Electric vehicle thermal runaway early warning method based on charging data

Abstract

The invention discloses an early warning method for thermal runaway of a battery pack of an electric vehicle, which relates to the technical field of battery management and is characterized in that the battery pack is charged through an improved multi-stage constant-current charging strategy, and the acquired data of the battery pack in the charging stage are analyzed; according to the method, information related to thermal runaway of the battery is extracted from the angles of capacity degradation, voltage abnormity and the like to serve as abnormal state indexes, an abnormity grading strategy based on a fuzzy weighted neural network is constructed, the abnormal state indexes of the single bodies serve as input to obtain the abnormal grades of the single bodies, and corresponding treatment measures are taken for different abnormal grades. Therefore, efficient and accurate electric vehicle thermal runaway early warning is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of battery management, and in particular to a thermal runaway early warning method for an electric vehicle based on charging data. Background Art

[0002] The continuous improvement of electric vehicle range and output power has posed a great challenge to the safety of lithium-ion batteries. Improper operation during use, such as electrical abuse, thermal abuse, and mechanical abuse, will accelerate the failure process of the battery, further leading to thermal runaway of the battery, seriously threatening the personal and property safety of users. The existing thermal runaway warning methods currently rely mainly on the on-board battery management system to set the battery's overvoltage threshold, undervoltage threshold, and overtemperature threshold. However, since the abnormal changes in battery voltage and temperature are not obvious before thermal runaway occurs, the traditional threshold-based warning method has an extremely high false alarm rate and cannot reserve enough disposal time for users. Summary of the invention

[0003] The present invention aims to solve the deficiencies of the above-mentioned prior art and proposes a thermal runaway warning method for an electric vehicle battery pack based on charging data, in order to extract abnormal information related to thermal runaway from the charging data and use it to divide the abnormal level of the battery, thereby achieving efficient and accurate thermal runaway warning for electric vehicles.

[0004] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme:

[0005] The electric vehicle thermal runaway early warning method based on charging data of the present invention is characterized in that it comprises the following steps:

[0006] S1: When charging an electric vehicle, a multi-level constant current charging strategy is used to obtain all charging data of the electric vehicle in a complete charging process;

[0007] S2: Based on the charging data, obtain N features related to thermal runaway of cell number i;

[0008] S3: Take the aging characteristics and abnormal characteristics of monomer i, a total of N characteristics, as abnormal state indicators { |n=1,2,…,N}, and input into the abnormal classification network FWNN based on fuzzy weighted neural network for processing to obtain the predicted abnormal state of monomer i , where the abnormal state includes: severe abnormal state, slight abnormal state and no abnormal state; Indicates the nth abnormal state indicator of monomer i;

[0009] S4: Based on the predicted abnormal state and the actual abnormal state of all monomers, a loss function is established to train the FWNN, and the parameters of the membership function and the weights of the rules are updated until the loss function converges, thereby obtaining the optimal monomer abnormality classification model, which is used to process the abnormal information extracted from the electric vehicle charging data and classify the abnormal levels of the monomers.

[0010] The electric vehicle thermal runaway warning method based on charging data described in the present invention is also characterized in that S2 includes the following steps:

[0011] S2.1: Based on the voltage change before and after the current switching during the complete charging process, use formula (1) to calculate the internal resistance R of the i-th cell i :

[0012] (1)

[0013] In formula (1), R i Represents the internal resistance of monomer i, V i,k and V i,k+1 I represents the voltage value of the i-th cell at time k before the current switching and time k+1 after the current switching, respectively. i,k and I i,k+1 Respectively represent the charging current value of cell i at time k before current switching and time k+1 after current switching;

[0014] S2.2: Select a fixed voltage range and use formula (2) to calculate the cumulative charging capacity ΔQ of cell number i in the voltage range i : (2)

[0015] In formula (2), Q i,end is the battery capacity of cell No. i at the end point of the voltage range, Q i,0 is the battery capacity of cell No. i at the starting point of the voltage interval;

[0016] S2.3: Select a voltage value as the starting voltage V of cell i i,0 , and use formula (3) to calculate the voltage change of monomer i after charging a fixed amount of electricity :

[0017] (3)

[0018] In formula (3), The termination voltage after the i-th cell is charged with a fixed amount of electricity;

[0019] S2.4: After using formula (4) to characterize the incremental capacity curve of monomer i and plotting it, extract the peak ICP in the incremental capacity curve iAnd the voltage corresponding to the peak value ICV i ;

[0020] (4)

[0021] In formula (4), Q i and V i They represent the capacity of the i-th cell and the voltage of the i-th cell respectively, and I represents the charging current of the cell; Indicates the time interval, represents the voltage of monomer i at time t, and d represents the differential;

[0022] S2.5: After plotting the temperature change curve of monomer i using equation (5), extract the peak value DTP in the temperature change curve i And the voltage DTV corresponding to the peak value i :

[0023] (5)

[0024] In formula (5), represents the temperature of monomer i, and L is the sampling interval; Indicates that monomer i is The temperature of the moment, Indicates that monomer i is The temperature at the moment, d represents the differential;

[0025] S2.6: Set ΔV i , ΔQ i , R i ,ICP i , ICV i , DTP i and DTV i As the aging feature of cell i, all cells in the battery pack are clustered using a clustering algorithm according to the aging feature, thereby obtaining clustered normal cells and capacity-degraded cells;

[0026] S2.7: Calculate the Z score of monomer number i at time k using formula (7) :

[0027] (7)

[0028] In formula (7), V i.k is the voltage of the ith cell at time k, U mean,k is the mean voltage at time k, U σ,k is the voltage standard deviation at time k;

[0029] S2.8: After filtering the Z score of cell number i during the charging process using the extended Kalman filter algorithm, the maximum Z score Z of cell number i during the charging process is counted. max,i ;

[0030] S2.9: Calculate the voltage sequence V of cell number i i With the average voltage series U mean The Euclidean distance ED i ;

[0031] (8)

[0032] In formula (8), is the voltage of the ith cell at time p, U mean,p is the mean voltage at time p;

[0033] S2.10: , Z max,i , E.D. i As an abnormal characteristic of monomer i.

[0034] Furthermore, S3 includes the following steps:

[0035] S3.1: Use the membership function to convert the nth abnormal index Mapped into fuzzy feature set A i,n ;

[0036] S3.2: Use equation (11) to establish the hth fuzzy rule R h , thus obtaining H fuzzy rules and forming a fuzzy rule base for abnormal classification;

[0037] (11)

[0038] In formula (11), A i,n,h is the nth abnormal state indicator x of the ith monomer i,n and the hth fuzzy rule R h The corresponding fuzzy feature set, y i,h is the i-th monomer and the h-th fuzzy rule R h The corresponding abnormal state, B i,h is the i-th monomer and the h-th fuzzy rule R h The corresponding fuzzy output;

[0039] S3.3: Using formula (12), we get the hth output result f of the FWNN rule layer of monomer i i,h :

[0040] (12)

[0041] In formula (12), ωh is the hth fuzzy rule R h The weight of is the abnormal status indicator of monomer i The degree of membership;

[0042] S3.4: Using formula (13) to obtain the predicted abnormal state of monomer i output by the aggregation layer of FWNN :

[0043] (13)

[0044] In formula (13), β i,h f i,h The corresponding weight.

[0045] The electronic device of the present invention includes a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the electric vehicle thermal runaway warning method, and the processor is configured to execute the program stored in the memory.

[0046] The present invention provides a computer-readable storage medium, and the computer-readable storage medium stores a computer program, which is characterized in that when the computer program is run by a processor, the steps of the electric vehicle thermal runaway warning method are executed.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. The present invention selects charging phase data containing rich battery pack inconsistency information as the analysis object, and extracts multiple features related to battery thermal runaway, such as battery internal resistance and temperature rise data; compared with the existing thermal runaway early warning method based on a single threshold, the present invention combines multiple abnormal features, which can effectively improve the data mining effect;

[0049] 2. The present invention combines different abnormal features and develops an abnormal classification network based on a fuzzy weighted neural network. By training the abnormal classification network, the weights of different abnormal features are optimized, thereby ensuring a high diagnosis rate and reducing the misdiagnosis of thermal runaway. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a schematic diagram of the overall method of the present invention;

[0051] Figure 2 This is a diagram of the multi-stage constant current charging strategy of the present invention. DETAILED DESCRIPTION

[0052] In this embodiment, a thermal runaway warning method for electric vehicle battery packs based on charging data is designed to achieve daily thermal runaway warnings to improve the safety of electric vehicles, and integrate multiple features related to battery failure to diagnose each single cell in the battery pack. In addition, in order to reduce the false alarm rate of the thermal runaway warning method, an abnormal classification strategy is established, and corresponding disposal measures are taken for different abnormal characteristics. Specifically, Figure 1 As shown, the method comprises the following steps:

[0053] S1: When charging an electric vehicle, a multi-level constant current charging strategy is used to obtain all charging data of the electric vehicle in a complete charging process;

[0054] like Figure 2 As shown in the figure, a charging process is divided into multiple constant current charging stages with different charging current values ​​according to the state of charge (SOC) and voltage, and the current change between each constant current charging stage is a step. In a constant current charging stage, the data sampling frequency of the battery management system is set to once every 30 seconds; within 10 seconds before and after the charging current is switched, the sampling frequency is increased to once every 1 second;

[0055] Upload the collected charging data to the cloud server. The main data include time, current, voltage of each cell, highest cell voltage, lowest cell voltage, battery pack SOC, and temperature of each cell.

[0056] Due to factors such as cost, usage space, and computing power, the on-board battery management system can only rely on a single temperature threshold and voltage threshold for thermal runaway warning, and is unable to integrate a large amount of historical data for comprehensive analysis; the collected battery pack operating status data is uploaded to the cloud server in real time, and with the help of the cloud's strong storage and computing capabilities, the battery's abnormal information can be efficiently mined to achieve timely thermal runaway warning; in addition, the cloud server can horizontally compare the health status of the battery packs of multiple electric vehicles, which helps to screen out abnormal batteries.

[0057] S2: Based on the charging data, multiple features related to thermal runaway are obtained;

[0058] S2.1: Based on the voltage change before and after the current switching during the complete charging process, use formula (14) to calculate the internal resistance R of the i-th cell i :

[0059] (14)

[0060] In formula (14), R i Represents the internal resistance of monomer i, V i,k and V i,k+1I represents the voltage value of the i-th cell at time k before the current switching and time k+1 after the current switching, respectively. i,k and I i,k+1 Respectively represent the charging current value of cell i at time k before current switching and time k+1 after current switching;

[0061] The sampling frequency of the vehicle battery management system generally does not exceed 0.1Hz. During the constant current charging stage, the sampling frequency will be further reduced to control the data storage space due to the relatively stable battery state. At this time, the voltage data is too sparse, and the calculated internal resistance value has a large error. The improved multi-level constant current charging strategy increases the sampling frequency to 1Hz within 10 seconds before and after the current value is switched, which improves the calculation accuracy of the battery internal resistance without significantly increasing the amount of collected data.

[0062] S2.2: Select a fixed voltage range and use equation (15) to calculate the cumulative charging capacity ΔQ of cell number i within the voltage range i : (15)

[0063] In formula (15), Q i,end is the battery capacity of cell i at the end point of the voltage range, Q i,0 is the battery capacity of cell i at the starting point of the voltage range;

[0064] S2.3: Select a voltage value as the starting voltage V of cell i i,0 , and use formula (16) to calculate the voltage change of monomer i after charging a fixed amount of electricity :

[0065] (16)

[0066] In formula (16), The termination voltage after the i-th cell is charged with a fixed amount of electricity;

[0067] S2.4: After the incremental capacity curve of monomer i is characterized by equation (17) and plotted, the peak ICP in the incremental capacity curve is extracted. i And the peak voltage ICV i ;

[0068] (17)

[0069] In formula (17), Q i and V i They represent the capacity of the i-th cell and the voltage of the i-th cell respectively, and I represents the charging current of the cell; Indicates the time interval, It represents the voltage of cell No. i at time t, and d represents the differential.

[0070] S2.5: After plotting the temperature change curve of monomer i using equation (18), extract the peak value DTP in the temperature change curve i And the voltage DTV corresponding to the peak value i :

[0071] (18)

[0072] In formula (18), represents the temperature of monomer i, and L is the sampling interval; Indicates that monomer i is The temperature of the moment, Indicates that monomer i is The temperature at the moment, d represents the differential.

[0073] S2.6: Set ΔV i , ΔQ i , R i ,ICP i , ICV i , DTP i and DTV i As the aging characteristics of the battery, after normalization processing, the corresponding characteristics are used as sample points of each monomer; the aging characteristics of all monomers in the battery pack are input into the clustering algorithm for clustering, and 2 clusters are selected to divide each monomer into a capacity-degraded monomer and a normal monomer;

[0074] The mean normalization method is selected to scale the features to the interval [0, 1] to avoid the influence of the magnitude of the features on the clustering effect. After obtaining the normalized feature sample points of each monomer, the agglomerative hierarchical clustering algorithm is selected for clustering. Agglomerative hierarchical clustering is a bottom-up clustering method similar to a dendrogram. First, all sample points are regarded as separate clusters, then the similarity between clusters is calculated, and clusters with high similarity are merged into larger clusters. After iterative processing, a nested clustering structure is constructed.

[0075] S2.7: Calculate the Z score of monomer number i at time k using formula (19) :

[0076] (19)

[0077] In formula (19), V i,k is the voltage of the ith cell at time k, U mean,k is the mean voltage at time k, U σ,k is the voltage standard deviation at time k;

[0078] S2.8: Z i,kThe extended Kalman filter (EKF) algorithm is used for filtering to reduce the impact of voltage measurement noise. The EKF-based filtering operation is as follows:

[0079] The state equation and observation equation are established as formula (20):

[0080] (20)

[0081] In formula (20), x i,k is the Z score of monomer i at time k, y i,k is the noisy measurement value of the i-th cell at time k, A k is the first-order identity matrix at time k, w k and v k represent the state noise and measurement noise at time k respectively.

[0082] The specific steps of the filtering operation are:

[0083] Initialize the state estimate using formula (21) The first observation :

[0084] (twenty one)

[0085] Update the prior state estimate using equation (22): and the state error covariance :

[0086] (twenty two)

[0087] In formula (22), represents the process noise covariance;

[0088] Use equation (23) to calculate the Kalman gain at time k: :

[0089] (twenty three)

[0090] In formula (23), represents the measurement noise covariance;

[0091] Update the posterior state estimate using equation (24): and the state error covariance :

[0092] (twenty four)

[0093] In order to obtain the Z of each monomer i After the noise reduction curve, the maximum Z score Z of the cell number i in the charging process is counted max,i ;

[0094] S2.9: Calculate the voltage sequence V of cell number i i With the average voltage series U mean The Euclidean distance ED i ;

[0095] (25)

[0096] In formula (25), is the voltage of the ith cell at time p, U mean,p is the mean voltage at time p;

[0097] S2.10: , Z max,i , E.D. i As an abnormal characteristic of monomer i.

[0098] S3: Take the aging characteristics and abnormal characteristics of monomer i, a total of N characteristics, as abnormal state indicators { |n=1,2,…,N}, and input into the abnormal classification network FWNN based on fuzzy weighted neural network for processing, and obtain the predicted abnormal state of monomer i, where the abnormal state includes: severe abnormal state, slight abnormal state and no abnormal state;

[0099] S3.1: Take the i-th cell as a normal cell or a capacity-degraded cell as a qualitative indicator, and take the i-th cell's internal resistance R i , maximum Z score Z max,i , Euclidean distance ED i As a quantitative indicator, the membership function is used to map the nth abnormal indicator into a fuzzy feature set A. i,n ; The membership function uses the Gaussian membership function;

[0100] (26)

[0101] In formula (26), c i is the center point of the membership function, σ i is the membership function width.

[0102] S3.2: Use equation (11) to establish the hth fuzzy rule R h , thus obtaining H fuzzy rules and forming a fuzzy rule base for abnormal classification;

[0103] (27)

[0104] In formula (27), A i,n,h is the nth abnormal state indicator x of the ith monomer i,n and the hth fuzzy rule R hThe corresponding fuzzy feature set, y i,h is the i-th monomer and the h-th fuzzy rule R h The corresponding abnormal state, B i,h is the i-th monomer and the h-th fuzzy rule R h The corresponding fuzzy output.

[0105] S3.3: Using formula (28), we can obtain the hth output result f of the FWNN rule layer of monomer i: i,h :

[0106] (12)

[0107] In formula (28), ω h is the hth fuzzy rule R h The weight of Abnormal status indicator The degree of membership;

[0108] The rule layer output is the result of the fuzzy inference system based on the input variables and fuzzy rules. It indicates the degree of response of each fuzzy rule to the current input. h Represents the importance of the rule in the entire system. h The larger the value, the stronger the response of the rule to the current input conditions, that is, the more appropriate the prediction for the current input.

[0109] S3.4: Using formula (29) to obtain the predicted abnormal state of monomer i output by the aggregation layer of FWNN :

[0110] (29)

[0111] In formula (29), β i,h f i,h The corresponding weights; the aggregation layer fuses the outputs of all rules to obtain the final output of the system. The abnormal status of the monomer is divided according to the final output.

[0112] S4: Based on the predicted abnormal state and the actual abnormal state of all monomers, a loss function is established to train the FWNN, and the parameters of the membership function and the weights of the rules are updated until the loss function converges, thereby obtaining the optimal monomer abnormality classification model, which is used to process the abnormal information extracted from the electric vehicle charging data and classify the abnormal levels of the monomers.

[0113] S5: Take corresponding disposal measures for the screened out slightly abnormal monomers and seriously abnormal monomers.

[0114] For slightly abnormal cells, send the electric vehicle related information and abnormal cell number to the electric vehicle maintenance company, continuously track the capacity degradation, voltage abnormality and abnormal temperature rise of the abnormal cell, and reduce the charging cut-off voltage of the electric vehicle to avoid accelerated degradation of the abnormal cell caused by overcharging;

[0115] For seriously abnormal cells, a thermal runaway warning will be immediately issued to the electric vehicle user. If the electric vehicle is in charging state at this time, the charging will be interrupted immediately. If the electric vehicle is in running state, the user will be reminded to unplug the electric vehicle and stop using it as soon as possible. At the same time, the electric vehicle maintenance company will recall the problematic battery pack as soon as possible.

[0116] In this embodiment, an electronic device includes a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the above method, and the processor is configured to execute the program stored in the memory.

[0117] In this embodiment, a computer-readable storage medium stores a computer program on the computer-readable storage medium, and the computer program executes the steps of the above method when executed by a processor.

Claims

1. A thermal runaway early warning method for electric vehicles based on charging data, characterized in that: The steps include: S1: When charging an electric vehicle, a multi-level constant current charging strategy is used to obtain all charging data of the electric vehicle in a complete charging process; S2: Based on the charging data, obtain N features related to thermal runaway of cell number i; S3: Take the aging characteristics and abnormal characteristics of monomer i, a total of N characteristics, as abnormal state indicators { |n=1,2,…,N}, and input into the abnormal classification network FWNN based on fuzzy weighted neural network for processing to obtain the predicted abnormal state of monomer i , where the abnormal state includes: severe abnormal state, slight abnormal state and no abnormal state; Indicates the nth abnormal state indicator of monomer i; S4: Based on the predicted abnormal state and the actual abnormal state of all monomers, a loss function is established to train the FWNN, and the parameters of the membership function and the weights of the rules are updated until the loss function converges, thereby obtaining the optimal monomer abnormality classification model, which is used to process the abnormal information extracted from the electric vehicle charging data and classify the abnormal levels of the monomers.

2. The electric vehicle thermal runaway early warning method based on charging data according to claim 1 is characterized in that: S2 includes the following steps: S2.1: Based on the voltage change before and after the current switching during the complete charging process, use formula (1) to calculate the internal resistance R of the i-th cell i : (1) In formula (1), R i Represents the internal resistance of monomer i, V i,k and V i,k+1 I represents the voltage value of the i-th cell at time k before the current switching and time k+1 after the current switching, respectively. i,k and I i,k+1 Respectively represent the charging current value of cell i at time k before current switching and time k+1 after current switching; S2.2: Select a fixed voltage range and use formula (2) to calculate the cumulative charging capacity ΔQ of cell number i in the voltage range i : (2) In formula (2), Q i,end is the battery capacity of cell No. i at the end point of the voltage range, Q i,0 is the battery capacity of cell No. i at the starting point of the voltage interval; S2.3: Select a voltage value as the starting voltage V of cell i i,0 , and use formula (3) to calculate the voltage change of monomer i after charging a fixed amount of electricity : (3) In formula (3), The termination voltage after the i-th cell is charged with a fixed amount of electricity; S2.4: After using formula (4) to characterize the incremental capacity curve of monomer i and plotting it, extract the peak ICP in the incremental capacity curve i And the peak voltage ICV i ; (4) In formula (4), Q i and V i They represent the capacity of the i-th cell and the voltage of the i-th cell respectively, and I represents the charging current of the cell; Indicates the time interval, represents the voltage of monomer i at time t, and d represents the differential; S2.5: After plotting the temperature change curve of monomer i using equation (5), extract the peak value DTP in the temperature change curve i And the voltage DTV corresponding to the peak value i : (5) In formula (5), represents the temperature of monomer i, and L is the sampling interval; Indicates that monomer i is The temperature of the moment, Indicates that monomer i is The temperature at the moment, d represents the differential; S2.6: Set ΔV i , ΔQ i , R i ,ICP i , ICV i , DTP i and DTV i As the aging characteristics of cell i, all cells in the battery pack are clustered using a clustering algorithm according to the aging characteristics, thereby obtaining clustered normal cells and capacity-degraded cells; S2.7: Calculate the Z score of monomer number i at time k using formula (7) : (7) In formula (7), V i.k is the voltage of the ith cell at time k, U mean,k is the mean voltage at time k, U σ,k is the voltage standard deviation at time k; S2.8: After filtering the Z score of cell number i during the charging process using the extended Kalman filter algorithm, the maximum Z score Z of cell number i during the charging process is counted. max,i ; S2.9: Calculate the voltage sequence V of cell number i i With the average voltage series U mean The Euclidean distance ED i ; (8) In formula (8), is the voltage of the ith cell at time p, U mean,p is the mean voltage at time p; S2.10: , Z max,i , E.D. i As an abnormal characteristic of monomer i.

3. The electric vehicle thermal runaway early warning method based on charging data according to claim 1 is characterized in that: S3 includes the following steps: S3.1: Use the membership function to convert the nth abnormal index Mapped into fuzzy feature set A i,n ; S3.2: Use equation (11) to establish the hth fuzzy rule R h , thus obtaining H fuzzy rules and forming a fuzzy rule base for abnormal classification; (11) In formula (11), A i,n,h is the nth abnormal state indicator x of the ith monomer i,n and the hth fuzzy rule R h The corresponding fuzzy feature set, y i,h is the i-th monomer and the h-th fuzzy rule R h The corresponding abnormal state, B i,h is the i-th monomer and the h-th fuzzy rule R h The corresponding fuzzy output; S3.3: Using formula (12), we get the hth output result f of the FWNN rule layer of monomer i i,h : (12) In formula (12), ω h is the hth fuzzy rule R h The weight of is the abnormal status indicator of monomer i The degree of membership; S3.4: Using formula (13) to obtain the predicted abnormal state of monomer i output by the aggregation layer of FWNN : (13) In formula (13), β i,h f i,h The corresponding weight.

4. An electronic device, comprising a memory and a processor, characterized in that: The memory is used to store a program that supports the processor to execute the electric vehicle thermal runaway warning method as described in any one of claims 1 to 3, and the processor is configured to execute the program stored in the memory.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the electric vehicle thermal runaway warning method described in any one of claims 1 to 3 are executed.

Images