A lithium ion battery life prediction method combining physical mechanism and machine learning

Abstract

The application discloses a lithium ion battery life prediction method based on the fusion of physical mechanism and machine learning, comprising the following steps: first charging and discharging cycle of the lithium ion battery and electrochemical characteristic measurement; construction of electrochemical characteristic curve and extraction of characteristic quantity information; construction of machine learning model and input of the characteristic quantity into the machine learning model to evaluate the life performance difference of the lithium ion battery at the time of leaving factory; construction of a physical degradation mechanism model to evaluate the life attenuation of the lithium ion battery caused by the charging and discharging history; and calculation of the residual life of the lithium ion battery based on the performance difference of the lithium ion battery at the time of leaving factory and the life attenuation in the charging and discharging process. The application is based on the fusion of the physical degradation mechanism model and the machine learning model, and can accurately and reliably predict the life of the lithium ion battery only according to the evaluation results of the two main factors of the initial inconsistency of the lithium ion battery at the time of leaving factory and the degradation of the charging and discharging cycle history, and is simple, efficient and reliable.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of lithium ion battery health state management and life evaluation, and particularly relates to a lithium ion battery fast charging life prediction method based on physical mechanism and machine learning fusion. BACKGROUND

[0002] In recent years, with the gradual expansion of the application field of lithium ion batteries, especially the wide application in the field of new energy vehicles, large-scale power stations and the like, the health problems of highly integrated lithium ion battery power supply systems are becoming more and more prominent. Among these problems, the accurate prediction of the remaining life of the lithium ion battery is the most core problem.

[0003] In the prior art, the accurate prediction of the remaining life of the lithium ion battery mainly depends on the detection of characteristic quantities in the entire loading history of the lithium ion battery, and the evolution trend of the measured characteristics is analyzed and modeled to extrapolate and predict the next life degradation trend of the lithium ion battery. However, the rigorous application conditions such as the massive data storage requirements, hardware resource requirements of data analysis, and economic factors of real-time monitoring generated by real-time monitoring and analysis of the lithium ion battery make it difficult to realize effective engineering application of the method.

[0004] In view of the above problems, it is necessary to develop a more simple, fast and high-precision lithium ion battery life prediction method to realize high-precision, fast and accurate lithium battery life evaluation and support health management and optimization design in the lithium ion battery integrated system. SUMMARY

[0005] The present application aims to provide a lithium ion battery life prediction method based on physical mechanism and machine learning fusion to solve the technical problem in the prior art that the storage and analysis method based on real-time monitoring data is not conducive to efficiently and accurately predicting the battery life.

[0006] To solve the above technical problems, the present application specifically provides the following technical solutions:

[0007] A lithium ion battery fast charging life prediction method based on physical mechanism and machine learning fusion, comprising the following steps:

[0008] Step 100: performing first charge-discharge cycle on the lithium ion battery out of the factory and performing electrochemical characteristic measurement;

[0009] Step 200: constructing an electrochemical characteristic curve of the first charge-discharge cycle process of the lithium ion battery according to the electrochemical measurement results and extracting characteristic quantities;

[0010] Step 300: constructing a machine learning model and inputting the extracted characteristic quantities into the machine learning model as input quantities to evaluate the initial life state of the lithium ion battery;

[0011] Step 400: based on the physical degradation mechanism of lithium ion battery charge and discharge cycle conditions, a physical mechanism degradation model is constructed to evaluate the life attenuation in the lithium ion charge and discharge cycle process;

[0012] Step 500: according to the initial life state of the battery and the life attenuation evaluation result in the charge and discharge cycle process, the remaining life of the lithium ion battery is calculated.

[0013] As a preferred scheme of the present application, in step 100, the electrochemical characteristic measurement index includes current value, voltage value, resistance value and corresponding measurement time.

[0014] As a preferred scheme of the present application, the electrochemical characteristic curve in step 200 includes current-time curve and voltage-time curve.

[0015] As a preferred scheme of the present application, the characteristic quantity in step 200 includes but is not limited to: fast charging time, charging voltage standard deviation, charging current standard deviation, constant current discharge voltage standard deviation, constant voltage discharge time, maximum capacity, ratio of maximum capacity to average fast charging current, initial resistance.

[0016] As a preferred scheme of the present application, the machine learning model constructed in step 300 has multiple, including but not limited to support vector machine, Gaussian process regression, multilayer artificial neural network model, and the optimal machine learning model is selected based on the accuracy of the initial performance prediction of the lithium ion battery by the machine learning model.

[0017] As a preferred scheme of the present application, in step 500, the algorithm for calculating the remaining life of the lithium ion battery is as follows:

[0018] Assuming that the estimated life of the lithium ion battery based on the performance factors and the charge and discharge mode factors is N, N is represented as N=f(c)g(ω);

[0019] Where f(c) represents the life attenuation caused by the charge and discharge mode c;

[0020] c specifically refers to the fast charging rate of the lithium ion battery;

[0021] g(ω) represents the evaluation of the performance difference of the lithium ion battery based on the machine learning model;

[0022] ω represents the characteristic quantity based on the initial charge and discharge cycle measurement data.

[0023] As a preferred scheme of the present application, the physical degradation mechanism refers to the attenuation of the maximum capacity of the lithium ion battery caused by the fast charging rate c.

[0024] As a preferred scheme of the present application, the initial life state of the battery refers to the cycle life difference of different batteries under the same charge and discharge conditions caused by the initial test performance difference of the lithium ion battery caused by factors such as material composition, manufacturing difference, production and storage environment cycle.

[0025] As a preferred scheme of the present application, the g (ω) is obtained by cycle life test of different batteries under the same charge and discharge conditions, and based on the current-time and voltage-time curves in the initial charge and discharge cycle, and the performance model of the different batteries corresponding to the cycle life training.

[0026] As a preferred scheme of the present application, the model f (c) is a life decay model obtained by life test under different charge and discharge cycle rates.

[0027] Compared with the prior art, the present application has the following beneficial effects:

[0028] Based on the fusion of physical degradation mechanism model and machine learning model, the present application separately evaluates the charge and discharge cycle history degradation factors and the initial life inconsistency at the time of factory delivery related to the cycle life of lithium ion battery, and gives the remaining life prediction of lithium ion battery according to the two separate evaluation results.

[0029] The present application does not depend on the monitoring of the historical degradation performance of lithium ion battery, and only needs to be based on the first charge and discharge cycle characteristic curve of lithium ion battery at the time of factory delivery and the preset charge and discharge mode, so as to accurately and reliably predict the life of lithium ion battery. BRIEF DESCRIPTION OF DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present application or the technical schemes in the prior art, the drawings needed to be used in the following embodiment or prior art description will be briefly introduced. Obviously, the drawings in the following description are only exemplary, and other implementation drawings can be obtained according to the provided drawings without creative labor for those skilled in the art.

[0031] Figure 1 The prediction method implementation step flow chart provided for the embodiment of the present application;

[0032] Figure 2 The specific prediction method flow chart provided for the embodiment of the present application;

[0033] Figure 3 The typical initial charge and discharge curve schematic diagram of lithium ion battery provided for the embodiment of the present application;

[0034] Figure 4 The precision comparison schematic diagram of a typical commercial lithium ion battery prediction effect provided for the embodiment of the present application. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] The lifespan of a rechargeable battery is closely related to its charge-discharge cycle count. However, because lithium-ion batteries are highly precise and complex devices, their performance is significantly affected by factors such as material composition, processing technology, and production and storage environments. This leads to significant performance differences between individual batteries, resulting in substantial variations in lifespan even under the same charge-discharge conditions. Therefore, initial battery performance evaluation is essential for achieving high-precision battery lifespan prediction. Adding an initial performance evaluation step can significantly improve the accuracy of lithium battery lifespan prediction.

[0037] This invention is based on the fusion of physical degradation mechanism model and machine learning model, and separately evaluates two main factors related to the cycle life of lithium-ion batteries: the degradation factors of charge and discharge cycle history and the inconsistency of initial life at the time of manufacture. Based on the results of the two separate evaluations, the remaining life of lithium-ion batteries is predicted.

[0038] like Figure 1 and Figure 2 As shown, this invention provides a lithium-ion battery lifetime prediction method that integrates physical mechanisms and machine learning, including the following steps:

[0039] Step 100: Measurement of electrochemical characteristics of lithium-ion battery during initial charge-discharge cycles.

[0040] The lithium-ion batteries were subjected to their first complete charge-discharge cycle and electrochemical characteristic measurements were performed. Electrochemical characteristic curves were then constructed based on the electrochemical characteristic measurement data.

[0041] Among them, the electrochemical characteristic measurement indicators include current value, voltage value, resistance value and corresponding measurement time.

[0042] Step 200: Extract feature information.

[0043] Based on the electrochemical characteristic curves of the first charge-discharge cycle measured experimentally, relevant characteristic quantities are extracted.

[0044] like Figure 3As shown in the middle, the electrochemical characteristic curve formed by the electrochemical evolution process in the initial charge-discharge cycle of the lithium ion battery based on the preset charge-discharge mode includes a current-time curve and a voltage-time curve.

[0045] The characteristic quantities closely related to the initial life performance of the lithium ion battery extracted from the electrochemical characteristic curve in the first charge-discharge cycle based on experimental measurement include but are not limited to: fast charging time, charging voltage standard deviation, charging current standard deviation, constant current discharge voltage standard deviation, constant voltage discharge time, maximum capacity, ratio of maximum capacity to average fast charging current, initial resistance.

[0046] Step 300: Evaluate the life inconsistency of the initial performance impact.

[0047] A machine learning model is constructed, and the extracted characteristic quantities are brought into the constructed machine learning model as input quantities to evaluate the initial life performance difference of the lithium ion battery.

[0048] Among them, the constructed machine learning model has multiple, including but not limited to support vector machine, Gaussian process regression, and multi-layer artificial neural network model, and the optimal machine learning model is selected based on the accuracy of the initial performance prediction of the lithium ion battery by the machine learning model.

[0049] The initial performance information of the lithium ion battery is embodied in the current and voltage curves that can be measured in a complete charge-discharge time sequence. To construct an initial performance model, it is necessary to first extract characteristic quantities that can reflect performance information from the curve information. The determination of these characteristic quantities is determined by the prediction effect of the battery life. The accuracy of the initial performance prediction of the lithium ion battery is verified based on the collocation combination of different characteristic quantities. The present embodiment provides a characteristic quantity combination that can accurately and reliably characterize the initial performance, which specifically includes the following eight specific characteristics that can be obtained from the current-time and voltage-time curves: fast charging time, charging voltage standard deviation, charging current standard deviation, constant current discharge voltage standard deviation, constant voltage discharge time, maximum capacity, ratio of maximum capacity to average fast charging current, and initial resistance. These characteristic quantities come from the current-time curve and voltage-time curve of the initial cycle of the battery, and can be easily obtained according to their names, such as standard deviation, average value, etc. Based on the influence of the initial performance prediction accuracy of multiple different machine learning models, a battery initial performance machine learning model based on support vector machine is established in the model.

[0050] Step 400: Evaluate the degradation life of the lithium ion battery during the cycle process.

[0051] Based on the physical degradation mechanism of the lithium ion battery charge-discharge cycle condition, a physical degradation model is constructed, the charge-discharge condition of the lithium ion battery to be evaluated is brought into the model, and the life attenuation during the lithium ion charge-discharge cycle process is evaluated.

[0052] The physical degradation mechanism refers to the different attenuation of the maximum capacity of the lithium ion battery caused by different fast charging rates c of the lithium ion battery. Generally, the faster the charging and discharging rate of the lithium battery, the faster the maximum capacity attenuation speed.

[0053] The life of lithium ions generally refers to the cycle number when the maximum capacity of the battery decays to 80% of the nominal capacity. The difference in cycle life of lithium batteries under the same charging and discharging conditions can be represented as the difference in performance of lithium batteries out of the factory. The life decay refers to the change in cycle life under different discharge rates. In this embodiment, the life decay is investigated to the decay of the maximum cycle number.

[0054] The initial life state of the battery in this embodiment refers to the significant difference in the cycle life of different batteries under the same charging and discharging conditions caused by the difference in the initial test performance of the lithium ion battery due to factors such as material composition, processing and manufacturing difference, production and storage environment period.

[0055] Step 500: Calculate the remaining life of the lithium ion battery.

[0056] According to the initial life state of the battery and the life decay evaluation results in the charging and discharging cycle process, the remaining life of the lithium ion battery is calculated. The algorithm for calculating the remaining life of the lithium ion battery is as follows:

[0057] Assuming that the estimated remaining life of the lithium ion battery based on the factors of the performance out of the factory and the charging and discharging mode is N, then N is represented as N=f(c)g(ω);

[0058] Where c specifically refers to the fast charging rate of the lithium ion battery;

[0059] f(c) represents the life decay caused by the charging and discharging mode c;

[0060] ω represents the characteristic quantity based on the initial charging and discharging cycle measurement data;

[0061] g(ω) represents the evaluation of the performance difference of the lithium ion battery out of the factory based on the machine learning model.

[0062] The establishment strategy of the model N=f(c)g(ω) is to consider and construct f(c) and g(ω) models separately, and finally form an integrated model that comprehensively considers the charging and discharging conditions and the performance difference out of the factory.

[0063] Wherein, the model f(c) is a life attenuation model obtained through life test under different charge and discharge cycle rates; g(w) is a factory performance model obtained through cycle life test of different batteries under the same charge and discharge conditions, and based on the current-time and voltage-time curves in the initial charge and discharge cycle, and the cycle life training corresponding to different batteries; the model g(w) = N / (f(c)) is obtained through the life results under the same charge and discharge cycle rate, and finally the life prediction model is obtained by integration.

[0064] The input parameters for training the g(w) model are the eight characteristic parameters given in step 300, the target parameter is the cycle life of the battery, and the specific model adopted is a machine learning model based on support vector machine. The initial performance difference of the battery is affected by multiple factors, and various factors are coupled and related to each other, and there is no specific and clear correlation, so it cannot be simply described as a certain index having a clear positive or negative correlation, but the model shows that the combination of the eight characteristic parameters given in step 300 can accurately depict the influence of the initial performance measurement results on the final life difference.

[0065] The prediction method provided by the embodiment of the present application is introduced and described by taking a commercial lithium ion battery A123 in the prior art as a sample, and the specific implementation is as follows:

[0066] The battery fast charging life prediction method provided by the embodiment of the present application is programmed, and the residual life of a typical commercial lithium ion battery A123 under different fast charging modes is predicted and calculated. The fast charging rate range of the test sample is between 3.6C and 8C, the initial charge and discharge characteristic curves of the battery ( Figure 3 ) and the cycle life when the capacity under different fast charging modes decays to 80% are obtained, and the physical mechanism model and the machine learning model for initial life evaluation are constructed based on the above.

[0067] Three different grouping strategies are selected to evaluate the life prediction of lithium ion batteries under different batteries and charge and discharge modes. As shown in Table 1 in Figure 4 The prediction effects of different prediction models in the three grouping strategies are given in Table 1 shown in

[0068] The embodiment of the present application does not depend on the monitoring of the historical degradation performance of the lithium ion battery, but only needs to be based on the first charge and discharge cycle characteristic curve of the lithium ion battery at the factory and the preset charge and discharge mode, so as to accurately and reliably predict the life of the lithium ion battery, which is simple and efficient.

[0069] The above examples are only exemplary embodiments of the present application, and are not intended to limit the present application, and the protection scope of the present application is defined by the claims. Those skilled in the art can make various modifications or equivalent replacements to the present application within the spirit and protection scope of the present application, and such modifications or equivalent replacements are also considered to fall within the protection scope of the present application.

Claims

1. A lithium-ion battery life prediction method based on the fusion of physical mechanism and machine learning, characterized in that, The method comprises the following steps: Step 100: performing first charge-discharge cycle and electrochemical characteristic measurement on the lithium ion battery out of the factory; Step 200: constructing electrochemical characteristic curve of the lithium ion battery during the first charge-discharge cycle and extracting characteristic quantity according to the electrochemical measurement result; Step 300: constructing machine learning model and inputting the extracted characteristic quantity into the machine learning model as input quantity to evaluate the initial life status of the lithium ion battery; Step 400: constructing physical mechanism degradation model based on the physical degradation mechanism of the lithium ion battery under the charge-discharge cycle condition to evaluate the life attenuation during the charge-discharge cycle of the lithium ion battery; Step 500: calculating the remaining life of the lithium ion battery according to the initial life status of the battery and the evaluation result of the life attenuation during the charge-discharge cycle; The algorithm for calculating the remaining life of the lithium ion battery is as follows: Assuming that the estimated life of the lithium ion battery based on the performance factors out of the factory and the charge-discharge mode factors is N, N is represented as N=f(c)g(ω); Wherein, f(c) represents the life attenuation caused by the charge-discharge mode; c specifically refers to the fast charging rate of the lithium ion battery; g(ω) represents the evaluation of the performance difference of the lithium ion battery out of the factory based on the machine learning model; ω represents the characteristic quantity based on the initial charge-discharge cycle measurement data.

2. The lithium-ion battery life prediction method of claim 1, wherein, In step 100, the electrochemical characteristic measurement indexes include current value, voltage value, resistance value and corresponding measurement time. 3.The physical mechanism and machine learning fusion-based lithium-ion battery life prediction method of claim 1, wherein, The electrochemical characteristic curve in step 200 includes current-time curve and voltage-time curve. 4.The physical mechanism and machine learning fusion-based lithium-ion battery life prediction method of claim 1, wherein, The characteristic quantity in step 200 includes but is not limited to: fast charging time, charge voltage standard deviation, charge current standard deviation, constant current discharge voltage standard deviation, constant voltage discharge time, maximum capacity, ratio of maximum capacity to average fast charging current, initial resistance.

5. The physical mechanism and machine learning fusion-based lithium-ion battery life prediction method of claim 1, wherein, The machine learning model constructed in step 300 has multiple, including but not limited to support vector machine, Gaussian process regression, multilayer artificial neural network model, and the optimal machine learning model is selected based on the accuracy of the initial performance prediction of the lithium ion battery by the machine learning model.

6. The physical mechanism and machine learning fusion-based lithium-ion battery life prediction method of claim 1, wherein, The physical degradation mechanism refers to the attenuation of the maximum capacity of the lithium ion battery caused by the fast charging rate c.

7. The physical mechanism and machine learning fusion lithium-ion battery life prediction method according to claim 5, characterized in that, The initial life status of the battery refers to the cycle life difference of different batteries under the same charge-discharge condition caused by the initial test performance difference of the lithium ion battery due to factors such as material composition, processing and manufacturing difference, production and storage environment cycle.

8. The physical mechanism and machine learning fusion lithium-ion battery life prediction method according to claim 7, characterized in that, The g(ω) is obtained by performing cycle life test on different batteries under the same charge-discharge condition, and based on the current-time and voltage-time curves in the initial charge-discharge cycle, and the out-of-factory performance model trained by different batteries corresponding to the cycle life.

9. The physical mechanism and machine learning fusion lithium-ion battery life prediction method according to claim 8, characterized by, f(c) is the life attenuation model obtained by life test under different charge-discharge cycle rates.

Images