Battery pack real-time voltage range estimation and anomaly classification method and system and vehicle

Through the real-time voltage range estimation method and abnormal classification strategy of battery packs based on the limit learning machine, the problem of battery pack voltage range estimation and abnormal detection is solved, and the accurate voltage range estimation and abnormal classification in the absence of data is realized, which improves the electric vehicle and the safety of energy storage systems.

CN118928146APending Publication Date: 2024-11-12SHANDONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410922821.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-10
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

There are safety hazards in actual operation of the battery pack, such as battery short circuit and overvoltage. It is difficult for the existing technology to accurately estimate the voltage range and abnormal classification of the battery pack, especially in the absence of data acquisition costs and data, and there are many types of abnormal causes, which is difficult to Effective detection and classification.

Method used

The real-time voltage range estimation method of battery packs based on the limit learning machine is adopted. There is no need for internal battery modeling. Only a normal discharge or charging process data is required to train an accurate model. The normal voltage range model is trained through the limit learning machine, combined with The voltage upper and lower limit curves extract abnormal characteristics and use cosine similarity to classify abnormalities.

Benefits of technology

It realizes accurate estimation of battery pack voltage range and abnormal classification without increasing hardware redundancy, improves the safety of electric vehicles and energy storage operations, and can effectively distinguish faults such as battery cell undervoltage and sensor sampling abnormalities, making it simple Excellent computing performance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118928146A_ABST
    Figure CN118928146A_ABST
Patent Text Reader

Abstract

The invention provides a battery pack real-time voltage range estimation and anomaly classification method and system and a vehicle, and the method comprises the steps: respectively carrying out the feature extraction of a voltage upper limit and the feature extraction of a voltage lower limit according to the determined upper and lower limit ranges of the voltage of a battery pack, and obtaining two groups of anomaly features; calculating the feature similarity between adjacent monomers in the two groups of abnormal features; and comparing the similarity of the two groups of characteristics with a set threshold value, and determining a voltage anomaly category according to a comparison result, so that the range of the normal working state of the battery pack can be accurately estimated in real time, a reliable safe reference voltage region is provided for safe operation of the battery pack, and anomaly detection and classification of the battery pack can be performed based on the range.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of voltage range estimation and abnormal classification of battery packs, and specifically relates to a method, system and vehicle for real-time voltage range estimation and abnormal classification of battery packs. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] The normal and safe operation of the battery pack is crucial for the driving and energy storage of electric vehicles. However, in the actual operation process, the battery pack will have various safety problems, such as battery short circuit, battery overvoltage, etc., which will lead to safety hazards in the operation of the battery pack. The battery pack voltage value is a parameter that is easy to collect and can intuitively reflect the status of the battery pack. Therefore, the ability to accurately and real-time estimate the battery pack voltage operating range can provide a guarantee for the normal operation of the battery pack. When the battery cell exceeds the normal operating voltage range, it can effectively distinguish the abnormal category and provide a reference for technical personnel in the industry.

[0004] As one of the core components of electric vehicles and energy storage, the safety and reliability of battery packs directly affect the performance and safety of vehicles and energy storage power stations. In actual working conditions, battery packs often have abnormalities, and the abnormal characteristics are difficult to capture. And due to the cost of data acquisition, the data that can be collected by battery packs are generally only voltage, current, temperature, etc. Among them, temperature data is generally the surface temperature data of the entire module, and it is not possible to specify the temperature of the battery cell. The battery packs in electric vehicles or energy storage power stations are generally connected in series, so the current data of each battery cell is consistent. Therefore, the voltage value is generally used for battery abnormality detection. The voltage value is an important and direct parameter reflecting the battery status. Therefore, being able to grasp the real-time working voltage range of the battery pack is an important indicator for evaluating the current battery pack status. In addition, battery abnormality detection is the primary link to provide a basis for the safe operation of electric vehicles and energy storage power stations, but the investigation of abnormal causes is also a field of battery safety research. There are many types of battery pack abnormal causes. In actual working conditions, undervoltage, overvoltage, sensor failure and other causes are most likely to occur. Therefore, abnormality classification without increasing hardware redundancy is an important research direction. Summary of the invention

[0005] In order to solve the above problems, the present invention proposes a battery pack real-time voltage range estimation, abnormality classification method, system and vehicle. The present invention can accurately and real-time estimate the range of normal working conditions of the battery pack, provide a reliable safety reference voltage area for the safe operation of the battery pack, and perform battery pack abnormality detection and classification based on the range.

[0006] According to some embodiments, the present invention adopts the following technical solutions:

[0007] In the first aspect, the present invention provides a method for estimating the real-time voltage range of a battery pack. The method does not require internal battery modeling and does not require a large amount of normal and fault data for training. In the absence of data, only a normal discharge or charging process data is needed to train an accurate estimation model, and the model is used to estimate the upper and lower voltage limits of the battery pack.

[0008] A method for estimating a real-time voltage range of a battery pack comprises the following steps:

[0009] Get battery pack voltage data;

[0010] Using the trained normal voltage range estimation model, estimate the battery pack voltage data to determine the upper and lower limits of the battery pack voltage;

[0011] The training process of the normal voltage range estimation model includes: preprocessing the historical voltage data of the battery pack;

[0012] Filter the data that meets the requirements from the preprocessed data as training data;

[0013] Calibrate the non-fault range voltage of the training data;

[0014] The normal voltage range estimation model based on extreme learning machine is trained using the calibrated training data.

[0015] As an optional implementation, the process of preprocessing the historical voltage data of the battery pack includes using a filtering algorithm to perform data smoothing and noise reduction.

[0016] As an optional implementation, data meeting the requirements is screened from the preprocessed data, and the process of using the training data includes screening at least one fully charged or fully discharged operating condition data whose completeness exceeds a set value, and no abnormality occurs in the cycle process.

[0017] As a further step, voltage data of at least N battery cells are randomly selected as training data, and the value of N is determined according to the number of battery cells in the battery pack.

[0018] As an optional implementation, the process of calibrating the non-fault range voltage of the training data includes:

[0019] For the trained battery pack voltage data U:

[0020]

[0021] Among them, u j,k represents the voltage value of the kth cell at the jth moment, v jis the voltage value of all monomers at the jth moment, and U represents the voltage matrix data;

[0022] Voltage upper limit sequence U up The calculation and calibration formula is:

[0023]

[0024] Among them, max(v j ) is the maximum voltage value of all monomers at the jth moment, The voltage upper limit margin coefficient is set to take into account the normal range of fluctuations in the actual working conditions of the battery;

[0025] Voltage lower limit sequence U down The calculation and calibration formula is:

[0026]

[0027] Among them, min(v j ) is the minimum voltage value of all cells at the jth moment, μ is the voltage lower limit margin coefficient, and the voltage upper limit and voltage lower limit margin coefficients are set according to the charge and discharge cycle conditions.

[0028] As an optional implementation, the process of training a normal voltage range estimation model based on an extreme learning machine includes:

[0029] Randomly initialize the connection weights W between the input layer and the hidden layer in and the bias vector b of the hidden layer, select the activation function;

[0030] Use the training data to calculate the output matrix H of the hidden layer: H = g(W in X+b), where W in is the connection weight between the input layer and the hidden layer, X is the training data sample, b is the bias vector, and g(*) is the activation function;

[0031] Using analytical methods, solve the connection weight W between the output layer and the hidden layer out :W out =Y×H + , where Y is the training output data, H + is the generalized inverse of the analytical method of H;

[0032] The voltage data of the battery pack is used as the input of the model, and the calibrated voltage upper and lower limits U up , U down As the output of the model, estimation model training is performed.

[0033] Correspondingly, a second aspect of the present invention provides a battery pack real-time voltage range estimation system, comprising:

[0034] A data acquisition module is configured to acquire battery pack voltage data;

[0035] A real-time voltage range estimation module is configured to estimate the battery pack voltage data using a trained normal voltage range estimation model to determine an upper and lower limit range of the battery pack voltage;

[0036] The model pre-training module is configured to pre-process the historical voltage data of the battery pack; select the data that meets the requirements from the pre-processed data as training data; calibrate the training data to the non-fault range voltage; and use the calibrated training data to train the normal voltage range estimation model based on the extreme learning machine.

[0037] The third aspect of the present invention provides a method for classifying real-time voltage anomalies of a battery pack. The method extracts abnormal features based on estimated upper and lower limit curves of voltage, and distinguishes the causes of abnormalities of the battery pack based on the extracted abnormal features. The method can effectively extract abnormal feature information and accurately distinguish faults such as battery cell undervoltage and sensor sampling abnormality.

[0038] A method for classifying abnormal real-time voltage of a battery pack comprises the following steps:

[0039] Based on the upper and lower limits of the battery pack voltage determined by the method provided in the first aspect, feature extraction of the upper voltage limit and feature extraction of the lower voltage limit are performed respectively to obtain two groups of abnormal features;

[0040] Calculate the feature similarity between adjacent monomers in two sets of abnormal features;

[0041] The similarities of the two groups of features are compared with the set thresholds, and the voltage anomaly category is determined based on the comparison results.

[0042] As an optional implementation, the process of extracting the feature of the upper voltage limit and the feature of the lower voltage limit respectively includes:

[0043] If the voltage sequence of the battery cell is U a Expressed as: U a =[u1,u2,…,u n ] T , where u1 represents the voltage sequence U a A voltage value of, T is the transpose;

[0044] Characteristic F based on upper voltage limit max The calculation formula is: max =U up -U a ;

[0045] Calculate the voltage upper limit characteristics of each cell in the battery pack and construct the voltage upper limit characteristic matrix Fmax ;

[0046] Characteristic F based on the voltage lower limit min The calculation formula is: min =U a -U down ;

[0047] Calculate the voltage lower limit characteristics of each cell in the battery pack and construct the voltage lower limit characteristic matrix F min .

[0048] As an optional implementation, the process of calculating the feature similarity between adjacent monomers in two groups of abnormal features includes: determining the feature similarity between adjacent monomers by calculating cosine similarity.

[0049] As an optional implementation, the threshold is determined according to the upper and lower limit characteristic values ​​of the voltage. When the voltage of a battery cell exceeds the corresponding upper limit threshold, the upper limit characteristic value of the battery cell has a critical characteristic point where the positive and negative values ​​change.

[0050] When a cell voltage exceeds the corresponding lower limit threshold, the lower limit characteristic value of the cell has a critical characteristic point where the positive and negative values ​​change;

[0051] If the corresponding cosine similarity value exceeds the corresponding threshold value, it is considered that there is an abnormality in the battery of the battery pack.

[0052] As an optional implementation, the similarities of the two sets of features are compared with the set threshold, and the process according to the comparison result includes:

[0053] If the cosine similarity of the voltage upper limit feature group exceeds the corresponding threshold and does not recover, and the cosine similarity of the voltage lower limit feature group is within the normal range, it is an overvoltage anomaly;

[0054] If the cosine similarity of the voltage lower limit feature group exceeds the corresponding threshold and does not recover, and the cosine similarity of the voltage upper limit feature group is within the normal range, it is an undervoltage anomaly;

[0055] If the cosine similarity of the voltage upper limit feature group exceeds the corresponding threshold, and the cosine similarity of the voltage lower limit feature group exceeds the corresponding threshold, but both return to the normal range, the sensor is abnormal.

[0056] In a fourth aspect, the present invention provides a battery pack real-time voltage anomaly classification system, comprising:

[0057] The feature extraction module is configured to extract the feature of the upper voltage limit and the feature of the lower voltage limit respectively based on the determined upper and lower limit ranges of the battery pack voltage, so as to obtain two groups of abnormal features;

[0058] A similarity calculation module is configured to calculate feature similarities between adjacent monomers in two sets of abnormal features;

[0059] The anomaly classification module is configured to compare the similarities of two sets of features with a set threshold, and determine the voltage anomaly category based on the comparison result.

[0060] In a fifth aspect, the present invention provides a battery pack control system, comprising the above-mentioned system or comprising a memory and a processor and computer instructions stored in the memory and running on the processor, wherein when the computer instructions are executed by the processor, the steps in the above-mentioned method are completed.

[0061] In a sixth aspect, the present invention provides an electric vehicle comprising the above-mentioned system or comprising a memory and a processor and computer instructions stored in the memory and running on the processor, wherein when the computer instructions are executed by the processor, the steps in the above-mentioned method are completed.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] The present invention innovatively proposes a real-time voltage range estimation method for a battery pack based on an extreme learning machine, and based on the range proposes a battery pack abnormal feature analysis and abnormality classification method, which can provide a reliable safety reference voltage area for the safe operation of the battery pack and classify the causes of abnormalities without adding hardware redundancy.

[0064] The present invention proposes a method for estimating the upper and lower limits of battery pack voltage. The method does not require internal battery modeling and does not require a large amount of normal and fault data for training. Only a small amount of data is needed to train an accurate estimation model.

[0065] Aiming at the problem that the abnormal characteristics of battery packs are hidden in the acquisition noise and it is difficult to directly detect abnormalities, the present invention proposes an abnormal characteristic extraction method based on the voltage upper and lower limit curves. The method can effectively extract abnormal characteristic information and can be combined with some fault location methods to perform abnormal detection and location. The method is simple to calculate and has good performance in battery pack data under actual working conditions.

[0066] The present invention proposes an abnormality classification strategy based on cosine similarity, which can effectively distinguish faults such as battery cell undervoltage and sensor sampling abnormality, and realize accurate abnormality classification.

[0067] The present invention can effectively improve the safety of electric vehicles during driving and energy storage operations.

[0068] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0070] Figure 1 is a specific flow chart of voltage upper and lower limit estimation according to an embodiment of the present invention;

[0071] Figure 2 It is a schematic diagram of anomaly detection according to an embodiment of the present invention. DETAILED DESCRIPTION

[0072] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0073] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0074] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0075] In the absence of conflict, the embodiments in this application and the features in the embodiments may be combined with each other.

[0076] Embodiment 1

[0077] A method for estimating the real-time voltage range of a battery pack, such as Figure 1 As shown, the following steps are included:

[0078] Step 1: Collect battery pack voltage data and construct a voltage matrix.

[0079] Step 2: Perform noise reduction on the dataset.

[0080] This embodiment uses Gaussian filtering to perform data smoothing and noise reduction.

[0081] Of course, in other embodiments, other preprocessing methods may be used for noise reduction.

[0082] Step 3: Training set division. Due to the complexity of actual operating conditions, not every charge and discharge cycle has full charge and discharge conditions. Most of the data is fragmentary, especially the data during the discharge process. In addition, in order to reduce the computational load consumption of the model, it is necessary to use lightweight data for training as much as possible.

[0083] In this embodiment, the training data needs to be screened first, mainly the charging and discharging cycles with relatively complete full charge or full discharge, and no abnormality occurs in the process. Such data is easy to obtain in actual working conditions.

[0084] Step 4: Normal voltage range calibration method in training set, that is, calibration method of training labels. Although battery packs are generally composed of the same type of cells, there are still slight inconsistencies in the battery pack. The impact of such inconsistency on the performance of the battery pack is generally negligible. That is, the cell parameters of the battery pack are concentrated in a smaller range, and the parameters of these cells are considered normal if they belong to this range at the current moment. However, this real-time normal voltage range is difficult to accurately obtain under actual working conditions, because the actual working conditions are full of randomness. Therefore, the present invention is obtained by calibration from a normal charge and discharge cycle, that is, by calibration from the training set selected in step 4.

[0085] Take a normal charge and discharge cycle as an example to illustrate the calibration method of the normal voltage range. Assume there is a set of battery pack voltage data U:

[0086]

[0087] Among them, u j,k represents the voltage value of the kth cell at the jth moment, v j is the voltage value of all monomers at the jth moment, and U represents the voltage matrix data.

[0088] Then the voltage upper limit sequence U of the battery pack in the training set is up It can be calculated and calibrated by the following formula

[0089]

[0090] Among them, max(v j ) is the maximum voltage value of all monomers at the jth moment, is the voltage upper limit margin coefficient, which is a real number slightly greater than 1. The margin coefficient is set to take into account the normal range of fluctuations in the actual working conditions of the battery.

[0091] Similarly, the voltage lower limit sequence U of the battery pack in the training set down It can be calculated by the following formula

[0092]

[0093] Among them, min(v j ) is the minimum voltage value of all cells at the jth moment, and μ is the voltage lower limit margin coefficient, which is a real number slightly less than 1. The margin coefficients of the voltage upper limit and the voltage lower limit can be set according to the complexity of the charge and discharge cycle conditions.

[0094] Step 5: Normal voltage range estimation model training based on extreme learning machine. Extreme Learning Machine (ELM) is a single hidden layer feedforward neural network, which has some unique features compared to traditional neural networks. The main idea of ​​ELM is to randomly initialize the connection weights between the input layer and the hidden layer and the bias of the hidden layer neurons during the training process, and then use analytical methods to directly solve the connection weights between the output layer and the hidden layer. This direct way of solving the weights makes the training process very fast and does not require iterative adjustments. In actual working conditions, lightweight models are also more needed to reduce the computational load.

[0095] Assume that there is a battery pack training voltage dataset containing N samples, each sample has d features, denoted as X = [x1, x2, ..., x N ] T , the corresponding output is Y = [y1,y2,…,y N ] T , the goal of building the model is to learn a single hidden layer neural network model to estimate the label data.

[0096] The ELM solution process can be briefly divided into the following steps:

[0097] Initialization parameters: Randomly initialize the connection weight W between the input layer and the hidden layer in and the bias vector b of the hidden layer, where W in The size of is L×d, and the size of b is L×1. Select the activation function g(z), which is usually a nonlinear function such as Sigmoid function, ReLU function, etc. In this paper, Sigmoid function is selected.

[0098] Calculate the hidden layer output: Use the training data set to calculate the output matrix H of the hidden layer. The calculation formula is as follows

[0099] H=g(W in X+b)

[0100] Among them, W in is the connection weight between the input layer and the hidden layer, X is the training set sample, b is the bias vector, and g(*) is the Sigmoid activation function.

[0101] Solving for the output weights:

[0102] Use analytical methods, such as Moore-Penrose generalized inverse, to directly solve the connection weight W between the output layer and the hidden layer. out The connection weight W between the output layer and the hidden layer out The calculation formula is

[0103] W out =Y×H +

[0104] Among them, Y is the training output data, H + is the Moore-Penrose generalized inverse of H.

[0105] Save model: Based on the training steps of the comprehensive extreme learning machine, the present invention uses the voltage data of the battery pack as the input of the model, and the calibrated voltage upper and lower limits U up , U down As the output of the model, the estimated model training is performed. The trained model is saved to facilitate the subsequent use of the obtained weights for model estimation or cloud deployment.

[0106] With its efficient training speed and simple implementation, ELM has advantages in processing large-scale data sets and applications with high real-time requirements.

[0107] Therefore, this embodiment uses this method to construct a real-time voltage range estimation method for a battery pack. The battery pack in actual working conditions is generally composed of different numbers of battery cells connected in series. Therefore, the number of inputs to the model can be adjusted according to actual conditions. The battery pack in an electric vehicle is generally composed of multiple battery cells. When training the model, considering the computational burden, 1 / 3 of the battery cell voltage data can be randomly selected as input training.

[0108] However, in order to ensure the training accuracy, this embodiment requires voltage data of at least 5 cells for training.

[0109] Step 6: Estimation of voltage upper and lower limits. Use the trained estimation model to estimate the voltage of the battery pack to be tested. The input of the model estimation is the voltage data of the battery pack to be tested.

[0110] Of course, in some embodiments, steps 2 to 5 in this embodiment are all preparatory work, which can be loaded or executed in advance on the execution subject (such as a processor, user-end device, etc.) to save the preprocessed model. In actual operation, it is only necessary to obtain voltage data and call the model to estimate the upper and lower limits of the voltage, that is, steps 1 and 6 of this embodiment.

[0111] Embodiment 2

[0112] Based on the method provided in the first embodiment, this embodiment provides a battery pack abnormality classification method based on the voltage upper and lower limit curves, such as Figure 2 As shown, the following steps are included:

[0113] Step 1: Feature extraction is divided into two parts, the first part is the feature extraction of the voltage upper limit, and the second part is the feature extraction of the voltage lower limit. That is, the abnormal features are divided into two groups.

[0114] Assume there is a cell voltage sequence U a , which is expressed as:

[0115] U a =[u1,u2,…,u n ] T

[0116] Where u1 represents the voltage sequence U a A voltage value of , T is the transpose.

[0117] Characteristic F based on upper voltage limit max The calculation formula is:

[0118] F max =U up -U a

[0119] If the battery pack has n cells, the above formula is used to calculate and construct the voltage upper limit characteristic matrix F max .

[0120] Similarly, based on the characteristic F of the voltage lower limit min The calculation formula is:

[0121] F min =U a -U down

[0122] If the battery pack has n cells, the above formula is used to calculate and construct the voltage lower limit characteristic matrix F min .

[0123] In summary, the establishment of F max 、F min After obtaining the characteristic matrix, the battery pack status analysis can be performed. max is the upper voltage limit U up Therefore, when the voltage of a cell exceeds the threshold T, the characteristic value of the cell will change from positive to negative, and there is a critical characteristic point where the positive and negative values ​​change. min The voltage lower limit U of each battery cell downTherefore, when a cell exceeds the threshold T, the characteristic value of the cell will also change from positive to negative, and there is also a critical characteristic point where the positive and negative values ​​change. This critical characteristic point can be used as a key feature for anomaly detection.

[0124] Of course, it should be noted that, for simplicity of expression, the thresholds are all expressed as T, but in actual application, the upper threshold and the lower threshold are not necessarily equal or the same.

[0125] Step 2: Anomaly classification method based on cosine similarity. Cosine similarity is a commonly used similarity measurement method, which is used to measure the directional similarity between two vectors. It is widely used in information retrieval, text similarity calculation, recommendation system and other fields.

[0126] This embodiment introduces cosine similarity as a basis for abnormality detection and classification of battery packs. The principle of abnormality calculation using cosine similarity is briefly described below.

[0127] Suppose there are two feature sequences F a and F b , this feature sequence is taken as the feature values ​​of two adjacent cells within a sliding window size. They are respectively expressed as:

[0128] F a =[a1,a2,…,a j ]

[0129] F b =[b1,b2,…,b j ]

[0130] In the formula, a j represents the characteristic value of cell a at time j, b j Represents the characteristic value of cell b at the jth moment.

[0131] The cosine similarity CS(F a ,F b ) is defined as their inner product divided by the product of their moduli, which can be expressed as the following formula:

[0132]

[0133] Among them, F a ·F b is the characteristic sequence F a and F b The dot product of ‖F a ‖ and ‖F b ‖ are the characteristic sequences F a and F b For F a ·F bThe sensitivity of the positive and negative values ​​of the value in a calculated sliding window is related to the input variable. a ·F b There will be changes from positive to negative values ​​during the calculation process.

[0134] Based on this, CS(F a ,F b ) will be used as the key parameter for battery pack anomaly detection and classification. By calculating two sets of abnormal features F max 、F min CS(F a ,F b ) value and combined with the threshold T to detect and classify battery pack anomalies. a ,F b ) value exceeds the threshold T, further abnormal classification judgment will be performed.

[0135] There are many reasons for battery pack abnormality. Undervoltage, overvoltage, sensor failure and other reasons are most likely to occur in actual working conditions. The present invention classifies battery pack abnormalities according to these three types of faults and classifies them based on two groups of abnormal features F max 、F min and CS(F a ,F b ) value proposes an anomaly classification strategy.

[0136] Overvoltage abnormality classification strategy: The voltage value of the battery cell with overvoltage abnormality in the battery pack will exceed that of the normal battery cell, and will increase. That is, it will exceed the proposed voltage upper limit value. In the abnormality detection result, the feature F based on the voltage upper limit max The alarm will be exceeded. a ·F b If it is a negative value, the alarm signal will not return to the normal threshold.

[0137] Undervoltage abnormal classification strategy: Undervoltage of the battery cell is generally caused by micro short circuit, aging, etc. Therefore, the voltage of this type of fault will be lower than that of other normal batteries in the battery pack, and will show a decreasing trend. That is, it will exceed the proposed voltage lower limit. In the abnormal detection result, the feature F based on the voltage lower limit will be min The alarm will be exceeded.

[0138] Sensor abnormality classification strategy: Sensor abnormality will cause the transmitted voltage value to drop suddenly, and this type of voltage mutation value is more sudden than the characteristics of micro-short-circuit cells, and some changes can reach 0.3V. But the sensor reading will return to normal immediately. This is due to F a ·F bThere is a change from positive to negative and then back to positive, so the alarm signal will return to the normal threshold. The voltage of the micro-short-circuited cell will gradually decrease, and the voltage value will be farther and farther away from the voltage value range of the other normal cells in the battery pack. a ·F b If the value is negative, the alarm signal will not return to the normal threshold. The same is true for the characteristics of battery cell overvoltage.

[0139] The specific anomaly classification strategy is shown in Table 1.

[0140] Table 1 Anomaly classification strategy

[0141]

[0142] Among them, CS max Represents the cosine similarity of the voltage upper limit feature group; CS min Represents the cosine similarity of the voltage lower limit feature group.

[0143] Embodiment 3

[0144] A battery pack real-time voltage range estimation system, comprising:

[0145] A data acquisition module is configured to acquire battery pack voltage data;

[0146] A real-time voltage range estimation module is configured to estimate the battery pack voltage data using a trained normal voltage range estimation model to determine an upper and lower limit range of the battery pack voltage;

[0147] The model pre-training module is configured to pre-process the historical voltage data of the battery pack; select the data that meets the requirements from the pre-processed data as training data; calibrate the training data to the non-fault range voltage; and use the calibrated training data to train the normal voltage range estimation model based on the extreme learning machine.

[0148] Embodiment 4

[0149] A battery pack real-time voltage abnormality classification system, comprising:

[0150] The feature extraction module is configured to extract the feature of the upper voltage limit and the feature of the lower voltage limit respectively based on the determined upper and lower limit ranges of the battery pack voltage, so as to obtain two groups of abnormal features;

[0151] A similarity calculation module is configured to calculate feature similarities between adjacent monomers in two sets of abnormal features;

[0152] The anomaly classification module is configured to compare the similarities of two sets of features with a set threshold, and determine the voltage anomaly category based on the comparison result.

[0153] Embodiment 5

[0154] A battery pack control system includes the system described in Example 3 or / and Example 4 or includes a memory and a processor and computer instructions stored in the memory and running on the processor, and when the computer instructions are executed by the processor, the steps in the method mentioned in Example 1 or Example 2 are completed.

[0155] Embodiment 6

[0156] An electric vehicle comprises the system described in Example 3, Example 4 or / and Example 5, or comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the steps in the method mentioned in Example 1 or Example 2 are completed.

[0157] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0158] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0159] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0160] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0161] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made by those skilled in the art within the spirit and principle of the present invention without creative labor shall be included in the protection scope of the present invention.

Claims

1. A method for estimating a real-time voltage range of a battery pack, characterized in that: The following steps are involved: Get battery pack voltage data; Using the trained normal voltage range estimation model, estimate the battery pack voltage data to determine the upper and lower limits of the battery pack voltage; The training process of the normal voltage range estimation model includes: preprocessing the historical voltage data of the battery pack; Filter the data that meets the requirements from the preprocessed data as training data; Calibrate the non-fault range voltage of the training data; The normal voltage range estimation model based on extreme learning machine is trained using the calibrated training data.

2. A method for estimating a real-time voltage range of a battery pack as claimed in claim 1, characterized in that: The process of preprocessing the battery pack historical voltage data includes using a filtering algorithm to perform data smoothing and noise reduction; The process of selecting data that meets the requirements from the preprocessed data as training data includes selecting at least one full-charge or full-discharge operating condition data whose completeness exceeds a set value, and no abnormality occurs in the cycle process; Or further, voltage data of at least N battery cells are randomly selected as training data, and the value of N is determined according to the number of battery cells in the battery pack.

3. A method for estimating a real-time voltage range of a battery pack as claimed in claim 1, characterized in that: The process of calibrating the non-fault range voltage of the training data includes: For the trained battery pack voltage data U: Among them, u j,k represents the voltage value of the kth cell at the jth moment, v j is the voltage value of all monomers at the jth moment, and U represents the voltage matrix data; Voltage upper limit sequence U up The calculation and calibration formula is: Among them, max(v j ) is the maximum voltage value of all monomers at the jth moment, The voltage upper limit margin coefficient is set to take into account the normal range of fluctuations in the actual working conditions of the battery; Voltage lower limit sequence U down The calculation and calibration formula is: Among them, min(v j ) is the minimum voltage value of all cells at the jth moment, μ is the voltage lower limit margin coefficient, and the voltage upper limit and voltage lower limit margin coefficients are set according to the charge and discharge cycle conditions.

4. A method for estimating a real-time voltage range of a battery pack as claimed in claim 1, characterized in that: The process of training the normal voltage range estimation model based on extreme learning machine includes: Randomly initialize the connection weights W between the input layer and the hidden layer in and the bias vector b of the hidden layer, select the activation function; Use the training data to calculate the output matrix H of the hidden layer: H = g(W in X+b), where W in is the connection weight between the input layer and the hidden layer, X is the training data sample, b is the bias vector, and g(*) is the activation function; Using analytical methods, solve the connection weight W between the output layer and the hidden layer out :W out =Y×H + , where Y is the training output data, H + is the generalized inverse of the analytical method of H; The voltage data of the battery pack is used as the input of the model, and the calibrated voltage upper and lower limits U up , U down As the output of the model, estimation model training is performed.

5. A battery pack real-time voltage range estimation system, characterized in that: include: A data acquisition module is configured to acquire battery pack voltage data; A real-time voltage range estimation module is configured to estimate the battery pack voltage data using a trained normal voltage range estimation model to determine an upper and lower limit range of the battery pack voltage; The model pre-training module is configured to pre-process the historical voltage data of the battery pack; and select data meeting the requirements from the pre-processed data as training data; Calibrate the non-fault range voltage of the training data; The normal voltage range estimation model based on extreme learning machine is trained using the calibrated training data.

6. A method for classifying real-time abnormal voltage of a battery pack, characterized in that: The following steps are involved: Based on the upper and lower limits of the battery pack voltage determined by the method provided in any one of claims 1 to 4, feature extraction of the upper voltage limit and feature extraction of the lower voltage limit are performed respectively to obtain two sets of abnormal features; Calculate the feature similarity between adjacent monomers in two sets of abnormal features; The similarities of the two groups of features are compared with the set thresholds, and the voltage anomaly category is determined based on the comparison results.

7. A method for classifying real-time abnormal voltage of a battery pack as claimed in claim 6, characterized in that: The process of extracting the features of the upper voltage limit and the lower voltage limit respectively includes: If the voltage sequence of the battery cell is U a Expressed as: U a =[u1,u2,…,u n ] T , where u1 represents the voltage sequence U a A voltage value of, T is the transpose; Characteristic F based on upper voltage limit max The calculation formula is: max =U up -U a ; Calculate the voltage upper limit characteristics of each cell in the battery pack and construct the voltage upper limit characteristic matrix F max ; Characteristic F based on the voltage lower limit min The calculation formula is: min =U a -U down ; Calculate the voltage lower limit characteristics of each cell in the battery pack and construct the voltage lower limit characteristic matrix F min .

8. A method for classifying real-time abnormal voltage of a battery pack as claimed in claim 6, characterized in that: The process of calculating the feature similarity between adjacent monomers in two sets of abnormal features includes: determining the feature similarity between adjacent monomers by calculating cosine similarity.

9. A method for classifying real-time abnormal voltage of a battery pack as claimed in claim 6, characterized in that: The threshold is determined according to the upper and lower characteristic values ​​of the voltage. When the voltage of a battery cell exceeds the corresponding upper threshold, the upper characteristic value of the battery cell has a critical characteristic point where the positive and negative values ​​change. When a cell voltage exceeds the corresponding lower limit threshold, the lower limit characteristic value of the cell has a critical characteristic point where the positive and negative values ​​change; If the corresponding cosine similarity value exceeds the corresponding threshold value, it is considered that there is an abnormality in the battery of the battery pack.

10. A method for classifying real-time abnormal voltage of a battery pack as claimed in claim 6, characterized in that: Compare the similarity of the two sets of features with the set threshold. The process based on the comparison results includes: If the cosine similarity of the voltage upper limit feature group exceeds the corresponding threshold and does not recover, and the cosine similarity of the voltage lower limit feature group is within the normal range, it is an overvoltage anomaly; If the cosine similarity of the voltage lower limit feature group exceeds the corresponding threshold and does not recover, and the cosine similarity of the voltage upper limit feature group is within the normal range, it is an undervoltage anomaly; If the cosine similarity of the voltage upper limit feature group exceeds the corresponding threshold, and the cosine similarity of the voltage lower limit feature group exceeds the corresponding threshold, but both return to the normal range, the sensor is abnormal.

11. A battery pack real-time voltage anomaly classification system, characterized in that: include: The feature extraction module is configured to extract the feature of the upper voltage limit and the feature of the lower voltage limit respectively based on the determined upper and lower limit ranges of the battery pack voltage, so as to obtain two groups of abnormal features; A similarity calculation module is configured to calculate feature similarities between adjacent monomers in two sets of abnormal features; The anomaly classification module is configured to compare the similarities of two sets of features with a set threshold, and determine the voltage anomaly category based on the comparison result.

12. A battery pack control system, characterized in that: A system comprising the system of claim 5 or 11 or comprising a memory and a processor and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the steps in the method of any one of claims 1-4 or 6-10 are completed.

13. An electric vehicle, characterized in that: A system comprising the system of claim 5, 11 or 12 or comprising a memory and a processor and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the steps in the method of any one of claims 1-4 or 6-10 are completed.