CN113821914B - Low-cost prediction method for cycle life of lithium ion battery - Google Patents

Abstract

To solve the problem of using the existingThe invention provides a low-cost prediction method for the cycle life of a lithium ion battery. The invention discovers the average ohmic internal resistance IR and the discharge capacity Q of the lithium ion battery monomer in the N (N is more than or equal to 100) th cycle process by screening characteristic parameters _d And a minimum temperature T _min The three characteristic parameters are subjected to dimension reduction and fusion to obtain a new characteristic

And the cycle life Y of the lithium ion battery ⁽ⁱ⁾ And a good linear relation is shown, so that a lithium ion battery cycle life prediction model is obtained by specially processing the three characteristic parameters and the cycle life. When the prediction model obtained by the invention is used for predicting the cycle life of the lithium ion battery to be tested, the cycle life of the lithium ion battery to be tested can be predicted more accurately only by adopting test data in the Nth (N is more than or equal to 100) cycle process of the lithium ion battery to be tested, so that the prediction cost is greatly reduced.

Description

A low-cost prediction method for the cycle life of lithium-ion batteries

technical field

The present invention relates to a method for testing the electrical condition of a battery, in particular to a low-cost prediction method of the cycle life of a lithium ion battery.

Background technique

As the traditional power system of underwater vehicle, the thermal power system cannot meet the requirements of large voyage depth due to high noise, poor concealment, and the influence of back pressure. In recent years, under the wave of the era that the country actively advocates new energy, lithium-ion batteries with excellent performance stand out and become a research hotspot in various industries. For the underwater field, the electric power system is a closed system, does not do any discharge, has no wake, and is not affected by back pressure, so it has become the object favored by researchers in the industry.

Lithium-ion batteries have the characteristics of high energy density and high power density. In order to maximize battery efficiency and maximize battery life, the battery is equipped with a battery management system (BMS) under normal use. Accurate estimation of battery state-of-health (SOH) is a key technology in BMS, and it is a reliable guarantee for the safe operation and work efficiency of battery systems, including battery cycle life parameters.

The battery cycle life can be obtained by testing, but because the lithium-ion battery itself has a long life, and in recent years, with the development of lithium-ion battery research and development technology has become more and more mature, the lithium-ion life has become longer and longer, and the number of cycles is even as high as tens of thousands of times. When using the existing cycle life test to obtain the battery cycle life, it often takes a long time, the cost is huge, and the prediction is difficult.

SUMMARY OF THE INVENTION

In order to solve the technical problems of long time and high cost in obtaining battery cycle life by using the existing cycle life test, the present invention provides a low-cost prediction method for the cycle life of a lithium ion battery.

In order to solve the above-mentioned technical problems, the technical scheme adopted in the present invention is:

A low-cost prediction method for the cycle life of a lithium-ion battery, comprising the following steps:

Step 1: Select m lithium-ion battery cells belonging to the same system, and carry out the cycle life test respectively; m≥60;

Step 2: For each lithium-ion battery cell, record the test parameters X ⁽ⁱ⁾ and cycle life Y ⁽ⁱ⁾ during the Nth cycle; X ⁽ⁱ⁾ = (IR ⁽ⁱ⁾ , Q _d ^{(i )} , T _min ⁽ⁱ⁾ ); i is the serial number of the lithium ion battery cell; IR ⁽ⁱ⁾ , Q _d ⁽ⁱ⁾ and T _min ⁽ⁱ⁾ are the Nth cycle process of the i th lithium ion battery cell, respectively Average ohmic internal resistance, discharge capacity and minimum temperature in ; N≥100;

对所述锂离子电池单体的循环寿命Y⁽ⁱ⁾进行离群点处理；Step 3: Perform dimensionality reduction processing on the X ^{( i)} , and fuse the three feature information in X (i) into one feature

performing outlier processing on the cycle life Y ⁽ⁱ⁾ of the lithium-ion battery cell;

和离群点处理后的电池循环寿命进行拟合，得到锂离子电池循环寿命预测模型；Step 4: Dimension reduction obtained from step 3

Fitting with the battery cycle life after outlier processing to obtain a lithium-ion battery cycle life prediction model;

Step 5: Input the parameters {IR, Q _d , T _min } measured in the Nth cycle of the lithium-ion battery to be tested into the lithium-ion battery cycle life prediction model, and then the cycle life of the lithium-ion battery to be tested can be obtained .

Further, N=100 in step 1.

Further, step 2 adopts the method of PCA (Principal Component Analysis, principal component analysis) to carry out dimension reduction processing on the X ⁽ⁱ⁾ , and its objective function is as follows:

In the formula,

w represents the direction vector, w=(w ₁ , w ₂ , w ₃ );

m represents the number of lithium-ion battery cells selected in step 1;

i represents the serial number of the lithium-ion battery cell;

表示均值归零化以后的特征，

represents the feature after mean zeroing,

Further, in step 3, the outlier processing adopts the boxplot analysis method, and its boundary value is defined as follows:

IQR=Q ₃ -Q ₁

l _u =Q ₃ +1.5IQR

l _l =Q ₁ -1.5IQR

in,

Q ₃ represents the upper-middle quarter value of the single cycle life Y ⁽ⁱ⁾ of the lithium-ion battery obtained in step 2;

Q ₁ represents the middle and lower quarter value of the single cycle life Y ⁽ⁱ⁾ of the lithium ion battery obtained in step 2;

IQR represents the interquartile range of the lithium-ion battery cell cycle life Y ⁽ⁱ⁾ obtained in step 2;

l _u is the upper limit used to filter out outliers;

l _l is the lower limit used to filter out outliers.

Further, in step 4, the least squares method is used for fitting, and its objective function is as follows:

in,

a, b are the coefficients of the linear regression equation;

n is the number of cells of the lithium-ion battery selected in step 1 after the outlier screening operation;

为X⁽ⁱ⁾经过降维后得到的特征；

is the feature obtained by X ⁽ⁱ⁾ after dimensionality reduction;

Y ⁽ⁱ⁾ is the cycle life of the i-th lithium-ion battery cell;

为所有锂离子电池单体经过降维后得到的特征的均值，

is the mean value of the features obtained after dimensionality reduction of all lithium-ion battery cells,

为经离群点处理后所得到的所有锂离子电池单体循环寿命的均值，

is the average value of the cycle life of all lithium-ion battery cells obtained after outlier processing,

Compared with the prior art, the beneficial effects of the present invention are as follows:

与该锂离子电池循环寿命Y⁽ⁱ⁾展现出较好的线性关系(而目前本领域技术人员通常认为电池容量衰减与循环次数呈非线性关系)，因此通过对这三个特征参数及循环寿命进行特殊处理，得到了锂离子电池循环寿命预测模型。在利用本发明得到的预测模型对待测锂离子电池的循环寿命进行预测时，只需采用待测锂离子电池第N(N≥100)次循环过程中的测试数据，即可较准确的预测其循环寿命，使得预测成本大大降低。1. The present invention analyzes the cycle life test data of at least 60 groups of lithium-ion battery cells, and finds the average ohmic internal resistance IR, discharge capacity during the Nth (N≥100) cycle of lithium-ion battery cells through feature parameter screening. The new features of the three feature parameters Q _d and minimum temperature T _min after dimensionality reduction and fusion

It shows a good linear relationship with the cycle life Y ⁽ⁱ⁾ of the lithium-ion battery (while those skilled in the art generally believe that the battery capacity decay has a nonlinear relationship with the number of cycles), so by comparing these three characteristic parameters and cycle life After special treatment, a prediction model for the cycle life of lithium-ion batteries was obtained. When using the prediction model obtained in the present invention to predict the cycle life of the lithium-ion battery to be tested, it is only necessary to use the test data during the Nth (N≥100) cycle of the lithium-ion battery to be tested to more accurately predict the cycle life of the lithium-ion battery to be tested. Cycle life, making the forecast cost greatly reduced.

When the data in the 100th cycle is used to establish the prediction model, the prediction accuracy can already meet the actual demand, and the modeling time is the shortest.

2. The prediction model obtained by the present invention is simple and intuitive, and at the same time, the prediction accuracy can basically meet the actual demand, which is of great significance for the sustainable and reliable development of the lithium ion battery.

3. The present invention uses PCA to perform dimensionality reduction processing on X ⁽ⁱ⁾ , which can ensure that the projection of the sample on the vector can retain as much original sample information as possible, and remove the noise generated during the test; Modeling after fusion can make the model more intuitive and easy to display.

4. Before fitting, the present invention uses the boxplot analysis method to process the cycle life Y ⁽ⁱ⁾ , eliminates outliers in the cycle life data caused by sensor noise, and improves the prediction accuracy.

5. The present invention adopts least squares for fitting, the model is simple and easy to understand, and the calculation amount is small.

Description of drawings

FIG. 1 is a schematic flow chart of the method of the present invention.

FIG. 2 is an example diagram of a prediction model obtained in an embodiment of the present invention.

Detailed ways

The present invention will be described in more detail below with reference to the accompanying drawings.

As shown in FIG. 1 , in the low-cost prediction method for the cycle life of a lithium-ion battery provided by the embodiment of the present invention, the specific steps are as follows:

Step 1: Select m lithium-ion battery cells belonging to the same system, and carry out cycle life tests on them respectively; m≥60 (m=124 in this embodiment; the systems mentioned here are distinguished by positive electrode materials, Those with the same positive electrode material are batteries of the same system, such as lithium iron phosphate batteries using lithium iron phosphate as the positive electrode material, and nickel cobalt lithium manganate batteries using lithium nickel cobalt manganate as the positive electrode material.

Step 2: For each lithium-ion battery cell selected in Step 1, store the test parameters X ⁽ⁱ⁾ and cycle life Y ⁽ⁱ⁾ during the Nth (in this embodiment, N=100) cycle process into In the database; X ⁽ⁱ⁾ = (IR ⁽ⁱ⁾ , Q _d ⁽ⁱ⁾ , T _min ⁽ⁱ⁾ ); where, IR ⁽ⁱ⁾ is the average of the i-th lithium-ion battery cell during the Nth cycle Ohmic internal resistance; Q _d ⁽ⁱ⁾ is the discharge amount during the Nth cycle of the i-th lithium-ion battery cell; T _min ⁽ⁱ⁾ is the minimum temperature during the Nth cycle of the i-th lithium-ion battery cell ; The battery cycle life Y ⁽ⁱ⁾ refers to the number of cycles of the lithium-ion battery when the battery capacity decays to 80% of the rated capacity.

Step 3: Data Processing:

为排除数据采集时存在的噪音，对锂离子电池单体的循环寿命Y⁽ⁱ⁾进行离群点处理。Perform dimensionality reduction processing on X ⁽ⁱ⁾ obtained in step 1, and fuse the three feature information IR ⁽ⁱ⁾ , Q _d ⁽ⁱ⁾ and T _min ^{(i) in X (i) into} one feature

In order to exclude the noise in the data collection, the cycle life Y ⁽ⁱ⁾ of the lithium-ion battery cell is processed as outliers.

In this embodiment, the PCA (Principal Component Analysis, principal component analysis) method is preferred to perform dimension reduction processing on X ⁽ⁱ⁾ . PCA is an optimization algorithm for data dimensionality reduction. Its goal is to find a unit vector in the feature space of the sample, so that the projection of the sample on this vector can retain as much original sample information as possible. Its objective function is as follows:

In the formula,

w represents a direction vector, and there are three characteristic information in X ⁽ⁱ⁾ of the present invention, so w is a three-dimensional vector, that is, w=(w ₁ , w ₂ , w ₃ );

m represents the total amount of samples, the number of lithium ion battery cells selected in step 1 in the present invention;

i represents the i-th sample, which is the serial number of the lithium-ion battery cell in the present invention;

表示均值归零化以后的特征，维数为m×3，计算方式如下：

Represents the feature after mean zeroing, the dimension is m×3, and the calculation method is as follows:

表示特征均值向量，计算方式如下：

Represents the feature mean vector, which is calculated as follows:

Therefore:

本实施例优选boxplot分析法对循环寿命Y⁽ⁱ⁾进行离群点处理，其边界值定义如下：

The preferred boxplot analysis method of the present embodiment performs outlier processing on the cycle life Y ⁽ⁱ⁾ , and its boundary value is defined as follows:

IQR=Q ₃ -Q ₁

l _u =Q ₃ +1.5IQR

l _l =Q ₁ -1.5IQR

in,

Q ₃ represents the upper-middle quarter value of the single cycle life Y ⁽ⁱ⁾ of the lithium-ion battery obtained in step 2;

Q ₁ represents the middle and lower quarter value of the single cycle life Y ⁽ⁱ⁾ of the lithium ion battery obtained in step 2;

IQR represents the interquartile range of the lithium-ion battery cell cycle life Y ⁽ⁱ⁾ obtained in step 2;

l _u is the upper limit of the boxplot, that is, the upper limit used to filter out outliers;

l _l is the lower limit of the boxplot, that is, the lower limit used to filter out outliers.

和离群点处理后的锂离子电池循环寿命进行拟合，得到锂离子电池循环寿命预测模型。Step 4: Dimension reduction obtained from step 3

Fitting with the cycle life of the lithium-ion battery after the outlier treatment is carried out to obtain a prediction model of the cycle life of the lithium-ion battery.

In this embodiment, the least squares method is preferred for fitting. The least squares method is a simple linear regression algorithm, and its objective function is as follows:

Among them, a and b represent the coefficients of the linear regression equation;

n is the number of cells of the lithium-ion battery selected in step 1 after the outlier screening operation;

表示X⁽ⁱ⁾经过降维后的特征；

Represents the features of X ⁽ⁱ⁾ after dimensionality reduction;

Y ⁽ⁱ⁾ represents the cycle life of the i-th lithium-ion battery cell;

表示所有锂离子电池单体经过降维后得到的特征的均值，计算方式如下：

Represents the mean value of the features obtained by all lithium-ion battery cells after dimensionality reduction, and is calculated as follows:

表示经离群点处理后所得到的所有锂离子电池单体循环寿命的均值，计算方式如下：

Represents the average value of the cycle life of all lithium-ion battery cells obtained after outlier processing. The calculation method is as follows:

The prediction model obtained in this embodiment is shown in Figure 2. It can be seen from the figure that the new feature X _re after dimension reduction by the PCA algorithm has a good linear relationship with the cycle life. The model is simple and intuitive, and it is convenient to apply.

Claims (5)

1. A low-cost prediction method for the cycle life of a lithium ion battery is characterized by comprising the following steps:

step 1: selecting m lithium ion battery monomers belonging to the same system, and respectively carrying out a cycle life test; m is more than or equal to 60;

step 2: recording the test parameter X of each lithium ion battery monomer in the Nth cycle process ⁽ⁱ⁾ And cycle life Y ⁽ⁱ⁾ ；X ⁽ⁱ⁾ ＝(IR ⁽ⁱ⁾ ，Q _d ⁽ⁱ⁾ ，T _min ⁽ⁱ⁾ ) (ii) a i is the serial number of the lithium ion battery monomer; IR ⁽ⁱ⁾ 、Q _d ⁽ⁱ⁾ And T _min ⁽ⁱ⁾ Respectively the average ohmic internal resistance, the discharge capacity and the minimum temperature of the ith lithium ion battery monomer in the Nth cycle process; n is more than or equal to 100;

and step 3: for the X ⁽ⁱ⁾ Performing dimensionality reduction treatment on the X ⁽ⁱ⁾ The three characteristic information in the system are fused into one characteristic

For the cycle life Y of the lithium ion battery monomer ⁽ⁱ⁾ Carrying out outlier treatment;

and 4, step 4: obtained by reducing the dimension in the step 3

Fitting the cycle life of the battery after the outlier treatment to obtain a lithium ion battery cycle life prediction model;

and 5: inputting lithium ions to be tested into the lithium ion battery cycle life prediction modelParameters measured in the Nth cycle of the battery { IR, Q } _d ，T _min And obtaining the cycle life of the lithium ion battery to be tested.

2. The lithium ion battery cycle life low-cost prediction method of claim 1, characterized in that: n in step 1 is 100.

3. The lithium ion battery cycle life low-cost prediction method of claim 1 or 2, characterized in that: step 2, performing Principal Component Analysis (PCA) on the X ⁽ⁱ⁾ And (4) performing dimension reduction treatment, wherein the objective function is as follows:

in the formula,

w represents a direction vector, w ═ w ₁ ，w ₂ ，w ₃ )；

m represents the number of the lithium ion battery monomers selected in the step 1;

i represents the number of the lithium ion battery cell;

representing the features after the mean has been zeroed,

4. the lithium ion battery cycle life low-cost prediction method of claim 1, characterized in that: in step 3, outlier processing adopts a boxplot analysis method, and the boundary values are defined as follows:

IQR＝Q ₃ -Q ₁

l _u ＝Q ₃ +1.5IQR

l _l ＝Q ₁ -1.5IQR

wherein,

Q ₃ shows the cycle life Y of the lithium ion battery monomer obtained in the step 2 ⁽ⁱ⁾ The middle upper quarter value of (1);

Q ₁ shows the cycle life Y of the lithium ion battery monomer obtained in the step 2 ⁽ⁱ⁾ The middle lower quarter value of (a);

IQR represents the cycle life Y of the lithium ion battery cell obtained in the step 2 ⁽ⁱ⁾ The quartile range of (d);

l _u is an upper limit value for screening outliers;

l _l is the lower limit used to screen out outliers.

5. The lithium ion battery cycle life low-cost prediction method of claim 4, characterized in that: and 4, fitting by adopting a least square method, wherein the target function is as follows:

wherein,

a and b are linear regression equation coefficients;

n is the number of the monomers of the lithium ion battery selected in the step 1 after the outlier screening operation;

is X ⁽ⁱ⁾ Obtaining the characteristics after dimensionality reduction;

Y ⁽ⁱ⁾ the cycle life of the ith lithium ion battery monomer is represented;

is the mean value of the characteristics obtained by dimension reduction of all lithium ion battery monomers,

is the average value of the cycle life of all lithium ion battery monomers obtained after the treatment of the outlier,

Images