Online identification method for equivalent circuit parameters of lithium-ion battery impedance spectrum based on data simulation and machine learning

Abstract

The present application relates to a kind of lithium ion battery impedance spectrum equivalent circuit parameter online identification method based on data simulation and machine learning, belong to battery detection technical field.The method includes: obtaining a small amount of battery first cycle EIS and corresponding AR-ECM parameter group of modified Randles equivalent circuit model by experiment, and determine the parameter rate matrix of AR-ECM parameter group;According to the AR-ECM parameter group of battery first cycle and parameter rate matrix, AR-ECM dataset is obtained by interpolation, AR-ECM parameter set is brought into mathematical model to generate EIS curve, and EIS and AR-ECM parameter comprehensive database are obtained;A EIS curve is measured, it is divided into high frequency part and low frequency part, and the Euclidean distance of its and corresponding EIS curve in database is calculated, according to distance, several EIS curves and their corresponding AR-ECM parameter group are selected, and GPR model is trained respectively combining the EIS characteristics of each parameter, then the data characteristics of measured EIS are input into corresponding GPR model, and the high / low frequency parameters in AR-ECM that affect EIS are estimated respectively.The AR-ECM parameter of EIS can be automatically identified stably, accurately and quickly by the present application.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of battery detection, and relates to an online identification method for equivalent circuit parameters of lithium ion battery impedance spectrum based on data simulation and machine learning. BACKGROUND

[0002] Lithium ion batteries are widely used in the field of electric vehicles due to their high energy density and low cost. In order to obtain more comprehensive information of lithium ion battery characteristics without affecting the internal electrochemical process of lithium ion battery, an efficient and non-destructive technology, electrochemical impedance spectroscopy (EIS), is used to study lithium ion battery. It can reflect and study various electrochemical processes of the battery, including solid electrolyte interface (SEI) growth, charge transfer between electrolyte and electrode, and low-frequency diffusion and adsorption process on the electrode, etc.

[0003] In order to reduce the dimension of EIS data and better explain, equivalent circuit model (ECM) is widely used to simulate impedance spectrum. However, the traditional method of fitting EIS to identify ECM parameters, ZView and CNLS, has defects. It has been proved through extensive practice that the ECM parameters fitted by ZView are more authoritative and accurate. However, this method needs to be manually adjusted parameters, thereby increasing the demand for time and experience, hindering the rapid and automated analysis of large-scale EIS data. At the same time, many studies have proposed various CNLS methods, aiming to automatically and quickly fit ECM parameters, and even use the obtained parameters for further research. However, due to the high sensitivity of CNLS method to initial guess, its robustness is relatively low, especially when fitting high-dimensional model parameters, it is easy to converge to local optimal solution, and there is a big consistency problem between the ECM parameters fitted by ZView and the parameters obtained by CNLS method. In the process of parameter identification, researchers have been plagued by the problem of high dimension of identified parameters. At present, there is no reasonable and general dimension reduction method, and the problem of too high dimension of EIS equivalent battery parameters in parameter identification has always plagued researchers. Therefore, the method of automatically, quickly and accurately identifying high-dimensional ECM parameters based on EIS online is of great significance to promote the research of EIS in the field of lithium ion battery. SUMMARY

[0004] Therefore, the present application aims to provide a lithium-ion battery impedance spectrum equivalent circuit parameter online identification method based on data simulation and machine learning, so as to automatically, quickly and accurately identify ECM parameters online based on EIS. In the field of machine learning, "data simulation" generally refers to generating synthetic data through models or rules. In this context, data simulation is to generate synthetic data similar to real data through theoretical or mathematical models. These synthetic data can be used to train machine learning models to supplement or replace real data, especially when real data is insufficient or difficult to obtain, to simulate battery behavior. In the present application, virtual battery impedance spectrum data can be generated through an equivalent circuit mathematical model for training and evaluation of the machine learning model.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0006] A lithium-ion battery impedance spectrum equivalent circuit parameter online identification method based on data simulation and machine learning, the method comprising the following steps:

[0007] S1, obtaining a small amount of EIS of the first cycle of the battery and the corresponding Adopted Randles equivalent circuit model (AR-ECM) parameter set through experiments, and determining the parameter rate matrix of the Adopted Randles equivalent circuit model AR-ECM;

[0008] S2, selecting several reference AR-ECM parameter sets of the first cycle of the battery, combining them with each other, multiplying the first linear interpolation result with the determined parameter rate matrix, and then performing second linear interpolation to obtain an AR-ECM data set, and bringing the AR-ECM parameter set into a mathematical model to generate a corresponding EIS curve, thereby obtaining a high / low frequency EIS and AR-ECM parameter comprehensive database;

[0009] S3, measuring an EIS curve, dividing it into a high frequency part and a low frequency part, calculating the Euclidean distance between the measured EIS curve and the corresponding EIS curve in the database, selecting several EIS curves and their corresponding AR-ECM parameter sets according to the distance, and extracting the corresponding EIS data features for training the GPR model corresponding to each parameter according to the characteristic engineering of each parameter, and extracting the parameter data features from the measured EIS and inputting them into the trained Gaussian Process Regression (GPR) model to estimate the parameters in the AR-ECM that affect the high frequency and low frequency of the EIS curve.

[0010] Further, in step S1, based on the parameter characteristics of the AR-ECM parameter set, the parameters affecting the high-frequency part of the EIS are defined as high-frequency parameters, and the parameters affecting the low-frequency straight-line part of the EIS are defined as low-frequency parameters, and a high-frequency parameter rate matrix and a low-frequency parameter rate matrix are constructed for them respectively, wherein the column in the high-frequency parameter rate matrix and the low-frequency parameter rate matrix represents the parameter type, and the row in the matrix represents the parameter rate under the corresponding parameter type; wherein the parameter types include: inductance L, ohmic resistance R ohm , polarization resistance R sei representing the solid electrolyte interface film, parameter CPE in the constant phase element CPE impedance formula t1 representing the solid electrolyte interface film, index CPE in the constant phase element CPE impedance formula p1 , charge transfer resistance R ct , diffusion resistance R w , index W in the Warburg impedance formula p , parameter CPE in the constant phase element CPE impedance formula representing the double-layer capacitance t2 , index CPE in the constant phase element CPE impedance formula representing the double-layer capacitance p2 .

[0011] Further, in step S2, constructing a lithium ion battery high / low-frequency EIS and AR-ECM parameter comprehensive database includes the following steps:

[0012] S21, selecting N groups of reference AR-ECM parameters to combine with each other, and then performing first linear interpolation;

[0013] S22, multiplying the AR-ECM parameters obtained by the first interpolation with the high / low-frequency parameter rate matrix respectively, and performing equal-interval linear interpolation in the middle to generate a high / low-frequency AR-ECM parameter set, and the repeated items in the parameter set are removed;

[0014] S23, bringing the AR-ECM parameter set obtained by the second interpolation into the mathematical model to generate the corresponding EIS curve, and forming a high / low-frequency EIS and AR-ECM parameter comprehensive database one by one; wherein the mathematical model is represented as follows:

[0015]

[0016] wherein Z(ω, paras) represents the impedance in the EIS when the frequency is ω, paras is the AR-ECM parameter, ω is the corresponding angular frequency used to collect the EIS, Z w represents the Warburg impedance, and W T is a constant.

[0017] Further, in step S3, the following steps are included:

[0018] S31, measure an EIS curve on a real vehicle, upload it to the cloud, and then divide the EIS into a high-frequency part and a low-frequency part;

[0019] S32, calculate the Euclidean distance of each EIS curve in the high / low frequency database, respectively, and select the nearest several EIS curves and the corresponding AR-ECM parameter groups as training data;

[0020]

[0021] wherein Z i and represent the feature values of the target feature data and the simulation data features in the database, respectively, and n is the dimension of the feature data;

[0022] S33, and according to the EIS data features corresponding to each AR-ECM parameter, extract the corresponding data features from the several EIS curves selected by the Euclidean distance and use them to train the GPR model corresponding to each parameter;

[0023] S34, according to the feature engineering of the AR-ECM parameters, extract the EIS data features corresponding to each parameter from the measured EIS and input them into the trained GPR model of each parameter, and estimate the parameters affecting the high-frequency and low-frequency of the EIS curve in the AR-ECM, respectively.

[0024] Further, in step S3, the EIS data features corresponding to each parameter are determined by feature engineering, wherein the contents of the feature engineering include:

[0025] ① The impedance data in the ultrahigh frequency part of the EIS is selected as the feature of the inductance L and the ohmic resistance R ohm ;

[0026] ② The impedance data in the two semicircular parts of the high-frequency region of the EIS is selected as the feature of the parallel network of the resistance R and the constant phase element CPE;

[0027] ③ The impedance data in the low-frequency straight-line region of the EIS is selected as the feature of the Warburg impedance.

[0028] The beneficial effects of the present application are:

[0029] The application provides a dimension reduction method for high-dimensional equivalent circuit model parameters. By analyzing the chemical mechanism inside the battery corresponding to EIS at different frequencies, the AR-ECM model with 11-dimensional parameters is reduced to two low-dimensional models with only 6-dimensional variable parameters and 2-dimensional variable parameters, respectively. Based on the data of the first cycle of the battery and the AR-ECM mathematical model, an EIS&AR-ECM synthetic database is constructed. The application has stability and accuracy in the rapid automatic identification of AR-ECM parameters. The method of the application has the speed and automatic identification ability of CNLS and the accuracy comparable to ZView, and effectively avoids the low-frequency distortion problem encountered by ZSimpWin.

[0030] Other advantages, objects, and features of the application will be set forth in part in the following specification taken in conjunction with the accompanying drawings, and in part will become apparent to those skilled in the art from a consideration of the following specification and drawings. The objects and other advantages of the application will be realized and attained by means of the instrumentalities and combinations pointed out in the following specification. BRIEF DESCRIPTION OF DRAWINGS

[0031] In order to make the objects, technical solutions and advantages of the application clearer, the preferred detailed description of the application will be made below in combination with the drawings, in which:

[0032] Figure 1 The overall framework flowchart of the online identification method of lithium ion battery impedance spectrum equivalent circuit parameters based on data simulation and machine learning of the application;

[0033] Figure 2 The algorithm flowchart of the online identification method of lithium ion battery impedance spectrum equivalent circuit parameters based on data simulation and machine learning of the application;

[0034] Figure 3 The AR-ECM equivalent circuit diagram used in the application;

[0035] Figure 4 The Nyquist diagram of the electrochemical impedance spectrum of the lithium ion battery;

[0036] Figure 5 The simulated electrochemical impedance spectrum diagram in the electrochemical impedance spectrum and improved Randles equivalent circuit parameter comprehensive database, wherein Figure 5 (a) EIS generated from the high-frequency AR-ECM parameter data set, Figure 5 (b) EIS generated from the low-frequency AR-ECM parameter data set; the block diagram in the upper right corner is a two-dimensional diagram corresponding to the three-dimensional diagram of the simulated EIS;

[0037] Figure 6 The parameter identification result schematic diagram under the embodiment of the application, whereinFigure 6 (a) is an estimate of the ohmic internal resistance R om Figure 6 (b) is an estimate of the polarisation resistance R sei Figure 6 (c) is an estimate of the parameter CPE t1 Figure 6 (d) is an estimate of the exponent CPE p1 Figure 6 (e) is an estimate of the charge transfer resistance R ct Figure 6 (f) is an estimate of the diffusion resistance R w Figure 6 (g) is an estimate of the exponent W p Figure 6 (h) is an estimate of the parameter CPE t2 Figure 6 (i) is an estimate of the exponent CPE p2

[0038] Figure 7 is a schematic diagram of the EIS curve fitting result under the embodiment of the present application. DETAILED DESCRIPTION

[0039] The present application will be described in more detail below by way of specific embodiments, and those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in the specification. The present application can also be implemented or applied by means of other different specific embodiments, and each detail in the specification can be modified or changed based on different views and applications without departing from the spirit of the present application. It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present application in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0040] The accompanying drawings are only used for exemplary illustration, and the representation is only a schematic diagram, not a physical diagram, and should not be understood as a limitation on the present application; in order to better illustrate the embodiments of the present application, some components in the drawings can be omitted, enlarged or reduced, and do not represent the size of the actual product; it can be understood by those skilled in the art that some known structures and their descriptions in the drawings can be omitted.

[0041] ​​​​​​​​​The same or similar reference numerals in the drawings of the embodiments of the present application correspond to the same or similar components; in the description of the present application, it is understood that if the orientations or positional relationships indicated by the terms "upper", "lower", "left", "right", "front", "back", etc. are based on the orientations or positional relationships shown in the drawings, they are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore the terms describing the positional relationships in the drawings are only used for exemplary illustration, and cannot be understood as a limitation on the present application, for those skilled in the art, the specific meanings of the above terms can be understood according to the specific circumstances.

[0042] Please refer to Figures 1 to 7 , a lithium ion battery impedance spectrum equivalent circuit parameter online identification method based on data simulation and machine learning.

[0043] Embodiments

[0044] As Figure 1 The overall framework flowchart of the lithium ion battery impedance spectrum equivalent circuit parameter online identification method based on data simulation and machine learning of the present application is shown, and the algorithm flowchart of the lithium ion battery impedance spectrum equivalent circuit parameter online identification method based on data simulation and machine learning of the present application is shown. Figure 2 The overall framework flowchart of the lithium ion battery impedance spectrum equivalent circuit parameter online identification method based on data simulation and machine learning of the present application is shown, and the algorithm flowchart of the lithium ion battery impedance spectrum equivalent circuit parameter online identification method based on data simulation and machine learning of the present application is shown.

[0045] S1, obtaining a small amount of EIS of the first cycle of the battery and the corresponding modified Randles equivalent circuit model (AR-ECM) parameter group through experiment, and determining the element parameter rate matrix of the modified Randles equivalent circuit model (AR-ECM);

[0046] S2, selecting several reference AR-ECM parameter groups of the first cycle of the battery, combining them with each other, multiplying the first linear interpolation result with the determined parameter rate matrix, and then performing the second linear interpolation to obtain an AR-ECM data set, and bringing the AR-ECM parameter set into a mathematical model to generate a corresponding EIS curve, thereby obtaining a high / low frequency EIS and AR-ECM parameter comprehensive database;

[0047] S3, a measured EIS curve is divided into a high-frequency part and a low-frequency part, and the Euclidean distance between the measured EIS curve and the corresponding EIS curve in the database is calculated, a plurality of EIS curves and their corresponding AR-ECM parameter sets are selected according to the distance, and the corresponding EIS data features are extracted according to the characteristics of each parameter to train the GPR model corresponding to each parameter. The parameter data features are extracted from the measured EIS and input into the trained GPR model, and the parameters in the AR-ECM that affect the high-frequency and low-frequency of the EIS curve are estimated respectively.

[0048] In step S1, the four lithium ion batteries are used to obtain the first cycle EIS and the corresponding reference AR-ECM parameter set, and a large number of reasonable AR-ECM parameter sets are generated as shown in the modified Randles equivalent circuit model (AR-ECM) parameter set. Figure 3 Figure 3 The mathematical model of the circuit is used to generate a large number of corresponding simulation EIS. However, in the case of W T is fixed (i.e. W T is constant in the method), there are still 10 parameters in the AR-ECM, and in the case of high-dimensional parameters, a large number of AR-ECM parameter sets will be generated, which will cause a dimension disaster. In order to overcome this problem, the AR-ECM initial parameter set is multiplied by a certain rate and interpolated between them to generate a large number of reasonable AR-ECM parameters.

[0049] Even so, the generated AR-ECM parameter set is very large, and further analysis of the characteristics of the 10 parameters is needed to reduce the dimension of the parameters. The 10 parameters are: inductance L, ohmic resistance R ohm , polarization resistance R sei characterizing the solid electrolyte interface film, parameter CPE t1 in the impedance formula of the constant phase element CPE characterizing the solid electrolyte interface film, exponent CPE p1 in the impedance formula of the constant phase element CPE characterizing the solid electrolyte interface film, charge transfer resistance R ct , diffusion resistance R w , exponent W p in the Warburg impedance formula, parameter CPE t2 in the impedance formula of the constant phase element CPE characterizing the double-layer capacitance, and exponent CPE p2 in the impedance formula of the constant phase element CPE characterizing the double-layer capacitance. The impedance and parameter value of each parameter are shown in Table 1:

[0050] Table 1

[0051]

[0052] ​Since the low-frequency linear portion of EIS is basically only affected by the Warburg impedance, this embodiment divides the AR-ECM parameters into two parts: eight parameters that affect the high-frequency portion of EIS and two parameters that affect the Warburg impedance of EIS, and constructs high-frequency and low-frequency multiplication matrices respectively.

[0053] For high-frequency parameter multipliers, such as Figure 4 The Nyquist plot of the electrochemical impedance spectroscopy (EIS) of the lithium-ion battery shown is only relevant to ultra-high frequencies; a small variation factor is sufficient. During battery aging, the internal resistance typically increases by a maximum of three times, with resistance factors set to 0.8, 1.5, 2, and 3. However, for the ohmic internal resistance R... ohm As mentioned earlier, it only relates to the intersection of the EIS curve and the real axis of the Nyquist plot, and does not affect the shape of the EIS curve; setting it to 1.5 or 3 is sufficient. The parameters of the constant phase element (CPE) t This is not a critical parameter; only the magnifications of 1.5 and 2.5 are needed. Another parameter for constant-phase elements is CPE. p The non-ideal properties of CPE are described, when CPE p When CPE = 1, CPE behaves like an ideal capacitor, while when CPE = 1... p When the capacitance is 0, the CPE behaves similarly to an ideal resistor. Studies have shown that increasing the surface roughness of the electrode makes the CPE behave more like an ideal capacitor. p Closer to 1. In this embodiment, CPE is used to describe the solid electrolyte interface layer and the double-layer capacitance, respectively. This embodiment assumes that the solid electrolyte interface layer will more closely resemble the characteristics of an ideal resistor as the SEI film grows, while the reference value of CPE... p1 It does indeed exhibit the characteristic of decreasing with age, therefore, for CPE p1 Setting two ratios of 0.75 and 0.9, a double-layer capacitor caused by a rough or porous surface is generally considered to have a CPE. p A constant-phase element with values ​​between 0.9 and 1, thus CPE p2 Simply interpolate between the initial value and 1. Warburg impedance affects the low-frequency straight-line portion, so the magnification of the relevant parameters is set to 1.

[0054] For low-frequency parameter ratios, only the inductance L ratio remains constant during the high-frequency range; all other parameter ratios are set to 1. Of the two parameters of the Warburg impedance, W... p The slope of the low-frequency linear portion of EIS on the Nyquist plot is affected, but in reality, the slope of the linear portion of the EIS in lithium batteries does not change much during aging. Therefore, W p The multiplier only needs to be set around 1. For the diffusion resistor R...w The embodiment still considers that the resistance will not increase more than 3 times during the aging process, thereby setting several reasonable multiples.

[0055] In summary, the high-frequency and low-frequency parameter multiples are shown in Table 2 as the amplification factor of the AR-ECM parameters. The high-frequency parameter multiple matrix is a 256 row x 10 column matrix, and the low-frequency parameter multiple matrix is a 25 row x 10 column matrix.

[0056] Table 2

[0057]

[0058] The embodiment in step S2 of constructing the battery impedance spectrum and AR-ECM parameter comprehensive database specifically includes the following steps:

[0059] S21, the selected 4 groups of reference AR-ECM parameters are combined with each other, and then a first small amount of linear interpolation is performed. Since there are 4 groups of parameters, they can be combined into combinations. 2 groups of parameters are inserted at equal intervals between each combination, so a total of 12 groups of parameters are inserted. In combination with the initial reference parameters, 16 groups of AR-ECM parameters can be obtained after the first interpolation.

[0060] S22, multiply the 16 groups of AR-ECM parameters obtained by the first interpolation with the high / low-frequency parameter multiple matrix, and linearly interpolate 40 groups at equal intervals in the middle, generate 2 AR-ECM parameter sets of high / low frequency, and remove the repeated items in the parameter set.

[0061] S23, the AR-ECM parameter set obtained by the second interpolation is brought into the mathematical model to generate the corresponding EIS curve, and the two are one-to-one corresponding to form the high / low-frequency EIS and AR-ECM parameter comprehensive database.

[0062]

[0063] where paras is the AR-ECM parameter, and ω is the angular frequency corresponding to the 60 frequencies used to collect EIS in the experiment.

[0064] The following Table 3 also gives the number of AR-ECM parameter groups in the process of constructing the database, W T is a constant 72.88.

[0065] Table 3

[0066]

[0067] Figure 5 The simulated EIS in the electrochemical impedance spectrum and AR-ECM parameter comprehensive database is shown. Figure 5(a) is a simulated EIS generated from the high-frequency AR-ECM parameter dataset. Figure 5 (b) is a simulated EIS generated from the low-frequency AR-ECM parameter dataset. The box plot in the upper right corner is the two-dimensional plot corresponding to the three-dimensional plot of the simulated EIS. It should be noted that the database contains a large amount of EIS curve data, so this embodiment uniformly selects 40 curves from it for display.

[0068] This embodiment also performs feature engineering on the EIS curve in step S3, which includes:

[0069] ① The impedance data in the ultrahigh frequency part of the EIS is selected as the features of the inductance (L) and ohmic resistance (R ohm ).

[0070] ② The impedance data in the two semicircular parts of the high-frequency region of the EIS is selected as the features of the two parallel networks (resistance and constant phase element R / CPE). At the same time, considering the possible influence of L and R ohm on the mid-high frequency impedance data (L affects the imaginary part, and R ohm affects the real part), this embodiment adopts a strategy to reduce this influence. Specifically, this embodiment performs relative changes on all selected mid-high frequency features in the EIS, that is, by subtracting a high-frequency feature from each mid-high frequency feature. This method helps to highlight the features of the parallel network, reduce the influence of L and R ohm , and enhance the distinguishability of the features

[0071] ③ Select the impedance data in the low-frequency straight-line region of the EIS as the features of the W S impedance. Similarly, in order to reduce the influence of other parameters, this embodiment performs relative changes on all selected low-frequency features in the EIS, that is, by subtracting a mid-frequency feature from each low-frequency feature.

[0072] Feature engineering determines the part of the electrochemical impedance spectrum data corresponding to each parameter feature.

[0073] Step S3 specifically includes the following steps:

[0074] S31, measure an EIS curve on a real vehicle, upload it to the cloud, and then divide the EIS into a high-frequency part and a low-frequency part;

[0075] S32, calculate the Euclidean distance of each EIS curve with the EIS curves in the high / low frequency database, and select the nearest 250 EIS curves and the corresponding AR-ECM parameter groups as training data.

[0076]

[0077] where Z i and respectively, where and denote the feature values of the target feature data and the simulated feature data in the database respectively, and n is the dimension of the feature data.

[0078] S33, and according to the EIS data features corresponding to each AR-ECM parameter, the corresponding data features are extracted from the 250 EISs and used for training the GPR model corresponding to each parameter;

[0079] S34, the EIS data features corresponding to each parameter are extracted from the measured EIS and input into the trained GPR model, and the parameters affecting the high frequency and the low frequency of the EIS curve in the AR-ECM are estimated respectively.

[0080] In summary, there is currently a lack of automatic, fast and accurate online methods to identify ECM parameters based on EIS. In order to further promote the practical application of EIS, the embodiment proposes an ECM parameter identification method based on EIS. The method provides the following contributions:

[0081] The present application proposes a dimension reduction method for high-dimensional equivalent circuit model parameters. By analyzing the corresponding chemical mechanism inside the battery under different frequencies, the AR-ECM model with 11-dimensional parameters is reduced to two low-dimensional models with only 6-dimensional and 2-dimensional variable parameters respectively. Based on a small amount of data of the first cycle of the battery and the AR-ECM mathematical model, the present embodiment constructs an EIS&AR-ECM synthetic database.

[0082] The present embodiment also verifies the stability and accuracy of the fast and automatic identification of AR-ECM parameters. The method of the present application has the speed and automatic recognition ability of CNLS and the accuracy comparable to ZView, while effectively avoiding the low-frequency distortion problem encountered by ZSimpWin. Figure 6 The estimation results of the AR-ECM parameters for the EIS&AR-ECM synthetic database generated based on four battery first cycle parameter groups at 25℃, wherein Figure 6 (a) is the estimation result of the ohmic resistance R ohm , Figure 6 (b) is the estimation result of the polarization resistance R sei representing the solid electrolyte interface film, Figure 6 (c) is the estimation result of the parameter CPE t1 in the impedance formula of the constant phase element CPE representing the solid electrolyte interface film, Figure 6 (d) is the estimation result of the index CPE p1 in the impedance formula of the constant phase element CPE representing the solid electrolyte interface film, Figure 6 (e) is the estimation result of the charge transfer resistance R ct , Figure 6 (f) is the estimation result of the diffusion resistance Rw The estimation results Figure 6 (g) represents the exponent W in the Warburg impedance formula. p The estimation results, Figure 6 (h) is the parameter CPE in the constant phase angle element impedance formula characterizing double-layer capacitance. pt2 The estimation results, Figure 6 (i) is the exponent CPE in the impedance formula of the constant phase angle element characterizing the double-layer capacitance. p2 The estimation results are shown in Table 4 below. Figure 7 The error between the estimated results of each parameter and the reference values ​​fitted by ZView is the largest average percentage error (MAPE) of only 5.9%.

[0083] Table 4

[0084]

[0085] like ​ The diagram shows the EIS curve fitting results. "Measured" refers to the experimentally measured EIS, and "Simulated" refers to the EIS generated by substituting the predicted AR-ECM parameters into the mathematical model. In a test conducted at 25°C for 309 cycles on 8 batteries, the mean absolute percentage error between the actual EIS curve and the EIS curve fitted by the method of this invention was 2.4%, with a root mean square error of 0.0094 Ω. Compared to the CNLS method's mean absolute percentage error of 20.68% between the identified parameters and the reference values ​​fitted by ZView, the method of this invention shows a significantly smaller error in estimating AR-ECM parameters, with a maximum mean absolute percentage error of only 5.9% compared to the reference values. Furthermore, the method of this invention outperforms the CNLS method at 35°C and 45°C.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An online identification method of lithium-ion battery impedance spectrum equivalent circuit parameters based on data simulation and machine learning, characterized in that: The method comprises the following steps: S1, obtaining a plurality of EIS of the first cycle of the battery and corresponding AR-ECM parameter sets through experiments, and determining a parameter ratio matrix of the AR-ECM parameter sets; S2, selecting a plurality of reference AR-ECM parameter sets of the first cycle of the battery, combining each other, multiplying the first linear interpolation result with the determined parameter ratio matrix, and performing second linear interpolation to obtain an AR-ECM data set, and bringing the AR-ECM parameter set into a mathematical model to generate a corresponding EIS curve, thereby obtaining a high / low frequency EIS and AR-ECM parameter comprehensive database; S3, measuring an EIS curve, dividing the EIS curve into a high frequency part and a low frequency part, calculating the Euclidean distance between the EIS curve and the corresponding EIS curve in the database, selecting a plurality of EIS curves and corresponding AR-ECM parameter sets according to the distance, and combining the characteristics of each parameter to extract corresponding EIS data features for training the GPR model corresponding to each parameter, extracting parameter data features from the measured EIS, and inputting the parameter data features into the trained GPR model to estimate the parameters in the AR-ECM that affect the high frequency and low frequency of the EIS curve; In step S1, based on the characteristics of the obtained AR-ECM parameter sets, the parameters affecting the high frequency part of the EIS are defined as high frequency parameters, and the parameters affecting the low frequency straight line part of the EIS are defined as low frequency parameters, and a high frequency parameter ratio matrix and a low frequency parameter ratio matrix are constructed for them respectively, Wherein, the list in the high-frequency parameter multiplier matrix and the low-frequency parameter multiplier matrix represents a parameter type, and the row in the matrix represents a parameter multiplier under the corresponding parameter type; wherein, the parameter type includes: inductance 、 , polarization resistance representing a solid electrolyte interface film , parameter in a constant phase element CPE impedance formula representing a solid electrolyte interface film , index in a constant phase element CPE impedance formula representing a solid electrolyte interface film , charge transfer resistance , diffusion resistance , index in a Warburg impedance formula , parameter in a constant phase element CPE impedance formula representing a double-layer capacitance , index in a constant phase element CPE impedance formula representing a double-layer capacitance ; In step S2, constructing a lithium ion battery high / low frequency EIS and AR-ECM parameter comprehensive database comprises the following steps: S21, selecting N groups of reference AR-ECM parameters and combining each other, and then performing first linear interpolation; S22, multiplying the AR-ECM parameters obtained by the first interpolation with the high / low frequency parameter ratio matrix, and performing equal-interval linear interpolation in the middle to generate high / low frequency AR-ECM parameter sets, and removing the repeated items in the parameter sets; S23, bringing the high / low frequency AR-ECM parameter sets obtained by the second interpolation into a mathematical model to generate a corresponding EIS curve, and forming a high / low frequency EIS and AR-ECM parameter comprehensive database one by one; wherein the mathematical model is represented as follows: wherein represents the impedance in EIS at frequency , paras are the AR-ECM parameters, is the corresponding angular frequency for the frequency at which EIS is collected, represents the Warburg impedance, is a constant.

2. The method of claim 1, wherein the method is characterized by: In step S3, the following steps are included: S31, measuring an EIS curve on the vehicle, uploading the EIS curve to the cloud, and then dividing the EIS curve into a high frequency part and a low frequency part; S32, calculating the Euclidean distance between the EIS curve and the EIS curves in the high / low frequency database respectively, and selecting a plurality of EIS curves and corresponding AR-ECM parameter sets as training data; wherein and respectively represent the characteristic values of the target characteristic data and the simulation characteristic data in the database, n is the dimension of the characteristic data; S33, and according to the EIS data features corresponding to each AR-ECM parameter, extracting corresponding data features from the plurality of EIS curves selected by the Euclidean distance and using the data features to train the GPR model corresponding to each parameter; S34, according to the feature engineering of AR-ECM parameters, the EIS data features corresponding to each parameter are extracted from the measured EIS and input into the GPR model of each parameter trained, and the parameters affecting the high frequency and low frequency of the EIS curve in the AR-ECM are estimated respectively.

3. The method of claim 2, wherein the method is characterized by: In step S3, the EIS data features corresponding to each parameter are determined by feature engineering, wherein the content of feature engineering includes: The impedance data in the ultra-high frequency part of the EIS is selected as the inductance L and the ohmic resistance R ohm characteristics; ②The impedance data in the two semicircular parts of the medium-high frequency region of EIS are selected as the features of the parallel network of resistor R and constant phase element CPE; ③The impedance data in the low-frequency straight-line region of EIS is selected as the feature of Warburg impedance.

Images