CN115219918A - Lithium ion battery life prediction method based on capacity decline combined model - Google Patents

Lithium ion battery life prediction method based on capacity decline combined model Download PDF

Info

Publication number
CN115219918A
CN115219918A CN202210835590.4A CN202210835590A CN115219918A CN 115219918 A CN115219918 A CN 115219918A CN 202210835590 A CN202210835590 A CN 202210835590A CN 115219918 A CN115219918 A CN 115219918A
Authority
CN
China
Prior art keywords
model
capacity
battery
lithium ion
ion battery
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.)
Pending
Application number
CN202210835590.4A
Other languages
Chinese (zh)
Inventor
熊永莲
魏颖
尚瑾
荣文毅
朱玉成
肖杰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Yancheng Institute of Technology
Original Assignee
Yancheng Institute of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Yancheng Institute of Technology filed Critical Yancheng Institute of Technology
Priority to CN202210835590.4A priority Critical patent/CN115219918A/en
Publication of CN115219918A publication Critical patent/CN115219918A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/382—Arrangements for monitoring battery or accumulator variables, e.g. SoC
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/392—Determining battery ageing or deterioration, e.g. state of health
    • 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 discloses a lithium ion battery service life prediction method based on a capacity decline combined model, and belongs to the field of lithium ion battery service life detection. The method comprises the following steps: performing regression characteristic analysis on the basic empirical model according to battery capacity decline experimental data, establishing a capacity decline combined model for representing the relation between the battery capacity and the charge-discharge cycle times, and verifying the fitting degree of the model by using multiple groups of experimental data; updating of the model parameters is achieved based on an improved particle filter algorithm, the predicted value of the battery capacity is calculated, and finally the remaining service life of the lithium ion battery is calculated. The lithium ion battery life prediction method has the advantages that the established combined model has stronger fitting capability on the capacity decline of the battery, a high-precision life prediction result can be obtained, the rapid estimation of the remaining service life of the lithium ion battery can be realized, and the development of the battery echelon utilization industry is facilitated.

Description

Lithium ion battery life prediction method based on capacity decline combined model
Technical Field
The invention relates to the field of lithium ion battery service life detection, in particular to a lithium ion battery service life prediction method based on a capacity decline combined model.
Background
Because the lithium ion battery has the advantages of high energy density, high power density and the like, along with the progress of battery technology, the application field of the lithium ion battery gradually extends from consumer batteries on electronic products to power batteries on hybrid electric vehicles and pure electric vehicles and large energy storage batteries in power grid power stations. The aging degree of the lithium ion battery greatly affects the working performance of power supply facilities of the lithium ion battery, and if the aging degree of the battery cannot be known in time, the electric automobile and the energy storage facilities cannot run normally. The method is limited by the prior art, and the aging degree of the lithium ion battery in use is difficult to directly measure, so that the establishment of an accurate battery health and residual service life prediction mechanism is particularly important for battery management and echelon utilization. At present, methods related to predicting the remaining life of a lithium ion battery mainly include methods based on models, data driving and fusion of the two.
The lithium ion battery model is divided into a degradation mechanism model and an empirical degradation model. Although the degradation mechanism model can accurately describe the internal reactions of different batteries in the degradation process, the degradation mechanism model only describes the operation rule of the batteries in a short time and does not consider time-varying factors, so that the model needs to be corrected when the degradation mechanism model is used for life prediction at present. Compared with a degradation mechanism model, the empirical degradation model ignores the complicated electrochemical performance in the battery, and has certain advantages in model complexity and calculation complexity, but the model lacks enough theoretical support, and the working condition used for modeling is greatly different from that used in practical application, so that the generalization capability of the obtained model is poor.
Since the degradation of the lithium ion battery is a nonlinear, time-varying dynamic electrochemical process, although the mechanism analysis is clear in terms of physical meaning and concept, the precise modeling of the lithium ion battery still needs to involve a large number of parameters and complex calculations, so that the data-driven-based method is gradually a research hotspot in recent years. The method does not consider the electrochemical reaction and the failure mechanism inside the lithium ion battery, depends on a large amount of data, is influenced by initial parameters, and overcomes the problems of poor dynamic precision and poor general adaptability, wherein the data comprises the influence of the environment on the battery.
Disclosure of Invention
The invention aims to overcome the technical problem of low battery life prediction precision caused by low fitting degree of an empirical degradation model of a lithium ion battery in the prior art, and provides a lithium ion battery life prediction method based on a capacity decline combined model.
In order to achieve the purpose, the invention is realized by the following technical scheme: a service life prediction method of a vehicle power battery based on a capacity fading model comprises the following steps:
performing regression characteristic analysis on a basic empirical model according to capacity fading experimental data of the lithium ion battery, and establishing an improved polynomial model;
step two, fusing and recombining the basic experience model and the improved polynomial model, and establishing a combined model for representing the relationship between the battery capacity and the charge-discharge cycle number:
C＝α 1 *k 3 +α 2 *k 2 +α 3 *k+b 3 *exp(b 4 *k)
wherein C is the battery capacity, k is the number of charge and discharge cycles of the battery under standard conditions, and alpha 1 ，α 2 ，α 3 ，b 3 ，b 4 Parameters in the model;
updating the parameters of the capacity fading combination model based on the improved particle filter algorithm;
and step four, calculating to obtain the capacity predicted value and the residual service life of the lithium ion battery.
Preferably, the polynomial model in the step one is to perform parameter fitting of a capacity change curve before the battery failure threshold value on the quadratic polynomial model and the cubic polynomial model respectively according to the battery capacity decline experimental data. The expressions of the quadratic polynomial model and the cubic polynomial model are respectively as follows:
C＝α 1 *k 2 +α 2 *k+α 3
C＝α 1 *k 3 +α 2 *k 2 +α 3 *k+α 4
wherein C is the battery capacity, k is the number of charge and discharge cycles of the battery under standard conditions, and alpha 1 ，α 2 ，α 3 ，α 4 Are parameters in the model.
Preferably, the battery capacity degradation data in the step one are obtained by performing charge and discharge cycle experiments on the lithium ion battery under standard conditions.
Preferably, in the second step, a second-order exponential model is selected for the basic empirical model, and a nonlinear least square method is adopted for fitting; and selecting a cubic polynomial model for the polynomial model, and fitting by adopting a linear least square method. The expression of the second order exponential model is:
C＝b 1 *exp(b 2 *k)+b 3 *exp(b 4 *k)
in the formula: c is the battery capacity, k is the number of charge and discharge cycles of the battery under standard conditions, b 1 ，b 2 ，b 3 ，b 4 Are parameters in the model.
Preferably, the decision coefficient will be corrected
Figure BDA0003747922310000031
And Root Mean Square Error (RMSE) as an evaluation criterion for verifying the validity of the model.
Preferably, the specific operation of step two is: firstly, according to the capacity decline data of the lithium ion battery, the battery capacity decline process is equally divided into two parts, regression characteristic analysis is carried out on the capacity decline data of each part, each item of the model with better fitting effect is selected as a key item, and the key item is combined to establish the capacity decline combination model of the lithium ion battery.
Preferably, the second step further includes combining two variable terms in the second-order exponential model with two variable terms and two constant terms in the third-order polynomial model, respectively, to obtain the lithium ion battery capacity fading combined model.
Preferably, the battery capacity degradation combination model is established by taking an information criterion of the akage battery (AIC) as an auxiliary evaluation criterion.
Preferably, in the third step, the prior probability density of the system state at the next moment is obtained by predicting the state equation on the basis of the known state at a certain moment, and then the corresponding posterior probability density is calculated by correcting the known actual measurement value.
Preferably, the predicted value of the capacity of the battery obtained in the fourth step, the number of cycles that are passed when the capacity of the battery decays to the failure threshold value is the remaining service life of the battery.
Advantageous effects
The invention provides a lithium ion battery service life prediction method based on a capacity fading combined model. Compared with the prior art, the method has the following beneficial effects:
the invention discloses a lithium ion battery service life prediction method based on a capacity decline combined model, which aims to enable the battery capacity decline model to better reflect the battery capacity decline characteristics so as to improve the prediction precision of the battery service life. The combined model established by the invention not only can accurately reflect the characteristic of the battery capacity decline, but also can verify the reliability of the model by correcting the decision coefficient and the root mean square error. Therefore, the high-precision prediction of the residual life of the lithium ion battery can be realized, and the development of the battery echelon use industry is facilitated.
Meanwhile, in the process of predicting the service life of the lithium ion battery, only the correlation between the battery capacity and the charging and discharging cycle number needs to be described, and the complicated electrochemical reaction in the battery does not need to be analyzed, so that the error introduction probability is low in the process of determining the model parameters, the operation speed can be increased, and the rapid prediction of the residual service life of the lithium ion battery is realized.
Drawings
FIG. 1 is a flow chart for predicting the remaining service life of a lithium ion battery;
FIG. 2 is a graph of a battery capacity fade fit based on a quadratic polynomial model;
FIG. 3 is a battery capacity fade fit graph based on a cubic polynomial model;
FIG. 4 is a battery capacity fade fit graph based on a second order exponential model;
fig. 5 shows the RUL prediction results of the lithium ion battery at different prediction starting points.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that, in this document, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
The invention is described in further detail below with reference to the accompanying drawings:
example 1
In this embodiment, a cylindrical lithium ion battery with a model number of 18650 and a rated capacity of 2.0A · h is used as a research object, and as shown in fig. 1, a lithium ion battery life prediction method based on a capacity fading combination model is established, specifically as follows:
1. reading and preprocessing the capacity decline data of the lithium ion battery
The battery capacity decline data required by the method is obtained by performing charge-discharge circulation (1C charge-discharge, constant current charging cut-off voltage is 4.2V and constant current charging cut-off current is 0.04 A.h) on a 18650 type cylindrical lithium ion battery under a standard condition by adopting a Wuhan blue electricity CT3002K type battery test system, after the data are obtained, firstly preprocessing the data, eliminating abnormal data, and finally selecting 4 groups of battery capacity decline data (A3, A4, A5 and A6) for research.
2. Selecting an empirical model of the capacity decline of the lithium ion battery
At present, a second-order exponential model and a polynomial model are widely applied to battery capacity degradation prediction as basic empirical models, wherein common polynomial models include a second-order polynomial model and a third-order polynomial model, and in order to select a more suitable polynomial model to predict the service life of a lithium ion battery, the fitting result of battery capacity degradation data is analyzed and discussed through the two polynomial models.
Expressions of the quadratic polynomial model and the cubic polynomial model are respectively shown as formula (1) and formula (2):
C＝α 1 *k 2 +α 2 *k+α 3 (1)
C＝α 1 *k 3 +α 2 *k 2 +α 3 *k+α 4 (2)
in formulae (1) and (2): alpha is alpha 1 ，α 2 ，α 3 ，α 4 K is a parameter to be determined in the model, k is the charge-discharge cycle number of the battery under the standard condition, and C is the battery capacity.
Aiming at the characteristics of the polynomial model, a linear least square method is adopted to carry out parameter fitting on the polynomial model, and the fitting effect is shown in fig. 2 and 3. In fig. 2 and 3, the charge and discharge data of the battery under the standard conditions are fitted with a quadratic polynomial model and a cubic polynomial model, respectively, by a linear least square method. The abscissa is the number of charge and discharge cycles of the battery, the ordinate is the capacity of the battery, and the white discrete points in the graph represent measured data during the charge and discharge of the battery.
The results of fitting the battery capacity decay parameters before the failure threshold based on the quadratic polynomial model and the cubic polynomial model are compared, as shown in table 1:
TABLE 1 comparison of fitting results before failure thresholds for different polynomial models
Figure BDA0003747922310000061
Remarking: in the context of table 1, the following,
Figure BDA0003747922310000062
determining coefficients for the correction; RMSE is the root mean square error of the curve fit.
Figure BDA0003747922310000063
Is an effective measuring tool for fitting effect, and the value range is [0,1 ]]The closer to 1, the better the data fitting effect of the model is, and the higher the confidence degree is; the RMSE is an effective tool capable of representing the deviation of two groups of data, and can reflect the fitting precision, and the fitting effect is better when the RMSE approaches 0.
As can be seen from Table 1, the coefficients are determined by correction
Figure BDA0003747922310000064
The cubic polynomial model has stronger characterization capability in fitting the battery capacity decline data compared with the quadratic polynomial model reflected by the root mean square error RMSE, so the cubic polynomial model is selected for the combined model.
In this embodiment, when the fitting effect of the quadratic polynomial model and the cubic polynomial model is compared, the correction decision coefficient is selected because the independent variables of the two models are different in number
Figure BDA0003747922310000071
But not the correlation coefficient R 2 。
3. Carrying out regression characteristic analysis on the capacity regression empirical model, and establishing a combined model
The expression of the second order exponential model is shown in equation (3):
C＝b 1 *exp(b 2 *k)+b 3 *exp(b 4 *k) (3)
in formula (3): c is the battery capacity, k is the number of charge and discharge cycles of the battery under standard conditions, b 1 ，b 2 ，b 3 ，b 4 Are the parameters to be determined in the model.
And aiming at the characteristics of the second-order exponential model, fitting by adopting a nonlinear least square method. The fitting results are shown in fig. 4. In fig. 4, the charge and discharge data of the battery under the standard condition is fitted by using a second-order exponential model, and the method used is a nonlinear least square method. The abscissa is the number of charge and discharge cycles of the battery, the ordinate is the capacity of the battery, and the white discrete points in the graph represent measured data during the charge and discharge of the battery.
The fitting results of the battery capacity regression data before the failure threshold based on the cubic polynomial model and the second-order exponential model are compared, as shown in table 2:
TABLE 2 comparison of fitting results before different model failure thresholds
Figure BDA0003747922310000072
As can be seen from table 2, the fitting effect of the two models on the global capacity fading of the group a battery is not constant, and the whole model has a changing trend. In order to ensure that the battery model has a globally stable fitting effect on capacity decline and realize accurate prediction of the battery RUL, a combination model with high matching degree on global data is established on the basis of a cubic polynomial model and a second-order bi-exponential model.
The traditional method for establishing a combined model generally weights each model correspondingly so as to comprehensively consider the fitting effect of various models, but for a cubic polynomial model and a second-order exponential model, if a new combined model is established in a weighted manner, 8 model parameters to be identified are generated, and the probability of error introduction in the fitting process is high due to excessive model parameters, and the process is not easy to implement.
The method comprises the steps of firstly, equally dividing the whole process from the decline of the battery capacity to the end of the service life into two parts, namely a part I and a part II, carrying out regression characteristic analysis by adopting a cubic polynomial model and a second-order exponential model, and establishing a combined model with better fitting effect by finding out key terms in each model.
The fitting results of the third-order polynomial model and the second-order exponential model in the sections I and II are shown in Table 3:
TABLE 3 comparison of the fitting results of different models in each part
Figure BDA0003747922310000081
As can be seen from Table 3, the fitting effect of the cubic polynomial model and the second-order exponential model in section I is not much different, while in section II, the matching degree of the second-order bi-exponential model is better. Therefore, the second-order exponential model can better reflect the characteristic of the battery capacity decline, so that the fitting capacity of the battery model to partial capacity decline data is improved in a combined mode on the basis of the two models.
Through the analysis of capacity fading data, two variable terms in a second-order exponential model are respectively combined with two variable terms and constant terms in a third-order polynomial model in pairs to obtain 4 different capacity fading combination models, such as formulas (4) to (7):
C 1 ＝α 1 *k 3 +α 2 *k 2 +α 3 *k+b 3 *exp(b 4 *k) (4)
C 2 ＝b 1 *exp(b 2 *k)+α 4 (5)
C 3 ＝α 1 *k 3 +α 2 *k 2 +α 3 *k+b 1 *exp(b 2 *k) (6)
C 4 ＝b 3 *exp(b 4 *k)+α 4 (7)
in the above formula: c is the battery capacity, k is the number of charge and discharge cycles of the battery under standard conditions, alpha 1 ，α 2 ，α 3 ，α 4 ，b 1 ，b 2 ，b 3 ，b 4 Are parameters in the model.
Since the number of unknown parameters of the 4 kinds of combination models is different, the correction decision coefficient is selected here
Figure BDA0003747922310000091
As an evaluation criterion for the effectiveness of the model.
Meanwhile, the Akabane Information Criterion (AIC) is introduced as an auxiliary judgment standard and used for comparing the fitting degrees of different models, and the expression is as follows:
AIC＝-2ln(L)+2t (8)
in equation (8), ln (L) is the maximum value of the log-likelihood function, and t is the number of parameters in the model.
Determining coefficients using correction
Figure BDA0003747922310000092
The Chichi information criterion (AIC) was used to optimize the 4 combination patterns described above, and the results are shown in Table 4:
TABLE 4 comparison of four combination models
Figure BDA0003747922310000093
Figure BDA0003747922310000101
As can be seen from Table 4, C 1 The model having more approximation to 1
Figure BDA0003747922310000102
And smaller AIC values, indicating C compared to the other three models 1 The model has stronger fitting ability to the data and is less likely to have an overfitting phenomenon. To verify C 1 The reliability of the model was compared with the cubic polynomial model and the second-order exponential model, and the results are shown in table 5:
TABLE 5 comparison of fitting results for three models
Figure BDA0003747922310000103
As can be seen from Table 5, C 1 Model in the a-battery capacity fade data fit,
Figure BDA0003747922310000104
closer to 1 and the RMSE value is minimal. According to
Figure BDA0003747922310000105
The more approaching to 1 and the more approaching to 0 the RMSE value, the better the fitting effect of the model, and it can be seen that C is proposed 1 The model is more suitable for predicting the RUL of the lithium ion battery.
4. Updating model parameters based on improved particle filter algorithm
In order to ensure the accuracy of the battery capacity decline prediction, parameters of the model are continuously updated to ensure the robustness of the model. Particle Filters (PFs) are widely used to solve such non-linear problems due to their non-parametric and non-gaussian characteristics.
In view of the problem that the RUL prediction precision is low due to the fact that particle diversity loss phenomenon exists in the process of predicting the remaining service life (RUL) of the lithium ion battery by the conventional PF algorithm, the linear optimization resampling thought is introduced, and the battery RUL prediction method based on the improved particle filtering is established.
The method comprises the following specific steps:
step 1: and reading the lithium ion battery capacity degradation data for prediction.
Step 2: setting a prediction starting point T, for example, T =100 indicates that the battery capacity deterioration data before the 100 th cycle is regarded as history data, and the battery capacity value is predicted for each cycle from the 100 th cycle.
And 3, step 3: and selecting the parameters of the capacity fading model as system state variables, and establishing a state equation and an observation equation.
The state equation is:
X(k)＝[v 1 (k)，v 2 (k)，v 3 (k)，v 4 (k)，v 5 (k)] (9)
Figure BDA0003747922310000111
the observation equation is:
Q(k)＝v 1 *k 3 +v 2 *k 2 +v 3 *k+v 4 *exp(v 5 *k)+μ(k)，μ(k)～N(0，σ μ ) (11)
in the formula, v 1 (k+1)，…，v 5 (k + 1) is a battery capacity fading model parameter; k =1,2, \8230, n is the number of charge and discharge cycles of the battery; process noise
Figure BDA0003747922310000112
Is Gaussian white noise, the mean value is 0, and the variance is
Figure BDA0003747922310000113
Y (k) is the capacity of the battery after k cycles; observation of noise mu k Is a mean of 0 and a variance of σ μ White gaussian noise.
And 4, step 4: setting relevant parameters in RUL prediction process, including number of particles N, process noise
Figure BDA0003747922310000114
Of the observed noise mu, the covariance Q of k Covariance R, battery failure threshold Q thres 。
And 5, step 5: using a prediction before the starting point TThe battery capacity decline data and the data after the battery capacity decline data are subjected to state tracking, and the battery capacity decline model parameter v is continuously updated 1，k ，v 2，k ，v 3，k ，v 4，k And v 5，k Thereby obtaining the predicted value of the battery capacity at the predicted starting point T and each time after the starting point T
Figure BDA0003747922310000121
Calculating the weight of each particle:
Figure BDA0003747922310000122
in the formula:
Figure BDA0003747922310000123
is the ith particle at the kth moment;
Figure BDA0003747922310000124
the weight of the ith particle at the kth moment;
Figure BDA0003747922310000125
a likelihood probability density function of a capacity prediction value corresponding to the ith particle at the kth time.
Normalization weight:
Figure BDA0003747922310000126
updating the particle weight:
Figure BDA0003747922310000127
x n ＝x α +HS(x α -x s ) (15)
in the formula, x n Are new particles; x is a radical of a fluorine atom α The particles with larger weight; x is the number of s For the comparison of weightSmall particles; s is (x) α -x s ) A suitable step size of; h is the corresponding step coefficient. If the weight of the particles satisfies the formula (14), the particles are classified as a discarded group, if the weight of the particles does not satisfy the formula (14), the particles are classified as a copied group, then the formula (15) is utilized to carry out linear optimization combination to generate new particles, a new particle set is obtained, and the parameters of the battery capacity decline model are updated
Figure BDA0003747922310000128
In the formula (I), the compound is shown in the specification,
Figure BDA0003747922310000131
δ (·) is a dirac function.
5. Calculating the predicted value of the capacity and the remaining service life of the lithium ion battery
Predicting the battery capacity:
Figure BDA0003747922310000132
and 6, a step of: judging the predicted value of the battery capacity at the kth time
Figure BDA0003747922310000133
Whether or not threshold Q is reached thres If the threshold is reached, the battery RUL is calculated and the algorithm ends. Otherwise, repeating the 5 th step and the 6 th step until the step
Figure BDA0003747922310000134
Reach the failure threshold Q thres . The number of charge and discharge cycles at the time of battery failure can be obtained from equation (18), and the probability density distribution of battery RUL can be obtained from equation (19).
Figure BDA0003747922310000135
Figure BDA0003747922310000136
In the formula:
Figure BDA0003747922310000137
the RUL predicted value is the RUL predicted value when the battery circulates for k times; l is k The real RUL value at k cycles of the battery.
On the basis of the technical scheme, the 250 th, 400 th and 550 th cycles of the A4, A5 and A6 batteries are respectively selected as prediction starting points; initial values of model parameters are set to v 1 ＝-1.31*10 -8 ，v 2 ＝-6.94*10 -6 ，v 3 ＝0.00121，v 4 ＝1.132，v 5 ＝-2.71*10 -4 (ii) a The number of sampling particles N =800; the failure threshold is 80 percent of rated capacity, namely 1.6 A.h; process noise
Figure BDA0003747922310000138
Covariance Q =0.0001; observation of noise mu k Covariance R =0.0001. The predicted results for the A4, A5, A6 cells are obtained, as shown in table 10.
TABLE 6 prediction results and errors for battery RUL at different prediction starting points
Figure BDA0003747922310000141
As can be seen from table 10, the total of 9 sets of prediction results are obtained from 3 batteries at 3 different prediction starting points, and the results with the real RUL values falling in the 95% confidence interval are 7 sets, accounting for 77.8% of the total prediction results, so that the model can obtain excellent prediction performance in terms of uncertainty expression of the prediction results. In addition, averaging the relative errors in table 1 resulted in an average relative error of 2.3%, indicating that the model has a high prediction accuracy.
The lithium ion battery RUL prediction method based on the capacity decline combined model provided by the invention can utilize the state tracking capability of the improved PF algorithm, flexibly set the parameters of the model and improve the precision of the prediction result; meanwhile, the probability density distribution of the RUL prediction result of the lithium ion battery can be given, and the uncertainty expression capability is realized.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (10)

1. A lithium ion battery life prediction method based on a capacity decline combined model is characterized by comprising the following steps:
performing regression characteristic analysis on a basic empirical model according to capacity fading experimental data of the lithium ion battery, and selecting a polynomial model for combination;
and step two, fusing and recombining the basic experience model and the polynomial model selected in the step one, and establishing a capacity fading combination model for representing the relation between the battery capacity and the charge-discharge cycle times:
C＝α 1 *k 3 +α 2 *k 2 +α 3 *k+b 3 *exp(b 4 *k)
wherein C is the battery capacity, k is the number of charge and discharge cycles of the battery under standard conditions, and alpha 1 ，α 2 ，α 3 ，b 3 ，b 4 Parameters in the model;
updating the parameters of the capacity fading combination model based on the improved particle filter algorithm;
and step four, calculating to obtain the capacity predicted value and the remaining service life of the lithium ion battery.
2. The lithium ion battery life prediction method based on the capacity fading combination model as claimed in claim 1, wherein the polynomial model in the step one is to perform parameter fitting of the capacity change curve before the battery failure threshold value to the quadratic polynomial model and the cubic polynomial model respectively according to the battery capacity fading experimental data. The expressions of the quadratic polynomial model and the cubic polynomial model are respectively as follows:
C＝α 1 *k 2 +α 2 *k+α 3
C＝α 1 *k 3 +α 2 *k 2 +α 3 *k+α 4
wherein C is the battery capacity, k is the number of charge and discharge cycles of the battery under standard conditions, and alpha 1 ，α 2 ，α 3 ，α 4 Are parameters in the model.
3. The method according to claim 1 or 2, wherein the battery capacity degradation data obtained in the first step is obtained by performing charge and discharge cycle experiments on the lithium ion battery under standard conditions.
4. The lithium ion battery life prediction method based on the capacity fading combination model as claimed in claim 1, wherein the decision coefficient is corrected
Figure FDA0003747922300000021
And Root Mean Square Error (RMSE) as an evaluation criterion for verifying the validity of the model.
5. The lithium ion battery life prediction method based on the capacity fading combination model according to claim 1, characterized in that in the second step, a second-order exponential model is selected for the basic empirical model, and fitting is performed by adopting a nonlinear least square method; and selecting a cubic polynomial model for the polynomial model, and fitting by adopting a linear least square method. The expression of the second order exponential model is:
C＝b 1 *exp(b 2 *k)+b 3 *exp(b 4 *k)
in the formula: c is the battery capacity, k is the number of charge and discharge cycles of the battery under standard conditions, b 1 ，b 2 ，b 3 ，b 4 Are parameters in the model.
6. The lithium ion battery life prediction method based on the capacity fading combination model as claimed in claim 1, wherein the specific operation of the second step is: firstly, according to the capacity decline data of the lithium ion battery, the battery capacity decline process is equally divided into two parts, regression characteristic analysis is carried out on the capacity decline data of each part, each item of the model with better fitting effect is selected as a key item, and the key item is combined to establish the capacity decline combination model of the lithium ion battery.
7. The lithium ion battery life prediction method based on the capacity fading combination model as claimed in claim 1, wherein the second step further comprises pairwise combining two variable terms in the second-order exponential model with the variable terms and constant terms in the third-order polynomial model, respectively, to obtain the lithium ion battery capacity fading combination model.
8. The lithium ion battery life prediction method based on the capacity fading combination model as claimed in claim 1, wherein the battery capacity fading combination model is established by taking an information criterion of red battery (AIC) as an auxiliary evaluation criterion.
9. The lithium ion battery life prediction method based on the capacity fading combination model according to claim 1, characterized in that in step three, the prior probability density of the system state at the next moment is obtained by predicting through a state equation on the basis of the known state at a certain moment, and then the corresponding posterior probability density is calculated by correcting through the known actual measurement value.
10. The lithium ion battery life prediction method based on the capacity fading combination model as claimed in claim 1, wherein the number of cycles that the battery capacity fades to the failure threshold is the remaining service life of the battery, according to the capacity prediction value of the battery obtained in step four.
CN202210835590.4A 2022-07-15 2022-07-15 Lithium ion battery life prediction method based on capacity decline combined model Pending CN115219918A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210835590.4A CN115219918A (en) 2022-07-15 2022-07-15 Lithium ion battery life prediction method based on capacity decline combined model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210835590.4A CN115219918A (en) 2022-07-15 2022-07-15 Lithium ion battery life prediction method based on capacity decline combined model

Publications (1)

Publication Number Publication Date
CN115219918A true CN115219918A (en) 2022-10-21

Family

ID=83612778

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210835590.4A Pending CN115219918A (en) 2022-07-15 2022-07-15 Lithium ion battery life prediction method based on capacity decline combined model

Country Status (1)

Country Link
CN (1) CN115219918A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115656862A (en) * 2022-12-09 2023-01-31 深圳海润新能源科技有限公司 Life prediction method, battery management system, electric device, and storage medium
CN117233630A (en) * 2023-11-16 2023-12-15 深圳屹艮科技有限公司 Method and device for predicting service life of lithium ion battery and computer equipment

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115656862A (en) * 2022-12-09 2023-01-31 深圳海润新能源科技有限公司 Life prediction method, battery management system, electric device, and storage medium
CN115656862B (en) * 2022-12-09 2023-03-10 深圳海润新能源科技有限公司 Life prediction method, battery management system, electric device, and storage medium
CN117233630A (en) * 2023-11-16 2023-12-15 深圳屹艮科技有限公司 Method and device for predicting service life of lithium ion battery and computer equipment
CN117233630B (en) * 2023-11-16 2024-03-15 深圳屹艮科技有限公司 Method and device for predicting service life of lithium ion battery and computer equipment

Similar Documents

Publication Publication Date Title
CN110208704B (en) Lithium battery modeling method and system based on voltage hysteresis effect
CN111398833B (en) Battery health state assessment method
Xiong et al. A novel practical state of charge estimation method: an adaptive improved ampere‐hour method based on composite correction factor
Vasebi et al. A novel combined battery model for state-of-charge estimation in lead-acid batteries based on extended Kalman filter for hybrid electric vehicle applications
CN108037460B (en) Real-time capacity evaluation method for lithium ion batteries produced in batches
CN110596593A (en) Lithium ion battery SOC estimation method based on intelligent adaptive extended Kalman filtering
CN107741568B (en) Lithium battery SOC estimation method based on state transition optimization RBF neural network
CN115219918A (en) Lithium ion battery life prediction method based on capacity decline combined model
CN105301509A (en) Combined estimation method for lithium ion battery state of charge, state of health and state of function
CN111856282B (en) Vehicle-mounted lithium battery state estimation method based on improved genetic unscented Kalman filtering
CN107783057B (en) Power battery SOC estimation method and estimation system
CN108535661B (en) Power battery health state online estimation method based on model error spectrum
CN111913109B (en) Method and device for predicting peak power of battery
Takyi-Aninakwa et al. An optimized long short-term memory-weighted fading extended Kalman filtering model with wide temperature adaptation for the state of charge estimation of lithium-ion batteries
CN111220920A (en) Decommissioned lithium ion battery charge state calculation method based on H infinity unscented Kalman filtering algorithm
CN112989690A (en) Multi-time scale state of charge estimation method for lithium battery of hybrid electric vehicle
CN111044924B (en) Method and system for determining residual capacity of all-condition battery
CN113785209B (en) Method for detecting abnormal battery cell
Chen et al. State of health estimation for lithium-ion battery based on long short term memory networks
CN113484771A (en) Method for estimating wide-temperature full-life SOC and capacity of lithium ion battery
Wang et al. Lithium-ion battery security guaranteeing method study based on the state of charge estimation
Xin et al. A novel state of charge estimation method for ternary lithium batteries based on system function and extended kalman filter
CN113296010A (en) Battery health state online evaluation method based on differential voltage analysis
CN116736171A (en) Lithium ion battery health state estimation method based on data driving
CN116449219A (en) Lithium battery model parameter identification method for improving VFFRLS

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