Retired battery health degree prediction method based on multi-factor coupling

By acquiring and integrating the environmental, electrical and operating characteristics of retired lithium batteries and inputting machine learning models, the problem of insufficient accuracy and robustness of lithium batteries in the existing technology is solved, and more accurate and robust health prediction is achieved. .

CN118962461APending Publication Date: 2024-11-15SHENZHEN JIECHENG NICKEL COBALT NEW ENERGY TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411085271.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-08
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing health status evaluation methods of retired lithium batteries fail to make full use of the multi-dimensional information during the battery charging and discharging process, resulting in limited evaluation accuracy and generalization capabilities and poor robustness.

Method used

Using a multi-factor coupling-based health prediction method for retired batteries, the characteristics with the greatest correlation with DC internal resistance are determined by acquiring environmental, electrical and operating characteristics, and the characteristics of the fusion are fused, and the fusion characteristics are input into the machine learning model to improve the prediction. Accuracy and robustness.

Benefits of technology

Through multi-dimensional feature fusion, the health status of retired batteries can be more accurately reflected, the error caused by a single feature can be reduced, and the robustness of the prediction model can be enhanced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118962461A_ABST
    Figure CN118962461A_ABST
Patent Text Reader

Abstract

The invention provides a retired battery health degree prediction method based on multi-factor coupling. The retired battery health degree prediction method comprises the following steps: S1, acquiring environment characteristics, electrical characteristics and operation characteristics of a retired battery; s2, determining a feature Pa having the maximum correlation with the DC internal resistance R of the decommissioned battery in the electrical features and a feature Qa having the maximum correlation with the DC internal resistance R of the decommissioned battery in the operation features; s3, obtaining features pa and qa according to the features Pa and Qa; s4, performing feature fusion on all the environment features, the features pa and qa to obtain a fusion feature G; and S5, inputting the fusion feature G into a machine learning model to obtain the predicted SOH of the retired battery. According to the method, the multi-dimensional features can be fused, so that the accuracy of predicting the health degree of the retired battery is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of retired battery health, and in particular to a retired battery health prediction method based on multi-factor coupling. Background Art

[0002] With the popularity of electric vehicles and mobile devices, a large number of lithium-ion batteries enter the retirement stage after use. These retired batteries still have high capacity and can be used for cascade utilization, but accurate assessment of their state of health (SOH) is the key to achieving cascade utilization.

[0003] At present, the evaluation of the health status of retired lithium batteries mainly relies on model methods, data-driven methods and hybrid methods. However, existing evaluation methods often fail to fully utilize the multi-dimensional information of the battery charging and discharging process, resulting in limited evaluation accuracy and generalization ability, and poor robustness.

[0004] Therefore, without solving the above technical problems, it is urgent to propose a new technical means. Summary of the invention

[0005] In order to improve the robustness of the prediction model, the present invention proposes a retired battery health prediction method based on multi-factor coupling.

[0006] The present invention provides a method for predicting the health of retired batteries based on multi-factor coupling, comprising the following steps:

[0007] S1. Obtain the environmental characteristics, electrical characteristics and operating characteristics of retired batteries;

[0008] S2. Determine the electrical characteristics P that has the greatest correlation with the DC internal resistance R of the retired battery a The characteristic Q that is most correlated with the DC internal resistance R of the retired battery among the operating characteristics a ;

[0009] S3. According to feature P a and Q a Get feature p a and q a ;

[0010] S4. For all environmental features, features p a and q a Perform feature fusion to obtain fused feature G;

[0011] S5. Input the fused feature G into the machine learning model to obtain the predicted state of health SOH of the retired battery.

[0012] Furthermore, the environmental characteristics include: temperature T and humidity S of the scene where the retired battery was used before retirement.

[0013] Furthermore, the fusion feature G is obtained by the following formula:

[0014]

[0015] α+β+γ+δ=1

[0016] Among them, α, β, γ and δ represent weight parameters, R represents the DC internal resistance of the retired battery, COV(R,T) represents the covariance of the DC resistance R of the retired battery and the temperature T of the scene where the retired battery was used before retirement, σ R represents the standard deviation of the DC resistance R of retired batteries, σ T represents the standard deviation of the temperature T of the target scenario where the retired battery will be used, and COV(R,S) represents the covariance of the DC resistance R of the retired battery and the humidity S of the scenario where the retired battery was used before retirement.

[0017] Further, the electrical characteristics include: a current I of the retired battery when it is charged and discharged before retirement, a voltage U of the retired battery when it is charged and discharged before retirement, and a charge and discharge depth D of the retired battery when it is charged and discharged before retirement.

[0018] Furthermore, when the feature P with the largest correlation a When the current I of the retired battery during charging and discharging before retirement is I, the characteristic p is obtained by the following method a :

[0019] Obtain the current value I of the retired battery when it is charged and discharged before retirement, and determine the number of charge and discharge times i below the preset current range and above the preset current range a , according to the number i a Calculate feature p a :

[0020]

[0021] Wherein, M represents the total number of charge and discharge times of the retired battery before retirement.

[0022] Furthermore, when the feature P with the largest correlation a When U is the voltage of the retired battery when it is charged and discharged before retirement, the characteristic p is obtained by the following method a :

[0023] Obtain the voltage value U of the retired battery when it is charged and discharged before retirement, and determine the number of charge and discharge times u below the preset voltage range and above the preset voltage range a , according to the number u a Calculate feature p a :

[0024]

[0025] Wherein, M represents the total number of charge and discharge times of the retired battery before retirement.

[0026] Furthermore, when the feature P with the largest correlation a When the charge and discharge depth D of the retired battery before retirement is charged and discharged, the characteristic p is obtained by the following method a :

[0027] Obtain the charge and discharge depth D of the retired battery before retirement, and determine the number of discharges d that is higher than the preset charge and discharge depth range a , according to the number of discharges d a Calculate feature p a :

[0028]

[0029] Among them, μ represents the impact factor of full discharge, and L represents the total number of discharges of the retired battery before retirement.

[0030] Furthermore, the operating characteristics include: the total number of times K of overcharging and overdischarging of the retired battery and the number of cycles H of the retired battery.

[0031] Furthermore, when the most relevant feature Q a When K is the total number of overcharge and overdischarge of retired batteries, the characteristic q is obtained by the following method a :

[0032] K=K 1 +K 2

[0033]

[0034] Among them, K 1 Indicates the number of overcharges, K 2 Indicates the number of over-discharges, K 21 It indicates the number of full discharges during over-discharge, μ indicates the influence factor of full discharge, and N indicates the total number of charge and discharge times of the preset retired battery.

[0035] Furthermore, when the most relevant feature Q a When H is the cycle number of the retired battery, the characteristic q is obtained by the following method a :

[0036]

[0037] Wherein, N represents the total number of charge and discharge times of the preset retired battery.

[0038] Beneficial effects of the present invention: The present invention can more comprehensively reflect the status of retired batteries by integrating features of multiple dimensions, help reduce the errors caused by a single feature, more accurately predict the health status of retired batteries, and enhance the robustness of the prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0040] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0041] The present invention is further described below in conjunction with the accompanying drawings:

[0042] The present invention provides a method for predicting the health of retired batteries based on multi-factor coupling, comprising the following steps:

[0043] S1. Obtain the environmental characteristics, electrical characteristics and operating characteristics of retired batteries;

[0044] S2. Determine the electrical characteristics P that has the greatest correlation with the DC internal resistance R of the retired battery a The characteristic Q that is most correlated with the DC internal resistance R of the retired battery among the operating characteristics a ;

[0045] S3. According to feature P a and Q a Get feature p a and q a ;

[0046] S4. For all environmental features, features p a and q a Perform feature fusion to obtain fused feature G;

[0047] S5. Input the fused feature G into the machine learning model to obtain the predicted state of health SOH of the retired battery. Through the above method, features with higher correlation can be fused to better and more comprehensively reflect the health status of the retired battery.

[0048] In this embodiment, in step S1, environmental characteristics, electrical characteristics, and operating characteristics of the retired battery are obtained.

[0049] Among them, the environmental characteristics include: the temperature T and humidity S of the scene where the retired battery was used before retirement.

[0050] The electrical characteristics include: the current I of the retired battery when it is charged and discharged before retirement, the voltage U of the retired battery when it is charged and discharged before retirement, and the charge and discharge depth D of the retired battery when it is charged and discharged before retirement.

[0051] The operating characteristics include: the total number of overcharge and overdischarge of the retired battery K and the number of cycles H of the retired battery.

[0052] The above features are all factors that will affect the health of the battery. Obtaining the above factors can provide a basis for subsequent feature fusion.

[0053] In this embodiment, in step S2, the feature P that has the greatest correlation with the DC internal resistance R of the retired battery among the electrical features is determined. a The characteristic Q that is most correlated with the DC internal resistance R of the retired battery among the operating characteristics a ;

[0054] If all environmental characteristics, electrical characteristics and operating characteristics are directly used for fusion, the amount of calculation and complexity will increase, and the three characteristics of the electrical characteristics have a correlation in their effects on the health. Only one or two characteristics can characterize the effect of the electrical characteristics on the battery health. Under the premise of ensuring the accuracy of the prediction, in order to reduce the complexity of the calculation, this application selects one electrical feature for feature fusion. The selection principle of the operating characteristics is the same as that of the electrical characteristics. The feature with the greatest correlation with the DC internal resistance R among the electrical characteristics and the operating characteristics is selected for feature fusion because the DC internal resistance is one of the key factors reflecting the health of retired batteries, and the health of retired batteries can be directly evaluated based on the DC internal resistance.

[0055] There are many methods for determining the correlation between electrical characteristics and operating characteristics and DC internal resistance, including but not limited to: Pearson correlation coefficient analysis, Spearman rank correlation coefficient analysis and canonical correlation analysis. Any correlation analysis method suitable for characteristic parameters is selected for analysis, and the correlation determination process is not described in detail here.

[0056] In this embodiment, in step S3, according to the feature P with the greatest correlation a and Q a Get feature p a and q a .

[0057] When the feature P with the largest correlation a When the current I of the retired battery during charging and discharging before retirement is I, the characteristic p is obtained by the following method a :

[0058] Obtain the current value I of the retired battery when it is charged and discharged before retirement, and determine the number of charge and discharge times i below the preset current range and above the preset current range a , according to the number i a Calculate feature p a :

[0059]

[0060] Among them, M represents the total number of charge and discharge times of the retired battery before retirement. In the above, when the charge and discharge current is lower than or higher than the preset current range, it will have serious consequences on the health of the battery. The preset current range is determined according to the battery model. Charging and discharging within the preset current range has little effect on the health of the retired battery.

[0061] When the feature P with the largest correlation a When U is the voltage of the retired battery when it is charged and discharged before retirement, the characteristic p is obtained by the following method a :

[0062] Obtain the voltage value U of the retired battery when it is charged and discharged before retirement, and determine the number of charge and discharge times u below the preset voltage range and above the preset voltage range a , according to the number u a Calculate feature p a :

[0063]

[0064] Among them, M represents the total number of charge and discharge times of the retired battery before retirement. In the above, when the charge and discharge voltage value U is lower than or exceeds the preset voltage range, it will have serious consequences on the health of the battery. The preset voltage range is determined according to the battery model. Charging and discharging within the preset voltage range has little effect on the health of the retired battery.

[0065] When the feature P with the largest correlation a When the charge and discharge depth D of the retired battery before retirement is charged and discharged, the characteristic p is obtained by the following method a :

[0066] Obtain the charge and discharge depth D of the retired battery before retirement, and determine the number of discharges d that is higher than the preset charge and discharge depth range a , according to the number of discharges d a Calculate feature p a :

[0067]

[0068] Among them, μ represents the impact factor of full discharge, L represents the total number of discharges of the retired battery before retirement. In the above, when the charge and discharge depth D is higher than the preset charge and discharge depth, it will cause serious damage to the battery health. Under normal circumstances, the preset charge and discharge depth is 1-80%. The charge and discharge depth exceeding 80% will accelerate battery aging, especially 100% discharge (full discharge) of the battery, which may also cause observable capacity loss, which will seriously affect the battery health.

[0069] When the most relevant feature Q a When K is the total number of overcharge and overdischarge of retired batteries, the characteristic q is obtained by the following method a :

[0070] K=K 1 +K 2

[0071]

[0072] Among them, K 1 Indicates the number of overcharges, K 2 Indicates the number of over-discharges, K 21 represents the number of complete discharges during overdischarge, μ represents the influencing factor of complete discharge, N represents the total number of charge and discharge times of the preset retired battery, N is estimated based on the capacity, conventional charge and discharge current and voltage of the retired battery when it leaves the factory, and the estimation method is in the prior art and will not be elaborated here; in the above, overcharge mainly damages the battery by causing thermal runaway and chemical reactions, while overdischarge causes a sharp decay of battery capacity through physical and chemical changes; the more overcharge and overdischarge times, the lower the health of the battery and the shorter its lifespan; especially complete discharge during overdischarge will deepen the decay of battery capacity.

[0073] When the most relevant feature Q a When H is the cycle number of the retired battery, the characteristic q is obtained by the following method a :

[0074]

[0075] Wherein, N represents the total number of charge and discharge times of the preset retired battery. In the above, as the number of charge and discharge times increases, the chemical reaction inside the battery and the aging of the materials will cause the battery performance to gradually decline, thereby affecting its health status.

[0076] In this embodiment, in step S4, all environmental features and features p a and q a Perform feature fusion to obtain fused feature G;

[0077] The fusion feature G is obtained by the following formula:

[0078]

[0079] α+β+γ+δ=1

[0080] Among them, α, β, γ and δ represent weight parameters, R represents the DC internal resistance of the retired battery, COV(R,T) represents the covariance of the DC resistance R of the retired battery and the temperature T of the scene where the retired battery was used before retirement, σ Rrepresents the standard deviation of the DC resistance R of retired batteries, σ T The standard deviation of the temperature T of the target scene where the retired battery will be used is represented by COV(R,S), which represents the covariance of the DC resistance R of the retired battery and the humidity S of the scene where the retired battery was used before retirement. Through the above formula, multi-dimensional features are integrated together to increase the robustness of the model and avoid the errors caused by using a single feature for prediction.

[0081] In this embodiment, in step S5, the fusion feature G is input into the machine learning model to obtain the predicted state of health SOH of the retired battery;

[0082] The machine learning model uses models that can perform numerical predictions, including but not limited to: multi-layer perceptron, neural network, and decision tree;

[0083] The fusion feature G of the retired battery with known health is input into the machine learning model for training, and the fusion feature of the retired battery whose health is to be predicted is input into the trained machine learning model to obtain the predicted health of the retired battery. The machine learning model has strong adaptability, can quickly output a large amount of data and make efficient predictions, and can also capture the complex relationship between the data. Using fusion features for prediction can improve the robustness of the model.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A retired battery health prediction method based on multi-factor coupling, characterized by: The following steps are involved: S1. Obtain the environmental characteristics, electrical characteristics and operating characteristics of retired batteries; S2. Determine the electrical characteristics P that has the greatest correlation with the DC internal resistance R of the retired battery a The characteristic Q that is most correlated with the DC internal resistance R of the retired battery among the operating characteristics a ; S3. According to feature P a and Q a Get feature p a and q a ; S4. For all environmental features, features p a and q a Perform feature fusion to obtain fused feature G; S5. Input the fused feature G into the machine learning model to obtain the predicted state of health SOH of the retired battery.

2. The retired battery health prediction method based on multi-factor coupling according to claim 1 is characterized in that: Environmental characteristics include: temperature T and humidity S of the scene where the retired battery was used before retirement.

3. The retired battery health prediction method based on multi-factor coupling according to claim 2 is characterized in that: The fusion feature G is obtained by the following formula: α+β+γ+δ=1 Among them, α, β, γ and δ represent weight parameters, R represents the DC internal resistance of the retired battery, COV(R,T) represents the covariance of the DC resistance R of the retired battery and the temperature T of the scene where the retired battery was used before retirement, σ R represents the standard deviation of the DC resistance R of retired batteries, σ T represents the standard deviation of the temperature T of the target scenario where the retired battery will be used, and COV(R,S) represents the covariance of the DC resistance R of the retired battery and the humidity S of the scenario where the retired battery was used before retirement.

4. The retired battery health prediction method based on multi-factor coupling according to claim 3 is characterized by: The electrical characteristics include: the current I of the retired battery when it is charged and discharged before retirement, the voltage U of the retired battery when it is charged and discharged before retirement, and the charge and discharge depth D of the retired battery when it is charged and discharged before retirement.

5. The retired battery health prediction method based on multi-factor coupling according to claim 4 is characterized in that: When the feature P with the largest correlation a When the current I of the retired battery during charging and discharging before retirement is I, the characteristic p is obtained by the following method a : Obtain the current value I of the retired battery when it is charged and discharged before retirement, and determine the number of charge and discharge times i below the preset current range and above the preset current range a , according to the number i a Calculate feature p a : Wherein, M represents the total number of charge and discharge times of the retired battery before retirement.

6. The retired battery health prediction method based on multi-factor coupling according to claim 4 is characterized in that: When the feature P with the largest correlation a When U is the voltage of the retired battery when it is charged and discharged before retirement, the characteristic p is obtained by the following method a : Obtain the voltage value U of the retired battery when it is charged and discharged before retirement, and determine the number of charge and discharge times u below the preset voltage range and above the preset voltage range a , according to the number u a Calculate feature P a : Wherein, M represents the total number of charge and discharge times of the retired battery before retirement.

7. The retired battery health prediction method based on multi-factor coupling according to claim 4 is characterized in that: When the feature P with the largest correlation a When the charge and discharge depth D of the retired battery before retirement is charged and discharged, the characteristic p is obtained by the following method a : Obtain the charge and discharge depth D of the retired battery before retirement, and determine the number of discharges d that is higher than the preset charge and discharge depth range a , according to the number of discharges d a Calculate feature p a : Among them, μ represents the impact factor of full discharge, and L represents the total number of discharges of the retired battery before retirement.

8. The retired battery health prediction method based on multi-factor coupling according to claim 3 is characterized by: The operating characteristics include: the total number of overcharge and overdischarge of the retired battery K and the number of cycles H of the retired battery.

9. The retired battery health prediction method based on multi-factor coupling according to claim 8 is characterized in that: When the most relevant feature Q a When K is the total number of overcharge and overdischarge of retired batteries, the characteristic q is obtained by the following method a : K=K1+K2 Among them, K1 represents the number of overcharges, K2 represents the number of overdischarges, and K 21 It indicates the number of full discharges during over-discharge, μ indicates the influence factor of full discharge, and N indicates the total number of charge and discharge times of the preset retired battery.

10. The retired battery health prediction method based on multi-factor coupling according to claim 8 is characterized in that: When the most relevant feature Q a When H is the cycle number of the retired battery, the characteristic q is obtained by the following method a : Wherein, N represents the total number of charge and discharge times of the preset retired battery.