CN111965559B - On-line estimation method for SOH of lithium ion battery - Google Patents

Abstract

The invention discloses an on-line estimation method for SOH of a lithium ion battery, which comprises the following steps: step 1, collecting voltage, current, time and capacity of an electric automobile battery in the constant voltage charging process and current, time and capacity of the electric automobile battery in the constant current discharging process in an off-line manner; step 2, extracting constant voltage charging time adopted in the step 1 as a health index; step 3, preprocessing health index data by adopting a Laida criterion; step 4, establishing a mapping relation between the health index and the capacity by adopting a BP neural network algorithm; and 5, acquiring on-line signals, extracting health indexes, estimating the capacity by adopting a BP neural network algorithm, and calculating the SOH value of the battery. The invention solves the problems that the existing method for estimating the SOH of the battery of the electric automobile takes the capacity or the internal resistance as a health index, the online estimation application cannot be realized, the parameters of the filtering method are acquired in a fixed charge and discharge mode, and the dynamic change characteristic of the electric automobile in the actual working process cannot be fully considered.

Description

A Lithium-ion Battery SOH Online Estimation Method

technical field

The invention belongs to the technical field of lithium batteries and relates to an online SOH estimation method for lithium ion batteries.

Background technique

Due to its high specific energy, high working voltage, wide temperature range, low self-discharge rate, long cycle life and good safety, lithium-ion batteries are widely used in electric vehicles and various fields, and gradually become the key and key to many important fields in the future. support. However, the performance degradation of lithium-ion batteries during use will affect the normal use of equipment and even cause serious failures. Usually, the state of health (SOH) of lithium-ion batteries refers to the ratio of the measured capacity to the rated capacity. Lithium-ion batteries SOH estimation can optimize the battery usage strategy and prolong the battery performance attenuation cycle; in addition, it can replace the bad battery in time, reduce the cost of battery use, and improve the life of the electric vehicle. Therefore, the SOH estimation of lithium-ion batteries is particularly important.

Internal resistance, capacity, charge-discharge behavior, and cycle times are the objects of research on SOH estimation of lithium-ion batteries, but these studies are usually used for offline SOH estimation method verification. Some scholars propose to extract the health indicator (HI) from the charge-discharge behavior and cycle times, and establish a mapping relationship between HI and SOH for real-time online SOH estimation. However, the charge-discharge behavior and cycle times of electric vehicles vary in actual operation, which is not suitable for online SOH estimation of lithium-ion batteries for electric vehicles.

The discharge process of electric vehicles is random, but the charging process of electric vehicles is regular. Electric vehicle charging is generally divided into three stages: constant current charging, constant voltage charging and floating charging. The characteristics of short constant current charging time are not obvious enough, and The charging starting points of its electric vehicles are different, and the monitoring information has large errors. There is almost no change in the floating charging characteristics, so only the constant voltage charging stage information is completely preserved, the initial state is the same as the terminal state, and is not affected by the previous incomplete discharge process.

In the existing electric vehicle battery SOH estimation methods, most of the methods that use capacity or internal resistance as the health index cannot be used for online estimation, and the parameter acquisition of the filtering method requires a fixed charging and discharging mode, which cannot fully consider the dynamic changes in the actual working process of electric vehicles. problem of characteristics.

Contents of the invention

The purpose of the present invention is to provide a lithium-ion battery SOH online estimation method, which solves the problem that the existing electric vehicle battery SOH estimation method uses capacity or internal resistance as the health index, cannot realize online estimation application, and the acquisition of filtering method parameters requires a fixed charging method. The discharge mode cannot fully consider the dynamic changing characteristics of the electric vehicle in the actual working process.

The technical scheme adopted in the present invention is an online estimation method for lithium-ion battery SOH, which is specifically implemented according to the following steps:

Step 1, offline collection of the voltage, current, time during the constant voltage charging process of the electric vehicle battery and the current, time and capacity during the constant current discharge process;

Step 2, extracting the constant voltage charging time taken in step 1 as a health indicator;

Step 3, using the Raida criterion to preprocess the health indicator data;

Step 4, using BP neural network algorithm to establish the mapping relationship between health indicators and capacity;

Step 5, collect online signals, extract health indicators, use BP neural network algorithm to estimate capacity, and calculate battery SOH value.

The present invention is also characterized in that,

Step 1 is specifically:

Use voltage and current sensors and time timers to collect voltage, current and time during constant voltage charging of lithium-ion batteries for electric vehicles;

After the electric vehicle is fully charged, the current sensor and time timer are used to collect the current and time during the constant current discharge process of the lithium ion battery of the electric vehicle, and the battery capacity is obtained by multiplying the current and time during the constant current discharge process.

Step 2 is specifically:

Calculate the constant voltage charging time health index time series, where the corresponding constant voltage discharge time series in the i-th cycle, that is, the i-th cycle health index HI _i is:

HI _i ＝t _CVCend (i)-t _CVCstart (i), i＝1,2,3,... (1)

In the formula, _tCVCstart and _tCVCend are the start and end times of constant voltage charging in the i-th cycle, respectively.

Step 3 is specifically:

Step 3.1, first calculate the standard deviation σ of the health index data, the calculation formula is

是健康指标的算术平均值；In the formula, n is the total number of health indicators,

is the arithmetic mean of health indicators;

In step 3.2, according to the Raida criterion, the health index data is judged and corrected. The Raida criterion is based on the following:

是健康指标估计值；In the formula, R _n is the residual error,

is the estimated value of the health indicator;

If the residual is greater than 3σ, the health index data is gross error and should be discarded or compensated; if the residual is less than or equal to 3σ, the health index data is normal data and should be retained.

Step 4 is specifically:

Step 4.1, use the Pearson Pearson correlation coefficient to analyze the correlation between the health index data and the capacity, and calculate the correlation coefficient between the health index data and the capacity. If the correlation coefficient is greater than 0.6, there is a strong correlation, and the health index time series is valid;

Step 4.2, use the time series of health indicators proved effective in step 4.1 as the input of the BP neural network algorithm, the lithium-ion battery capacity data is the output of the BP neural network algorithm, establish the relationship between the health indicators and the lithium-ion battery capacity, and train to obtain network parameters ω _ij , ω _jk , a _j and b _k , specifically:

In the formula, C _k is the battery capacity sequence, i is the number of nodes in the input layer, j is the number of nodes in the hidden layer, k is the number of nodes in the output layer, and the number of nodes in the input layer and the output layer are equal, so i=k , ω _ij is the connection weight between the input layer and the hidden layer, a _j is the connection threshold between the input layer and the hidden layer, ω _jk is the connection weight between the input layer and the hidden layer, b _k is the input layer and the hidden layer Connection threshold.

In step 4.1, the Pearson Pearson correlation coefficient is used to analyze the correlation between the health index data and the capacity, and the correlation coefficient between the health index data and the capacity is calculated. The calculation formula of the correlation coefficient r is:

In the formula, N is the total number of charge and discharge times, C _i is the battery capacity of the i-th cycle, and HI _i is the health index of the i-th cycle.

The specific process of step 5 is:

Step 5.1, using voltage and current sensors and a time timer online to collect the charging voltage, current and time during the constant voltage charging of the electric vehicle;

Step 5.2, according to the method of step 2, extract the constant voltage charging time in step 5.1 as the health index, and obtain the time series HI′ _i of the current health index;

In step 5.3, input the time series HI′ _i of the current health index obtained in step 5.2 into the relationship between the health index and lithium-ion battery capacity established in step 4.2, and perform capacity estimation to obtain the current battery capacity sequence C′ _k :

Step 5.4, calculate the SOH of the current battery through the SOH definition formula:

C _N is the nominal capacity of the battery, and C' _k is the current battery capacity sequence.

The beneficial effect of the present invention is

The present invention can effectively prevent the filtering method from being affected by the number of battery cycles and the fixed charging and discharging mode, and solve the problem that most of the existing electric vehicle battery SOH estimation methods that use capacity or internal resistance as health indicators cannot realize online estimation applications, and the parameters of the filtering method Acquisition requires a fixed charging and discharging mode, which cannot fully consider the dynamic changing characteristics of electric vehicles in the actual working process. The present invention proposes an online lithium-ion battery SOH estimation method suitable for the actual working conditions of electric vehicles. First, the measurable parameters of the actual charging process of electric vehicles are used to propose constant-voltage charging time as a health indicator, and the Raida criterion method is used to carry out Health index preprocessing; secondly, the Pearson correlation coefficient health index and battery capacity analysis, and the BP neural network algorithm is used to establish the mapping relationship between the proposed health index and the actual battery capacity; finally, the BP neural network algorithm is used for SOH estimation.

Description of drawings

Fig. 1 is the overall flowchart of a kind of lithium-ion battery SOH online estimation method of the present invention;

Fig. 2 is a lithium-ion battery SOH online estimation method of the present invention, a lithium-ion battery data acquisition diagram under the line;

Fig. 3 is a kind of lithium ion battery SOH online estimation method in the present invention, online lithium ion battery health index HI extracting method figure;

Fig. 4 is a flow chart of processing abnormal points of the health index HI in a lithium-ion battery SOH online estimation method of the present invention;

Fig. 5 is a health index HI and lithium ion battery capacity mapping model figure in a kind of lithium ion battery SOH online estimation method of the present invention;

Fig. 6 is an online lithium-ion battery SOH estimation model diagram in a lithium-ion battery SOH online estimation method of the present invention;

Fig. 7 is a comparison diagram of the abnormal point unprocessed health index HI and the lithium-ion battery in a lithium-ion battery SOH online estimation method of the present invention;

Fig. 8 is a comparison diagram of the health index HI and the lithium ion battery after abnormal point processing in the online estimation method of lithium ion battery SOH of the present invention;

FIG. 9 is a simulation diagram of online lithium-ion battery SOH estimation in an online lithium-ion battery SOH estimation method according to the present invention.

Detailed ways

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

A kind of lithium-ion battery SOH online estimation method of the present invention, its flow process is as shown in Figure 1, specifically implements according to the following steps:

Step 1, offline collection of the voltage, current, time during the constant voltage charging process of the electric vehicle battery, and the current, time and capacity during the constant current discharge process; as shown in Figure 2, specifically:

Use voltage and current sensors and time timers to collect voltage, current and time during constant voltage charging of lithium-ion batteries for electric vehicles;

After the electric vehicle is fully charged, the current sensor and time timer are used to collect the current and time during the constant current discharge process of the lithium ion battery of the electric vehicle, and the battery capacity is obtained by multiplying the current and time during the constant current discharge process.

Step 2, extract the constant voltage charging time taken in step 1 as a health indicator; as shown in Figure 3, specifically:

Calculate the constant voltage charging time health index time series, where the corresponding constant voltage discharge time series in the i-th cycle, that is, the i-th cycle health index HI _i is:

HI _i ＝t _CVCend (i)-t _CVCstart (i), i＝1,2,3,... (1)

In the formula, _tCVCstart and _tCVCend are the start and end times of constant voltage charging in the i-th cycle, respectively.

Step 3, using the Raida criterion to preprocess the health indicator data; as shown in Figure 4, specifically:

Step 3.1, first calculate the standard deviation σ of the health index data, the calculation formula is

是健康指标的算术平均值；In the formula, n is the total number of health indicators, generally not less than 20 times,

is the arithmetic mean of health indicators;

In step 3.2, according to the Raida criterion, the health index data is judged and corrected. The Raida criterion is based on the following:

是健康指标估计值；In the formula, R _n is the residual error,

is the estimated value of the health indicator;

If the residual is greater than 3σ, the health index data is gross error and should be discarded or compensated; if the residual is less than or equal to 3σ, the health index data is normal data and should be retained.

Step 4, use the BP neural network algorithm to establish the mapping relationship between health indicators and capacity; as shown in Figure 5, specifically:

Step 4.1, use the Pearson Pearson correlation coefficient to analyze the correlation between the health index data and the capacity, and calculate the correlation coefficient between the health index data and the capacity. If the correlation coefficient is greater than 0.6, there is a strong correlation, and the health index time series is valid; Among them, the Pearson Pearson correlation coefficient is used to analyze the correlation between the health index data and the capacity, and the correlation coefficient between the health index data and the capacity is calculated. The calculation formula of the correlation coefficient r is:

In the formula, N is the total number of charging and discharging, C _i is the battery capacity of the i-th cycle, and HI _i is the health index of the i-th cycle;

Step 4.2, use the time series of health indicators proved effective in step 4.1 as the input of the BP neural network algorithm, and the lithium-ion battery capacity data is the output of the BP neural network algorithm, establish the relationship between the health indicators and the lithium-ion battery capacity, and train to obtain network parameters ω _ij , ω _jk , a _j and b _k , specifically:

In the formula, C _k is the battery capacity sequence, i is the number of nodes in the input layer, j is the number of nodes in the hidden layer, k is the number of nodes in the output layer, and the number of nodes in the input layer and the output layer are equal, so i=k , ω _ij is the connection weight between the input layer and the hidden layer, a _j is the connection threshold between the input layer and the hidden layer, ω _jk is the connection weight between the input layer and the hidden layer, b _k is the input layer and the hidden layer Connection threshold.

Step 5, collect online signals, extract health indicators, use BP neural network algorithm to estimate capacity, and calculate battery SOH value, as shown in Figure 6, the specific process is as follows:

Step 5.1, using voltage and current sensors and a time timer online to collect the charging voltage, current and time during the constant voltage charging of the electric vehicle;

Step 5.2, according to the method of step 2, extract the constant voltage charging time in step 5.1 as the health index, and obtain the time series HI′ _i of the current health index;

In step 5.3, input the time series HI′ _i of the current health index obtained in step 5.2 into the relationship between the health index and lithium-ion battery capacity established in step 4.2, and perform capacity estimation to obtain the current battery capacity sequence C′ _k :

Step 5.4, calculate the SOH of the current battery through the SOH definition formula:

C _N is the nominal capacity of the battery, and C' _k is the current battery capacity sequence.

Example

The data set used in the present invention comes from the lithium-ion battery test bed built by the NASA PCoE Research Center. The battery is a commercially available 18650Li-ion battery with a rated capacity of 2Ah. Lithium-ion cells (B0005, B0006 and B0007) were tested at room temperature through 3 different operating characteristics (charge, discharge and impedance). Charge in 1.5A constant current (CC) mode until the battery voltage reaches 4.2V, then charge in constant voltage (CV) mode until the charge current drops to 20mA.

The constant voltage charging voltage, current and time, and Li-ion battery capacity are extracted from the data. Figure 7 shows the health indicator HI and capacity comparison chart. It can be seen that there will be errors in the actual extraction process and measurement, and the error points need to be corrected. Therefore, this paper uses the Raida criterion method to correct the data. After correction, as shown in Figure 8, the error points have been corrected, and the constant voltage charging There is a strong negative correlation between the time health index and the battery capacity, and the correlation is shown in Table 1.

Table 1 Correlation between health indicator HI and battery capacity

Lithium Ion Battery Before outlier correction After outlier correction B0005 -0.8551 -0.9803 B0006 -0.8634 -0.9474 B0007 -0.8380 -0.9632

The BP neural network algorithm is used to establish the mapping between the three battery health indicators HI and the battery capacity, and the online lithium-ion battery SOH estimation verification is carried out using B0005 as an example. The results are shown in Figure 9. The maximum absolute error is 0.0749, the minimum absolute error is 3.2730e-07, the average absolute error is 0.0180, and the root mean square error is 0.0252. The overall SOH estimation error of the invention is less than 2%, and is applicable to the actual working conditions of electric vehicles.

Claims (1)

1. The on-line estimation method for the SOH of the lithium ion battery is characterized by comprising the following steps of:

step 1, collecting voltage, current, time and capacity of an electric automobile battery in the constant voltage charging process and current, time and capacity of the electric automobile battery in the constant current discharging process in an off-line manner;

the step 1 specifically comprises the following steps:

collecting voltage, current and time in the constant voltage charging process of the lithium ion battery of the electric automobile by adopting a voltage and current sensor and a time timer;

after the electric automobile is fully charged, acquiring current and time in the constant-current discharging process of the lithium ion battery of the electric automobile by adopting a current sensor and a time meter, and multiplying the current and time in the constant-current discharging process to obtain the battery capacity;

step 2, extracting constant voltage charging time adopted in the step 1 as a health index;

the step 2 specifically comprises the following steps:

calculating constant voltage charging time health index time sequence, wherein the corresponding constant voltage discharging time sequence in the ith period is i.e. the health index HI of the ith period _i The method comprises the following steps:

HI _i ＝t _CVCend (i)-t _CVCstart (i),i＝1,2,3,...(1)

wherein t is _CVCstart And t _CVCend Respectively the start time and the stop time of constant voltage charging in the ith period;

step 3, preprocessing health index data by adopting a Laida criterion;

the step 3 specifically comprises the following steps:

step 3.1, firstly, calculating the standard deviation sigma of the health index data, wherein the calculation formula is as follows

Wherein n is the total number of times of the health index,

is the arithmetic mean of the health index;

and 3.2, judging and correcting the health index data according to the Laida criterion, wherein the Laida criterion is judged according to the following steps:

wherein R is _n Is the residual error that is present in the sample,

is a health index estimated value;

if the residual error is larger than 3 sigma, the health index data is coarse error and should be discarded or compensated, and if the residual error is smaller than or equal to 3 sigma, the health index data is normal data and is reserved;

step 4, establishing a mapping relation between the health index and the capacity by adopting a BP neural network algorithm;

the step 4 specifically comprises the following steps:

step 4.1, carrying out correlation analysis between the health index data and the capacity by adopting the pearson correlation coefficient, calculating the correlation coefficient of the health index data and the capacity, and if the correlation coefficient is larger than 0.6, carrying out strong correlation, and enabling the health index time sequence to be effective;

step 4.2, taking the effective health index time sequence proved in step 4.1 as the input of the BP neural network algorithm, taking the lithium ion battery capacity data as the output of the BP neural network algorithm, establishing the relation between the health index and the lithium ion battery capacity, and training to obtain the network parameter omega _ij 、ω _jk 、a _j And b _k The method specifically comprises the following steps:

wherein C is _k For the battery capacity sequence, i is the number of nodes of an input layer, j is the number of nodes of an hidden layer, and k is the number of nodes of an output layer, wherein the number of the nodes of the input layer is equal to the number of the nodes of the output layer, so i=k, omega _ij For connecting weights between input layer and hidden layer, a _j For input layer and hidden layer connection threshold, ω _jk B, connecting weights for the input layer and the hidden layer _k Connecting thresholds for the input layer and the hidden layer;

in the step 4.1, the pearson correlation coefficient is adopted to perform correlation analysis between the health index data and the capacity, and the correlation coefficient between the health index data and the capacity is calculated, wherein the calculation formula of the correlation coefficient r is as follows:

wherein N is the total charge and discharge times, C _i HI for the i-th cycle battery capacity _i Is the ith periodic health index;

step 5, collecting on-line signals, extracting health indexes, estimating capacity by adopting a BP neural network algorithm, and calculating a battery SOH value;

the specific process of the step 5 is as follows:

step 5.1, collecting charging voltage, current and time in the constant voltage charging process of the electric automobile by adopting a voltage and current sensor and a time meter on line;

step 5.2, extracting the constant voltage charging time in step 5.1 as a health index according to the method of step 2 to obtain a time sequence HI 'of the current health index' _i ；

Step 5.3, time series HI 'of the current health index obtained in the step 5.2' _i Inputting the current battery capacity sequence C 'obtained by performing capacity estimation in the relation between the health index and the lithium ion battery capacity established in the step 4.2' _k ：

Step 5.4, calculating the SOH of the current battery according to the SOH definition formula:

C _N is the nominal capacity of the battery, C' _k Is the current battery capacity sequence.

Images