A method for joint health state estimation and life prediction of lithium batteries

Abstract

The present invention discloses a method for combined health state estimation and life prediction of lithium batteries, comprising: 1. in the normal operation stage of lithium battery aging, pulse testing the lithium battery is performed, and the health state and remaining life of the lithium battery are predicted by establishing a Thevenin model; 2. after the lithium battery enters the deep aging stage, impedance spectrum data is obtained by performing an EIS test on the lithium battery, and the Thevenin model is modified to a variable-order RC model; 3. the health state of the deep aging stage is estimated using the impedance spectrum data and the parameters identified by the variable-order RC model, respectively. When the prediction error of the two is less than an error threshold, it is considered that a battery health state and remaining life prediction model suitable for the deep aging stage is successfully established; otherwise, the order of the variable-order RC model is adjusted to reduce the error. The present invention can ensure the reliability of lithium battery life prediction estimation throughout the entire life cycle, thereby effectively reducing the control complexity and cost of the test working conditions.

Description

Technical Field

[0001] The present invention relates to the field of lithium battery recycling and battery management system (BMS) applications, and more specifically, to a method for combined health state estimation and life prediction of lithium batteries. Background Art

[0002] With the rapid development of electric vehicles and renewable energy, the performance and safety of lithium batteries, one of the key energy storage devices, are attracting increasing attention. However, due to problems such as capacity fade, increased internal resistance, and thermal runaway, accurate assessment of lithium battery health and life prediction technologies are crucial research areas to ensure their safety and reliability.

[0003] Currently, there are multiple methods for evaluating the health status and lifespan prediction of lithium batteries, including model-based methods, methods based on current-voltage characteristics, and data-driven estimation methods. Among them, the method based on establishing an equivalent circuit model is considered to be a relatively accurate and reliable evaluation method. It can infer the lithium battery mechanism by monitoring the changes in relevant physical quantities of the battery. The relevant parameters of the model are believed to have a strong correlation with the state of the lithium battery. Therefore, reliable model parameters can be used for learning and training to predict the health status.

[0004] However, existing equivalent circuit modeling methods still have numerous drawbacks and shortcomings, such as requiring prior assumptions. The distributed relaxation time model (DRT model) can decompose the different polarization links within a battery, helping to establish a highly accurate battery model. However, establishing the DRT model requires a large amount of EIS test data, which presents testing difficulties and long test times (relative to the method of the present invention) in real-world applications (large batteries, large modules, BMSs, charging stations, etc.). Summary of the Invention

[0005] In order to avoid the shortcomings of the above-mentioned existing technologies, the present invention provides a method for joint health status estimation and life prediction of lithium batteries, so as to correct the battery model through a small number of EIS tests, realize SOH estimation and RUL prediction in two different stages, and ensure the reliability of lithium battery life prediction estimation throughout the entire life cycle, thereby effectively reducing the control complexity and cost of the test conditions.

[0006] The present invention adopts the following technical solutions to solve the technical problems:

[0007] The present invention provides a method for combined health status estimation and life prediction of lithium batteries, which comprises the following steps:

[0008] Step S1: During the normal operation phase of the aging lithium battery, the health status and remaining life of the lithium battery are predicted by establishing a Thevenin model:

[0009] Step S1.1, perform the kth pulse test on the lithium battery to be tested in the normal aging operation stage using the pulse current I, collect the voltage and current signals of the lithium battery, and thus obtain the kth current sampling sequence of the lithium battery and the kth voltage sampling sequence ,in, Indicates the moment when the charging pulse is applied to the lithium battery, Indicates the time when the lithium battery is at rest after the charging pulse ends. represents the current data of the lithium battery at the nth moment of the kth pulse test, represents the current data of the lithium battery at the mth moment of the kth pulse test, represents the voltage data of the lithium battery at the nth moment of the kth pulse test, represents the voltage data of the lithium battery at the mth moment of the kth pulse test; M represents the test time of the lithium battery;

[0010] Step S1.2, completely charge and discharge the lithium battery to detect the health status of the lithium battery in the normal aging operation stage;

[0011] Step S1.3: Take the kth voltage sampling sequence The first n moments The mean value of the kth open circuit voltage of the lithium battery , using the kth voltage sampling sequence Voltage data at the nth moment , voltage data at the n+1th moment , voltage data at the mth moment , voltage data at the m+1th moment and the pulse current I to calculate the kth DC internal resistance of the lithium battery;

[0012] Step S1.4: Construct the Thevenin model and convert the kth current sampling sequence As the input of the Thevenin model, the kth voltage sampling sequence As the output of the Thevenin model, the Thevenin model is identified for the kth time by the least square method, and the polarization resistance and polarization capacitance corresponding to the concentration polarization link and the polarization resistance and polarization capacitance corresponding to the electrochemical polarization link in the Thevenin model under the kth identification are obtained;

[0013] Step S1.5: Repeat the process of steps S1.2-S1.4 to perform K aging tests and parameter identification on the lithium battery to obtain the health status sequence of the lithium battery. And the open circuit voltage sequence corresponding to each health state , DC internal resistance series , concentration polarization resistance series , concentration polarization capacitance series , electrochemical polarization resistance series , electrochemical polarization capacitance series ,in, represents the kth health state of the lithium battery in the normal operation stage of aging, represents the kth open circuit voltage of the lithium battery in the normal operation stage of aging, represents the kth concentration polarization resistance of the lithium battery in the normal aging operation stage, represents the kth electrochemical polarization resistance of the lithium battery in the normal operation stage of aging, represents the kth concentration polarization capacitance of the lithium battery in the normal operation stage of aging, represents the kth electrochemical polarization capacitance of the lithium battery during the normal aging operation stage;

[0014] Step S1.6: 、 、 、 、 and As the input of the neural network, the SOH is used as the output of the neural network, so as to train the neural network and obtain a battery health status prediction model of the lithium battery in the normal aging operation stage;

[0015] Step S1.7: extract a number of consecutive health status sequences from the SOH periodically and use them as input to a time series neural network. The next health status of each extracted health status sequence is used as output, thereby training the time series neural network to obtain a lithium battery remaining life (RUL) prediction model.

[0016] Step S2: After the lithium battery enters the deep aging stage, an EIS test is performed on the lithium battery to obtain impedance spectrum data, and the Thevenin model is changed to a variable-order RC model to establish a battery health status prediction model suitable for the deep aging stage:

[0017] Step S2.1: Perform electrochemical impedance spectroscopy on the lithium battery that has entered the deep aging stage to obtain impedance spectrum data, including: frequency distribution And the corresponding real part of the impedance and the imaginary part of the impedance ,in, represents the hth test frequency, Indicates the hth test frequency The real part of the impedance under Indicates the hth test frequency The imaginary part of the impedance;

[0018] Step S2.2: Based on the impedance spectrum data, a relaxation time distribution model of the lithium battery is established, and the parameters of the DRT model are identified using a regularization method to obtain the DRT distribution function. ,in, represents the relaxation time constant;

[0019] Step S2.3: Based on the impedance spectrum data, the Thevenin model is modified to obtain the polarization resistance and polarization capacitance of each polarization link of the variable-order RC model:

[0020] Step S2.4, completely charge and discharge the lithium battery to detect the health status of the lithium battery in the normal aging operation stage;

[0021] Step S2.5: Perform K aging tests and parameter identification on the lithium battery in the deep aging stage according to the process of steps 2.1 to 2.4 to obtain the health state sequence and polarization resistance sequence of the lithium battery in the deep aging stage. , polarized capacitor sequence ,in, Indicates the kth test Polarization resistance of the order polarization link, Indicates the kth test Polarization capacitance of the order polarization link; is the specified order;

[0022] Step S2.6: As the input of another neural network, the health status sequence of the lithium battery in the deep aging stage is used as the output of the corresponding neural network, so as to train the corresponding neural network and obtain a battery health status model based on the impedance spectrum data;

[0023] Step S2.7, polarization resistance sequence and polarization capacitance series As the input of the third neural network, the health status sequence of the lithium battery in the deep aging stage is used as the output of the third neural network, so as to train the third neural network and obtain a health status prediction model of the lithium battery in the deep aging stage;

[0024] Step S2.8: Compare the health status prediction value output by the battery health status model based on the impedance spectrum data with the health status prediction value output by the health status prediction model of the lithium battery in the deep aging stage to obtain the prediction error. ;

[0025] Step S2.9: When the error is less than the threshold, it means that the health status prediction model of lithium battery in the deep aging stage is suitable for the health status prediction in the deep aging stage. Otherwise, change the order After that, the new polarization resistance and new polarization capacitance are obtained again according to the process of step 2.3, and used as the input of the third neural network in step S2.7, and executed sequentially until until it is less than the error threshold.

[0026] The method for combined health status estimation and life prediction of lithium batteries according to the present invention is also characterized in that step S2.3 includes the following steps:

[0027] Step S2.3.1: Change the Thevenin model to a specified order The variable-order RC model is fitted using impedance spectrum data to obtain the initial parameters of the variable-order RC model;

[0028] Step S2.3.2, performing a pulse test on the lithium battery in the deep aging stage using a pulse current I, and collecting voltage and current signals of the lithium battery, thereby obtaining a current sampling sequence and a voltage sampling sequence of the lithium battery in the deep aging stage;

[0029] Step S2.3.3: Perform a pulse test on the lithium battery according to the process of step 1.1, use the current sampling sequence of the lithium battery in the deep aging stage as the input of the variable-order RC model, and use the voltage sampling sequence of the lithium battery in the deep aging stage as the output of the variable-order RC model, thereby identifying the variable-order RC model through the least squares method to obtain the polarization resistance of each polarization link of the variable-order RC model. and polarization capacitance ;in, Indicates the The polarization resistance of the order, Indicates the polarization capacitance of the order.

[0030] An electronic device of the present invention includes a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the method, and the processor is configured to execute the program stored in the memory.

[0031] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program executes the steps of the method when executed by a processor.

[0032] Compared with the existing technology, the beneficial effects of the present invention are embodied in:

[0033] 1. The method of the present invention realizes the joint health status estimation and life prediction of lithium batteries in stages. In the normal operation stage, health status estimation and remaining life prediction are performed only through time domain data. In the deep aging stage, a small number of EIS tests are used to correct the health status estimation and remaining life prediction. This solves the shortcoming that EIS tests cannot be performed on lithium batteries regularly under actual operating conditions, and has stronger adaptability and practicality.

[0034] 2. The method of the present invention performs a lithium battery pulse test during the normal operation stage of the lithium battery, and uses the parameters of the Thevenin model to estimate the health state and predict the remaining life. The error is small during the normal operation stage, thereby achieving a more economical, simpler, more reliable and rapid parameter identification.

[0035] 3. In the deep aging stage, the method of the present invention uses EIS test condition data to estimate the health state. The established DRT model has high physical significance and rich information, which is of guiding significance for model determination, parameter identification, and health state estimation. Its high-precision characteristics enable the health state prediction results based on impedance spectrum data to be used for feedback correction of the variable-order RC model parameter identification results, solving the problem of error accumulation in the health state estimation of the Thevenin model parameters after deep aging of lithium batteries, and improving the reliability of the health state and remaining life prediction within the entire life cycle.

[0036] 4. The method of the present invention performs an EIS test on the lithium battery when it enters the deep aging stage to complete the correction of the health state estimate, changes the Thevenin model into a variable-order RC model, and then estimates the health state based on the parameters of the variable-order RC model. The SOH change trend is predicted by predicting the change trend of the model parameters, and the lithium battery life prediction is further realized, thereby effectively reducing the control complexity and cost of the test conditions. It has strong practical application value and can be applied to battery residual value evaluation, battery classification and screening, BMS design and battery cascade utilization and other occasions. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a basic schematic diagram of the normal operation stage and deep aging stage in the present invention;

[0038] Figure 2 This is the voltage and current response curve under the pulse test of the present invention;

[0039] Figure 3 The Thevenin model is obtained in the normal operation stage of the present invention;

[0040] Figure 4 It is the variable-order RC model used in the deep aging stage of the present invention;

[0041] Figure 5This is a flow chart of the hierarchical two-stage SOH estimation and RUL prediction combined method in the present invention. DETAILED DESCRIPTION

[0042] In this embodiment, a method for combined health status estimation and life prediction of lithium batteries is mainly implemented by modeling lithium batteries in different stages and predicting the remaining life (RUL) trend of lithium batteries. Figure 1 As shown, the battery can be divided into a normal operating phase and a deep aging phase. During the normal operating phase, a Thevenin model is established and predicted using pulse time-domain signals. After the battery ages to a specific state of health or has operated for a specific period of time, it enters the deep aging phase. At this stage, the battery's state of health and life predictions derived from the Thevenin model parameters are subject to error. To correct for this error, EIS testing is performed on the lithium battery and a high-precision relaxation time-scale model is established to predict the battery's current state of health (SOH). The Thevenin model is then modified to a variable-order RC model, thereby establishing a battery state of health and remaining life prediction model suitable for the deep aging phase. The lithium battery life prediction method of the present invention uses pulse signals to build a model and establish a relationship with the health status during the normal operation stage of the lithium battery, and predicts the attenuation trend of the lithium battery life by predicting the change of model parameters, thereby reducing the testing cost during the normal operation stage. When the lithium battery enters the deep aging stage, the EIS test is performed to correct the battery model to a variable-order RC model, and a battery health status and remaining life prediction model for the deep aging stage is established. This can ensure the reliability of the lithium battery life prediction estimation within the entire life cycle, and can effectively reduce the control complexity and cost of the test conditions. It has strong practical application value and can be applied to battery residual value assessment, battery classification and screening, BMS design, and battery cascade utilization. Specifically, if Figure 5 As shown, the method includes the following steps:

[0043] Step S1: During the normal operation phase of the aging lithium battery, the health status and remaining life of the lithium battery are predicted by establishing a Thevenin model:

[0044] Step S1.1, use the pulse current I to perform the kth pulse test on the lithium battery to be tested in the normal aging and normal operation stage, the pulse signal is as follows: Figure 2 As shown in the figure: the lithium battery to be tested is first left to stand for a period of time, a current of magnitude I is applied at time n, and the lithium battery is left to stand at time m until the end of time M. The voltage and current signals of the lithium battery are collected to obtain the kth current sampling sequence of the lithium battery and the kth voltage sampling sequence ,in, Indicates the moment when the charging pulse is applied to the lithium battery, Indicates the time when the lithium battery is at rest after the charging pulse ends. represents the current data of the lithium battery at the nth moment of the kth pulse test, represents the current data of the lithium battery at the mth moment of the kth pulse test, represents the voltage data of the lithium battery at the nth moment of the kth pulse test, represents the voltage data of the lithium battery at the mth moment in the kth pulse test; M represents the test time of the lithium battery.

[0045] Step S1.2, completely charge and discharge the lithium battery to detect the health status of the lithium battery in the normal aging operation stage;

[0046] Step S1.3: Take the kth voltage sampling sequence The first n moments The mean value of the kth open circuit voltage of the lithium battery , using the kth voltage sampling sequence Voltage data at the nth moment , voltage data at the n+1th moment , voltage data at the mth moment , voltage data at the m+1th moment and the pulse current I to calculate the kth DC internal resistance of the lithium battery; the open circuit voltage and DC resistance The calculation formulas are as follows:

[0047] (1)

[0048] (2)

[0049] Since the test time is short and the impact on the battery is small, it can be considered that the state of the lithium battery has not been changed. The entire test process can be and Considered to be a constant.

[0050] Step S1.4, construct Figure 3 Thevenin model shown in the figure, and the kth current sampling sequence As the input of the Thevenin model, the kth voltage sampling sequence As the output of the Thevenin model, the Thevenin model is identified for the kth time by the least squares method to obtain the kth polarization resistance corresponding to the concentration polarization link in the Thevenin model. , the kth polarized capacitor The kth polarization resistance corresponding to the electrochemical polarization link , the kth polarized capacitor ;

[0051] The expression of the Thevenin model of lithium battery in the s domain is as follows:

[0052] (3)

[0053] In formula (3), It is the kth current sampling sequence Pull-type transformation yields, is the kth voltage sampling sequence Perform a Laplace transformation to obtain , where s represents a complex variable in the s domain.

[0054] Then write the transfer function as formula (4):

[0055] (4)

[0056] In formula (4), represents the time constant of the kth concentration polarization link, and 、 represents the time constant of the kth electrochemical polarization link, and .

[0057] The system is transformed from the s domain to the z domain using bilinear transformation as shown in Equation (5):

[0058] (5)

[0059] In formula (5), z represents the complex variable in the z domain;

[0060] The discretized transfer function of the system is as follows:

[0061] (6)

[0062] In formula (6), is the first intermediate variable, is the second intermediate variable, is the third intermediate variable, is the fourth intermediate variable, is the fifth intermediate variable.

[0063] Convert Equation (6) into a differential equation as shown in Equation (7):

[0064] (7)

[0065] In formula (7), represents the system output in the form of a difference equation, For the System output value, For the System output value, For the system output values, and . is the kth voltage sampling sequence The value of the lth moment, is the kth current sampling sequence The lth moment value of is the kth current sampling sequence No. A moment value, is the kth current sampling sequence No. Then, the intermediate variable sequence is obtained by fitting through least squares fitting or other fitting algorithms. The transformation relationship from the z domain to the s domain is as follows:

[0066] (8)

[0067] In formula (8), T represents the discrete sampling time of the system;

[0068] Substituting formula (8) into formula (6), it can be written as formula (9):

[0069] (9)

[0070] Comparing Equation (4) and Equation (9), we can obtain the relationship shown in Equation (10):

[0071] (10)

[0072] According to formula (10), the five parameters of the Thevenin model in the kth pulse test can be calculated: 、 、 、 and .

[0073] Step S1.5: Repeat the process of steps S1.2-S1.4 to perform K aging tests and parameter identification on the lithium battery to obtain the health status sequence of the lithium battery. And the open circuit voltage sequence corresponding to each health state , DC internal resistance series , concentration polarization resistance series , concentration polarization capacitance series , electrochemical polarization resistance series , electrochemical polarization capacitance series ,in, represents the kth health state of the lithium battery in the normal operation stage of aging, represents the kth open circuit voltage of the lithium battery in the normal operation stage of aging, represents the kth concentration polarization resistance of the lithium battery in the normal aging operation stage, represents the kth electrochemical polarization resistance of the lithium battery in the normal operation stage of aging, represents the kth concentration polarization capacitance of the lithium battery in the normal operation stage of aging, It represents the kth electrochemical polarization capacitance of the lithium battery during the normal operation stage of aging.

[0074] Step S1.6: 、 、 、 、 and As the input of the neural network, the SOH is used as the output of the neural network, so as to train the neural network and obtain the SOH prediction model of the lithium battery in the normal aging operation stage.

[0075] Step S1.7: extract a number of consecutive health status sequences from the SOH periodically and use them as input to a time series neural network. The next health status of each extracted health status sequence is used as output, thereby training the time series neural network to obtain a lithium battery remaining life prediction model.

[0076] From SOH sequence data Extract multiple sequence data As the input of the time series neural network, the SOH value of the next cycle of each part of the sequence data As the output, the time series neural network is trained to predict the remaining battery life, where and The health status of the lithium battery detected for the qth and xth times respectively.

[0077] Step S2: After the lithium battery enters the deep aging stage, an EIS test is performed on the lithium battery to obtain impedance spectrum data, and the Thevenin model is changed to a variable-order RC model to establish a battery health status and remaining life prediction model suitable for the deep aging stage:

[0078] Step S2.1: Perform electrochemical impedance spectroscopy (EIS) on the lithium battery that has entered the deep aging stage to obtain impedance spectrum data, including: frequency distribution And the corresponding real part of the impedance and the imaginary part of the impedance ,in, represents the hth test frequency, Indicates the hth test frequency The real part of the impedance under Indicates the hth test frequency The imaginary part of the impedance.

[0079] Step S2.2: Based on the impedance spectrum data, a relaxation time distribution model (DRT model) of the lithium battery is established, and the parameters of the DRT model are identified using a regularization method to obtain the DRT distribution function. ,in, represents the relaxation time constant.

[0080] The expression of the DRT model is as follows:

[0081] (11)

[0082] In formula (11), is the total impedance of the DRT model, is the ohmic internal resistance of the lithium battery, and f is the frequency.

[0083] In actual EIS testing, only discrete series impedance can be obtained. ,in, is the frequency corresponding to the nth frequency point; therefore, the fitting solution The process is to minimize the sum of squared errors of formula (12):

[0084] (12)

[0085] In formula (12), Represents the optimization objective function, N represents the total number of test frequency points, Frequency The actual impedance corresponding to Frequency The corresponding experimentally measured impedance, The actual impedance The real part of The actual impedance The imaginary part of The actual impedance The real part of The actual impedance The imaginary part of the interpolation optimization problem of formula (12) can be solved by Bayesian optimization and other optimization algorithms. Then the square root formula (11) can be used to calculate .

[0086] Step S2.3: Based on the impedance spectrum data, the Thevenin model is modified:

[0087] Step S2.3.1, change the Thevenin model to Figure 4 Specified order shown The variable-order RC model is fitted using impedance spectrum data to obtain the initial parameters of the variable-order RC model;

[0088] Step S2.3.2, using Figure 2 The pulse current I shown is used to perform a pulse test on a lithium battery in the deep aging stage, and collect voltage and current signals of the lithium battery, thereby obtaining a current sampling sequence and a voltage sampling sequence of the lithium battery in the deep aging stage;

[0089] Step S2.3.3: Use the current sampling sequence of the lithium battery in the deep aging stage as the input of the variable-order RC model, and use the voltage sampling sequence of the lithium battery in the deep aging stage as the output of the variable-order RC model, so as to identify the variable-order RC model through the least squares method and obtain the polarization resistance of each polarization link of the variable-order RC model. and polarization capacitance ;in, Indicates the The polarization resistance of the order, Indicates the polarization capacitance of the order.

[0090] Step S2.4, completely charge and discharge the lithium battery to detect the health status of the lithium battery in the normal aging operation stage;

[0091] Step S2.5: Perform K aging tests and parameter identification on the lithium battery in the deep aging stage according to the process of steps 2.1 to 2.4 to obtain the health state sequence and polarization resistance sequence of the lithium battery in the deep aging stage. , polarized capacitor sequence ,in Indicates the kth test The polarization resistance of the order polarization link, where Indicates the kth test Polarization capacitance of the order polarization link.

[0092] Step S2.6, based on As the input of another neural network, the health status sequence of the lithium battery in the deep aging stage is used as the output of the corresponding neural network, so as to train the corresponding neural network and obtain a battery health status model based on the impedance spectrum data;

[0093] Step S2.7, polarization resistance sequence and polarization capacitance series As the input of the third neural network, the health status sequence of the lithium battery in the deep aging stage is used as the output of the third neural network, so as to train the third neural network and obtain a health status prediction model of the lithium battery in the deep aging stage.

[0094] Step S2.8: Compare the health status prediction value output by the battery health status model based on the impedance spectrum data with the health status prediction value output by the health status prediction model of the lithium battery in the deep aging stage to obtain the prediction error. ;

[0095] Step S2.9: When the error is less than the threshold, it means that the health status prediction model of lithium battery in the deep aging stage is suitable for the health status prediction in the deep aging stage. Otherwise, change the order After that, the new polarization resistance and new polarization capacitance are obtained again according to the process of step 2.3, and used as the input of the third neural network in step S2.7, and executed sequentially until until it is less than the error threshold.

[0096] In this embodiment, an electronic device includes a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the above method, and the processor is configured to execute the program stored in the memory.

[0097] In this embodiment, a computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above method are executed.

Claims

1. A method for joint health status estimation and life prediction of lithium batteries, characterized in that: The steps include: Step S1: During the normal operation phase of the aging lithium battery, the health status and remaining life of the lithium battery are predicted by establishing a Thevenin model: Step S1.1, perform the kth pulse test on the lithium battery to be tested in the normal aging operation stage using the pulse current I, collect the voltage and current signals of the lithium battery, and thus obtain the kth current sampling sequence of the lithium battery and the kth voltage sampling sequence ,in, Indicates the moment when the charging pulse is applied to the lithium battery, Indicates the time when the lithium battery is at rest after the charging pulse ends. represents the current data of the lithium battery at the nth moment of the kth pulse test, represents the current data of the lithium battery at the mth moment of the kth pulse test, represents the voltage data of the lithium battery at the nth moment of the kth pulse test, represents the voltage data of the lithium battery at the mth moment of the kth pulse test; M represents the test time of the lithium battery; Step S1.2, completely charge and discharge the lithium battery to detect the health status of the lithium battery in the normal aging operation stage; Step S1.3: Take the kth voltage sampling sequence The first n moments The mean value of the kth open circuit voltage of the lithium battery , using the kth voltage sampling sequence Voltage data at the nth moment , voltage data at the n+1th moment , voltage data at the mth moment , voltage data at the m+1th moment and the pulse current I to calculate the kth DC internal resistance of the lithium battery; Step S1.4: Construct the Thevenin model and convert the kth current sampling sequence As the input of the Thevenin model, the kth voltage sampling sequence As the output of the Thevenin model, the Thevenin model is identified for the kth time by the least square method, and the polarization resistance and polarization capacitance corresponding to the concentration polarization link and the polarization resistance and polarization capacitance corresponding to the electrochemical polarization link in the Thevenin model under the kth identification are obtained; Step S1.5: Repeat the process of steps S1.2-S1.4 to perform K aging tests and parameter identification on the lithium battery to obtain the health status sequence of the lithium battery. And the open circuit voltage sequence corresponding to each health state , DC internal resistance series , concentration polarization resistance series , concentration polarization capacitance series , electrochemical polarization resistance series , electrochemical polarization capacitance series ,in, represents the kth health state of the lithium battery in the normal operation stage of aging, represents the kth open circuit voltage of the lithium battery in the normal operation stage of aging, represents the kth concentration polarization resistance of the lithium battery in the normal aging operation stage, represents the kth electrochemical polarization resistance of the lithium battery in the normal operation stage of aging, represents the kth concentration polarization capacitance of the lithium battery in the normal operation stage of aging, represents the kth electrochemical polarization capacitance of the lithium battery during the normal aging operation stage; Step S1.6: 、 、 、 、 and As the input of the neural network, the SOH is used as the output of the neural network, so as to train the neural network and obtain a battery health status prediction model of the lithium battery in the normal aging operation stage; Step S1.7: extract a number of consecutive health status sequences from the SOH periodically and use them as input to a time series neural network. The next health status of each extracted health status sequence is used as output, thereby training the time series neural network to obtain a lithium battery remaining life (RUL) prediction model. Step S2: After the lithium battery enters the deep aging stage, an EIS test is performed on the lithium battery to obtain impedance spectrum data, and the Thevenin model is changed to a variable-order RC model to establish a battery health status prediction model suitable for the deep aging stage: Step S2.1: Perform electrochemical impedance spectroscopy on the lithium battery that has entered the deep aging stage to obtain impedance spectrum data, including: frequency distribution And the corresponding real part of the impedance and the imaginary part of the impedance ,in, represents the hth test frequency, Indicates the hth test frequency The real part of the impedance under Indicates the hth test frequency The imaginary part of the impedance; Step S2.2: Based on the impedance spectrum data, establish the relaxation time distribution model of the lithium battery, and use the regularization method to identify the parameters of the DRT model to obtain the DRT distribution function ,in, represents the relaxation time constant; Step S2.3: Based on the impedance spectrum data, the Thevenin model is modified to obtain the polarization resistance and polarization capacitance of each polarization link of the variable-order RC model: Step S2.4, completely charge and discharge the lithium battery to detect the health status of the lithium battery in the aging and normal operation stage; Step S2.5: Perform K aging tests and parameter identification on the lithium battery in the deep aging stage according to the process of steps 2.1 to 2.4 to obtain the health state sequence and polarization resistance sequence of the lithium battery in the deep aging stage. , polarized capacitor sequence ,in, Indicates the kth test Polarization resistance of the order polarization link, Indicates the kth test Polarization capacitance of the order polarization link; is the specified order; Step S2.6: As the input of another neural network, the health status sequence of the lithium battery in the deep aging stage is used as the output of the corresponding neural network, so as to train the corresponding neural network and obtain a battery health status model based on the impedance spectrum data; Step S2.7, polarization resistance sequence and polarization capacitance series As the input of the third neural network, the health status sequence of the lithium battery in the deep aging stage is used as the output of the third neural network, so as to train the third neural network and obtain a health status prediction model of the lithium battery in the deep aging stage; Step S2.8: Compare the health status prediction value output by the battery health status model based on the impedance spectrum data with the health status prediction value output by the health status prediction model of the lithium battery in the deep aging stage to obtain the prediction error. ; Step S2.9: When the error is less than the threshold, it means that the health status prediction model of lithium battery in the deep aging stage is suitable for the health status prediction in the deep aging stage. Otherwise, change the order After that, the new polarization resistance and new polarization capacitance are obtained again according to the process of step 2.3, and used as the input of the third neural network in step S2.7, and executed sequentially until until it is less than the error threshold.

2. A lithium battery joint health status estimation and life prediction method according to claim 1, characterized in that: Step S2.3 includes the following steps: Step S2.3.1: Change the Thevenin model to a specified order The variable-order RC model is fitted using impedance spectrum data to obtain the initial parameters of the variable-order RC model; Step S2.3.2, performing a pulse test on the lithium battery in the deep aging stage using a pulse current I, and collecting voltage and current signals of the lithium battery, thereby obtaining a current sampling sequence and a voltage sampling sequence of the lithium battery in the deep aging stage; Step S2.3.3: Perform a pulse test on the lithium battery according to the process of step 1.1, use the current sampling sequence of the lithium battery in the deep aging stage as the input of the variable-order RC model, and use the voltage sampling sequence of the lithium battery in the deep aging stage as the output of the variable-order RC model, thereby identifying the variable-order RC model through the least squares method to obtain the polarization resistance of each polarization link of the variable-order RC model. and polarization capacitance ;in, Indicates the The polarization resistance of the order, Indicates the polarization capacitance of the order.

3. An electronic device comprising a memory and a processor, characterized in that: The memory is used to store a program that supports a processor to execute the method according to claim 1 or 2, and the processor is configured to execute the program stored in the memory.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to claim 1 or 2 are performed.

Images