Battery fault identification method for real vehicle working condition

Abstract

The invention relates to a battery fault identification method for real vehicle working conditions, and belongs to the technical field of batteries. The method comprises the following steps: S1, collecting battery operation data of the electric vehicle, including operation data of each monomer in a battery pack, and establishing a battery operation database; s2, according to the collected battery data, utilizing a correlation coefficient method to screen out features having high correlation with the single voltage; s3, establishing an accurate voltage estimation model; s4, real-time power battery operation data of the electric vehicle are input into the model trained in the S3, and whether a fault occurs or not is judged based on the residual error of the real-time voltage and the model estimation voltage; s5, when it is detected that a fault occurs, designing a sliding window to extract sensitive features of the fault, and constructing a fault feature two-dimensional diagram; and S6, identifying whether the detected fault is the battery fault or the sensor reading is abnormal by using an unsupervised clustering algorithm. According to the invention, the risk of false alarm and false alarm can be greatly reduced.

Description

Technical Field

[0001] The invention belongs to the technical field of batteries and relates to a battery fault identification method oriented to actual vehicle working conditions. Background Art

[0002] With the rapid development of new energy vehicles, the safety and reliability of battery systems have become increasingly important issues. During long-term use, battery systems are prone to self-discharge, internal short circuits and other faults due to material defects, extreme working conditions and improper operation. If these faults cannot be detected and handled in time, they will seriously affect the safety of vehicle operation and even cause safety accidents.

[0003] Currently, battery management systems (BMS) are usually equipped with voltage sensors to monitor the voltage of battery cells and identify potential faults by comparing the differences between the cell voltages. However, in practical applications, this fault detection method based on voltage differences has the following limitations:

[0004] Complexity of working conditions: During the actual operation of electric vehicles, they are affected by various factors, such as driving habits, road conditions, ambient temperature, etc., which leads to complex and changeable operating conditions of the battery. This makes it difficult for fault detection methods based on voltage differences to adapt to different working conditions and is prone to false alarms or missed alarms.

[0005] Abnormal sensor readings: Sensor readings in the battery management system may be affected by noise interference or faults, resulting in abnormal voltage data. Fault detection methods based on voltage differences cannot distinguish such voltage differences caused by abnormal sensor readings, and are prone to misjudge abnormal sensor readings as battery faults, resulting in false alarms.

[0006] Therefore, developing a battery fault identification method that can adapt to actual vehicle operating conditions and effectively distinguish between battery failures and abnormal sensor readings is of great significance for ensuring the safe operation of new energy vehicles. Summary of the invention

[0007] In view of this, an object of the present invention is to provide a battery fault identification method oriented to actual vehicle operating conditions.

[0008] In order to achieve the above object, the present invention provides the following technical solutions:

[0009] A battery fault identification method for actual vehicle working conditions includes the following steps:

[0010] S1: Collect battery operation data of electric vehicles, including battery pack parameters and cell voltages of all battery cells and the temperature of temperature probes near them, and establish a battery operation database;

[0011] S2: Screening out features with high correlation with the cell voltage using a correlation analysis method according to the battery operation database;

[0012] S3: Building an accurate voltage estimation model based on the features screened in S2;

[0013] S4: Input the real-time power battery operation data of the electric vehicle into the voltage estimation model trained in S3, and determine whether a fault occurs based on the residual of the real-time voltage and the model estimated voltage;

[0014] S5: When a fault is detected in S4, a sliding window is designed to extract sensitive features of the fault and construct a two-dimensional fault feature map;

[0015] S6: Based on the fault feature two-dimensional graph constructed in S5, an unsupervised clustering algorithm is used to identify whether the detected fault is a battery fault or an abnormal sensor reading.

[0016] Furthermore, in S2, a correlation analysis is performed using the Pearson correlation coefficient or the grayscale correlation coefficient to screen out features whose correlation coefficient with the monomer voltage is greater than a preset value.

[0017] Furthermore, in S3, a nonlinear regression algorithm is selected as the voltage estimation model, including long short-term memory neural network, convolutional neural network, Gaussian process regression or relevance vector machine regression.

[0018] Furthermore, in S4, the preset fault threshold is 0.1V, 0.15V or 0.2V.

[0019] Furthermore, in S5, the fault sensitive features extracted include voltage change rate and average voltage difference.

[0020] Furthermore, in S6, a density-based clustering algorithm with noise, a Gaussian mixture algorithm, a K-means algorithm or an unsupervised clustering algorithm of fuzzy clustering is used to perform fault identification.

[0021] Furthermore, in S1, the collected battery operation data also includes the SOC, internal resistance, SOH or discharge rate of the battery pack.

[0022] Furthermore, in S3, a linear regression algorithm of support vector machine regression or ridge regression is used as a voltage estimation model.

[0023] Furthermore, in S5, the constructed two-dimensional fault feature graph includes features in two dimensions: voltage change rate and average voltage difference.

[0024] Furthermore, in S6, the faults are divided into two categories: battery faults and abnormal sensor readings according to the clustering results.

[0025] The beneficial effects of the present invention are:

[0026] (1) By establishing an accurate voltage estimation model and comparing it with the real-time voltage, the abnormality of battery cell voltage can be effectively identified and the occurrence of missed alarms and false alarms can be reduced.

[0027] (2) By collecting and analyzing data such as battery pack parameters and single cell voltage, the established features can adapt to different driving habits, road conditions, ambient temperature and other operating conditions, thereby improving the robustness of fault detection.

[0028] (3) By extracting fault-sensitive features and performing cluster analysis, it is possible to effectively distinguish between battery failure and abnormal sensor readings, thus avoiding false alarms caused by abnormal sensor readings.

[0029] (4) By timely detecting battery faults, further development of the fault can be avoided, thereby improving the safety of the battery system and ensuring the safe operation of the vehicle.

[0030] (5) By promptly detecting and handling battery failures, damage to the battery caused by the failure can be avoided, thereby extending the battery life.

[0031] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0033] Figure 1 A flow chart of the present invention as a whole;

[0034] Figure 2 are cell voltage curves of a normal cell and a faulty cell in an embodiment of the present invention;

[0035] Figure 3 The safety risk assessment results of the normal monomer and the faulty monomer in the embodiment of the present invention;

[0036] Figure 4 For fault identification based on different algorithm parameters. DETAILED DESCRIPTION

[0037] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0038] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0039] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0040] See also Figure 1 , a battery fault identification method for actual vehicle conditions can be divided into the following steps:

[0041] S1: Collect the battery operation data of electric vehicles, including the operation data of each cell in the battery pack, and establish a battery operation database;

[0042] S2: Based on the collected battery data, the correlation coefficient method is used to screen out features that have a high correlation with the single cell voltage;

[0043] S3: Building an accurate voltage estimation model based on the features extracted in S2;

[0044] S4: Input the real-time power battery operation data of the electric vehicle into the model trained in S3, and determine whether a fault occurs based on the residual of the real-time voltage and the model estimated voltage;

[0045] S5: When a fault is detected in S4, a sliding window is designed to extract sensitive features of the fault and construct a two-dimensional fault feature map;

[0046] S6: Based on the fault feature two-dimensional graph constructed in S5, an unsupervised clustering algorithm is used to identify whether the detected fault is a battery fault or an abnormal sensor reading.

[0047] As an optional embodiment, the above S1 specifically includes S11-S13:

[0048] S11: Collect the battery pack operation data of electric vehicles, including battery pack parameters such as time, current, voltage, temperature, and insulation resistance;

[0049] S12: Collecting battery operation data of electric vehicles, including the cell voltage of all battery cells and the temperature of each temperature probe placed near the cell;

[0050] S13: Establish a power battery database based on the collected battery data.

[0051] As an optional embodiment, the above S2 specifically includes S21-S22:

[0052] S21: Preliminary screening of data theoretically related to monomer voltage as features;

[0053] S22: Using the correlation coefficient method, screen out features that are highly correlated with the monomer voltage.

[0054] As an optional embodiment, the data theoretically related to the single cell voltage in S21 include total current, battery pack voltage, average single cell voltage, SOC, battery pack temperature, internal resistance, SOH, discharge rate, etc.

[0055] As an optional embodiment, the correlation coefficient method in S22 may adopt the Pearson correlation coefficient, and its formula is as follows:

[0056]

[0057] As an optional embodiment, the feature with high correlation with the monomer voltage in S22 may select a feature with a Pearson correlation coefficient greater than 0.9 as a high correlation feature.

[0058] As an optional embodiment, the above S3 specifically includes S31-S32:

[0059] S31: Select a nonlinear regression algorithm as the target machine learning algorithm;

[0060] S32: Based on the screened characteristics of the fault-free stage and the single cell voltage data, an accurate voltage estimation model is established.

[0061] As an optional embodiment, a nonlinear regression algorithm in S31 may adopt a long short-term memory neural network. The construction process of the long short-term memory neural network is as follows:

[0062] Long short-term memory neural network is a variant of recursive neural network, which avoids the problem of gradient vanishing and gradient exploding that recursive neural network is prone to in the training process. In addition, long short-term memory neural network can retain historical learning information and achieve more accurate future predictions. The basic structure of neurons in long short-term memory neural network can be expressed as:

[0063] i t =σ(W Xt X t +W hi h t-1 +b i )

[0064] f t =σ(W Xf X t +W hf h t-1 +b f )

[0065] o t =σ(W Xo X t +W ho h t-1 +b o )

[0066]

[0067] Where i, f, o and c represent the input gate, forget gate, output gate and memory unit respectively; subscript t represents the time step; h is the hidden state; W and b are the weights and biases of different neurons; represents the element-wise product operator; σ and tanh are activation functions. In this work, the input X is composed of SOC, battery pack voltage and average cell voltage.

[0068] Usually, the fully connected layer needs to be concatenated to the long short-term memory layer for output prediction. A fully connected layer can be represented as:

[0069]

[0070] Among them, y and h represent output and input respectively; w and b represent the weight and variance of the i-th neuron in the previous layer respectively; N represents the number of neurons in the previous layer used to transmit information.

[0071] As an optional embodiment, the above S4 specifically includes S41-S42:

[0072] S41: Based on the trained voltage estimation model, real-time battery operation data is input, and real-time estimated voltage is output;

[0073] S42: Calculate the residual between the real-time voltage and the estimated voltage, design a threshold based on the voltage residual, and determine whether a fault occurs.

[0074] As an optional embodiment, the fault threshold in S42 may be set to 0.15V.

[0075] As an optional embodiment, the above S5 specifically includes S51-S52:

[0076] S51: extracting fault sensitive features of the detected faulty part, including voltage change rate, average voltage difference, etc.;

[0077] S52: Construct a two-dimensional fault feature map.

[0078] As an optional embodiment, the calculation formula of the fault sensitivity feature in S51 is as follows:

[0079] By statistically analyzing the voltage of the faulty part, the average absolute value of the voltage change rate aavcr and the voltage difference aavd are extracted as sensitive features of the voltage based on the designed data sliding window dw. Their calculation formulas are as follows:

[0080]

[0081] Wherein, v represents the voltage of the battery cell; the subscript t represents the time when the fault is detected; si represents the sampling period of the battery management system; is the voltage estimated by the LSTM neural network.

[0082] As an optional embodiment, the above S6 specifically includes S61-S62:

[0083] S61: Select an unsupervised clustering algorithm as the target machine learning algorithm for fault identification;

[0084] S62: Clustering the fault feature two-dimensional graph to identify whether the detected fault is a battery fault or an abnormal sensor reading.

[0085] As an optional embodiment, the unsupervised clustering algorithm in S61 may adopt a density-based clustering algorithm with noise (Density-Based Spatial Clustering of Applications with Noise, DBSCAN).

[0086] DBSCAN is a density-based clustering algorithm. Its performance is closely related to the two parameters of the minimum density reachable point MinPts and the cluster radius ε, and it has the advantage of not having to specify the number of clusters in advance. The clustering process of the DBSCAN algorithm is as follows

[0087] 1) Given a domain radius: ε and the minimum number of points within the domain radius to become a core object: MinPts.

[0088] 2) Starting from any point p, mark it as "visited", check whether it is a core point (that is, p's ε neighborhood has at least MinPts objects), if not a core point, mark it as a noise point. Otherwise, create a new cluster C for p, and put all objects in p's ε neighborhood into the candidate set N.

[0089] 3) Iteratively add objects in N that do not belong to other clusters to C. In this process, for the object p′ marked as "unvisited" in N, mark it as "visited" and check its ε neighborhood. If p′ is also a core object, all objects in p′'s ε neighborhood are added to N. Continue to add objects to C until C cannot be expanded, that is, until N is empty. At this point, cluster C is completely generated.

[0090] 4) Randomly select the next unvisited object from the remaining objects and repeat the process of 3) until all objects have been visited.

[0091] Through the above solution, battery faults can be effectively identified to avoid false alarms and misreporting caused by abnormal sensor readings.

[0092] In order to illustrate the effectiveness of the present invention, this embodiment prepares the battery pack operation data of two electric vehicles, such as Figure 2 As shown in the figure, the No. 2 and No. 49 cells of vehicle 1 had self-discharge faults and internal short circuit faults respectively, and the cell voltages of the 11 cells No. 87, No. 90, and No. 100-108 were abnormal due to sensor reading problems, as shown in the figure. Figure 2 As shown in the upper part; the cell voltage of cells 83-89 of vehicle 2 is abnormal due to sensor reading problems, such as Figure 2 The bottom half is shown. After testing, the fault detection methods all detected battery cells with abnormal voltage data, but failed to distinguish whether it was a battery failure or an abnormal sensor reading. Figure 3 It is a two-dimensional graph of features extracted based on different time windows; Figure 4The F1 scores of fault identification based on different algorithm parameters are all greater than 0.96. When identifying voltage anomalies, all voltage anomaly clusters that are concentrated with the battery fault center are considered to be battery faults. In contrast, clusters classified as "other" are considered to indicate abnormal sensor readings. Therefore, it is shown that the present invention can accurately and effectively identify battery faults and abnormal sensor readings, and is universal for different vehicles, which can greatly reduce the risk of false alarms and missed alarms.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A battery fault identification method for actual vehicle working conditions, characterized by: The following steps are involved: S1: Collect battery operation data of electric vehicles, including battery pack parameters and cell voltages of all battery cells and the temperature of temperature probes near them, and establish a battery operation database; S2: Screening out features with high correlation with the cell voltage using a correlation analysis method according to the battery operation database; S3: Building an accurate voltage estimation model based on the features screened in S2; S4: Input the real-time power battery operation data of the electric vehicle into the voltage estimation model trained in S3, and determine whether a fault occurs based on the residual of the real-time voltage and the model estimated voltage; S5: When a fault is detected in S4, a sliding window is designed to extract sensitive features of the fault and construct a two-dimensional fault feature map; S6: Based on the fault feature two-dimensional graph constructed in S5, an unsupervised clustering algorithm is used to identify whether the detected fault is a battery fault or an abnormal sensor reading.

2. The battery fault identification method for actual vehicle working conditions according to claim 1 is characterized in that: In S2, a correlation analysis is performed using the Pearson correlation coefficient or the grayscale correlation coefficient to screen out features whose correlation coefficient with the monomer voltage is greater than a preset value.

3. The battery fault identification method for actual vehicle working conditions according to claim 1 is characterized in that: In S3, a nonlinear regression algorithm is selected as the voltage estimation model, including long short-term memory neural network, convolutional neural network, Gaussian process regression or relevance vector machine regression.

4. The battery fault identification method for actual vehicle working conditions according to claim 1 is characterized in that: In S4, the preset fault threshold is 0.1V, 0.15V or 0.2V.

5. The method for identifying battery faults in a real vehicle operating condition according to claim 1 is characterized in that: In S5, the fault sensitive features extracted include the voltage change rate and the average voltage difference.

6. A battery fault identification method for actual vehicle working conditions according to claim 1, characterized in that: In S6, a density-based clustering algorithm with noise, a Gaussian mixture algorithm, a K-means algorithm or an unsupervised clustering algorithm of fuzzy clustering is used to perform fault identification.

7. The method for identifying battery faults in a real vehicle operating condition according to claim 1 is characterized in that: In S1, the collected battery operation data also includes the SOC, internal resistance, SOH or discharge rate of the battery pack.

8. The battery fault identification method for actual vehicle working conditions according to claim 1 is characterized in that: In S3, a linear regression algorithm of support vector machine regression or ridge regression is used as a voltage estimation model.

9. The battery fault identification method for actual vehicle working conditions according to claim 1 is characterized in that: In S5, the constructed two-dimensional fault feature graph includes features in two dimensions: voltage change rate and average voltage difference.

10. The battery fault identification method for actual vehicle working conditions according to claim 1 is characterized in that: In S6, the faults are divided into two categories: battery faults and abnormal sensor readings according to the clustering results.

Images