F2, and F3) and integrates them into a feature point to characterize the state of the battery pack, and can score whether the battery pack has an internal short circuit; then the DBSCAN unsupervised clustering method is used to more accurately find the abnormal fragments, which not only overcomes the problem of false alarms caused by partial sensor errors when using methods such as only a single feature, but also takes into account the warning speed.

[0130] The use of the present invention can provide valuable decision-making time for the BMS system, greatly improving the safety performance of the BMS system and thermal management system. Even if an internal short circuit in the battery pack of an electric vehicle causes the thermal runaway of the entire vehicle to be uncontrollable, the deployment of this patent can also greatly increase the escape time for users.

[0131] A person skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the relevant hardware through a program, and the program can be stored in a computer-readable storage medium, which can include: ROM, RAM, disk or CD, etc.

[0132] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A battery pack short circuit fault detection method based on multi-feature fusion, characterized in that: include: Real-time interval sampling of voltage data, current data and probe temperature data of each cell in the battery pack; According to the voltage data of each cell in the battery pack, the average maximum voltage difference F1 is calculated; According to the voltage data of each cell in the battery pack, after decentralization processing, the maximum decentralized voltage deviation F2 is calculated; Calculate the weighted absolute current-temperature rise ratio F3 based on the current data and probe temperature data of each cell in the battery pack; Integrate the average maximum voltage difference F1, the maximum decentralized voltage deviation F2 and the weighted absolute current-temperature rise ratio F3 into feature points, and map the feature points into data points; The DBSCAN clustering algorithm is used to cluster the data points obtained by mapping to obtain the abnormal data points, and the short-circuited battery cells in the battery pack are obtained according to the abnormal data points.

2. The method for detecting short-circuit faults in a battery pack based on multi-feature fusion according to claim 1, characterized in that: The method for calculating the average maximum voltage difference is specifically as follows: in a sampling segment, the maximum voltage difference between each cell is solved according to the voltage data of each cell in the battery pack, and the maximum voltage difference between each battery cell is averaged to obtain the average maximum voltage difference.

3. The method for detecting short-circuit faults in a battery pack based on multi-feature fusion according to claim 1, characterized in that: The calculation process of the average maximum voltage difference is specifically as follows: In the sampling segment, the voltage matrix V of each battery cell in the time window is obtained; Calculate the maximum battery cell voltage difference V between each battery cell in each row of the voltage matrix V i_max_diff ; According to the V of each row in the voltage matrix i_max_diff , calculate the average value of the maximum voltage difference and obtain the average maximum voltage difference.

4. The method for detecting short-circuit faults in a battery pack based on multi-feature fusion according to claim 1, characterized in that: The calculation process of the maximum decentralized voltage deviation is: In the sampling segment, the voltage matrix V of each battery cell in the time window is obtained; Calculate the relative difference between the voltage matrix V and its row mean to obtain the row mean difference matrix V diff ; Calculate the row mean difference matrix V diff and the voltage matrix V column mean V col_mean The relative difference of the monomer voltage decentralization matrix V is obtained dece_volt ; According to the monomer voltage decentralization matrix, the average voltage mean (V dece_volt (j)[1:10]) minus the average voltage mean (V dece_volt (j)[M-9:M]), the maximum decentralized voltage deviation V is calculated max_dece_diff .

5. According to claim 1, the method for detecting short-circuit faults in a battery pack based on multi-feature fusion is characterized in that: The calculation process of the weighted absolute current-temperature rise ratio is: According to the current data of each monomer in the battery pack, the absolute current average value between two frames of current in the time window is calculated; According to the probe temperature data of each cell in the battery pack, the probe temperature difference between every two frames in the time window is calculated, and the maximum probe temperature difference between every two frames is calculated according to the probe temperature difference; According to the absolute current average value and the maximum temperature difference of the probe, a weighted absolute current-temperature rise ratio and WS are calculated; The weighted absolute current-temperature rise ratio is calculated based on the weighted absolute current-temperature rise ratio and WS and the sliding segment length.

6. The method for detecting short-circuit faults in a battery pack based on multi-feature fusion according to claim 1, characterized in that: The DBSCAN clustering algorithm is used to cluster the mapped data points. The clustering process is specifically as follows: Set the DBSCAN hyperparameter ε to define the radius of the neighborhood MinPts; S601: Mark all data points as unprocessed; S602: For each unprocessed point P, obtain all points in the ε neighborhood of point P, record as set N, then check the core points, if the number of points in set N is greater than or equal to MinPts, mark point P as a core point, otherwise temporarily mark the data point as a boundary point; S603: For each point Q in the set N, if point Q is an unprocessed point, obtain the ε neighborhood N(Q) of point Q. If the number of points in the neighborhood N(Q) is greater than or equal to MinPts, mark point Q as a core point, and add all points in the neighborhood N(Q) to the current cluster; if point Q does not belong to any cluster, temporarily mark it as a boundary point; S604: Repeat S602 and S603 until all data points are processed, and the data points that are not added to any cluster are determined as boundary points, which are abnormal data points.

7. The method for detecting short-circuit faults in a battery pack based on multi-feature fusion according to claim 1, characterized in that: The average maximum voltage difference F1, the maximum decentralized voltage deviation F2 and the weighted absolute current-temperature rise ratio F3 are normalized, and the normalized features are integrated into feature points.

8. A battery pack internal short circuit fault detection system based on multi-feature fusion, characterized in that the system comprises: A data acquisition module, used to respond to a fault detection request and collect voltage data, current data and probe temperature data of each cell in the battery pack by sampling in real time and at intervals; A data processing module is used to process the voltage data, current data and probe temperature data of each monomer to obtain the average maximum voltage difference F1, the maximum decentralized voltage deviation F2 and the weighted absolute current-temperature rise ratio F3, and normalize them to obtain the corresponding characteristics after normalization; A data integration module is used to integrate and map the corresponding features after normalization into data points; The clustering module is used to cluster the mapped data points using the DBSCAN clustering algorithm to obtain the abnormal data points; The output module is used to output the short circuit fault detection result in the battery pack according to each abnormal data point.

9. An electronic device, comprising: one or more processors; A memory for storing one or more instructions, wherein when the one or more instructions are executed by the one or more processors, the one or more processors are enabled to implement the method according to any one of claims 1 to 7. 10 . A computer-readable storage medium having executable instructions stored thereon, wherein when the executable instructions are executed by a processor, the processor is enabled to implement the method according to any one of claims 1 to 7.

Images