CN111967194B - Battery classification method based on cloud historical data - Google Patents
Battery classification method based on cloud historical data Download PDFInfo
- Publication number
- CN111967194B CN111967194B CN202010871109.8A CN202010871109A CN111967194B CN 111967194 B CN111967194 B CN 111967194B CN 202010871109 A CN202010871109 A CN 202010871109A CN 111967194 B CN111967194 B CN 111967194B
- Authority
- CN
- China
- Prior art keywords
- battery
- data
- capacity
- internal resistance
- current
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
- Y02E60/10—Energy storage using batteries
Abstract
The invention provides a battery classification method based on cloud historical data, which comprises the following steps: step 1, collecting historical big data of a battery; step 2, data cleaning is carried out on the historical big data to obtain voltage data and current data; step 3, calculating the battery monomer capacity and the battery internal resistance to obtain the battery monomer capacity and the battery internal resistance of m samples, wherein the battery monomer capacity and the battery internal resistance are used as a sample set a to be classified; step 4, normalizing the data in the sample set a to be classified to obtain a normalized sample data set A; step 5, selecting k class clusters in the sample data set A, and initializing cluster centers of the k class clusters to obtain an initial cluster center B; step 6, calculating the category of each sample by adopting a Euclidean distance formula, introducing a weight factor during calculation, calculating a new centroid mu by using a category set C j Updating the cluster center; step 7, repeating the step 6 until mu j And stopping iteration when the threshold sigma requirement is not changed, the threshold sigma requirement is met or the maximum iteration number N is reached, and obtaining a clustering result.
Description
Technical Field
The invention belongs to the technical field of battery management, and particularly relates to a battery classification method based on cloud historical data.
Background
With the continuous improvement of sales of new energy automobiles in China, the development strategy of the new energy automobiles has been raised to the national strategy. In 2018, the sales of new energy automobiles in China is about 100 ten thousand, and in 2020, the production capacity of pure electric automobiles and plug-in hybrid electric automobiles is estimated to be 200 ten thousand, and the accumulated sales exceeds 500 ten thousand. Due to the inconsistency among batteries caused in the use process, consistency sorting is required before the retired lithium ion batteries are used in a cascade such as energy storage application. However, the existing retired lithium battery sorting method is mainly based on field capacity test, so that the existing retired lithium battery sorting method has the problems of low efficiency, high sorting cost and the like. The problem of consistency sorting with high speed and high economical efficiency is to be solved.
Disclosure of Invention
The present invention is made to solve the above problems, and an object of the present invention is to provide a battery classification method based on cloud history data.
The invention provides a battery classification method based on cloud historical data, which has the characteristics that: step 1, acquiring historical big data of a battery by adopting a battery management system BMS; step 2, data cleaning is carried out on the historical big data to obtain voltage data and current data; step 3, calculating the battery cell capacity and the battery internal resistance through the voltage data and the current data respectively to obtain the battery cell capacity and the battery internal resistance of m samples, wherein the battery cell capacity and the battery internal resistance are recorded as a sample set a= { x to be classified 1 ,x 2 ,……,x m And x is i Is a two-dimensional array formed by internal resistance and capacity, namely x i =[x C,i x r,i ]And i=1, 2, … …, m; step 4, carrying out data standardization on the battery monomer capacity and the battery internal resistance to obtain a standardized sample data set, wherein the sample data set is A= { X 1 ,X 2 ,……,X m And normalized formula:step 5, selecting k clustered clusters from the sample data set A, and initializing the clustering centers of the k clustered clusters, namely selecting k samples from the sample data set A as an initial clustering center, wherein the initial clustering center is B= { mu 1 ,μ 2 ,……,μ k -a }; step 6, calculating the category of each sample by using Euclidean distance formula, and calculating each element X of the data set A i And centroid B j Euclidean distance d between ij Introducing weight factors, calculating to obtain classification result, and marking the classification result as classification set C, C= { C 1 ,C 2 ,……,C k Then calculate the new centroid mu j Updating the cluster center; step 7, repeating the step 6 until mu j Stopping iteration when the threshold sigma requirement is not changed, or the maximum iteration number N is reached, obtaining a clustering result, wherein x is C,i And x r,i Battery capacity and battery internal resistance of the i-th sample to be classified, < >>And->Respectively, minimum and maximum values of battery capacity in the ith sample to be classified, +.>And->Respectively the minimum value and the maximum value of the internal resistance of the battery in the ith sample to be classified, mu k =[μ C,k μ r,k ],μ C,k Volume data, μ representing the kth sample r,k Internal resistance data representing the kth sample.
The battery classification method based on cloud historical data provided by the invention can also have the following characteristics: the calculation steps of the battery cell capacity shown in the step 3 are as follows: step 3-1, screening out a complete charging engineering data graph from voltage data and current data; step 3-2, performing preliminary estimation on the capacity of the battery pack based on the charging process by adopting an safety integral method to obtain a preliminary estimation result Q P The method comprises the steps of carrying out a first treatment on the surface of the Step 3-3, adopting fuzzy Kalman filtering to perform preliminary estimation result Q P Performing optimization correction to obtain an optimal estimation result Q of the battery capacity sys The method comprises the steps of carrying out a first treatment on the surface of the Step 3-4, calculating the residual charge capacity RCC of each monomer based on the voltage data of the cloud monomer i And residual discharge capacity RDC i Then the capacity Q of each battery monomer is calculated according to the following formula i ,Q i =Q sys +RCC i +RDC i In which Q i Battery capacity for the ith cell, RCC i RDC for the residual charge capacity of the ith monomer i The residual discharge capacity of the i-th monomer. .
The battery classification method based on cloud historical data provided by the invention can also have the following characteristics: wherein, in step 3-2The formula of the safety point is:wherein t is 1 Indicating the charge start time of the charging section, t 2 Indicating the charge end time of the charging section, I (t) indicating the current value at time t in the charging section, the remaining charge capacity RCC in step 3-4 i And residual discharge capacity RDC i All are calculated by interpolation method.
The battery classification method based on cloud historical data provided by the invention can also have the following characteristics: the calculation step of the internal resistance of the battery shown in the step 3 is as follows: step 3-a, current data and voltage data are imported into MATLAB for screening, and a period of time when the battery is in a multi-stage constant current charging state is obtained; step 3-b, screening out micro-segments of voltage-current jump meeting the charging current condition in the time period; step 3-c, extracting the voltage and the current of the two stages before and after the charging jump from the micro-segment, respectively marked as U A 、U B 、I A 、I B The method comprises the steps of carrying out a first treatment on the surface of the And 3-d, calculating to obtain the battery resistance r by adopting the ratio of the voltage difference and the current difference of the two stages.
The battery classification method based on cloud historical data provided by the invention can also have the following characteristics: in step 3-b, the charging current conditions are as follows:
for the front and rear current values of current jump in the same charging section, the kth charging current value I k Less than-75A, the (k+1) th charging current value I k+1 Greater than-60A and less than-30A, and there is a period of mid-SOC where the current switches, where the internal resistance represents an average level of the internal resistance of the battery,
in the step 3-c, the specific values are as follows:
for the kth and the kth+1th current values in the same charging section, each cell voltage value corresponding to the k value is recorded as U A The current value is recorded as I A Each monomer voltage value corresponding to the k+1 value is recorded as U B The current value is recorded as I B 。
The invention provides an electric based on cloud historical dataThe pool classification method may further have the feature that: wherein the internal resistance of the battery in the step 3-c is
The battery classification method based on cloud historical data provided by the invention can also have the following characteristics: wherein the weight factor in the step 6 is the echelon utilization scene coefficient delta, 0 < delta < 1, and the Euclidean distance formula after the weight factor is introduced is
The battery classification method based on cloud historical data provided by the invention can also have the following characteristics: the calculation formula of the new centroid in the step 6 is as follows:and j=1, 2, … …, k.
Effects and effects of the invention
According to the battery classification method based on cloud historical data, external characteristic data such as time, voltage, current and temperature of a battery in a plurality of charging and discharging processes are obtained through a battery management system BMS, cloud big data or experiments and the like on an automobile; secondly, selecting a valid data segment according to the historical data; thirdly, according to the cloud big data of the electric automobile, a fuzzy Kalman filtering algorithm, a residual charge power RCC algorithm and a residual discharge power RDC algorithm are fused to estimate the monomer capacity offline; thirdly, estimating ohmic internal resistance according to voltage-current jump caused by charge-discharge state switching in the effective charge segment data; and finally, classifying a large number of cloud monomers by using a K-means algorithm according to the internal resistance and the capacity of the battery monomers obtained by cloud computing. By the battery classification method based on cloud historical data, comprehensive classification of vehicle-mounted monomers of different electric vehicles can be realized, the sorting scale is enlarged, the service time and cost of a factory building are reduced, the consistency of the recombined monomers is effectively ensured, and the safety of echelon utilization scenes is ensured.
Therefore, the battery classification method based on cloud historical data utilizes a large amount of historical data to estimate the capacity and the internal resistance of the vehicle-mounted battery and process noise on line. Under the condition of different data sources, single-body-level capacity internal resistance monitoring is carried out on the power battery, data support is rapidly selected when the power battery is utilized in a echelon mode, and then cloud classification is carried out on single power battery through a K-means algorithm.
Drawings
FIG. 1 is a flow chart of a battery classification method based on cloud history data in an embodiment of the invention;
FIG. 2 is a diagram of fuzzy control rules selected in an embodiment of the present invention;
FIG. 3 is a schematic diagram of a monomer remaining charge amount estimation and a monomer remaining discharge amount estimation according to an embodiment of the present invention;
FIG. 4 is a schematic representation of monomer capacity results obtained directly based on an ampere-hour integral estimate in an embodiment of the invention;
FIG. 5 is a graph showing the results of adding fuzzy Kalman filtering to the monomer capacity in the embodiment of the invention.
FIG. 6 is a schematic diagram of a charging segment generated by importing data into MATLAB after purging in an embodiment of the invention;
FIG. 7 is a graphical illustration of voltage-current jumps during a multi-stage constant current charging period in an embodiment of the invention;
FIG. 8 is a graph showing the K-means classification results of steps 4 to 7 in the embodiment of the present invention.
Detailed Description
In order to make the technical means and effects of the present invention easy to understand, the present invention will be specifically described with reference to the following examples and the accompanying drawings.
Examples:
as shown in fig. 1, the present embodiment provides a battery classification method based on cloud historical data, which includes the following steps:
and step 1, acquiring historical big data of the battery by adopting a battery management system BMS.
In this embodiment, the historical big data are all a large amount of data of each parameter collected by the battery management system BMS of the vehicle-mounted battery in the historical time period, and then the data are obtained through the internet of vehicles platform, so that the follow-up processing is performed.
And step 2, data cleaning is carried out on the historical big data to obtain voltage data and current data.
In this embodiment, a large amount of useless data is removed by using a Python algorithm, and first generation data such as voltage, current, temperature and the like required for battery life evaluation is obtained.
Wherein x is C,i And x r,i The battery capacity and the battery internal resistance of the i-th sample to be classified are respectively,and->Respectively, minimum and maximum values of battery capacity in the ith sample to be classified, +.>And->Respectively the minimum value and the maximum value of the internal resistance of the battery in the ith sample to be classified.
In this embodiment, the calculation steps of the battery cell capacity are as follows:
and 3-1, screening out a complete charging engineering data graph from the voltage data and the current data, as shown in fig. 7.
Step 3-2, performing preliminary estimation on the capacity of the battery pack based on the charging process by adopting an safety integral method to obtain a preliminary estimation result Q P And the equation of the time safety integral is:wherein t is 1 Indicating the charge start time of the charging section, t 2 The charging end time of the charging section is represented, and I (t) represents the current value at time t in the charging section.
Step 3-3, adopting fuzzy Kalman filtering to perform preliminary estimation result Q P Performing optimization correction to obtain an optimal estimation result Q of the battery capacity sys The method is characterized by comprising the following steps:
the state equation and the system output equation of the discrete form of the fuzzy kalman filter are shown as follows:
x i+1 =x i +ω i
y i =x i +υ i
wherein x is i For the capacity Q to be estimated sys As a system state vector, it cannot be directly obtained by measurement; y is i For calculating the capacity value Q P As a measurable system output; w (w) i 、v i Input noise and output noise, respectively, which are not available by measurement, w is due to small variation in battery capacity between single charge intervals i The filtering requirement can be met by taking a smaller value. For v i The value of (2) is determined using fuzzy logic ideas.
The iterative formula of Kalman filtering is as follows:
state estimation time update:
error covariance time update:
kalman gain update:
state estimation measurement update:
error covariance measurement update:
wherein, i is more than or equal to 2,is the optimal result after the last state correction, +.>Is the result of using the last state prediction, +.>Is the optimal result after the current state is corrected. />Is->Corresponding covariance,/>Is->Corresponding covariance,/>Is the covariance of the current state update. L (L) i Is the current state kalman gain. Q, R are respectively input noise w i Covariance of (v) and input noise v i Is a covariance of (c). And obtaining a capacity estimation result which accords with the real attenuation trend of the battery pack of the electric automobile through a fuzzy Kalman filtering process.
Further, when the initial value of the fuzzy kalman filter is set as:P 0 =1,Q=0.03 2 when R is controlled by a fuzzy rule as shown in FIG. 2, where e Q Representing the preliminary estimation result Q P Is a relative error of (a).
Step 3-4, calculating the residual charge capacity RCC of each monomer based on the voltage data of the cloud monomer i And residual discharge capacity RDC i Then the capacity Q of each battery monomer is calculated according to the following formula i ,Q i =Q sys +RCC i +RDC i In which Q i Battery capacity for the ith cell, RCC i RDC for the residual charge capacity of the ith monomer i The residual discharge capacity of the i-th monomer.
In this embodiment, the fuzzy Kalman corrected battery pack capacity Q sys The obtained estimation results of the capacities of the battery cells are shown in fig. 5, the horizontal axis represents the capacity, the vertical axis represents the electric quantity, and the points with the same color represent the estimation results of the capacities of different cells in the same charging process. Compared with the single capacity result directly obtained based on the ampere-hour integral estimation of the battery pack capacity in fig. 4, the single battery capacity estimation result obtained by the method provided by the embodiment is more in accordance with the actual capacity attenuation trend, and the estimation result is more accurate.
Further, based on CCVC consistency assumption, translating other monomer voltage curves based on the monomer voltage curve with the highest charge termination voltage, and calculating RCC of each monomer by interpolation method i The method specifically comprises the following steps:
based on cloud charging voltage data, meterThe RCC of each monomer was calculated. It is assumed that in a battery pack having n cells, cell j (1. Ltoreq.j. Ltoreq.n) is the first cell charged to a charge cutoff voltage, and the charge cutoff time is t CE And after that the battery management system BMS stops charging to prevent overcharge. Based on CCVC consistency assumption, CCVC translation of monomer j can obtain CCVC of other monomer i (i not equal to j), at time t CE Thereafter, if the monomers i can be charged individually, the time t is from charging CE The time until the monomer i is charged to the charge cutoff voltage is the remaining charge time (Δt) of the monomer i i,C ) Then the end point P of the CCVC curve of monomer i i The time of the interpolated point on CCVC of monomer j is t CE -Δt i,C The RCC for monomer I can be calculated from the following formula:
in this embodiment, fig. 3 (a) is a charge-discharge voltage-time diagram, fig. 3 (b) is a partial enlarged view of a charge voltage curve, and fig. 3 (c) is a partial enlarged view of a discharge voltage curve.
When the voltage curve of the primary charged cell in a battery pack having 4 cells is as shown in FIG. 3, cell 1 is the first cell to charge to a charge cutoff voltage, and the charge cutoff time is t CE And after that the battery management system BMS stops charging to prevent overcharge. Taking CCVC of the monomer 1 as a benchmark, and shifting the CCVC of the monomer 1 to the right according to the CCVC consistency assumption can obtain CCVC of the monomer 4. At time t CE Thereafter, if monomer 4 can be charged alone, then its subsequent CCVC is as time t CE To t CE +Δt 4,C Shown in phantom in the monomer 4.Δt (delta t) 4,C The remaining charge time of monomer 4, also t CE Time and by combining t CE The voltage of time instant cell 4 is interpolated to the difference in the corresponding time instant on CCVC of cell 1 as indicated by the circle in fig. 3 (b). If the charge current is I, the RCC of monomer 4 can be calculated by the following formula:
similarly, the voltage curves of other monomers are translated by taking the monomer voltage curve with the lowest discharge end voltage as a reference, and the RDC of each monomer is calculated by interpolation i The method specifically comprises the following steps:
based on cloud monomer charging voltage data, each monomer RDC is calculated. Suppose that in a battery pack having n cells, cell j (1. Ltoreq.j. Ltoreq.n) is the first cell to discharge to a discharge cutoff voltage, and the discharge cutoff time is t DE And after that the BMS stops discharging to prevent overdischarge. Based on CCVC consistency assumption, CCVC translation of monomer j can obtain CCVC of other monomer i (i not equal to j), at time t DE Thereafter, if the monomer i can be discharged alone, the discharge is stopped for a time t DE The time until the cell i is discharged to the discharge cutoff voltage is the remaining discharge time (Δt) of the cell i i,D ) Then the starting point S of the CCVC curve of monomer i i The time of the interpolated point on CCVC of monomer j is t DE +Δt i,D The RDC of monomer I can be calculated from the formula:
when the voltage curve of the primary charged cell in the battery pack having 4 cells is as shown in FIG. 3, cell 4 is the first cell to discharge to the discharge cut-off voltage, and the discharge cut-off time is t DE And after that the BMS stops discharging to prevent overdischarge. CCVC of monomer 4 was used as a reference. According to CCVC consistency theory, CCVC of the monomer 4 is shifted leftwards to obtain CCVC of the monomer 1. At time t DE Thereafter, if monomer 1 alone can be discharged, then its subsequent CCVC is as time t DE -Δt 1,D To t DE Is shown in phantom in the monomer 1 of (2). Δt (delta t) 1,D Is the residual discharge time of monomer 1, also t DE Time and by combining t DE The voltage of time instant 1 is interpolated onto CCVC of cell 4 as the difference in the corresponding time instants shown by the circles in fig. 3 (c). If the charging current is I,the RDC of monomer 1 can be calculated from the formula:
in this embodiment, the calculation step of the internal resistance of the battery in step 3 is as follows:
and step 3-a, current data and voltage data are imported into MATLAB for screening, and the period that the battery is in the multi-stage constant current charging state is obtained, as shown in fig. 6.
Step 3-b, screening out micro-segments of voltage-current jump meeting the charging current conditions in the time period, wherein the charging current conditions are as follows:
for the front and rear current values of current jump in the same charging section, the kth charging current value I k Less than-75A, the (k+1) th charging current value I k+1 Greater than-60A and less than-30A, and there is a period of mid-SOC where the current switches, where the internal resistance represents an average level of the internal resistance of the battery.
Step 3-c, extracting the voltage and the current of the two stages before and after the charging jump from the micro-segment, respectively marked as U A 、U B 、I A 、I B The specific values are as follows:
for the kth and the kth+1th current values in the same charging section, each cell voltage value corresponding to the k value is recorded as U A The current value is recorded as I A Each monomer voltage value corresponding to the k+1 value is recorded as U B The current value is recorded as I B As shown in fig. 7.
Step 3-d, calculating to obtain the battery resistance r by adopting the ratio of the voltage difference and the current difference of the two stages, wherein the formula isThe battery resistance is an internal resistance predicted value corresponding to the current jump time of the battery, and meanwhile, the life condition of the battery at the moment can be predicted according to the resistance value because the resistance value can basically represent the average level of the internal resistance of the battery.
Step 4, for the battery monomerThe capacity and the internal resistance of the battery are subjected to data standardization to obtain a standardized sample data set, wherein the sample data set is A= { X 1 ,X 2 ,……,X m And normalized formula:
step 5, selecting k clustered clusters from the sample data set A, and initializing the clustering centers of the k clustered clusters, namely selecting k samples from the sample data set A as an initial clustering center, wherein the initial clustering center is B= { mu 1 ,μ 2 ,……,μ k }。
Wherein mu k =[μ C,k μ r,k ],μ C,k Volume data, μ representing the kth sample r,k Internal resistance data representing the kth sample.
Step 6, calculating the category of each sample by using Euclidean distance formula, and calculating each element X of the data set A i And centroid B j Euclidean distance d between ij Introducing weight factors, calculating to obtain classification result, and marking the classification result as classification set C, C= { C 1 ,C 2 ,……,C k Then calculate the new centroid mu j To update the cluster center.
In this embodiment, the weight factor is a gradient utilization scene coefficient delta, 0 < delta < 1, and the Euclidean distance formula after the weight factor is introduced is
Step 7, repeating the step 6 until mu j Terminating the iteration when the threshold sigma requirement is not changed, the threshold sigma requirement is met or the maximum iteration number N is reached, and obtaining a clustering result as shown in figure 8, wherein the threshold sigma and the iteration number N are equal toThe actual battery performance differences are relevant.
In this embodiment, fig. 8 (a) shows a clustering result when the scene coefficient δ is 0.1, fig. 8 (b) shows a clustering result when the scene coefficient δ is 0.5, and fig. 8 (c) shows a clustering result when the scene coefficient δ is 0.9.
As can be seen from fig. 8 (a), when δ is 0.1, the retired lithium batteries of the same class are arranged horizontally, and the internal resistance of the same class of batteries is high in uniformity and the capacity of the same class of batteries is poor in uniformity, so that the power density of the classified batteries is high in uniformity, and the classified batteries are suitable for power application scenarios with high power requirements.
As can be seen from fig. 8 (b), when δ is 0.5, the classification criteria are turned to the batteries with the same capacity and internal resistance, and the batteries with relatively high consistency between the capacity and the internal resistance are classified into one type, so that the classified batteries are suitable for application scenarios requiring both the capacity and the internal resistance.
As can be seen from fig. 8 (c), when δ is 0.9, the retired batteries of the same class are vertically arranged, and the classification standard is also that the consistency of the capacities is high, so that the batteries with higher capacities are more focused on classifying the batteries into one class, and therefore, the classified batteries are suitable for energy scenes with more severe requirements on the capacities.
Effects and effects of the examples
According to the battery classification method based on cloud historical data, external characteristic data, such as time, voltage, current, temperature and the like, of a battery in a multi-charge and discharge process are obtained through a battery management system BMS, cloud big data or experiments and the like on an automobile; secondly, selecting a valid data segment according to the historical data; thirdly, according to the cloud big data of the electric automobile, a fuzzy Kalman filtering algorithm, a residual charge power RCC algorithm and a residual discharge power RDC algorithm are fused to estimate the monomer capacity offline; thirdly, estimating ohmic internal resistance according to voltage-current jump caused by charge-discharge state switching in the effective charge segment data; and finally, classifying a large number of cloud monomers by using a K-means algorithm according to the internal resistance and the capacity of the battery monomers obtained by cloud computing. By the battery classification method based on cloud historical data, comprehensive classification of vehicle-mounted monomers of different electric vehicles can be realized, the sorting scale is enlarged, the service time and cost of a factory building are reduced, the consistency of the recombined monomers is effectively ensured, and the safety of echelon utilization scenes is ensured.
Further, the method of the embodiment mainly calculates the capacity and the internal resistance of the battery unit by a method for offline estimating the battery unit capacity based on the cloud big data of the electric vehicle and a method for online estimating the internal resistance of the power battery and detecting the health state of the vehicle, and finally classifies the battery unit by a K-means algorithm, wherein the accuracy of the capacity estimation of the battery unit depends on the battery unit capacity Q regarding the method for offline estimating the battery unit capacity based on the cloud big data of the electric vehicle sys Estimation accuracy and residual charge capacity RCC of cell i And residual discharge capacity RDC i Calculation accuracy of (2), RCC i And RDC i The calculation accuracy mainly depends on the sampling accuracy of cloud data, Q sys The estimation precision depends on the precision of the capacitance value calculated by the start-end SOC and the ampere-hour integration method of the selected charging section; in addition, the start SOC (SOC) of the selected charge data segment is evaluated by fuzzy synthesis min ) Terminating SOC (SOC) max ) Capacity result Q of ampere-hour integral calculation P Selecting proper output noise for Kalman filtering to obtain an estimation result conforming to the capacity attenuation trend of the battery pack of the electric automobile, combining all the monomer voltage data of the cloud, and calculating the RCC of each monomer based on the consistency assumption of the charge monomer voltage curve CCVC i And RDC i Finally, summing to obtain the capacity Q of each monomer i The error of capacity estimation is effectively reduced, and battery pack monomer capacity estimation based on electric automobile cloud data is achieved.
Therefore, the battery classification method based on cloud historical data in the embodiment utilizes a large amount of historical data, performs capacity and internal resistance estimation and noise processing on the vehicle-mounted battery on line, monitors the capacity and internal resistance of the power battery in a single-body level under the condition of different data sources, provides rapid data classification support for the power battery when the power battery is used in a echelon, and further performs cloud classification on the power battery through a K-means algorithm.
The above embodiments are preferred examples of the present invention, and are not intended to limit the scope of the present invention.
Claims (7)
1. The battery classification method based on cloud historical data is characterized by comprising the following steps of:
step 1, acquiring historical big data of a battery by adopting a battery management system BMS;
step 2, data cleaning is carried out on the historical big data to obtain voltage data and current data;
step 3, calculating the battery cell capacity and the battery internal resistance through the voltage data and the current data respectively to obtain the battery cell capacity and the battery internal resistance of m samples, wherein the battery cell capacity and the battery internal resistance are recorded as a sample set a = { x to be classified 1 ,x 2 ,……,x m And x is i Is a two-dimensional array formed by internal resistance and capacity, namely x i =[x C,i x r,i ]And i=1, 2, … …, m;
step 4, carrying out data standardization on the battery monomer capacity and the battery internal resistance to obtain a standardized sample data set, wherein the sample data set is A= { X 1 ,X 2 ,……,X m And normalized formula:
step 5, selecting k clustered clusters in the sample data set a, and initializing the clustering centers of k clustered clusters, namely selecting k samples from the sample data set a as an initial clustering center, wherein the initial clustering center is b= { mu 1 ,μ 2 ,……,μ k };
Step 6, calculating the category of each sample by using Euclidean distance formula, and calculating each element X of the data set A i And centroid B j Euclidean distance d between ij Introducing weight factors to calculate and obtain classification result, and recording the classification result as classificationSet C, c= { C 1 ,C 2 ,……,C k Then calculate the new centroid mu j Updating the cluster center;
step 7, repeating the step 6 until mu j Stopping iteration when the threshold sigma requirement is not changed, the threshold sigma requirement is met or the maximum iteration number N is reached, obtaining a clustering result,
wherein x is C,i And x r,i The battery capacity and the battery internal resistance of the i-th sample to be classified are respectively,and->Respectively, minimum and maximum values of battery capacity in the ith sample to be classified, +.>And->Respectively the minimum value and the maximum value of the internal resistance of the battery in the ith sample to be classified, mu k =[μ C,k μ r,k ],μ C,k Volume data, μ representing the kth sample r,k Internal resistance data representing the kth sample.
2. The battery classification method based on cloud history data according to claim 1, wherein:
the step of calculating the battery cell capacity in the step 3 is as follows:
step 3-1, screening out a complete charging engineering data graph from the voltage data and the current data;
step 3-2, performing preliminary estimation on the capacity of the battery pack based on the charging process by adopting an safety integral method to obtain a preliminary estimation result Q P ;
Step 3-3, adopting fuzzy Kalman filtering pairThe preliminary estimation result Q P Performing optimization correction to obtain an optimal estimation result Q of the battery capacity sys ;
Step 3-4, calculating the residual charge capacity RCC of each monomer based on the voltage data of the cloud monomer i And residual discharge capacity RDC i Then the capacity Q of each battery cell is calculated according to the following formula i ,
Q i =Q sys +RCC i +RDC i
In which Q i Battery capacity for the ith cell, RCC i RDC for the residual charge capacity of the ith monomer i The residual discharge capacity of the i-th monomer.
3. The battery classification method based on cloud history data according to claim 2, wherein:
wherein t is 1 Indicating the charge start time of the charging section, t 2 Indicating the charge end time of the charging section, I (t) indicating the current value at time t in the charging section,
the remaining charge capacity RCC in the step 3-4 i And the residual discharge capacity RDC i All are calculated by interpolation method.
4. The battery classification method based on cloud history data according to claim 1, wherein:
wherein, the calculation step of the internal resistance of the battery in the step 3 is as follows:
step 3-a, the current data and the voltage data are imported into MATLAB for screening, and a period of time when the battery is in a multi-stage constant current charging state is obtained;
step 3-b, screening out micro-segments of voltage-current jump meeting the charging current condition in the period;
step 3-c, extracting the voltage and the current of the two stages before and after the charging jump from the micro-segment, and respectively marking as U A 、U B 、I A 、I B ;
And 3-d, calculating the battery resistance r by adopting the ratio of the voltage difference and the current difference of the two stages.
5. The battery classification method based on cloud history data as claimed in claim 4, wherein:
in the step 3-b, the charging current conditions are as follows:
for the front and rear current values of current jump in the same charging section, the kth charging current value I k Less than-75A, the (k+1) th charging current value I k+1 Greater than-60A and less than-30A, and there is a period of mid-SOC where the current switches, where the internal resistance represents an average level of the internal resistance of the battery,
in the step 3-c, the specific values are as follows:
for the kth and the kth+1th current values in the same charging section, each cell voltage value corresponding to the k value is recorded as U A The current value is recorded as I A Each monomer voltage value corresponding to the k+1 value is recorded as U B The current value is recorded as I B 。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010871109.8A CN111967194B (en) | 2020-08-26 | 2020-08-26 | Battery classification method based on cloud historical data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010871109.8A CN111967194B (en) | 2020-08-26 | 2020-08-26 | Battery classification method based on cloud historical data |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111967194A CN111967194A (en) | 2020-11-20 |
CN111967194B true CN111967194B (en) | 2023-06-20 |
Family
ID=73390529
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010871109.8A Active CN111967194B (en) | 2020-08-26 | 2020-08-26 | Battery classification method based on cloud historical data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111967194B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112615070B (en) * | 2020-12-15 | 2022-03-11 | 中国人民解放军陆军勤务学院 | Battery energy storage integration system and method |
CN113083739B (en) * | 2021-04-07 | 2023-03-21 | 东软睿驰汽车技术(沈阳)有限公司 | Battery cell sorting method and device and computer equipment |
CN115144778A (en) * | 2022-09-02 | 2022-10-04 | 力高(山东)新能源技术有限公司 | Method for estimating internal resistance of battery by big data |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110135478A (en) * | 2019-04-29 | 2019-08-16 | 上海理工大学 | Consider that echelon utilizes the retired lithium battery classification method of scene |
CN110752410A (en) * | 2019-10-30 | 2020-02-04 | 上海理工大学 | Method for rapidly sorting and recombining retired lithium batteries |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9869724B2 (en) * | 2013-07-24 | 2018-01-16 | Rohm Co., Ltd. | Power management system |
-
2020
- 2020-08-26 CN CN202010871109.8A patent/CN111967194B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110135478A (en) * | 2019-04-29 | 2019-08-16 | 上海理工大学 | Consider that echelon utilizes the retired lithium battery classification method of scene |
CN110752410A (en) * | 2019-10-30 | 2020-02-04 | 上海理工大学 | Method for rapidly sorting and recombining retired lithium batteries |
Also Published As
Publication number | Publication date |
---|---|
CN111967194A (en) | 2020-11-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111967194B (en) | Battery classification method based on cloud historical data | |
CN110068774B (en) | Lithium battery health state estimation method and device and storage medium | |
CN109633474B (en) | Lithium ion battery residual life prediction method | |
CN110146822B (en) | Vehicle power battery capacity online estimation method based on constant-current charging process | |
CN110596593A (en) | Lithium ion battery SOC estimation method based on intelligent adaptive extended Kalman filtering | |
EP4019321A1 (en) | Battery managing method and device, and vehicle | |
CN107741568B (en) | Lithium battery SOC estimation method based on state transition optimization RBF neural network | |
CN107132490B (en) | Method for estimating state of charge of lithium battery pack | |
CN111007399A (en) | Lithium battery state of charge prediction method based on improved generation countermeasure network | |
CN107024663A (en) | The lithium battery screening technique clustered based on charging curve feature KPCA | |
CN109633456B (en) | Power lithium battery pack SOC estimation method based on segmented voltage identification method | |
WO2023221589A1 (en) | Method, apparatus and system for detecting battery health state | |
CN113567873B (en) | Battery capacity and SOC estimation method based on single charge curve difference in battery pack | |
CN113064089A (en) | Internal resistance detection method, device, medium and system of power battery | |
CN110850298A (en) | Lithium battery SOH estimation method and system based on data driving | |
CN112462282A (en) | Method for determining real-time state of charge of battery pack based on mechanism model | |
Park et al. | Capacity estimation of Li-ion batteries using constant current charging voltage with multilayer perceptron | |
CN112698217B (en) | Battery monomer capacity estimation method based on particle swarm optimization algorithm | |
CN113125967A (en) | Lithium battery SOE calculation method based on temperature rise prediction | |
CN111781529A (en) | Battery pack monomer capacity estimation method and device based on electric vehicle cloud data | |
CN116953547A (en) | Energy storage battery health evaluation method, system, equipment and storage medium | |
CN106815406B (en) | Power battery SOC estimation method based on feature model | |
CN116068402A (en) | New energy automobile lithium battery state prediction method, device, equipment and storage medium | |
CN112162198B (en) | Battery health diagnosis system and method suitable for hybrid vehicle | |
CN112946480B (en) | Lithium battery circuit model simplification method for improving SOC estimation real-time performance |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |