CN113900033B - Lithium battery online service life prediction method based on charging data spatial distribution characteristics - Google Patents

Abstract

The invention discloses an online lithium battery service life prediction method based on charging data spatial distribution characteristics. The invention comprises the following steps: collecting charging voltage and current data and cycle life of a brand new lithium battery in a preset charging and discharging cycle interval; calculating corresponding spatial distribution characteristics; repeating the steps to collect and calculate each lithium battery, obtaining the cycle life and the corresponding spatial distribution characteristics of each lithium battery, and forming a training set; training the lithium battery life prediction regression model to obtain a trained lithium battery life prediction regression model; during online prediction, charging voltage and current data of the lithium battery to be predicted in a preset charging and discharging cycle interval are collected, corresponding spatial distribution characteristics are obtained through calculation and input into a regression model for prediction, and the cycle life of the lithium battery to be predicted at present is output. The invention realizes the accurate prediction of the service life of the lithium battery and improves the reliability and safety of the lithium battery.

Description

Online life prediction method of lithium battery based on spatial distribution characteristics of charging data

technical field

The invention belongs to a lithium battery online life prediction method in the lithium battery application field, and particularly relates to a lithium battery online life prediction method based on the spatial distribution characteristics of charging data.

Background technique

Lithium batteries have the advantages of low cost, high energy density, and long cycle life, and are widely used in stationary, portable, and transportation fields. Life prediction technology plays an important role in the safe operation, predictive maintenance and secondary use of lithium batteries. However, due to the complex aging mechanism of lithium batteries, and the aging path is affected by many factors in the design, production and application process, it is difficult to achieve simple, simple, and reliable performance under complex aging paths, extensive equipment variability, and variable dynamic operating conditions. Fast and accurate lithium battery life prediction has become a huge challenge. In addition, for large lithium battery packs consisting of thousands of cells, due to the inevitable existence of various intrinsic and extrinsic differences between cells, a separate life prediction for each cell is required, which will bring huge data Storage burden, computational burden, and cost burden. At the same time, since the discharge patterns of lithium batteries are random in practical applications, it is not possible to extract features based on specific discharge tests for online life prediction of lithium batteries. In general, the charging mode of lithium batteries is more fixed than the discharging mode, so extracting features from charging data to predict the life of lithium batteries is a more ideal choice. In addition, a large number of current lithium battery life prediction methods rely on accurate capacity measurement, which requires the complete charging process data of the lithium battery state of charge from zero, which is not in line with the actual use of lithium batteries. Therefore, it is of great significance to design and develop better lithium battery life prediction methods based on charging data and independent of capacity measurement.

SUMMARY OF THE INVENTION

In order to solve the deficiencies of the prior art, the present invention proposes an online life prediction method for lithium batteries based on the spatial distribution characteristics of charging data.

The scheme adopted in the present invention is:

The present invention includes the following steps:

1) Collect the charging voltage and current data of the new lithium battery in the same constant current charging mode stage in the preset charge-discharge cycle interval, and simultaneously collect the cycle life of the lithium battery in the preset charge-discharge cycle interval;

2) According to all the charging voltage and current data collected by the current lithium battery in the preset charge-discharge cycle interval, calculate and obtain the spatial distribution characteristics of the current lithium battery in the preset charge-discharge cycle interval;

3) Repeat steps 1)-2) to collect and calculate each lithium battery, obtain the cycle life and corresponding spatial distribution characteristics of each lithium battery in the preset charge-discharge cycle interval, and form a training set;

4) Train the lithium battery life prediction regression model based on the training set, and obtain the lithium battery life prediction regression model after training;

5) During online prediction, collect the charging voltage and current data of the new lithium battery to be predicted in the same constant current charging mode stage in the preset charge-discharge cycle interval, and calculate and obtain the space of the lithium battery to be predicted in the preset charge-discharge cycle interval. Distribution features, input the obtained spatial distribution features into the trained lithium battery life prediction regression model for prediction, and output the current cycle life of the lithium battery to be predicted.

Described step 2) is specifically:

According to the charging voltage and current data of the current lithium battery in the preset charging and discharging cycle interval, draw a two-dimensional coordinate axis space with the charging voltage as the horizontal axis and the charging current as the vertical axis, and draw the current lithium battery in the preset charging and discharging cycle interval. All charging voltage and current data are plotted in the two-dimensional coordinate axis space. According to the range of charging voltage and current, the horizontal axis and vertical axis are respectively divided into m equidistant voltage sub-intervals and n equidistant current sub-intervals to obtain m× n two-dimensional subspaces; count the distribution points of all the charging voltage and current data accumulated in the preset charge-discharge cycle interval of the current lithium battery in each two-dimensional subspace and use it as the spatial distribution characteristics of the current lithium battery.

The preset charge-discharge cycle interval in the step 5) is the same as the preset charge-discharge cycle interval in the step 1), and the same constant current charging mode stage in the step 5) is the same as the same constant current charging mode stage in the step 1).

For the lithium battery life prediction regression model, a machine learning regression model is selected.

The beneficial effects of the present invention are:

The invention solves the problem that the online life prediction of the lithium battery depends on a specific discharge mode and accurate capacity measurement in practical applications. The spatial distribution characteristics of charging voltage and current data are applied to the online life prediction of lithium batteries. Only by collecting the charging voltage and current data of the lithium battery in the same constant current charging mode stage in the preset charging and discharging cycle interval and obtaining statistics in each two. The number distribution characteristics of the dimensional subspace, and then accurately predict the cycle life of lithium batteries, without relying on specific discharge patterns and accurate capacity measurements that do not exist in actual lithium battery applications, and can be used for online life prediction of lithium batteries in different practical application scenarios. Better safe operation, predictive maintenance and secondary use of lithium batteries in practical applications.

Description of drawings

FIG. 1 is a flowchart of an online life prediction method for lithium batteries based on the spatial distribution characteristics of charging data proposed by the present invention.

FIG. 2 is a schematic diagram of the same constant current charging mode stage of a lithium battery selected in an embodiment of the present invention.

3 is a schematic diagram of the spatial distribution of all charging voltage and current data of the lithium battery in the same constant current charging mode stage in the 10th and 200th charge-discharge cycle intervals according to an embodiment of the present invention.

FIG. 4 is a graph showing the difference between the lifespan of a lithium battery predicted based on the spatial distribution characteristics of charging voltage and current data and the lifespan of an actual lithium battery in an embodiment of the present invention.

Detailed ways

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

The present invention includes the following steps:

1) Collect the charging voltage and current data of the new lithium battery in the same constant current charging mode stage in each charge-discharge cycle in the preset charge-discharge cycle interval, and simultaneously collect the lithium battery in each charge-discharge cycle in the preset charge-discharge cycle interval The preset charge-discharge cycle interval is specifically: the minimum value of the preset charge-discharge cycle interval is one of the 1st to 20th charge-discharge cycles, and the maximum value is one of the 50th and above charge-discharge cycles. In specific implementation, the minimum value is the 10th charge-discharge cycle, and the maximum value is the 200th charge-discharge cycle. In specific implementation, the same constant current charging mode stage is shown in Figure 2, Figure 2 (a) is the voltage change curve in the same constant current charging mode stage, Figure 2 (b) is the current change in the same constant current charging mode stage curve. The cycle life of the lithium battery is obtained from the experiment, and the lithium battery is continuously charged and discharged. When the actual total capacity of the lithium battery decays to 80% of the initial total capacity of the lithium battery, the current number of charge and discharge cycles performed is lithium Cycle life of the battery.

2) According to all the charging voltage and current data collected by the current lithium battery in the preset charge-discharge cycle interval, calculate and obtain the spatial distribution characteristics of the current lithium battery in the preset charge-discharge cycle interval;

Step 2) is specifically:

According to all the charging voltage and current data collected in the preset charging and discharging cycle interval of the current lithium battery, draw a two-dimensional coordinate axis space with the charging voltage as the horizontal axis and the charging current as the vertical axis, and plot the current lithium battery in the preset charging and discharging cycle. The charging voltage and current data in the interval are plotted in the two-dimensional coordinate axis space. According to the range of the charging voltage and current, the horizontal axis and the vertical axis are divided into m equidistant voltage sub-intervals and n equidistant current sub-intervals, respectively, to obtain m. ×n two-dimensional subspaces; count the number of points distributed in each two-dimensional subspace of all the charging voltage and current data accumulated in the current lithium battery in the preset charge-discharge cycle interval and use it as the spatial distribution characteristics of the current lithium battery. Among them, the points falling on the lower boundary of current and voltage belong to the current two-dimensional subspace, and the points falling on the upper boundary of current and voltage do not belong to the current two-dimensional subspace.

3) Repeat steps 1)-2) to collect and calculate each lithium battery, obtain the cycle life and corresponding spatial distribution characteristics of each lithium battery in the preset charge-discharge cycle interval, and form a training set;

When implemented, all lithium batteries are of the same model. As shown in Table 1 and Table 2, the horizontal axis is divided into 19 equidistant voltage subintervals, the vertical axis is divided into 12 equidistant current subintervals, and 228 two-dimensional subspaces are obtained, as shown in Figure 3.

Table 1 Voltage sub-interval division (V) in the constant current charging stage

3.40-3.41 3.41-3.42 3.42-3.43 3.43-3.44 3.44-3.45 3.45-3.46 3.46-3.47 3.47-3.48 3.48-3.49 3.49-3.50 3.50-3.51 3.51-3.52 3.52-3.53 3.53-3.54 3.54-3.55 3.55-3.56 3.56-3.57 3.57-3.58 3.58-3.59

Table 2 Current sub-interval division in constant current charging stage (A)

0.9970-0.9975 0.9975-0.9980 0.9980-0.9985 0.9985-0.9990 0.9990-0.9995 0.9995-1.0000 1.0000-1.0005 1.0005-1.0010 1.0010-1.0015 1.0015-1.0020 1.0020-1.0025 1.0025-1.0030

4) Train the lithium battery life prediction regression model based on the training set, and obtain the lithium battery life prediction regression model after training; the lithium battery life prediction regression model selects the machine learning regression model. In the specific implementation, the XGBoost model is used.

5) During online prediction, collect the charging voltage and current data in the same constant current charging mode stage in the preset charge-discharge cycle interval of the brand-new to-be-predicted lithium battery, and calculate and obtain the space of the to-be-predicted lithium battery in the preset charge-discharge cycle interval. Distribution features, input the obtained spatial distribution features into the trained lithium battery life prediction regression model for prediction, and output the current cycle life of the lithium battery to be predicted. The deviation between the life expectancy of the lithium battery predicted based on the spatial distribution characteristics of the charging voltage and current data and the life of the actual lithium battery is shown in Fig. 4 .

The number of preset charge-discharge cycle intervals in step 5) is the same as the number of different preset charge-discharge cycle intervals in step 1), and the same constant current charging mode stage in step 5) is the same as that in step 1).

Claims (3)

1. A lithium battery online service life prediction method based on charging data spatial distribution characteristics is characterized by comprising the following steps:

1) collecting charging voltage and current data of a brand new lithium battery at the same constant current charging mode stage in a preset charging and discharging cycle interval, and simultaneously collecting the cycle life of the lithium battery in the preset charging and discharging cycle interval;

2) calculating and obtaining the spatial distribution characteristics of the current lithium battery in a preset charge-discharge cycle interval according to all charge voltage and current data collected by the current lithium battery in the preset charge-discharge cycle interval;

3) repeating the steps 1) -2) to collect and calculate each lithium battery, obtaining the cycle life and the corresponding spatial distribution characteristics of each lithium battery in a preset charge-discharge cycle interval, and forming a training set;

4) training the lithium battery life prediction regression model based on a training set to obtain a trained lithium battery life prediction regression model;

5) during online prediction, collecting charging voltage and current data of a lithium battery to be predicted at the same constant-current charging mode stage in a preset charging and discharging cycle interval, calculating to obtain spatial distribution characteristics of the lithium battery to be predicted in the preset charging and discharging cycle interval, inputting the obtained spatial distribution characteristics into a trained lithium battery life prediction regression model for prediction, and outputting the current cycle life of the lithium battery to be predicted;

the step 2) is specifically as follows:

according to charging voltage and current data of the current lithium battery in a preset charging and discharging cycle interval, drawing a two-dimensional coordinate axis space by taking the charging voltage as a horizontal axis and the charging current as a vertical axis, drawing all the charging voltage and current data of the current lithium battery in the preset charging and discharging cycle interval in the two-dimensional coordinate axis space, and dividing the horizontal axis and the vertical axis into m equidistant voltage sub-intervals and n equidistant current sub-intervals according to the ranges of the charging voltage and the current to obtain m multiplied by n two-dimensional sub-spaces; counting the number of points of all charging voltage and current data accumulated in a preset charging and discharging cycle interval of the current lithium battery in each two-dimensional subspace, and taking the points as the spatial distribution characteristics of the current lithium battery.

2. The method for predicting the online service life of the lithium battery based on the spatial distribution characteristics of the charging data as claimed in claim 1, wherein the preset charging and discharging cycle interval in the step 5) is the same as the preset charging and discharging cycle interval in the step 1), and the same constant current charging mode stage in the step 5) is the same as the same constant current charging mode stage in the step 1).

3. The lithium battery online life prediction method based on the charge data spatial distribution characteristics as claimed in claim 1, wherein the lithium battery life prediction regression model is a machine learning regression model.

Images