CN115586452A - Lithium ion battery health state estimation method based on novel health characteristics - Google Patents
Lithium ion battery health state estimation method based on novel health characteristics Download PDFInfo
- Publication number
- CN115586452A CN115586452A CN202211374561.9A CN202211374561A CN115586452A CN 115586452 A CN115586452 A CN 115586452A CN 202211374561 A CN202211374561 A CN 202211374561A CN 115586452 A CN115586452 A CN 115586452A
- Authority
- CN
- China
- Prior art keywords
- health
- voltage
- novel
- time interval
- optimized
- 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
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]
- G01R31/392—Determining battery ageing or deterioration, e.g. state of health
-
- 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]
- G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Secondary Cells (AREA)
- Tests Of Electric Status Of Batteries (AREA)
- Charge And Discharge Circuits For Batteries Or The Like (AREA)
Abstract
The invention relates to a lithium ion battery health state estimation method based on novel health characteristics, which comprises the steps of firstly, acquiring a charging voltage curve under the condition of constant-current constant-voltage charging, and taking the voltage difference with equal initial voltage and equal time intervals as the novel health characteristics; then, the initial voltage and the time interval are two parameters which influence the novel health characteristics, so that the initial voltage and the time interval are optimized, and an optimized initial voltage and a plurality of optimized time intervals are obtained from the same charging voltage curve; combining the optimized starting voltage with each optimized time interval to obtain a plurality of novel health characteristics extracted from the same charging voltage curve; and finally, constructing a lithium battery health state estimation model based on a multiple linear regression model, training the model by using the extracted novel health characteristics and the corresponding health state values, and using the trained model for lithium battery health state estimation. The method is small in calculation amount, greatly improves the estimation speed on the premise of ensuring the estimation precision, and can be applied to the condition that the calculation power of the current battery management system is insufficient.
Description
Technical Field
The invention belongs to the technical field of health state estimation of a battery management system, and particularly relates to a lithium ion battery health state estimation method based on novel health characteristics.
Background
With the development of lithium ion battery technology, the electric automobile industry is rapidly developed. During the long-term cycle use of the lithium ion battery, the performance and the capacity are degraded, and fire or explosion can be caused by improper operation. In order to ensure reliable and safe operation of a lithium Battery, a Battery Management System (BMS) needs to monitor the State of Health (SOH) of the Battery, which reflects the Battery's ability to currently store and supply energy/power relative to the beginning of its life, and the available capacity of the Battery is typically used to characterize the State of Health. When the SOH decays below a certain level, it is determined that the battery has reached its maximum service life and needs to be replaced. Therefore, evaluating the state of health of the battery is of great importance for safe operation of the system.
On one hand, in the existing lithium ion battery SOH estimation research, the development and application of an intelligent algorithm greatly improve the estimation precision. However, these algorithms are high in complexity, and especially, the algorithms based on deep learning require more computing resources, and since the current battery management system mainly includes a microcontroller and has a limited memory space for computing, these intelligent algorithms are difficult to deploy in the battery management system. On the other hand, most algorithms directly take complete charging data (SOH from 0% to 100% of charging data) as input or perform feature extraction, however, in the actual use process of the battery, the initial charging state and the ending charging state of the charging are difficult to fix due to the influence of working conditions, and the obtained charging data is often a fragment, so that the estimation accuracy is difficult to guarantee. For the two reasons mentioned above, simplifying the model and reducing the amount of computation without losing accuracy are important factors in the ability of the estimation algorithm to be deployed in a battery management system.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to solve the technical problem of providing a lithium ion battery health state estimation method based on novel health characteristics.
The technical scheme adopted by the invention for solving the technical problems is as follows: the invention provides a lithium ion battery health state estimation method based on novel health characteristics, which specifically comprises the following steps:
step one, acquiring a charging voltage curve under a constant-current constant-voltage charging condition, selecting a voltage difference with equal initial voltage and equal time intervals as a novel health characteristic, and expressing the formula as follows:
ΔV＝V(t+N)-V 0 (t) (1)
wherein, V 0 (t) represents a starting voltage, V (t + N) represents a charging voltage at time t + N, N represents a time interval;
step two, the linear relation between the novel health characteristics and the health state of the lithium battery can be known through analysis, and a correlation coefficient method is used for verification;
step three, because the charging and discharging platforms of different lithium batteries are different, the interval where the novel health characteristics on the charging voltage curve and the health state of the lithium batteries are in a linear relation is different, in order to determine the interval in the linear relation, the initial voltage and the time interval are optimized, and an optimized initial voltage and a plurality of optimized time intervals are obtained from the same charging voltage curve; combining the optimized starting voltage with each optimized time interval to obtain a plurality of novel health characteristics extracted from the same charging voltage curve;
step 3-1, optimizing the initial voltage: setting a selection interval of initial voltage, traversing voltage values in the interval at a certain interval, roughly selecting a time interval at the moment, calculating a Pearson correlation coefficient between novel health characteristics obtained by different initial voltages and time intervals and a lithium battery health state, and taking a voltage corresponding to the maximum absolute value of the Pearson correlation coefficient as an optimized initial voltage;
step 3-2, optimizing time intervals: setting a time interval, wherein the time interval comprises a plurality of time intervals, and combining each time interval with the optimized initial voltage to obtain a novel health characteristic set; performing OLS regression analysis by taking the novel health feature set as input and the lithium battery health state as output to obtain the significance of each novel health feature in a linear relation with the lithium battery health state; removing the novel health features with the minimum significance, then carrying out OLS regression analysis on the remaining features and the health state, repeating the process, gradually removing the novel health features with the smaller significance until the inspection values of the remaining novel health features are smaller than a set threshold value, and obtaining the time interval corresponding to the remaining novel health features as the optimization time interval;
selecting a multiple linear regression model to construct a lithium battery health state estimation model, wherein the expression is as follows:
y≈w T x+b (2)
wherein y = { y = 1 ,y 2 ,…,y i ,…,y m The state of health matrix of the lithium battery corresponding to the m charging voltage curves is used, and x = (x) 11 ,x 12 ,…,x 1n ),(x 21 ,x 22 ,…,x 2n ),…,(x m1 ,x m2 ,…,x mn ) Is a new health feature matrix corresponding to y, (x) i1 ,x i2 ,…,x in ) For the ith charging voltage curve at the optimum time interval N 1 ,N 2 ,…,N n The method comprises the following steps that (1) a novel health characteristic matrix is formed, i belongs to m, b is a constant coefficient matrix, T is a matrix transposition, w is a characteristic coefficient matrix, and the novel health characteristic matrix is determined by a least square method according to a formula (3);
wherein | · | purple sweet 2 Represents a 2-norm;
inputting the novel health characteristics extracted from the charging voltage curve in the third step and the corresponding health state values into a lithium battery health state estimation model, training the model, and using the trained model for lithium battery health state estimation.
Further, for lithium batteries of the same model, the health state estimation can be carried out by using the formula (2) after charging voltage data under the same charging condition is obtained; for lithium batteries of different models or charging voltage data obtained by changing charging conditions, the initial voltage and the time interval need to be re-optimized, and then the lithium battery health state estimation model needs to be retrained.
Compared with the prior art, the invention has the beneficial effects that:
1. the voltage difference of equal initial voltage at equal time intervals is used as a health characteristic, only the interval in a linear relation needs to be found from a charging voltage curve according to the linear relation between the characteristic and the health state, voltage data on the whole curve is not selected, the calculation amount is small, a multiple linear regression model is selected as an estimation model, complex calculation and training of a neural network are not needed, and experimental results show that on the premise of ensuring the estimation accuracy, compared with a KRR method, the method disclosed by the invention has the advantages that the estimation speed is improved by more than 2 times, the method can be deployed in a battery management system, and the method can be applied to the condition that the calculation force of the current battery management system is insufficient.
2. In order to extract the characteristics which are in a remarkable linear relation with the health state, two influence parameters of the initial voltage and the time interval of the health characteristics are optimized, and the parameter optimization is positioned before model training, so that the optimized parameters can be directly used for the health state estimation of the lithium batteries with the same type and the same charging condition, the secondary optimization is not needed, and the estimation speed is further improved. For different types of batteries or changing charging conditions, parameters need to be re-optimized, and the model needs to be retrained for state of health estimation.
Drawings
FIG. 1 is an overall flow diagram of the present invention;
FIG. 2 (a) is a graph of the charging voltage of lithium batteries at different SOH in the aging data set of Oxford university over time;
FIG. 2 (b) is a graph of lithium battery charging voltage versus time at different SOHs for a NASA random use data set;
FIG. 3 (a) is an actual capacity fade curve of each lithium battery at different cycle times in the aging data set of Oxford university;
FIG. 3 (b) is the actual capacity decay curve for each lithium cell in the NASA random use dataset at different cycle times;
FIG. 4 is a fit curve of the novel health characteristics and health status of each lithium battery in the aging data set of Oxford university;
FIG. 5 (a) is a plot of Pearson correlation coefficient versus initial voltage for each lithium cell in the Oxford university Battery aging data set;
FIG. 5 (b) is a plot of Pearson correlation coefficient versus initial voltage for each lithium cell in the NASA random use data set;
fig. 6 (a) is the state of health estimation result for battery No. 4 in the aging dataset of oxford university;
fig. 6 (b) is the state of health estimation result of battery No. 8 in the aging data set of oxford university.
Detailed Description
The technical method of the present invention is further described in detail with reference to the drawings and the detailed description, but the scope of the present invention is not limited thereto.
The invention provides a lithium ion battery health state estimation method (a method for short) based on novel health characteristics, which comprises the following specific steps:
step one, processing and analyzing aging data of a lithium battery: acquiring a charging Voltage curve under the constant-current constant-Voltage charging condition, and selecting a Voltage Difference at Equal time Intervals (VDEISV) of the Same initial Voltage as a novel health characteristic; the constant-current constant-voltage charging condition is that the constant-current charging is carried out until the cut-off voltage is reached, and then the constant-voltage charging is carried out until the current is reduced to 0.01A;
in the embodiment, an aging data set of oxford university and a data set randomly used by NASA are taken as examples for analysis, the two data sets record charging voltage data in the use process of a lithium battery, and the battery capacity is measured by a discharge period ampere-hour integration method after every other period of circulation to obtain actual SOH, so that subsequent analysis and verification can be performed; fig. 2 (a) and (b) are graphs showing the change of the charging voltage of the lithium battery under different SOHs in two data sets with time, the SOH is represented by a gradual change color from light to deep along with the decay of the service life of the battery, and it can be found from the graphs that the change of the charging voltage curve of the lithium battery is relatively stable due to the fixed charging mode, and the change of the charging voltage curve of the lithium battery is regularly changed along with the aging of the lithium battery, which is specifically represented as follows: when the lithium battery is short in service time, namely the SOH of the battery is large (a light-colored curve part), the process that the charging voltage rises to the cut-off voltage (4.2V) is slow, and the constant-current charging time is long, so that the essence that the internal chemical reaction time is longer and the capacity is larger when the battery is newer is also met; when the lithium battery is used for a period of time, the SOH is reduced (part of a dark curve), the process of increasing the charging voltage to the cut-off voltage is accelerated, the constant-current charging time is obviously shortened, and the capacity of the lithium battery is reduced at the moment; as can be seen from the analysis of the charging voltage curve, a method of directly performing analysis processing on the charging voltage curve to perform SOH estimation is possible.
8 Kokam lithium cobaltate ion bagged batteries are collected in an aging data set of the Oxford university, the rated capacity of the batteries is 740 mA.h, all data are obtained by repeatedly carrying out 2C (1.48A) constant current discharge and recharging on the lithium ion batteries under the condition of constant environmental temperature of 40 ℃, and the capacity is attenuated to 0.43 A.h at the minimum; in the charging process, the voltage acquisition frequency is 1Hz, 519 complete charging voltage curves are obtained, and fig. 3 (a) is an actual capacity attenuation curve of each lithium battery in the data set under different cycle times. And (4) removing the latter parts of the capacity mutation of the Cell2 and the Cell5 according to the graph to obtain a charging voltage curve 490 which can be used for SOH estimation. The battery used by the NASA random use data set is an LG Chem 18650 cylindrical battery, the rated capacity is 2.1 A.h, and the conventional working voltage range is 3.2-4.2V. Although each group of experiments in the data set adopt different random working conditions during discharging, the same constant-current and constant-voltage working conditions are adopted during charging. In this embodiment, eight batteries tested at room temperature are selected for analysis, namely RW13, RW14, RW15, RW16 subjected to random discharge at a lower current and RW17, RW18, RW19, RW20 subjected to random discharge at a higher current, and standard charging and discharging are performed after each 50 cycles, wherein the discharging condition is that 1A constant current is discharged to 3.2V for calculating the actual capacity, the charging condition is 2A constant current charging to 4.2V, and the constant voltage charging is performed until the current is reduced to 0.01A, so that a total of 203 charging voltage curves under constant charging conditions are obtained; the NASA random use data set considers that a standard charge and discharge test is performed after every 50 random discharge cycles, and after the 450 th cycle, the first sampling voltage value reaches 4.0V, and the subsequent cycle curve is cut off. That is, 80 usable curve data of 8 batteries are obtained, and fig. 3 (b) is an actual capacity fading curve of different lithium batteries in the data set under different cycle times.
The voltage difference value that increases with charging voltage curve along with time is as the measurement characteristic of curve self, combines together with the time interval of waiting to charge the voltage difference, constructs novel healthy characteristic, selects the voltage difference Δ V of waiting time interval such as initial voltage from charging voltage curve promptly as novel healthy characteristic, and its expression is:
ΔV＝V(t+N)-V 0 (t) (1)
wherein, V 0 (t) represents the starting voltage, i.e. the charging voltage at time t; v (t + N) represents the charging voltage at time t + N, N representing the time interval;
analyzing the relationship between the novel health characteristics and the lithium battery health state through curve fitting, and verifying by using a correlation coefficient method;
according to the characteristics of the characteristics, two parameters, namely initial voltage and time interval, are required to be determined for extracting the health state of the lithium battery, so that the initial voltage is roughly selected to be 3.8V, the time interval is 600Hz, and curve fitting is carried out on the novel health characteristics and the corresponding health state of the lithium battery; fig. 4 is a fitting curve of the novel health characteristics and the health status of each lithium battery in the aging data set of oxford university, and it is obvious from the figure that there is a significant linear relationship between the two;
the relation between the novel health characteristics and the health state is verified through the Pearson correlation coefficient, the absolute values of the Pearson correlation coefficient are all above 0.99, and therefore the novel health characteristics and the health state of the lithium battery are in a linear relation after verification.
Step three, because the charging and discharging platforms of different lithium batteries are different, the interval of the novel health characteristics on the charging voltage curve and the lithium battery health state in a significant linear relation is different, so in order to accurately determine the interval in the linear relation, the initial voltage and the time interval need to be optimized to obtain the optimized initial voltage and the optimized time interval; combining the optimized initial voltage of the same charging voltage curve with different optimized time intervals to obtain a plurality of novel health characteristics extracted from the same charging voltage curve, and using the novel health characteristics for training a lithium battery health state estimation model;
step 3-1, initial Voltage V 0 The optimization of (2): setting the starting voltage V 0 The voltage value in the interval is traversed at certain intervals, the time interval N at the moment is roughly selected, and different initial voltages V are calculated 0 And acquiring a Pearson correlation coefficient between the novel health characteristics and the lithium battery health state corresponding to the time interval N, taking the maximum Pearson correlation coefficient (presenting negative correlation, namely having the minimum negative value) as an optimization result, and taking the voltage corresponding to the maximum Pearson correlation coefficient as an initial voltage V with the most obvious influence on linear relation 0 ；
3-1-1, wherein the voltage selection interval of the aging data set of the Oxford university is 3.75-3.81V, every 0.01 unit of voltage is taken as an initial voltage, the time interval N is 600Hz, and the change curve of the Pearson correlation coefficient along with the initial voltage is shown in a graph 5 (a), so that the optimized initial voltage is 3.8V;
3-1-2, the voltage selection interval of the NASA random use data set is 3.87-4.02V, the selected time interval N is 4, the charging voltage sampling interval is 60s, and the actual time is 240s; the pearson correlation coefficient is plotted against the initial voltage as shown in fig. 5 (b), and it can be seen that the optimal initial voltage is 3.89V.
Step 3-2, optimizing the time interval N: setting a time interval, wherein the time interval comprises a plurality of time intervals in total, and each time interval and the optimized initial voltage V 0 The combination can obtain a novel health characteristic, so that a novel health characteristic set is obtained; taking the novel health characteristic set as input, taking the lithium battery health state as output, and performing regression analysis by using an OLS (ordinary least square method) to obtain the linear relation between each characteristic in the novel health characteristic set and the lithium battery health stateThe significance is high; removing the novel health features with the minimum significance, then carrying out OLS regression analysis on the remaining features and the health state, repeating the process, gradually removing the novel health features with the smaller significance until the inspection values of the remaining novel health features are smaller than a set threshold value, and obtaining the time interval corresponding to the remaining novel health features as the optimization time interval;
3-2-1, wherein the time interval of the aging data set of the Oxford university is 100-1000, the interval is 100, so that the time interval is 10, and the first OLS regression analysis result is shown in a table 1; setting the threshold value to be 0.05, and rejecting the time intervals 300 and 500 according to a gradual rejection principle to finally obtain the optimized time intervals of 100, 200, 400, 600, 700, 800, 900 and 1000.
TABLE 1 OLS analysis results of the aging dataset of Oxford university
| |
100 | 200 | 300 | 400 | 500 |
| P>|t| | 0.000 | 0.006 | 0.143 | 0.001 | 0.007 |
| |
600 | 700 | 800 | 900 | 1000 |
| P>|t| | 0.034 | 0.002 | 0.019 | 0.008 | 0.000 |
And 3-2-2, performing a standard charge and discharge test by the NASA random use data set after every 50 random discharge cycles, wherein after the 450 th cycle, the first sampling voltage value reaches 4.0V, and the subsequent cycle curve is cut off. OLS analysis was performed using 80 curve data of 8 cells, with voltage sampling recording intervals of 60s, N of 1 sampling interval, time interval of 1-10, and first OLS regression analysis results as shown in Table 2. Considering that too large data sampling interval will result in larger results of the OLS analysis, the threshold is selected to be 0.2, and the optimal time intervals are 1, 2, 3, 5, 6, 7, and 9.
TABLE 2 OLS analysis results of NASA random use dataset
| |
1 | 2 | 3 | 4 | 5 |
| P>|t| | 0.093 | 0.077 | 0.223 | 0.887 | 0.301 |
| |
6 | 7 | 8 | 9 | 10 |
| P>|t| | 0.062 | 0.284 | 0.349 | 0.310 | 0.366 |
Step four, considering the problem of insufficient computing resources faced by the current battery management system, and meanwhile, according to the relationship between the novel health characteristics obtained by analysis in the step two and the health state of the lithium battery, selecting a Multiple Linear Regression (MLR) model as a lithium battery health state estimation model, wherein the expression is as follows:
y≈w T x+b (2)
wherein y = { y = 1 ,y 2 ,…,y i ,…,y m Is a lithium battery health state matrix corresponding to m charging voltage curves, and x ={(x 11 ,x 12 ,…,x 1n ),(x 21 ,x 22 ,…,x 2n ),…,(x m1 ,x m2 ,…,x mm ) Is a new health feature matrix corresponding to y, (x) i1 ,x i2 ,…,x in ) For the ith charging voltage curve, N is the optimum time interval 1 ,N 2 ,…,N n The method comprises the following steps that (1) a novel health characteristic matrix is formed, i belongs to m, b is a constant coefficient matrix, T is a matrix transposition, w is a characteristic coefficient matrix, and the novel health characteristic matrix is determined by a least square method according to a formula (3);
wherein | · | purple sweet 2 Representing a second order norm;
inputting the novel health characteristics extracted from the charging voltage curve in the third step and the corresponding health state values into a lithium battery health state estimation model, training the model, and using the trained model for lithium battery health state estimation.
For lithium batteries of the same model, the health state estimation can be carried out by using the formula (2) after the charging voltage data under the same charging condition is obtained; for lithium batteries of different models or charging voltage data obtained by changing charging conditions, the optimal initial voltage and time interval need to be determined again according to the third step, and the lithium battery health state estimation model needs to be trained again according to the fourth step.
In order to verify the effectiveness of the method, the method is quantitatively evaluated by selecting Mean Absolute Error (MAE) and maximum Error (MAX Error, MAX) as evaluation indexes, and compared with the existing Kernel Ridge Regression (KRR) method in estimation accuracy, and meanwhile, computational power comparison is carried out through model training and estimated time. The expressions of the two evaluation indexes are respectively:
MAX＝max|y j -y′ j | (5)
in the formula: m is the number of samples to be measured, y j 、y′ j The SOH estimation result and the actual value are respectively represented.
For comparison with the results obtained by the Kernel Ridge Regression (KRR) method, the data of batteries No. 4 and No. 8 in the aging data set of oxford university were used as test samples, and the remaining 6 batteries were used as training samples. The NASA random use data set compares only global evaluation results because of different used battery data, taking RW16 (low rate discharge), RW20 (high rate discharge) battery data as test samples and the remaining 6 batteries as training samples. A comparison of the estimates of the two data sets is shown in tables 3 and 4, where mlr represents the method of the invention.
TABLE 3 SOH estimation results for the aging dataset of Oxford university
TABLE 4 NASA stochastic usage data SOH estimation results evaluation
The health state estimation results of the battery No. 4 and the battery No. 8 in the aging data set of the university of Oxford are shown in fig. 6 (a) and (b), and as can be seen from fig. 6, the estimation results of the two methods are basically consistent, and the reliability of the method is verified. As can be seen from table 3, compared with the KRR method, the method of the present invention not only reduces the computational complexity, increases the computation speed by more than two times, but also does not lose much precision, has a difference of only 0.01% between the MAX errors, reduces the global MAE, and can be applied to the situation of insufficient computational power of the current battery management system. Table 4 shows that for the NASA random use data set, the method of the present invention still obtains good estimation results under the conditions that the battery charging voltage sampling interval is large and the data calculation is not very accurate. The global mean absolute value error (MAE) is 1.65%, the maximum error (MAX) is 3.80%, the speed is improved, and the precision is also partially improved. The method shows strong generalization capability.
Nothing in this specification is said to apply to the prior art.
Claims (2)
1. A lithium ion battery health state estimation method based on novel health characteristics is characterized by comprising the following steps:
step one, acquiring a charging voltage curve under a constant-current constant-voltage charging condition, selecting a voltage difference with equal initial voltage and equal time intervals as a novel health characteristic, and expressing the formula as follows:
ΔV＝V(t+N)-V 0 (t) (1)
wherein, V 0 (t) represents a starting voltage, V (t + N) represents a charging voltage at time t + N, N represents a time interval;
step two, the linear relation between the novel health characteristics and the health state of the lithium battery can be known through analysis, and a correlation coefficient method is used for verification;
step three, because the charging and discharging platforms of different lithium batteries are different, the interval where the novel health characteristics on the charging voltage curve and the health state of the lithium batteries are in a linear relation is different, in order to determine the interval in the linear relation, the initial voltage and the time interval are optimized, and an optimized initial voltage and a plurality of optimized time intervals are obtained from the same charging voltage curve; combining the optimized starting voltage with each optimized time interval to obtain a plurality of novel health characteristics extracted from the same charging voltage curve;
step 3-1, optimizing the initial voltage: setting a selection interval of initial voltage, traversing voltage values in the interval at a certain interval, roughly selecting a time interval at the moment, calculating a Pearson correlation coefficient between novel health characteristics obtained by different initial voltages and time intervals and a lithium battery health state, and taking a voltage corresponding to the maximum absolute value of the Pearson correlation coefficient as an optimized initial voltage;
step 3-2, optimizing time intervals: setting a time interval, wherein the time interval comprises a plurality of time intervals, and combining each time interval with the optimized initial voltage to obtain a novel health characteristic set; performing OLS regression analysis by taking the novel health characteristic set as input and the lithium battery health state as output to obtain the significance of each novel health characteristic in a linear relation with the lithium battery health state; removing the novel health features with the minimum significance, then carrying out OLS regression analysis on the remaining features and the health state, repeating the process, gradually removing the novel health features with the smaller significance until the inspection values of the remaining novel health features are smaller than a set threshold value, and obtaining the time interval corresponding to the remaining novel health features as the optimization time interval;
selecting a multiple linear regression model to construct a lithium battery health state estimation model, wherein the expression is as follows:
y≈w T x+b (2)
wherein y = { y = 1 ，y 2 ，…，y i ，…，y m The matrix is a lithium battery health state matrix corresponding to m charging voltage curves, and x = { (x) 11 ，x 12 ，…，x 1n )，(x 21 ，x 22 ，…，x 2n )，…，(x m1 ，x m2 ，…，x mn ) Y is a new health feature matrix corresponding to (x) i1 ，x i2 ，…，x in ) For the ith charging voltage curve, N is the optimum time interval 1 ，N 2 ，…，N n The method comprises the following steps that (1) a novel health characteristic matrix is formed, i belongs to m, b is a constant coefficient matrix, T is a matrix transposition, w is a characteristic coefficient matrix, and the novel health characteristic matrix is determined by a least square method according to a formula (3);
wherein | · | purple sweet 2 Represents a 2-norm;
inputting the novel health characteristics extracted from the charging voltage curve in the third step and the corresponding health state values into a lithium battery health state estimation model, training the model, and using the trained model for lithium battery health state estimation.
2. The lithium ion battery state of health estimation method based on novel health characteristics according to claim 1, wherein for lithium batteries of the same type, obtaining charging voltage data under the same charging conditions enables state of health estimation using equation (2); for lithium batteries of different models or charging voltage data obtained by changing charging conditions, the initial voltage and the time interval need to be optimized again, and then the lithium battery health state estimation model needs to be trained again.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211374561.9A CN115586452A (en) | 2022-11-03 | 2022-11-03 | Lithium ion battery health state estimation method based on novel health characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211374561.9A CN115586452A (en) | 2022-11-03 | 2022-11-03 | Lithium ion battery health state estimation method based on novel health characteristics |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115586452A true CN115586452A (en) | 2023-01-10 |
Family
ID=84781373
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211374561.9A Pending CN115586452A (en) | 2022-11-03 | 2022-11-03 | Lithium ion battery health state estimation method based on novel health characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115586452A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116908723A (en) * | 2023-06-08 | 2023-10-20 | 武汉亿纬储能有限公司 | Calculation method and device for battery cycle times |
| CN116973794A (en) * | 2023-09-06 | 2023-10-31 | 广东工业大学 | Lithium battery SOH estimation method based on incomplete charging voltage curve reconstruction |
-
2022
- 2022-11-03 CN CN202211374561.9A patent/CN115586452A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116908723A (en) * | 2023-06-08 | 2023-10-20 | 武汉亿纬储能有限公司 | Calculation method and device for battery cycle times |
| CN116973794A (en) * | 2023-09-06 | 2023-10-31 | 广东工业大学 | Lithium battery SOH estimation method based on incomplete charging voltage curve reconstruction |
| CN116973794B (en) * | 2023-09-06 | 2024-04-19 | 广东工业大学 | Lithium battery SOH estimation method based on incomplete charging voltage curve reconstruction |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108896914B (en) | Gradient lifting tree modeling and prediction method for health condition of lithium battery | |
| CN110146822B (en) | Vehicle power battery capacity online estimation method based on constant-current charging process | |
| CN115586452A (en) | Lithium ion battery health state estimation method based on novel health characteristics | |
| CN107741568B (en) | Lithium battery SOC estimation method based on state transition optimization RBF neural network | |
| CN112986830A (en) | Lithium battery capacity estimation method based on convolution time memory neural network | |
| CN112557907A (en) | SOC estimation method of electric vehicle lithium ion battery based on GRU-RNN | |
| 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 | |
| CN112684363A (en) | Lithium ion battery health state estimation method based on discharge process | |
| CN113189490B (en) | Lithium battery health state estimation method based on feature screening and Gaussian process regression | |
| CN112611976A (en) | Power battery state of health estimation method based on double differential curves | |
| CN116381541B (en) | Health assessment method and system for energy storage lithium battery system | |
| CN111832221A (en) | Lithium battery life prediction method based on feature screening | |
| CN113311337A (en) | Lithium ion battery residual life detection method based on relevance vector regression | |
| US20230349977A1 (en) | Method and apparatus for estimating state of health of battery | |
| CN115308606B (en) | Lithium ion battery health state estimation method based on adjacent characteristics | |
| CN115994441A (en) | Big data cloud platform online battery life prediction method based on mechanism information | |
| CN114545275A (en) | Indirect prediction method for remaining service life of lithium ion battery | |
| CN117054892B (en) | Evaluation method, device and management method for battery state of energy storage power station | |
| Lyu et al. | State-of-charge estimation of lithium-ion batteries based on deep neural network | |
| CN116224085A (en) | Lithium battery health state assessment method based on data driving | |
| CN116449244A (en) | Lithium battery aging track early prediction method and system based on single discharge curve | |
| CN116754960A (en) | Cloud SOH estimation and residual life prediction method, device and program for sodium ion battery | |
| CN116736171A (en) | Lithium ion battery health state estimation method based on data driving | |
| Kuang et al. | State-of-charge estimation hybrid method for lithium-ion batteries using BiGRU and AM co-modified Seq2Seq network and H-infinity filter | |
| CN113255215B (en) | Lithium battery health state estimation method based on voltage segments |
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 |