Lithium battery health state estimation method based on full charge current and double correlation

Abstract

A lithium battery health state estimation method based on full current and double correlation obtains the health state and full current sequence of each charge and discharge cycle of the lithium battery, extracts features and classifies them according to the full current sequence; calculates the gray correlation between the feature and the health state as the first correlation analysis, selects the feature with the highest gray correlation and the feature with the highest gray correlation in other classes, and determines the number of input features according to the maximum correlation interval method; the feature with the highest correlation and the feature with the highest correlation in other classes use the Pearson correlation coefficient analysis method to perform the second correlation analysis between the features, and construct the optimal input feature. Finally, the support vector machine is optimized by the sparrow search algorithm to construct a lithium battery health state estimation model; the health state estimation model is used to estimate the health state of the lithium battery. The present invention has the advantages of fast, effective and accurate estimation of health state, and provides a reference for the selection of input feature sequences.

Description

Technical Field

[0001] The present invention relates to the field of battery technology, and more specifically, to a method for estimating the health state of a lithium battery based on full current and dual correlation. Background Art

[0002] Lithium batteries have the advantages of high energy density, high safety, long cycle life, and no pollution, and are widely used in electric vehicles, energy storage systems, aerospace and other fields. As a power source, the health status of lithium batteries affects the safe operation and replacement of batteries. With the continuous charge and discharge cycle of lithium batteries, their storage capacity and fast charge and discharge capabilities will decline and weaken. The health status of lithium batteries is one of the key indicators to describe battery performance and reflects the degree of battery aging. Therefore, it is crucial to accurately estimate the health status of lithium batteries.

[0003] In practical applications, constant current and constant voltage charging is the most common charging method for lithium batteries. However, the state of charge of lithium batteries is generally not zero when charging begins, and it is difficult to obtain charging data starting from the zero state for battery health state estimation. Therefore, the full current sequence of the constant voltage charging stage is used to obtain features. After performing a double correlation analysis on the features, the optimal input feature sequence can be obtained, which can characterize the aging degree of lithium batteries from multiple angles to improve the estimation accuracy of the health state of lithium batteries. Support vector machine is a network with complete mathematical theory, small sample training error, fast convergence speed, and conversion of nonlinear problems into linear problems. The penalty coefficient and width factor of the support vector machine have an important influence on the prediction performance of the model, and manual adjustment is very time-consuming and troublesome.

[0004] The present invention proposes a method for estimating the health state of a lithium battery based on full current and double correlation. The method is: first, the health state and full current sequence of each cycle are obtained; then, the full current sequence of each charge and discharge cycle is statistically analyzed, 10 features are calculated and divided into 5 categories; then, each feature is subjected to a first correlation analysis with the health state sequence, and the number (dimension) of input features is determined by the maximum correlation interval method; combined with the determined number (dimension) of input features, double correlation analysis is performed to quickly screen out features with strong correlation, determine and construct the optimal input features, and finally estimate the health state of the lithium battery by a support vector machine optimized by the sparrow algorithm. The method for estimating the health state of a lithium battery based on full current and double correlation proposed by the present invention considers the source of lithium battery data: the full current sequence in the constant voltage charging stage. Because the charging process of lithium batteries in actual applications generally starts with a non-zero charge state, which is different from the charging process starting with a zero charge state in a laboratory environment. Using the full current sequence in the constant voltage charging stage to estimate the health state of lithium batteries is more in line with the actual application of lithium batteries, and has great guiding significance and reference value. Summary of the invention

[0005] In view of the current research on the health status estimation of lithium batteries, the voltage data of the whole stage of constant current and constant voltage charging starting from the battery charge state zero, or the current data is generally used to carry out state estimation. The present invention proposes a lithium battery health status estimation method based on full current and double correlation, which can not only quickly and effectively respond to the capacity decay, but also conform to the non-zero charge state charging process of actual application of lithium batteries.

[0006] In order to solve the above practical problems, the technical solution adopted by the present invention is:

[0007] The method for estimating the health status of a lithium battery based on full current and double correlation is characterized in that it specifically includes the following steps:

[0008] S1, obtain the health status and full current sequence of each charge and discharge cycle of the lithium battery;

[0009] S2. Perform statistical analysis on the full current sequence of each charge and discharge cycle, calculate the features from the full current sequence and classify them:

[0010] The specific contents of step S2 are as follows:

[0011] After obtaining the lithium battery charging cycle data, the data of the characteristic along the cycle charging and discharging process is [F r,1 ,F r,2 ,…,F r,n ], constant voltage charging time is t i =t i,e -t i,0 , constant voltage charging time accounts for Where r is the feature number, r = 1, ..., 10, t i is the total duration of constant voltage charging in the ith charge-discharge cycle, n is the number of charge-discharge cycles, e is the end time of constant voltage charging, and 0 is the start time of constant voltage charging;

[0012] The constant voltage charging current area is The constant voltage charging current area ratio is S i is the area enclosed by the charging current in the constant voltage stage of the i-th charge-discharge cycle, L is the charging time, I i,j is the jth sampled current value in the i-th charge and discharge cycle, N i is the total number of samples in the i-th charge and discharge cycle, S i ' is the area ratio of the charging current in the constant voltage stage of the i-th charge and discharge cycle;

[0013] The constant voltage charging current entropy is The constant voltage charging current entropy ratio is I(x)i is the charging current entropy during the constant voltage phase of the i-th charge-discharge cycle, p(x i,j ) is the probability of the jth sampling current value in the i-th charge and discharge cycle, I(x) i ' is the entropy ratio of the charging current in the constant voltage stage of the i-th charge and discharge cycle;

[0014] The constant voltage charging capacity is C i =C i,e -C i,0 , constant voltage charging capacity accounts for Among them, C i is the charging capacity in the constant voltage stage of the i-th charge-discharge cycle;

[0015] The constant voltage charging energy is E i =E i,e -E i,0 , the constant voltage charging energy ratio is E i is the energy charged in the constant voltage charging phase of the i-th charge-discharge cycle;

[0016] S3, performing a first correlation analysis on each feature and health status sequence;

[0017] The specific content of step S3 is as follows:

[0018] The first correlation analysis is as follows: for each feature [F r,1 ,F r,2 ,…,F r,n ] are respectively related to the health status sequence [H 1 ,H 2 ,…,H n ] to perform grey relational analysis and calculate the grey relational coefficient [ξ r (1),ξ r (2),…,ξ r (n)], and take the average value of the grey correlation coefficient R r As the grey correlation degree sequence between each feature and the health status of the lithium battery [R 1 ,R 2 ,…,R m ], where the grey correlation coefficient R r ∈[0,1] is each feature F r Grey correlation degree with the health status of lithium battery, y(n)=[H 1 ,H 2 ,…,H n ], x r (n)=[F r,1 ,F r,2 ,…,F r,n], ρ∈(0,1) is the resolution coefficient, grey relational degree m is the number of grey relational degree;

[0019] S4. According to the maximum correlation interval method, determine whether the number M of input features is greater than one. If it is greater than one, execute steps S5 and S6; otherwise, directly jump to step S7;

[0020] S5, select the feature with the highest correlation in each class, and select the feature with the highest correlation among all features;

[0021] S6, performing a second correlation analysis between the features with the highest correlation and the features with the highest correlation in other classes, and selecting input features based on the mutual correlation between the features;

[0022] The second correlation analysis is:

[0023] The feature with the highest grey correlation degree is subjected to Pearson correlation analysis with the features with the highest grey correlation degree in the other four categories, and four mutual correlations P are obtained. 1 , P 2 , P 3 , P 4 , sort the mutual correlation from high to low, select the features of the other 4 categories corresponding to the first M-1 mutual correlations, and add the feature with the highest grey correlation to form M features as input features

[0024] Among them, Pearson correlation analysis a is the feature with the highest grey correlation degree;

[0025] S7, using the input features and the health status sequence to iteratively find the penalty coefficient and kernel function width factor of the optimal support vector machine through the sparrow search algorithm, and establish an estimation model for the health status of the lithium battery;

[0026] The specific content of step S7 is as follows:

[0027] Take the first k, k=1, 2, ..., n-1 sample values ​​of the input feature and health status sequence and As the training set, take the k+1, k+2,…,n sample values and As a test set, the process of iteratively finding the penalty coefficient and kernel function width factor of the optimal support vector machine through the sparrow search algorithm is as follows: the root mean square error of the true value and the estimated value of the health state is used as the individual fitness function, the penalty coefficient and width factor are mapped to the two-dimensional position of each sparrow, the global optimal two-dimensional position is iteratively optimized, and the penalty coefficient and width factor of the optimal two-dimensional position are applied to establish the input feature and the estimation model of the health state of the lithium battery;

[0028] S8. Use the established lithium battery health status estimation model to estimate the health status of the lithium battery.

[0029] As a further improvement of the present invention, the specific contents of step S1 are as follows:

[0030] The full charge is the charging state of the constant voltage stage during the constant current and constant voltage charging process; the full charge current sequence is the charging current sequence I in the constant voltage stage. 1 ,I 2 ,…,I t ; The health status data sequence of lithium battery in the cyclic charge and discharge process is H 1 ,H 2 ,...,H n , the measured data Among them, H i is the health status data of lithium battery in the i-th, i=1,2,K,n charge and discharge process, n is the number of cycle charge and discharge, C i is the maximum discharge capacity of the lithium battery in the ith charge and discharge process, and C is the rated capacity of the lithium battery.

[0031] As a further improvement of the present invention, the full current sequence in step S2 calculates the following 10 characteristics and divides them into 5 categories:

[0032] (1) Constant voltage charging time and constant voltage charging time ratio;

[0033] (2) Constant voltage charging current area and constant voltage charging current area ratio;

[0034] (3) Constant voltage charging current entropy and constant voltage charging current entropy ratio;

[0035] (4) Constant voltage charging capacity and constant voltage charging capacity ratio;

[0036] (5) Constant voltage charging energy and constant voltage charging energy ratio.

[0037] As a further improvement of the present invention, the specific content of step S4 is as follows:

[0038] The maximum correlation interval method is: when R r ∈[0,0.6], add other new features, when R r∈(0.6,0.7], the number of input features M = 4, when R r ∈(0.7,0.8], the number of input features M = 3, when R r ∈(0.8,0.9], the number of input features M = 2, when R r ∈(0.9,1], the number of input features M = 1; the number of input features M is determined in this way.

[0039] As a further improvement of the present invention, the specific content of step S5 is as follows:

[0040] The grey correlation between each feature and the health status of the lithium battery [R 1 ,R 2 ,…,R m ] After arranging in descending order, select the feature with the highest grey correlation degree in each class, and select the feature with the highest grey correlation degree among all features.

[0041] As a further improvement of the present invention, the detailed steps of optimizing the penalty coefficient and width factor of the support vector machine by the sparrow search algorithm in step S7 are as follows;

[0042] (a) Read data, perform data preprocessing, and build a health status estimation model;

[0043] (b) Initialize the sparrow population parameters and determine the value range of the penalty coefficient and width factor;

[0044] (c) Divide the predator detectors and followers and randomly select vigilant ones to update their location information;

[0045] (d) Determine the fitness function of the sparrow search algorithm. According to the principle of the sparrow search algorithm, find the optimal function value, that is, determine the position of the best sparrow individual;

[0046] (e) Obtaining the optimal values ​​of penalty coefficient and width factor according to the best individual sparrow position;

[0047] (f) Repeat steps (c) to (e) to determine whether the termination condition, i.e., the maximum number of iterations, is met;

[0048] (g) Assign the optimal penalty coefficient and width factor to the support vector machine.

[0049] Compared with other inventions, the difference of the solution proposed by the present invention is:

[0050] 1. The data source of the present invention is: in actual application, the charging process of lithium batteries generally starts from a non-zero charge state, which is different from the charging process starting from a zero charge state in a laboratory environment. Using the full current sequence in the constant voltage charging stage to estimate the health status of lithium batteries is more in line with the actual application of lithium batteries.

[0051] 2. The research method of the present invention is: first, statistically analyze the full-charge current sequence to calculate 10 features and divide them into 5 categories; then perform the first correlation analysis, that is, perform grey correlation analysis on each feature and the health status sequence to obtain the feature with the maximum correlation and the feature with the largest correlation in the remaining 4 categories; determine the number (dimension) of input features by the maximum correlation interval method; then, based on the determined number (dimension) of input features, perform the second correlation analysis, that is, perform Pearson correlation analysis between the feature with the maximum correlation and the feature with the largest correlation in the remaining 4 categories to obtain 4 mutual correlations, and then determine and construct the optimal input feature; finally, bring in the support vector machine optimized by the sparrow algorithm to estimate the health status of the lithium battery. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flow chart of a method for estimating the health status of a lithium battery based on full current and double correlation provided by an embodiment of the present invention;

[0053] Figure 2 This is an aging diagram of a lithium battery health status along with charge and discharge cycles provided by an embodiment of the present invention;

[0054] Figure 3 is a full charge current sequence diagram of a lithium battery charging process provided by an embodiment of the present invention;

[0055] Figure 4 It is a comparison diagram of the lithium battery health state estimation method based on full current and double correlation provided by an embodiment of the present invention and the lithium battery health state estimation results of three comparative experiments;

[0056] Figure 5 It is a comparison diagram of health state estimation errors of a lithium battery health state estimation method based on full charge current and double correlation provided by an embodiment of the present invention and three comparative experiments. DETAILED DESCRIPTION

[0057] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0058] like Figure 1 FIG. 1 is a flow chart of a method for estimating the health status of a lithium battery based on full current and double correlation provided by an example of the present invention. Figure 1 The method shown comprises the following steps:

[0059] S1, obtain the health status and full current sequence of each charge and discharge cycle of the lithium battery;

[0060] In the embodiment of the present invention, in step S1, full charge is a charging state in the constant current and constant voltage charging process, and the full charge current sequence is the charging current sequence I in the constant voltage stage. 1 ,I 2 ,…,I t ; The health status data sequence of lithium battery in the cyclic charge and discharge process is H 1 ,H 2 ,...,H n , the measured data Among them, H i is the health status data of the lithium battery in the i-th (i=1,2,K,n) charge and discharge process, n is the number of cycle charge and discharge, C i is the maximum discharge capacity of the lithium battery in the i-th charge and discharge process, and C is the rated capacity of the lithium battery;

[0061] S2. Perform statistical analysis on the full current sequence of each charge and discharge cycle, and calculate the following 10 features from the full current sequence and divide them into 5 categories: (1) constant voltage charging time and constant voltage charging time ratio, (2) constant voltage charging current area and constant voltage charging current area ratio, (3) constant voltage charging current entropy and constant voltage charging current entropy ratio, (4) constant voltage charging capacity and constant voltage charging capacity ratio, (5) constant voltage charging energy and constant voltage charging energy ratio;

[0062] In the embodiment of the present invention, in step S2, after obtaining the charging cycle data of the lithium battery, 10 features are calculated from the current data curve of the constant voltage charging stage and divided into 5 categories: (1) constant voltage charging time and constant voltage charging time ratio, (2) constant voltage charging current area and constant voltage charging current area ratio, (3) constant voltage charging current entropy and constant voltage charging current entropy ratio, (4) constant voltage charging capacity and constant voltage charging capacity ratio, (5) constant voltage charging energy and constant voltage charging energy ratio;

[0063] The data of the characteristic along the cycle charge and discharge process is [F r,1 ,F r,2 ,…,F r,n ], constant voltage charging time is t i =t i,e -t i,0 , constant voltage charging time accounts for Where r is the feature number, r = 1, ..., 10, t i is the total duration of constant voltage charging in the ith charge-discharge cycle, n is the number of charge-discharge cycles, e is the end time of constant voltage charging, and 0 is the start time of constant voltage charging.

[0064] The constant voltage charging current area is The constant voltage charging current area ratio is S iis the area enclosed by the charging current in the constant voltage stage of the i-th charge-discharge cycle, L is the charging time, I i,j is the jth sampled current value in the i-th charge and discharge cycle, N i is the total number of samples in the i-th charge and discharge cycle, S i ' is the area ratio of the charging current in the constant voltage stage of the i-th charge and discharge cycle.

[0065] The constant voltage charging current entropy is The constant voltage charging current entropy ratio is I(x) i is the charging current entropy during the constant voltage phase of the i-th charge-discharge cycle, p(x i,j ) is the probability of the jth sampling current value in the i-th charge and discharge cycle. I(x) i ' is the entropy ratio of the charging current in the constant voltage stage of the i-th charge and discharge cycle.

[0066] The constant voltage charging capacity is C i =C i,e -C i,0 , constant voltage charging capacity accounts for Among them, C i is the charging capacity in the constant voltage stage of the i-th charge and discharge cycle.

[0067] The constant voltage charging energy is E i =E i,e -E i,0 , the constant voltage charging energy ratio is E i It is the energy charged in the constant voltage charging stage of the i-th charge and discharge cycle.

[0068] S3, performing a first correlation analysis on each feature and health status sequence;

[0069] In the embodiment of the present invention, in step S3, the first correlation analysis is: for each feature [F r,1 ,F r,2 ,…,F r,n ] are respectively related to the health status sequence [H 1 ,H 2 ,…,H n ] to perform grey relational analysis and calculate the grey relational coefficient [ξ r (1),ξ r (2),…,ξ r (n)], and take the average value of the grey correlation coefficient R r As the grey correlation degree sequence between each feature and the health status of the lithium battery [R 1 ,R 2 ,…,R m ]. Among them, the grey correlation coefficient R r ∈[0,1] is each feature F r Grey correlation degree with the health status of lithium battery, y(n)=[H 1 ,H 2 ,…,H n ], x r (n)=[F r,1 ,F r,2 ,…,F r,n ], ρ∈(0,1) is the resolution coefficient, grey relational degree m is the number of grey relational degree.

[0070] S4. According to the maximum correlation interval method, determine whether the number M of input features is greater than one. If it is greater, execute steps S5 and S6; otherwise, directly jump to step S7;

[0071] In the embodiment of the present invention, in step S4, the maximum correlation interval method is: when R r ∈[0,0.6], add other new features, when R r ∈(0.6,0.7], the number of input features M = 4, when R r ∈(0.7,0.8], the number of input features M = 3, when R r ∈(0.8,0.9], the number of input features M = 2, when R r ∈(0.9,1], the number of input features M = 1; the number of input features M is determined in this way.

[0072] S5, select the feature with the highest correlation in each class, and select the feature with the highest correlation among all features;

[0073] In the embodiment of the present invention, in step S5, the grey correlation degree [R 1 ,R 2 ,…,R m ] After arranging in descending order, select the feature with the highest correlation and the feature with the highest correlation in other classes.

[0074] S6, performing a second correlation analysis between the features with the highest correlation and the features with the highest correlation in other classes, and selecting input features based on the mutual correlation between the features;

[0075] In the embodiment of the present invention, in step S6, the second correlation analysis is: the feature with the highest grey correlation is subjected to Pearson correlation analysis with the features with the highest grey correlation in the other four categories, and four mutual correlations P are obtained. 1 , P 2 , P 3 , P 4, sort the mutual correlation from high to low, select the features of the other 4 categories corresponding to the first M-1 mutual correlations, and add the feature with the highest grey correlation to form M features as input features Among them, Pearson correlation analysis a is the feature with the highest grey correlation.

[0076] S7, using the input features and the health status sequence to iteratively find the penalty coefficient and kernel function width factor of the optimal support vector machine through the sparrow search algorithm, and establish an estimation model for the health status of the lithium battery;

[0077] In the embodiment of the present invention, in step S7, the first k (k=1, 2, ..., n-1) sample values ​​of the input feature and health status sequence are taken. and As the training set, take the k+1, k+2,…,n sample values and As a test set, the process of iteratively finding the penalty coefficient and kernel function width factor of the optimal support vector machine through the sparrow search algorithm is as follows: the root mean square error of the true value and the estimated value of the health state is used as the individual fitness function, the penalty coefficient and width factor are mapped to the two-dimensional position of each sparrow, the global optimal two-dimensional position is iteratively optimized, and the penalty coefficient and width factor of the optimal two-dimensional position are applied to establish the input feature and the estimation model of the health state of the lithium battery. The detailed steps of the sparrow search algorithm to optimize the penalty coefficient and width factor of the support vector machine are as follows.

[0078] (a) Read data, perform data preprocessing, and build a health status estimation model.

[0079] (b) Initialize the sparrow population parameters and determine the value range of the penalty coefficient and width factor.

[0080] (c) Divide the predator detectors and followers and randomly select vigilant ones to update their location information.

[0081] (d) Determine the fitness function of the sparrow search algorithm. According to the principle of the sparrow search algorithm, find the optimal function value, that is, determine the position of the best sparrow individual.

[0082] (e) Obtain the optimal values ​​of the penalty coefficient and width factor based on the best individual sparrow position.

[0083] (f) Repeat steps (c) to (e) to determine whether the termination condition, i.e., the maximum number of iterations, is met.

[0084] (g) Assign the optimal penalty coefficient and width factor to the support vector machine.

[0085] The sparrow search algorithm in step S7 is:

[0086] The sparrow search algorithm divides the sparrow population into finders, followers, and vigilant ones according to the strength of their foraging ability, and iteratively updates the location information of foraging finders, followers, and vigilant ones according to their individual fitness, thereby obtaining the location information of the global optimal ones. The sparrow search algorithm updates the position in the multidimensional solution space as follows.

[0087] The discoverer's location update status:

[0088]

[0089] In the formula Represents the position state; R 2 is the alert value, α and R 2 are all random numbers between 0 and 1; T max is the maximum number of iterations of the population; Q is a random number from a standard normal distribution; L is a 1×d unit vector; ST is the alert threshold.

[0090] Follower's location update status:

[0091]

[0092] In the formula represents the optimal position of the discoverer at the t+1th iteration; n is the number of sparrows in the population; A represents a 1×d vector, A + =A T (AA T ) -1

[0093] Vigilant's location update status:

[0094]

[0095] In the formula represents the global optimal position; represents the global worst position, β is a random number that follows a standard normal distribution; K is a random number between -1 and 1; f i is the individual fitness value, f g 、f w are the current optimal and worst fitness values ​​respectively. ε is a constant to prevent the denominator from being zero.

[0096] S8. Estimate the health state of the lithium battery using the established lithium battery health state estimation model;

[0097] In the embodiment of the present invention, in step S8, the k+1th, ..., nth charge-discharge cycle data of the input feature are The established lithium battery health status estimation model is introduced to estimate the health status of the lithium battery;

[0098] The following is an example to illustrate the process and estimation accuracy of the lithium battery health status estimation method based on full current and double correlation proposed by the present invention.

[0099] Example: 18650 lithium iron phosphate (LiFePO 4 ) The battery was subjected to a charge and discharge cycle experiment with a rated capacity of 2400mAh. During the charging phase, it was charged at a constant current of 1.2A. When the battery terminal voltage rose to 4.2V, the terminal voltage was kept constant at 4.2V for constant voltage charging. Charging was terminated when the charging current dropped to 48mA. After 2 minutes of storage, it was discharged at a constant current of 2.4A until the terminal voltage of the lithium battery dropped to 3V, and the discharge was terminated. The number of charge and discharge cycles was 290 times. Based on the health status of each cycle of charge and discharge of the lithium battery and the full current sequence, a double correlation analysis was performed to estimate the health status of the lithium battery, where Figure 2 The capacity decay process of the lithium battery health status along with the charge and discharge cycle. Figure 3 The full current sequence for the lithium battery charging cycle. The specific steps are as follows:

[0100] (1) After obtaining the lithium battery charging data, the full current sequence was statistically analyzed, and the following 10 characteristics were obtained and divided into 5 categories: (1) constant voltage charging time and constant voltage charging time ratio, (2) constant voltage charging current area and constant voltage charging current area ratio, (3) constant voltage charging current entropy and constant voltage charging current entropy ratio, (4) constant voltage charging capacity and constant voltage charging capacity ratio, (5) constant voltage charging energy and constant voltage charging energy ratio; and then the characteristic sequence of the lithium battery charging process of 290 times was obtained.

[0101] (2) Grey correlation analysis is performed on the 10 features and the health status sequence. The grey correlation is shown in Table 1. According to the maximum correlation interval method, the maximum correlation feature is the constant voltage charging current entropy, that is, F 5 = 0.8563∈(0.8,0.9), it can be determined that the number of input features M = 2. Execute steps (3) and (4).

[0102] Table 1

[0103] feature <![CDATA[F 1 ]]> <![CDATA[F 2 ]]> <![CDATA[F 3 ]]> <![CDATA[F 4 ]]> <![CDATA[F 5 ]]> <![CDATA[F 6 ]]> <![CDATA[F 7 ]]> <![CDATA[F 8 ]]> <![CDATA[F 9 ]]> <![CDATA[F 10 ]]> <![CDATA[r i ]]> 0.7025 0.7041 0.7097 0.6694 0.8563 0.8501 0.7097 0.6694 0.7095 0.6728

[0104] (3) Select the features with the highest correlation and the features with the highest correlation in other categories, namely, constant voltage charging time ratio, constant voltage charging current area, constant voltage charging current entropy, constant voltage charging capacity, and constant voltage charging energy; the grey correlation is shown in Table 2.

[0105] Table 2

[0106] feature <![CDATA[F 2 ]]> <![CDATA[F 3 ]]> <![CDATA[F 5 ]]> <![CDATA[F 7 ]]> <![CDATA[F 9 ]]> <![CDATA[r i ]]> 0.7041 0.7097 0.8563 0.7097 0.7095

[0107] (4) The Pearson correlation analysis results of the constant voltage charging current entropy with the highest grey correlation degree and the constant voltage charging time ratio, constant voltage charging current area, constant voltage charging capacity, and constant voltage charging energy are shown in Table 3.

[0108] Construct an input feature with the number of input features being two, namely, constant voltage charging current entropy and constant voltage charging time ratio as input features.

[0109] Table 3

[0110] feature <![CDATA[F 2 ]]> <![CDATA[F 3 ]]> <![CDATA[F 7 ]]> <![CDATA[F 9 ]]> <![CDATA[F 5 ]]> 0.9741 0.9735 0.9735 0.9734

[0111] (5) The constant voltage charging current entropy, constant voltage charging time, and lithium battery health status data of the same charging cycle are taken as a group of data. The 290 groups of lithium battery charging process data are divided into 1:1, that is, groups 1-145 are training sets, and groups 146-290 are test sets. The sparrow search algorithm is used to iteratively optimize the penalty coefficient and width factor. The penalty coefficient and width factor are mapped to the two-dimensional position information of the sparrow. After randomly initializing the two-dimensional position information, the iterative optimization begins. The population size and the number of iterations are 100 and 50 respectively; the penalty coefficient is 0-50, and the width factor is 0.0001-0.1. At the end of the iterative optimization, the penalty coefficient and the width factor are 7.4765 and 0.0013 respectively.

[0112] (6) The penalty coefficient and width factor after optimization are applied to establish a lithium battery health state estimation model, and the lithium battery health state estimation is performed based on the test set. At the same time, three groups of comparative experiments are set up to conduct comparative experiments with the method proposed in the present invention, as shown in Table 4. Figure 4 , Figure 5 They are the comparison of estimation results and error comparison of comparative experiments, and Table 5 shows the average absolute error and maximum error of the estimation results of lithium battery health status in comparative experiments.

[0113] Table 4

[0114]

[0115] Table 5

[0116]

[0117] From the error comparison of the lithium battery health state estimation results in Table 5, it can be seen that the average error and maximum error of the method proposed in the present invention are lower than those of comparative tests 1, 2, and 3, which shows that the input features composed of the constant voltage charging current entropy and the constant voltage charging time ratio obtained after the double correlation analysis can more comprehensively characterize the degradation of the lithium battery health state. This shows that the lithium battery health state estimation method based on full current and double correlation proposed in the present invention has the advantages of simple operation, small error, and high efficiency.

[0118] It should be pointed out that, according to the needs of implementation, the various steps / components described in this application can be split into more steps / components, and two or more steps / components or partial operations of steps / components can be combined into new steps / components to achieve the purpose of the present invention.

[0119] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.

Claims

1. A lithium battery health status estimation method based on full current and double correlation, characterized in that: The specific steps include: S1, obtaining the health status sequence and full current sequence of each cycle charge and discharge of the lithium battery; The specific contents of step S1 are as follows: The full charge is a charging state in the constant current and constant voltage charging process, which is carried out to the constant voltage stage; the full charge current sequence is the charging current sequence I1, I2, ..., I t ; The health status data sequence of lithium batteries during the cyclic charge and discharge process is H1, H2, …, H n , the measured data Among them, H i is the health status data of lithium battery in the i-th, i=1,2,K,n charge and discharge process, n is the number of cycle charge and discharge, C i is the maximum discharge capacity of the lithium battery in the i-th charge and discharge process, and C is the rated capacity of the lithium battery; S2. Perform statistical analysis on the full current sequence of each charge and discharge cycle, calculate the features from the full current sequence and classify them: The specific contents of step S2 are as follows: After obtaining the lithium battery charging cycle data, the data of the characteristic along the cycle charging and discharging process is [F r,1 ,F r,2 ,…,F r,n ], constant voltage charging time is t i =t i,e -t i,0 , constant voltage charging time accounts for Where r is the feature number, r = 1, ..., 10, t i is the total duration of constant voltage charging in the ith charge-discharge cycle, n is the number of charge-discharge cycles, e is the end time of constant voltage charging, and 0 is the start time of constant voltage charging; The constant voltage charging current area is The constant voltage charging current area ratio is S i is the area enclosed by the charging current in the constant voltage stage of the i-th charge-discharge cycle, L is the charging time, I i,j is the jth sampled current value in the i-th charge and discharge cycle, N i is the total number of samples in the i-th charge and discharge cycle, S i ' is the area ratio of the charging current in the constant voltage stage of the i-th charge and discharge cycle; The constant voltage charging current entropy is The constant voltage charging current entropy ratio is I(x) i is the charging current entropy during the constant voltage phase of the i-th charge-discharge cycle, p(x i,j ) is the probability of the jth sampling current value in the i-th charge and discharge cycle, I(x) i ' is the entropy ratio of the charging current in the constant voltage stage of the i-th charge and discharge cycle; The constant voltage charging capacity is C i =C i,e -C i,0 , constant voltage charging capacity accounts for Among them, C i is the charging capacity in the constant voltage stage of the i-th charge-discharge cycle; The constant voltage charging energy is E i =E i,e -E i,0 , the constant voltage charging energy ratio is E i is the energy charged in the constant voltage charging phase of the i-th charge-discharge cycle; The step S2 calculates the following 10 characteristics of the full current sequence and divides them into 5 categories: (1) Constant voltage charging time and constant voltage charging time ratio; (2) Constant voltage charging current area and constant voltage charging current area ratio; (3) Constant voltage charging current entropy and constant voltage charging current entropy ratio; (4) Constant voltage charging capacity and constant voltage charging capacity ratio; (5) Constant voltage charging energy and constant voltage charging energy ratio; S3, performing a first correlation analysis on each feature and health status sequence; The specific content of step S3 is as follows: The first correlation analysis is as follows: for each feature [F r,1 ,F r,2 ,…,F r,n ] are respectively related to the health state sequence [H1,H2,…,H n ] to perform grey relational analysis and calculate the grey relational coefficient [ξ r (1),ξ r (2),…,ξ r (n)], and take the average value of the grey correlation coefficient R r As the grey relational degree sequence [R1, R2, …, R m ], where the grey correlation coefficient R r ∈[0,1] is each feature F r Grey correlation degree with the health status of lithium battery, y(n)=[H1,H2,…,H n ], x r (n)=[F r,1 ,F r,2 ,…,F r,n ], ρ∈(0,1) is the resolution coefficient, grey relational degree m is the number of grey relational degree; S4. According to the maximum correlation interval method, determine whether the number M of input features is greater than one. If it is greater than one, execute steps S5 and S6; otherwise, directly jump to step S7; The specific content of step S4 is as follows: The maximum correlation interval method is: when R r ∈[0,0.6], add other new features, when R r ∈(0.6,0.7], the number of input features M = 4, when R r ∈(0.7,0.8], the number of input features M = 3, when R r ∈(0.8,0.9], the number of input features M = 2, when R r ∈(0.9,1], the number of input features M = 1; the number of input features M is determined in this way; S5, select the feature with the highest correlation in each class, and select the feature with the highest correlation among all features; S6, performing a second correlation analysis between the features with the highest correlation and the features with the highest correlation in other classes, and selecting input features based on the mutual correlation between the features; The second correlation analysis is: Perform Pearson correlation analysis on the feature with the highest grey correlation degree with the features with the highest grey correlation degree in the other four categories, and obtain four mutual correlations P1, P2, P3, and P4. Sort the mutual correlations from high to low, select the features of the other four categories corresponding to the first M-1 mutual correlations, and add the feature with the highest grey correlation degree to form M features as input features. Among them, Pearson correlation analysis a is the feature with the highest grey correlation degree; S7, using the input features and the health status sequence to iteratively find the penalty coefficient and kernel function width factor of the optimal support vector machine through the sparrow search algorithm, and establish an estimation model for the health status of the lithium battery; The specific content of step S7 is as follows: Take the first k, k=1, 2, ..., n-1 sample values ​​of the input feature and health status sequence and As the training set, take the k+1, k+2,…,n sample values and As a test set, the process of iteratively finding the penalty coefficient and kernel function width factor of the optimal support vector machine through the sparrow search algorithm is as follows: the root mean square error of the true value and the estimated value of the health state is used as the individual fitness function, the penalty coefficient and width factor are mapped to the two-dimensional position of each sparrow, the global optimal two-dimensional position is iteratively optimized, and the penalty coefficient and width factor of the optimal two-dimensional position are applied to establish the input feature and the estimation model of the health state of the lithium battery; S8. Use the established lithium battery health status estimation model to estimate the health status of the lithium battery.

2. The method for estimating the health status of a lithium battery based on full current and double correlation according to claim 1, characterized in that: The specific content of step S5 is as follows: The grey correlation between each feature and the health status of the lithium battery [R1, R2, …, R m ] After arranging in descending order, select the feature with the highest grey correlation degree in each class, and select the feature with the highest grey correlation degree among all features.

3. The method for estimating the health status of a lithium battery based on full current and double correlation according to claim 2, characterized in that: The detailed steps of step S7 of optimizing the penalty coefficient and width factor of the support vector machine by the sparrow search algorithm are as follows; (a) Read data, perform data preprocessing, and build a health status estimation model; (b) Initialize the sparrow population parameters and determine the value range of the penalty coefficient and width factor; (c) Divide the predator detectors and followers and randomly select vigilant ones to update their location information; (d) determining the fitness function of the sparrow search algorithm, and finding the optimal function value according to the principle of the sparrow search algorithm, that is, determining the position of the best sparrow individual; (e) Obtaining the optimal values ​​of penalty coefficient and width factor according to the best individual sparrow position; (f) Repeat steps (c) to (e) to determine whether the termination condition, i.e., the maximum number of iterations, is met; (g) Assign the optimal penalty coefficient and width factor to the support vector machine.

Images