Multi-working-condition power battery thermal runaway early warning method based on vehicle cloud system

Abstract

The invention provides a power battery thermal runaway safety early warning method based on vehicle cloud cooperation. The method is mainly characterized in that a real-time estimated vehicle-mounted terminal and a cloud control platform of which the state is updated regularly work cooperatively, thermal runaway early warning is realized, and the method comprises the following steps: step 1, a vehicle end collects real vehicle data, extracts related characteristics through a sub-calculation interval matrix at set intervals, and uploads the related characteristics to a cloud end in real time; step 2, the cloud end receives the characteristics calculated by the vehicle end, calculates the health degree of each single battery by using a characteristic set, and divides fault types; and step 3, the fault information calculated by the cloud end is sent to the vehicle end in real time through the server, and the vehicle end gives an alarm and reminds a driver of timely maintenance. And the real-time performance of thermal runaway early warning is improved through a vehicle-cloud cooperation method.

Description

Technical Field

[0001] The present invention belongs to the technical field of power batteries, and in particular relates to a method for early warning of thermal runaway of power batteries under multiple working conditions based on a vehicle-to-cloud system. Background Art

[0002] With growing environmental awareness, more people are realizing the seriousness of greenhouse gas emissions. Consequently, new energy vehicles have become a hot topic, explaining the growing popularity of low-emission vehicles such as electric vehicles (EVs), hybrid electric vehicles (HEVs), and fuel cell hydrogen electric vehicles (FCEVs). The application of power batteries in transportation has also become a hot topic. However, in recent years, frequent safety issues involving power battery fires due to thermal runaway have severely hampered the rapid, healthy, and sustainable development of the industry.

[0003] The battery management system (BMS) provides functions such as battery state estimation, thermal management, balancing management, and fault diagnosis, and is key to ensuring stable, efficient, and safe battery operation. However, BMSs have low storage capacity, limited computing power, and difficulty modeling onboard batteries, limiting them to simple functions such as performance monitoring and state estimation. Therefore, with the advent of the big data era, high-performance cloud-based computing platforms have been established to store data throughout the vehicle lifecycle and enable long-term monitoring. However, real-world vehicle operating conditions are complex and varied, and fault types are unknown. As data is uploaded, cloud servers require significant time to train algorithms on the raw data, potentially leading to serious accidents.

[0004] The application number is 202410607171.4, and the invention is titled "A multi-time-scale integrated method for diagnosing thermal runaway of lithium-ion batteries throughout their life cycle," including using an electrochemical impedance spectroscopy sampling chip for online real-time detection of electrochemical impedance spectroscopy to obtain EIS impedance data for each single cell in the battery pack; a clustering model on the vehicle side, based on the EIS impedance data, provides a short-time-scale, real-time alarm for thermal runaway of each single cell in the battery pack; and a battery management system on the vehicle side transmits the real-time collected battery data to a cloud platform for long-time-scale prediction of thermal runaway of each single cell in the battery pack in the cloud. On a short-time scale, the present invention relies less on parameterization and has higher theoretical accuracy; on a long-time scale, it highlights the prospect of accurately predicting future battery system thermal runaway failures based on past and current data.

[0005] However, this invention mainly combines short-time-scale real-time alarms with long-time-scale fault predictions. However, the short-time-scale alarms are completed independently on the vehicle side, and the long-time pre-test is based on the short-time predictions. However, the computing power of the on-board system is limited and cannot meet the needs of early alarms. The cloud can achieve long-time predictions, but the cloud computing requires a lot of time and cannot achieve real-time predictions. Summary of the Invention

[0006] The present invention aims to address the shortcomings of the aforementioned prior art by providing a multi-operating-condition thermal runaway warning method for power batteries based on vehicle-cloud collaboration. This method, through the collaborative work of the vehicle and the cloud, monitors vehicle safety in real time. The vehicle collects data in real time, selecting and calculating relevant features suitable for cloud-based fault diagnosis. Through vehicle-side feature engineering optimization and leveraging the computing power of the cloud, the health of the data uploaded by the vehicle is assessed and fault types are classified, thereby providing more accurate predictions and warnings for battery life.

[0007] To ensure data timeliness, this invention leverages the limited computing power of the BMS. The vehicle uses the collected data to calculate relevant features and upload them to the cloud. The cloud then uses a hybrid clustering method to score the battery pack's health and classify fault types. If a fault occurs, an immediate vehicle-side alarm is issued, providing early warning of thermal runaway.

[0008] The present invention adopts the following technical solutions:

[0009] A method for early warning of thermal runaway of a power battery based on vehicle-cloud collaboration, comprising:

[0010] S1. There is a vehicle data matrix A k×n , each sampling moment will generate a new row of data. To avoid data lag, feature recognition is required on the vehicle side. Extract the latest m rows of historical data into a matrix C m×n Internal data is used to calculate relevant features, including voltage and temperature differences, mean deviation, and sample entropy eigenvalues. An equivalent circuit model is then established to dynamically identify and correct open-circuit voltages. Internal resistance features are extracted online on the vehicle using recursive least squares with forgetting factors. All feature information is sent to the cloud for fault identification at intervals corresponding to the length of each matrix interval.

[0011] S2. Receive the feature matrix uploaded by the vehicle. First, adjust the sign of each feature so that its correlation with battery life is negative. Then, divide the seven features calculated on the vehicle into sub-feature sets of arbitrary dimensions. Use a hybrid clustering-based method on the cloud platform to calculate the health of individual batteries in real time, set a danger threshold, and identify any low-scoring batteries.

[0012] On the cloud platform, a hybrid clustering method is used to calculate the health of single cells in real time: several common clustering methods are selected, including fuzzy C-means clustering, K-means clustering, DBSCAN, and Gaussian mixture clustering. Different clustering methods are used based on different sub-features to divide all single cells into different categories:

[0013]

[0014] In the formula, cell i is the i-th monomer in the current category; a, b, and c are the number of monomers contained in the three categories Y1, Y2, and Y3 respectively.

[0015] Calculate the characteristic average of all monomers in each class:

[0016]

[0017] Where R ip is the pth eigenvalue of the i-th monomer; x is the number of features contained in the sub-feature set in the current clustering process, are the feature averages of categories Y1, Y2, and Y3 respectively.

[0018] The health of the monomer in the class with the largest feature average value is set to 0, and the health of the other classes is set to 1, which is the result of a single clustering. Let the number of times all sub-feature monomers i are in state 1 be F i , traverse n clustering methods and λ cycles, and the health degree S of monomer i in the fault matrix can be obtained i ;

[0019]

[0020] Where i is the number of single cells and s is the characteristic number.

[0021] Power battery failures that cause thermal runaway are categorized as internal or external short circuit failures, inconsistency failures, and abnormal resistance failures. Temperature-related characteristics, temperature and voltage-related characteristics, and temperature and internal resistance-related characteristics are evaluated as three types of indicators, and the failure types are classified in the cloud based on the evaluation results. Finally, the single-cell battery failure information calculated in the cloud is sent to the vehicle via the server in real time, which issues an alarm and reminds the driver to conduct prompt repairs.

[0022] S3. Use vehicle data that has experienced thermal runaway to verify the accuracy and feasibility of the algorithm.

[0023] Furthermore, the difference, average deviation and sample entropy eigenvalue of the voltage and temperature in the matrix extracted in step S1 include:

[0024] The difference d between the voltage and temperature of the data x of the i-th cell in the battery pack i, average deviation i And the sample entropy SampEn eigenvalues ​​are defined as follows:

[0025]

[0026] in are the maximum, minimum and average characteristics of the voltage and temperature of monomer i in the matrix; n is the total number of monomers; r is the measurement window of the original sequence; A α (r) and B α (r) represents the probability that two sequences match α and α+1 points respectively under the same r;

[0027] Furthermore, in step S1, an equivalent circuit model is established, and dynamically identifying and correcting the open circuit voltage includes:

[0028] The first-order Thevenin equivalent circuit model is selected as the research mathematical model. The state equation obtained according to the equivalent circuit model is:

[0029]

[0030] Where I is the current inside the circuit, here we use the total current of the vehicle, U t is the terminal voltage of the circuit, U0 is the voltage across R0, U oc is the open circuit voltage, U p Used to represent the voltage across an RC circuit.

[0031] Furthermore, the least squares method is used to identify the model parameters online, including the internal resistance R0, the polarization internal resistance R p and polarization capacitance C p , and based on the above circuit state equation, the specific implementation process is as follows:

[0032] Suppose the system equation at sampling time k is:

[0033]

[0034] Where ε is the noise variable, is the information vector, θ is the parameter vector, and y is the system output. According to the basic theory of the least squares algorithm, after deduction and transformation, the recursive expression can be obtained as follows:

[0035]

[0036] Where K is the gain, is the estimated value of the parameter vector θ; e is the prediction error of the terminal voltage; P is the covariance matrix; λ is the forgetting factor, which ranges from [0.95, 1].

[0037] According to the above expression, the terminal voltage of the equivalent circuit can be obtained

[0038]

[0039] Among them I L is the load current. s is the secondary variable of the transfer function, and its bilinear transformation is:

[0040]

[0041] Where q represents the discretization operator, T s Indicates the sampling time. The difference between the terminal voltage and the open circuit voltage E t for

[0042] E t =U t -U ocv

[0043] Get the transfer function:

[0044]

[0045] Through the subvariable s G s Perform discretization:

[0046]

[0047] Among them, the α1, α2, and α3 coefficients in the numerator and denominator are defined as:

[0048]

[0049] Among them, R0, R p 、C p are the internal resistance, polarization internal resistance and polarization capacitance of the equivalent circuit model respectively.

[0050] In adjacent sampling times, the open circuit voltage OCV can be regarded as a constant

[0051]

[0052] The formula for obtaining the real-time terminal voltage is:

[0053] U t，k =α1U t，k-1 +(1-α1)I OCV，k +α2I L，k +α3I L，k-1

[0054] Let τ = R p C p , the parameter identification results are:

[0055]

[0056] Based on the above results, the real-time calculated internal resistance characteristics can be extracted on the vehicle, and all characteristics, including the voltage and temperature difference, deviation and sample entropy, can be uploaded to the cloud platform to calculate whether there is a fault.

[0057] Furthermore, the evaluation results of the hybrid clustering in step S2 are divided into the following fault types in the cloud:

[0058] For all battery cells, health evaluation is performed based on temperature characteristics, voltage and temperature characteristics, and voltage and internal resistance characteristics, and health levels S1, S2, and S3 are obtained. When a cell has a health level lower than 60, it is considered a fault state, with the fault state set to 1 and the normal state set to 0. The method for classifying fault types is as follows:

[0059] when When only the temperature is abnormal, it is generally caused by thermal runaway transfer around the cell or abnormal thermal management system, which has nothing to do with the current cell i. There is no abnormality in the cell, and the battery system needs to be checked.

[0060] when When , it indicates that the cell consistency is poor and the battery module has an inconsistent fault.

[0061] when or The internal resistance of each cell is inconsistent or the internal resistance increases due to temperature rise, and the battery module has abnormal internal resistance failure.

[0062] when When the battery is detected, there is a micro short circuit or external short circuit fault in the single cell, and it is determined that the battery module has an internal or external short circuit fault.

[0063] Beneficial effects:

[0064] This invention utilizes a vehicle-cloud collaboration approach to convert basic vehicle parameters into eigenvalues ​​that can be directly utilized by cloud-based algorithms, thereby reducing the amount of data transmitted to the cloud and optimizing the real-time and accuracy of fault diagnosis. Vehicle-side processing simplifies data and reduces transmission latency, while advanced cloud-based algorithms utilize these eigenvalues ​​for precise analysis, enhancing the accuracy of fault identification. Compared to traditional long- and short-term warning methods, this invention not only reduces false alarms and missed alarms but also enables faster response to complex faults, improving the system's adaptability and scalability. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 This is the overall flow chart of the present invention, (a) is the interaction diagram between the vehicle and the cloud, (b) is the cloud flow chart;

[0066] Figure 2 Schematic diagram of a first-order RC equivalent circuit model in an embodiment of the present invention;

[0067] Figure 3 This is a schematic diagram of fault matrix division in an embodiment of the present invention;

[0068] Figure 4 The feature map extracted from the real vehicle data in the embodiment of the present invention;

[0069] Figure 5 This is a health scoring chart for a power battery vehicle in an embodiment of the present invention. DETAILED DESCRIPTION

[0070] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention are described clearly and completely below. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0071] like Figure 1-5 As shown, a method for early warning of thermal runaway of a power battery based on vehicle-cloud collaboration of the present invention includes:

[0072] S1. There is a vehicle data matrix A k×n , each sampling moment will generate a new row of data. To avoid data lag, feature recognition is required on the vehicle side. Extract the latest m rows of historical data into a matrix C m×n Internal data is used to calculate relevant features, including voltage and temperature differences, mean deviation, and sample entropy eigenvalues. An equivalent circuit model is then established to dynamically identify and correct open-circuit voltages. Internal resistance features are extracted online on the vehicle using recursive least squares with forgetting factors. All feature information is sent to the cloud for fault identification at intervals corresponding to the length of each matrix interval.

[0073] The difference, average deviation and sample entropy eigenvalue of the voltage and temperature in the matrix extracted in step S1 include:

[0074] The difference d between the voltage and temperature of the data x of the i-th cell in the battery pack i , average deviation i And the sample entropy SampEn eigenvalues ​​are defined as follows:

[0075]

[0076] in are the maximum, minimum and average characteristics of the voltage and temperature of monomer i in the matrix; n is the total number of monomers; r is the measurement window of the original sequence; A α (r) and B α(r) represents the probability that two sequences match α and α+1 points respectively under the same r;

[0077] In step S1, an equivalent circuit model is established, and dynamic identification and correction of the open circuit voltage include:

[0078] The equivalent circuit model is a first-order RC model. The first-order Thevenin equivalent circuit model is selected as the research mathematical model. The state equation obtained according to the equivalent circuit model is:

[0079]

[0080] Where I is the current inside the circuit, here we use the total current of the vehicle, U t is the terminal voltage of the circuit, U0 is the voltage across R0, U oc is the open circuit voltage, U p Used to represent the voltage across an RC circuit;

[0081] Traditional offline parameter identification will produce errors due to parameter changes in the actual working environment and is not suitable for use in actual vehicles. To address this problem, the least squares method is used to perform online identification of model parameters. The identification parameters include internal resistance R0, polarization internal resistance R p and polarization capacitance C p , and based on the above circuit state equation, the specific implementation process is as follows:

[0082] Suppose the system equation at sampling time k is:

[0083]

[0084] Where ε is the noise variable, is the information vector, θ is the parameter vector, and y is the system output. According to the basic theory of the least squares algorithm, after deduction and transformation, the recursive expression can be obtained as follows:

[0085]

[0086] Where K is the gain, is the estimated value of the parameter vector θ; e is the prediction error of the terminal voltage; P is the covariance matrix; λ is the forgetting factor, which ranges from [0.95, 1].

[0087] According to the above expression, the terminal voltage of the equivalent circuit can be obtained:

[0088]

[0089] Among them I L is the load current. s is the secondary variable of the transfer function, and its bilinear transformation is:

[0090]

[0091] Where q represents the discretization operator, T s Indicates the sampling time.

[0092] The difference between the terminal voltage and the open circuit voltage E t for

[0093] E t =U t -U ocv

[0094] Get the transfer function:

[0095]

[0096] Through the subvariable s G s Perform discretization:

[0097]

[0098] Among them, the α1, α2, and α3 coefficients in the numerator and denominator are defined as:

[0099]

[0100] Among them, R0, R p 、C p are the internal resistance, polarization internal resistance and polarization capacitance of the equivalent circuit model respectively.

[0101] In adjacent sampling times, the open circuit voltage OCV can be regarded as a constant

[0102] U OCV，k ≈U OCV,k-1

[0103] The formula for obtaining the real-time terminal voltage is:

[0104] U t，k =α1U t，k-1 +(1-α1)U OCV，k +α2I L，k +α3I L,k-1

[0105] Let τ = R p C p , the parameter identification results are:

[0106]

[0107] Based on the above results, the real-time calculated internal resistance characteristics can be extracted on the vehicle, and all characteristics, including the voltage and temperature difference, deviation and sample entropy, can be uploaded to the cloud platform to calculate whether there is a fault.

[0108] S2. Receive the feature matrix uploaded by the vehicle, first adjust each feature for normalization, and then adjust the feature sign so that its correlation with battery life is negative, such as Figure 4 As shown in the figure, the data source is a new energy vehicle that has experienced thermal runaway. The internal battery pack is composed of 96 single cells connected in series and parallel. From the results, it can be seen that the characteristic values ​​of most cells are below 0.5, while most characteristic values ​​of some cells are larger than those of other batteries, especially the voltage deviation values ​​of cells 5, 30, 51, and 94 are close to 1. These batteries may be abnormal.

[0109] On the cloud platform, a hybrid clustering method is used to calculate the health of single cells in real time: several common clustering methods are selected, including fuzzy C-means clustering, K-means clustering, DBSCAN, and Gaussian mixture clustering. Different clustering methods are used based on different sub-features to divide all single cells into different categories:

[0110]

[0111] In the formula, cell i is the i-th monomer in the current category; a, b, and c are the number of monomers contained in the three categories Y1, Y2, and Y3 respectively.

[0112] Calculate the characteristic average of all monomers in each class:

[0113]

[0114] Where R ip is the pth eigenvalue of the i-th monomer, x is the number of features contained in the sub-feature set in the current clustering process, are the feature averages of categories Y1, Y2, and Y3 respectively.

[0115] The health of the monomer in the class with the largest feature average value is set to 0, and the health of the other classes is set to 1, which is the result of a single clustering. Let the number of times all sub-feature monomers i are in state 1 be F i , traverse n clustering methods and λ cycles, and the health degree S of monomer i in the fault matrix can be obtained i ;

[0116]

[0117] Where i is the number of single cells and s is the characteristic number.

[0118] Power battery failures that cause thermal runaway are categorized as internal or external short circuit failures, inconsistency failures, and abnormal resistance failures. Temperature-related characteristics, temperature and voltage-related characteristics, and temperature and internal resistance-related characteristics are evaluated as three types of indicators. The evaluation results are then clustered and classified in the cloud. Finally, the single-cell battery failure information calculated in the cloud is sent to the vehicle via the server in real time, which issues an alarm and reminds the driver to conduct prompt repairs.

[0119] Furthermore, the evaluation results of the hybrid clustering in step S2 are divided into the following fault types in the cloud:

[0120] For all battery cells, health evaluation is performed based on temperature characteristics, voltage and temperature characteristics, and voltage and internal resistance characteristics, and health levels S1, S2, and S3 are obtained. When a cell has a health level lower than 60, it is considered a fault state, with the fault state set to 1 and the normal state set to 0. The method for classifying fault types is as follows:

[0121] when When only the temperature is abnormal, it is generally caused by thermal runaway transfer around the cell or abnormal thermal management system, which has nothing to do with the current cell i. There is no abnormality in the cell, and the battery system needs to be checked.

[0122] when When , it indicates that the cell consistency is poor and the battery module has an inconsistent fault.

[0123] when or The internal resistance of each cell is inconsistent or the internal resistance increases due to temperature rise, and the battery module has abnormal internal resistance failure.

[0124] when When the battery is detected, there is a micro short circuit or external short circuit fault in the single cell, and it is determined that the battery module has an internal or external short circuit fault.

[0125] S3. Use vehicle data that has experienced thermal runaway to verify the accuracy and feasibility of the algorithm.

[0126] See also Figure 5 , multiple single cells in the detected battery module have inconsistent faults, and all of them are related to thermal faults. The method of the present invention can classify the fault types and locate the faulty single cells.

[0127] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for early warning of thermal runaway of power batteries based on vehicle-cloud collaboration, characterized in that: include: S1. There is a vehicle data matrix A k×n , a new row of data is generated at each sampling moment. To avoid data lag, feature recognition is required on the vehicle side, and the latest m rows of historical data are extracted as a matrix C m×n The relevant features of the internal data are calculated, including the difference between voltage and temperature, average deviation and sample entropy eigenvalues, and an equivalent circuit model is established to dynamically identify and correct the open-circuit voltage. The recursive least squares method parameters with forgetting factors are used on the vehicle side to extract the internal resistance features online. All feature information is sent to the cloud for fault identification every time corresponding to the length of a matrix interval. S2. Receive the feature set uploaded by the vehicle, adjust the sign of each feature so that its correlation with the battery life is negative, and divide the 7 features calculated by the vehicle into sub-feature sets of any dimension, and use the hybrid clustering-based method on the cloud platform to calculate the health of the single battery in real time; The health of single cells is calculated in real time on the cloud platform using a hybrid clustering method: Several common clustering methods are selected, including fuzzy C-means clustering, K-means clustering, DBSCAN, and Gaussian mixture clustering. Different clustering methods are used based on different sub-features to divide all single cells into the following categories: In the formula, cell i is the i-th monomer in the current category; a, b, c are the number of monomers contained in the three categories Y1, Y2, and Y3 respectively; Calculate the average of the characteristics of all monomers in each class: In the formula, R ip is the pth eigenvalue of the ith monomer; x is the number of features contained in the sub-feature set in the current clustering process, are the feature averages of categories Y1, Y2, and Y3 respectively; Set the monomer health in the class with the largest feature average value to 0, and the rest to 1, which is the result of a single clustering. Let the number of times that the state of all sub-feature monomers i is 1 be F i , traverse n clustering methods and λ cycles to obtain the health S of monomer i in the fault matrix i ; Where i is the number of single cells and s is the characteristic number; The power battery failures that cause thermal runaway are divided into internal and external short circuit failures, inconsistency failures and abnormal resistance failures. The temperature-related characteristics, temperature and voltage-related characteristics, and temperature and internal resistance-related characteristics are used as three types of indicators to score the health level based on the hybrid clustering method. The fault types are divided in the cloud based on the evaluation results. Finally, the fault information calculated in the cloud is sent to the vehicle side in real time through the server, and the vehicle side issues an alarm and reminds the driver to repair it in time. S3. Use vehicle data that has experienced thermal runaway to verify the accuracy and feasibility of the algorithm.

2. The method according to claim 1, characterized in that Extracting the voltage and temperature difference, average deviation and sample entropy eigenvalue in the matrix in step S1 includes: the voltage and temperature difference d of the data x of the i-th battery cell in the battery pack i , average deviation i And the sample entropy SampEn eigenvalues ​​are defined as follows: in are the maximum, minimum and average characteristics of the voltage and temperature of monomer i in the matrix; n is the total number of monomers; r is the measurement window of the original sequence; A α (r) and B α (r) represents the probability that two sequences match α and α+1 points respectively under the same r.

3. The method according to claim 1, characterized in that In step S1, an equivalent circuit model is established, and the dynamic identification and correction of the open circuit voltage includes: The first-order Thevenin equivalent circuit model is selected as the research mathematical model. The state equation obtained according to the equivalent circuit model is: Where I is the current inside the circuit, here we use the total current of the vehicle, U t is the terminal voltage of the circuit, U0 is the voltage across R0, and U oc is the open circuit voltage, U p Used to represent the voltage across an RC circuit.

4. The method according to claim 1, characterized in that: Perform online parameter identification to extract internal resistance characteristics, including: The least square method is used to identify the model parameters online. The identification parameters include ohmic internal resistance R0, polarization internal resistance R p and polarization capacitance C p , and is carried out on the basis of the circuit state equation. The specific implementation process is as follows: Suppose the system equation at sampling time t is: Where ε is the noise variable, is the information vector, θ is the parameter vector, and y is the system output. According to the basic theory of the least squares algorithm, after derivation and transformation, the recursive expression is as follows: Where K is the gain, is the estimated value of the parameter vector θ; e is the prediction error of the terminal voltage; P is the covariance matrix; λ is the forgetting factor, which ranges from [0.95,1]; According to the above expression, the terminal voltage of the equivalent circuit is obtained: Among them I L is the load current. s is the secondary variable of the transfer function, and its bilinear transformation is: Where q represents the discretization operator, T s Indicates sampling time; The difference between the terminal voltage and the open circuit voltage is E t for: E t =U t -IN ocv Get the transfer function: Through the side variable s, G s Discretize it: Among them, the α1, α2, and α3 coefficients in the numerator and denominator are defined as: Among them, R0, R p , C p They are the equivalent circuit model ohmic internal resistance, polarization internal resistance and polarization capacitance respectively; During the adjacent sampling time, the open circuit voltage OCV is considered as a constant IN OCV,k ≈In OCV,k-1 The formula for obtaining the real-time terminal voltage is: U t，k =α1U t，k-1 +(1-α1)U OCV，k +α2I L,k +α3I L,k-1 Let τ = R p C p , the parameter identification result is: Based on the above results, the internal resistance characteristics calculated in real time are extracted on the vehicle, and all characteristics, including the difference between voltage and temperature, deviation and sample entropy, are uploaded to the cloud platform to calculate whether there is a fault.

5. The method according to claim 1, characterized in that The evaluation results of the hybrid clustering in step S2 are divided into fault types in the cloud, including: for all battery cells, health evaluation is performed based on temperature characteristics, voltage and temperature characteristics, voltage and internal resistance characteristics, and health levels S1, S2 and S3 are obtained. When a cell with a health level lower than 60 points is in a fault state, the fault state is set to 1 and the normal state is set to 0. The method for dividing the fault type is as follows: when When , only the temperature abnormality is caused by thermal runaway transmission around the cell or abnormality of the thermal management system, which has nothing to do with the current cell i. The cell has no abnormality and the battery system needs to be checked; when When the cell is inconsistent, the battery module has an inconsistent fault; when or The internal resistance of the single cell is inconsistent or the internal resistance increases due to temperature rise, and the battery module has abnormal internal resistance fault; when When the battery is detected, there is a micro short circuit or external short circuit fault in the single cell, and it is determined that the battery module has an internal or external short circuit fault.

Images