Lithium battery health state estimation method based on charging stage health characteristics
Technical Field
The invention relates to a lithium battery health state estimation method, in particular to a lithium battery health state estimation method based on health characteristics in a charging stage.
Background
With the increasing environmental and energy crisis in the world, lithium batteries have become the most potential energy storage devices at present due to their advantages of high energy density, long service life, low self-discharge rate, no memory effect, etc. The service environment of the lithium battery is complex, severe current change and environment change are easy to occur, the performance of the battery is reduced, the service life of the battery is shortened, and even safety accidents such as fire, explosion and the like are caused, so that the safe and stable operation of the whole system is endangered.
State of health (SOH) estimation of lithium batteries is a prerequisite and key to effective management of batteries. SOH is an estimate of the change in the ability of a battery in use to store and release electrical energy relative to a fresh battery, essentially reflecting the aging and degradation of the battery. The SOH parameter is often defined by the ratio of the change in capacity of the battery, as given by equation (1)Show, CnowIndicates the present capacity of the battery, CratedWhich represents the rated capacity of the battery, i.e. the capacity of a whole new battery. However, the lithium battery has various types of loads, the discharge working conditions are complex and changeable, the real-time monitoring of the discharge capacity is required to provide high requirements for the sensor, and the difficulty is high.
SOH=Cnow/Crated (1)
Currently, common methods for estimating the health status of lithium batteries can be classified into model-based methods and data-driven methods. The data driving method comprises algorithms such as a neural network, a support vector machine, a Bayesian network and a particle filter. The method can establish a relation model between the external health characteristics of the battery and the internal aging process without researching the complex internal aging mechanism of the battery and identifying model parameters, and realizes capacity monitoring. The health characteristics are generally extracted from charge/discharge current and voltage curves in charge/discharge cycle data of the lithium battery, and are commonly represented by open-circuit voltage, discharge voltage drop and the like. The selection of the health characteristics has a great influence on the accuracy of the estimation result and the application range of the method.
In order to estimate the health state of the lithium battery under complex working conditions, the method for only researching the health characteristics in the discharge stage and selecting the method depending on experience is far from enough, and further research is needed.
In recent years, in the research on the lithium battery health state estimation method, external health characteristics are often extracted from voltage and current curves in a discharging stage, and the problems of high measurement difficulty and narrow application range are often caused. If the open-circuit voltage can be obtained only by long-time standing of the battery, the discharge voltage drop is limited by the constant-current discharge working condition and is not suitable for the conditions of pulse discharge or discontinuous discharge and the like. In addition, the environmental temperature has a great influence on the physical and chemical reactions in the battery, which can seriously affect the degradation trend of the battery, but the current health feature extraction work has less research on temperature-related features.
Therefore, it is one of the problems to be urgently solved by the technical staff in the field to provide a method for estimating the health state of a lithium battery based on the health characteristics of the charging stage, which has the advantages of reasonable design, simple application, accurate and reliable estimation and judgment, low cost and high working efficiency.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, and aims to solve the problems that the existing health characteristics are high in measurement difficulty and narrow in application range and do not relate to the important factor of temperature. For the complicated changeable condition of the operating mode that discharges, the charging process of lithium cell is comparatively fixed, and generally can divide into Constant Current (CC) and Constant Voltage (CV) two stages of charging. The health characteristics in the charging stage are extracted, and the influence of the discharge working condition change can be effectively avoided.
On the other hand, the health characteristics are generally given directly according to experience, or only a single health characteristic is analyzed with capacity decline, and a relatively definite method for extracting the health characteristics is lacked. Aiming at the problem, the invention selects health characteristics which are easy to measure from charging current, voltage and temperature curves of the lithium battery, provides a set of complete and easy-to-follow health characteristic extraction method based on grey correlation analysis and local linear embedding, and realizes SOH estimation on the basis; the lithium battery health state estimation method based on the charging stage health characteristics is reasonable in design, simple and convenient to apply, accurate and reliable in estimation and judgment, low in cost and high in working efficiency.
In order to achieve the purpose, the invention adopts the technical scheme that: a lithium battery health state estimation method based on charging stage health characteristics is characterized by comprising the following implementation steps: the method comprises two stages of modeling and health state estimation:
the concrete steps of the modeling stage comprise:
step 1.1: on the basis of lithium battery cycle charging data, 14 charging health characteristics of the following 4 types are acquired from data curves of different charging stages, namely a constant-current charging stage and a constant-voltage charging stage:
(1) the time used in the different stages and their ratios: l is1Time of constant voltage charging phase, L2Charging for constant currentTime of stage, ratio L1/L2;
(2) The area of the region formed by the curves of different stages of current and temperature and the x axis is as follows: a. the1The area of the region formed by the current curve and the x axis in the constant current charging stage, A2The area of a region formed by a current curve and an x axis in a constant voltage charging stage is shown, and A is the area of the region formed by the current curve and the x axis in the whole charging stage; t is1The area of the region formed by the temperature curve of the constant current charging stage and the x axis, T2The area of a region formed by a temperature curve and an x axis in a constant voltage charging stage is shown, and T is the area of the region formed by the temperature curve and the x axis in the whole charging stage;
(3) maximum slope K of the voltage curve1Maximum slope K of the sum current curve2;
(4) The ratio of the corresponding temperature curve area to the current curve area in category (2): t is1/A1,T2/A2And T/A;
step 1.2: calculating gray correlation degrees of 14 health characteristics and SOH, extracting n charging health characteristics with high correlation degrees, and constructing a health characteristic matrix HF (HF) ([ HF) of m multiplied by n by taking a value of n according to the actual situation of the battery data1,HF2,…,HFi,…,HFn]TWherein HF isiA health signature data series in m dimensions, where m is the number of samples in the battery data set;
step 1.3: the health characteristic matrix constructed in the step 1.2 has a plurality of dimensions due to the influence of n, so that the data size is large, and the calculation load is increased; therefore, a Principal Component Analysis (PCA) of one of the dimensionality reduction algorithms is used for reducing the characteristic matrix from the m multiplied by n matrix to a healthy characteristic sequence HI of 1 multiplied by m under the condition of keeping most of original information, so that the calculation amount is reduced;
step 1.4: adopting a Relevance Vector Machine (RVM) as a regression model algorithm, inputting the health characteristic sequence HI obtained in the step 1.3 as an RVM model, outputting SOH as an RVM model, and training to obtain a regression model;
the specific steps of the SOH estimation stage comprise:
step 2.1: collecting the lithium battery charging health characteristics extracted in the step 1.2, and constructing a health characteristic matrix HF ═ HF1,HF2,…,HFi,…,HFn]T(ii) a Wherein HF isiA health signature data series in m dimensions, where m is the number of samples in the battery data set;
step 2.2: reducing the dimension of the feature matrix HF in the step 2.1 by using PCA to obtain a health feature sequence HI of 1 Xm;
step 2.3: and inputting the health characteristic HI in the step 2.2 into the RVM model established in the step 1.4 to obtain an SOH estimation result.
The invention has the beneficial effects that:
(1) the health characteristics of the charging stage are adopted, so that the method is applicable to batteries under all discharging working conditions;
(2) the proposed health characteristics are easy to collect, and no additional collection experiment is needed;
(3) the temperature, an important environmental factor influencing the aging of the battery, is introduced;
(4) provides a set of complete and clear health characteristic extraction method which is easy to follow.
In a word, the method is reasonable in design, simple and convenient to apply and accurate and reliable in estimation and judgment; the method can greatly improve the working efficiency, has very obvious application effect and very wide application prospect.
Drawings
FIG. 1 is a flow diagram of the modeling phase of the present invention;
FIG. 2 is a block diagram of a process for estimating the state of health of a lithium battery according to the present invention;
FIG. 3 is a diagram illustrating a state of health estimation result curve of a lithium battery according to the present invention based on the health characteristics of the charging stage.
Detailed Description
The following detailed description of specific embodiments, structures and features provided in accordance with the present invention can be read in conjunction with the accompanying drawings and the preferred embodiments as follows:
as shown in fig. 1-3, a method for estimating the state of health of a lithium battery based on the health characteristics of the charging phase; the lithium battery health state estimation objects in the following embodiments of the method are #30 to #32 lithium ion batteries in an open source database of the national aeronautics and astronautics administration (NASA); in the above, the #30 and #31 batteries were used as training set to establish a model, and #32 was used as a test set to perform SOH estimation.
The method comprises the following concrete implementation steps of firstly, modeling and health state estimation:
the concrete steps of the modeling stage comprise:
referring to fig. 1, step 1.1: on the basis of lithium battery cycle charging data, the following 4 types of 14 charging health characteristics are collected from data curves of different charging stages (including two stages of constant-current charging and constant-voltage charging):
(1) the time used in the different stages and their ratios: l is1Time of constant voltage charging phase, L2Time of constant current charging phase, ratio L1/L2;
(2) The area of the region formed by the curves of different stages of current and temperature and the x axis is as follows: a. the1Is the current curve of the constant current charging phase, A2The current curve of the constant voltage charging stage is shown, and A is the current curve of the whole charging stage; t is1Is the temperature curve of the constant current charging phase, T2The temperature curve of the constant-voltage charging stage is shown, and T is the temperature curve of the whole charging stage;
(3) maximum slope K of the voltage curve1Maximum slope K of the sum current curve2;
(4) The ratio of the corresponding temperature curve area to the current curve area in category (2): t is1/A1,T2/A2And T/A.
Step 1.2: the basic idea of grey correlation analysis is to study the similarity of sequence curves of different factors, i.e. the more similar the geometric shapes of the curves are, the greater the correlation between the factors is; conversely, the smaller the degree of association between the factors; when the two curves are identical, the grey correlation degree is equal to 1. Reference sequence X0={x0(k) 1,2, K, m, and the comparison sequence Xi={xi(k) The reference sequence and the comparison sequence can be calculated according to the formula (2) in the following mannerThe correlation coefficient between columns, where ρ is the resolution coefficient, here taken to be 0.5.
Calculating the average value of the correlation coefficients to obtain the gray correlation degree of the reference sequence and the comparison sequence:
for example, the gray correlation degree between the 14 health characteristics of batteries #30 and #31 and the SOH was calculated using the SOH value as a reference sequence and the health characteristic value as a comparison sequence, and the results are shown in table 1, and 8 characteristics a having high correlation values were selected1,A,L1,L1/L2,T,T1,T2/A2,K2Constructing a health feature matrix HFx=[A1,A,L1,L1/L2,T,T1,T2/A2,K2]T。
The preferred range of the n charging health characteristics extracted in the step 1.2 with higher correlation degree is 4-10, and the selection can be performed according to the correlation and comprehensiveness of the 4 types of 14 charging health characteristics in the lithium battery health state estimation, the simplicity and the simplicity of the estimation numerical operation and the accuracy of the estimation numerical operation.
TABLE 1 Grey correlation of health characteristics
Step 1.3: a dimension reduction algorithm is used: principal Component Analysis (PCA) of feature matrix HFxReducing vitamins to obtain health characteristic sequence HIx. Firstly, the health feature matrix HFxNormalized to X*And calculating a covariance matrix S:
x can be calculated by the formula (5)*Characteristic vector u ofiAnd X*Characteristic value λ ofiI is 1, K, x, and the matrix Z after dimensionality reduction is calculated by equation (6):
Sui=λiui (5)
Z=X*×S (6)
in order to determine the number of principal components selected in Z, the contribution rate of each principal component after dimensionality reduction can be calculated according to formula (7):
the results of calculating the contribution ratios of the respective principal components are shown in table 2; it can be seen that the contribution rate of the principal component 1 to the original data reaches over 99%, and most information of the original matrix is reserved; principal component 1 can be used as a dimension reduction result, namely, the health feature sequence HIx。
TABLE 2 principal Components contribution ratio
|
Principal component 1
|
Principal component 2
|
Principal component 3
|
#30
|
99.94873%
|
0.044322%
|
0.005443%
|
#31
|
99.88438%
|
0.109917%
|
0.005049% |
Step 1.4: adopting a Relevance Vector Machine (RVM) as a regression model algorithm, and carrying out the HI on the health characteristic sequence obtained in the step 1.3xAnd (4) as RVM model input, SOH as RVM model output, and training to obtain a regression model.
The specific steps of the SOH estimation stage comprise:
see fig. 2, step 2.1: collecting for example 8 charging health features A extracted in said step 1.21,A,L1,L1/L2,T,T1,T2/A2,K2Constructing a health feature matrix HFt=[A1,A,L1,L1/L2,T,T1,T2/A2,K2]T。
Step 2.2: using PCA to obtain the feature matrix HF obtained in step 2.1tReducing dimension, using principal component 1 obtained by PCA dimension reduction as health characteristic sequence HIt。
The dimensionality reduction algorithm can also adopt a local linear embedding or multidimensional scaling dimensionality reduction algorithm.
Step 2.3: using the health signature HI of step 2.2tThe RVM regression model established in step 1.4 is input to obtain the SOH estimation result, as shown in FIG. 3.
The regression model algorithm established in step 1.4 or step 2.3 may also adopt a support vector machine or a neural network algorithm.
The method of the invention has the main characteristics and principles that:
the method firstly determines the health characteristics of 4 types of charging stages, and selects the characteristics with high correlation degree to form a matrix on the basis of carrying out grey correlation degree analysis on the health characteristics and SOH. And (4) performing dimensionality reduction on the health characteristic matrix by using a PCA algorithm, and verifying dimensionality reduction effects according to the principal component contribution rate to obtain a final health characteristic sequence. And establishing an RVM regression model by taking the extracted health characteristic sequence as input to realize the estimation of the SOH of the lithium battery. Through the processes of analysis, calculation and the like, the implementation method effectively avoids the influence of the change of the battery discharge process on the selection of the health characteristics, provides a health characteristic selection and optimization process which is easy to follow, and provides a new idea with very obvious application effect for the selection of the health characteristics of the lithium battery SOH estimation.
The above detailed description of the method for estimating the state of health of a lithium battery based on the health characteristics of the charging phase with reference to the embodiments is illustrative and not restrictive; thus, variations and modifications can be made without departing from the general inventive concept and, therefore, the scope of the present invention is to be determined.