Online generation method for full-frequency electrochemical impedance spectroscopy of battery

Abstract

The invention provides a battery full-frequency electrochemical impedance spectroscopy on-line generation method, which comprises the following steps: S1, selecting a lithium ion battery as a test battery, performing an accelerated aging test, and performing a relaxation voltage test and a full-frequency electrochemical impedance spectroscopy test on the test battery; s2, identifying parameters of the fractional order equivalent circuit model offline by using impedance spectrum data, and forming a parameter set P; s3, using the relaxation voltage data to identify relaxation voltage model parameters in an off-line manner, and forming a parameter set theta; s4, taking the parameter set P and the parameter set theta as input, performing offline training on the neural network model, and obtaining a mapping relation model between the battery relaxation voltage model and the fractional order equivalent circuit model; s5, performing online application on the offline trained neural network model, and reconstructing a multi-frequency impedance spectrum curve of the battery; the method is low in calculation complexity, convenient and efficient, high in accuracy, capable of effectively reducing cost and getting rid of limitation and dependence of laboratory scenes, and wide in application scene.

Description

Technical Field

[0001] The present invention relates to the technical field of power battery management, and in particular to a method for online generation of a full-frequency electrochemical impedance spectrum of a battery. Background Art

[0002] Lithium-ion batteries have attracted widespread attention due to their excellent rate performance and cycle performance, and have been widely used in portable electronic devices, energy storage power stations, and electric transportation. However, as a high-energy electrochemical system, the safety issues of lithium-ion batteries have become an academic focus and technical difficulty. Therefore, battery internal state monitoring and fault diagnosis are crucial to ensure their safe and reliable operation. Since lithium-ion batteries exhibit strong frequency dependence, it is possible to analyze the internal state of the battery through non-destructive measurements in the frequency domain, such as electrochemical impedance spectroscopy (EIS). When sinusoidal currents or voltage signals of different frequencies are applied to the battery, the impedance at different frequencies is calculated by collecting the corresponding voltage or current signals, and the real and imaginary parts are plotted in the complex plane coordinate system according to the order of frequency from high to low to obtain EIS. Compared with the traditional method of directly placing sensors inside the battery to measure the internal state, EIS is more attractive due to the reduction in cost and system complexity. However, EIS measurement is extremely dependent on high-precision experimental equipment, and the equipment is costly, takes a long time to measure, and is not portable, so it is limited to laboratory scenarios. It is still a challenge to obtain full-frequency EIS online for vehicle batteries or mobile batteries.

[0003] The present invention proposes an online generation method of a battery full-frequency electrochemical impedance spectrum based on relaxation voltage. The relaxation voltage model has a simple structure, and the characteristic parameters can be fitted and calculated to extract. The fractional-order model can deeply describe the internal mechanism of the battery and has a clear relationship with the impedance spectrum, but the calculation amount is large and it is difficult to obtain online. The present invention proposes to use a machine learning algorithm to establish a complex mapping relationship model between the battery relaxation voltage model and the fractional-order equivalent circuit model. Summary of the invention

[0004] The purpose of the present invention is to provide a method for obtaining the electrochemical impedance spectroscopy of a lithium battery which is convenient, efficient, highly accurate and has a wide range of application scenarios.

[0005] To achieve the above object, the present invention proposes a method for online generation of full-frequency electrochemical impedance spectroscopy of a battery, characterized in that a mapping relationship between a battery relaxation voltage model and a fractional-order equivalent circuit model is established based on the relaxation voltage; the method comprises the following steps:

[0006] S1: Select a certain number of lithium-ion batteries as test batteries, and after accelerated aging test, perform relaxation voltage test and full-frequency electrochemical impedance spectroscopy test on the test batteries to obtain relaxation voltage data and impedance spectrum data;

[0007] S2: using the impedance spectrum data, offline identifying the parameters of the fractional-order equivalent circuit model, and sorting the RC component parameters from high to low according to the time constant to form a parameter set P;

[0008] S3: Using the relaxation voltage data, offline identifying relaxation voltage model parameters and forming a parameter set θ;

[0009] S4: Taking the parameter set P and the parameter set θ as input, the neural network model is trained offline to obtain a mapping relationship model between the battery relaxation voltage model and the fractional-order equivalent circuit model;

[0010] S5: Apply the offline trained neural network model online to reconstruct the multi-frequency impedance spectrum curve of the battery.

[0011] Furthermore, in step S1, the accelerated aging test is to quickly obtain test batteries with different aging degrees, and the relaxation voltage test is: applying the same current to the test battery and then immediately letting it stand to measure the output voltage response of the battery.

[0012] Furthermore, in step S1, the full-frequency electrochemical impedance spectroscopy test is: using an impedance spectroscopy instrument to measure the full-frequency impedance spectrum of the test battery to obtain an impedance spectrum curve.

[0013] The relaxation voltage test and the full-frequency electrochemical impedance spectroscopy test are performed under different SOCs and different aging degrees, and the data obtained from the test can improve the accuracy and robustness of the training model when the neural network model is trained offline.

[0014] Furthermore, in step S2, the method for offline identifying the fractional-order equivalent circuit model parameters is: finding the optimal fractional-order equivalent circuit model parameters through an optimization algorithm or a search algorithm to minimize the root mean square error between the model impedance and the measured impedance in the full frequency domain.

[0015] Further, in step S3, offline identification of relaxation voltage model parameters is specifically: offline identification or fitting of relaxation voltage model parameters through relaxation voltage data, the offline identification or fitting process is a process in which the root mean square error between the model voltage and the measured voltage is minimized, and the characteristic parameter set is obtained by equation calculation,

[0016] Furthermore, a parameter in the characteristic parameter set characterizes a variation rule of a time constant of a first-order RC model.

[0017] Furthermore, in step 5, the method for applying the offline trained neural network model online is: online collecting local relaxation signals, calculating characteristic parameters by fitting, and inputting them into the trained neural network model to obtain the fractional-order model parameter set at this moment, thereby reconstructing the multi-frequency impedance spectrum curve of the battery online.

[0018] Furthermore, in step 5, the fractional-order model parameter set is solved in the frequency domain to obtain a full-frequency impedance spectrum curve of the battery to be tested, thereby completing the reconstruction process.

[0019] Compared with the prior art, the advantages of the present invention are:

[0020] 1. The method of the present invention gets rid of the dependence on specialized high-precision instruments in the laboratory. It only needs to collect short-term battery relaxation voltage information to generate the impedance spectrum curve of the lithium-ion battery. It is convenient, efficient, and highly accurate, effectively reducing costs and getting rid of the limitations and dependence of laboratory scenarios, thus expanding the scope of application.

[0021] 2. Compared with the traditional method of using pulse current to obtain electrochemical impedance spectroscopy online, the method of the present invention does not need to generate ultra-high frequency current excitation signals according to Shannon sampling theorem, and has low computational complexity, convenience, high efficiency, strong operability and high accuracy.

[0022] 3. Compared with the traditional method of using artificial intelligence to obtain electrochemical impedance spectra online, the present invention does not require the collection of current signals and reduces the requirements for voltage signal collection, effectively reducing the cost of test hardware and having a wide range of application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A flowchart of a method for online generation of full-frequency electrochemical impedance spectroscopy of a battery proposed in an embodiment of the present invention;

[0024] Figure 2 This is a schematic diagram of the relaxation voltage model structure in an embodiment of the present invention;

[0025] Figure 3 A schematic diagram of the structure of a fractional-order equivalent circuit model in an embodiment of the present invention;

[0026] Figure 4 An overall schematic diagram of offline training and online application in an embodiment of the present invention;

[0027] Figure 5 These are the online reconstruction results of impedance spectra of batteries with different aging degrees and different acquisition requirements in this embodiment. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be further described below.

[0029] This embodiment provides a method for online generation of full-frequency electrochemical impedance spectrum of a battery. The method is divided into two parts: an offline training part and an online application part. The key is to establish a mapping relationship between a battery relaxation voltage model and a fractional-order equivalent circuit model based on relaxation voltage. The specific steps are as follows:

[0030] S1: Select a certain number of lithium-ion batteries as test batteries. After accelerated aging test, perform relaxation voltage test and full-frequency electrochemical impedance spectrum test on the test batteries to obtain relaxation voltage data and impedance spectrum data; wherein, the relaxation voltage test and the full-frequency electrochemical impedance spectrum test process are carried out at different SOCs and different aging degrees, and the data obtained in this way can improve the accuracy and robustness of the training model when the neural network model is trained offline. The relaxation voltage test is to apply the same current to the test battery and then let it stand immediately to measure the output voltage response of the battery; the full-frequency electrochemical impedance spectrum test is to use an impedance spectrum instrument to measure the full-frequency impedance spectrum of the test battery to obtain an impedance spectrum curve.

[0031] S2: Using impedance spectrum data, offline identify the parameters of the fractional-order equivalent circuit model, and sort the RC component parameters from high to low according to the time constant to form a parameter set P; wherein, the method for offline identifying the parameters of the fractional-order equivalent circuit model is to find the optimal fractional-order equivalent circuit model parameters through an optimization algorithm or a search algorithm to minimize the root mean square error between the model impedance and the measured impedance in the full frequency domain.

[0032] In this embodiment, the fractional-order equivalent circuit structure is as follows: Figure 3 As shown, its mathematical expression in the frequency domain is:

[0033]

[0034] In the formula, f is the frequency, and the meanings of other parameters are as follows Figure 3 As shown, where G is the impedance value, P FOM is the fractional-order model parameter, L is the UHF inductor, R is the UHF resistor, j is a complex unit, R ohm is the ohmic internal resistance, R SEI is the double layer polarization resistance, CPE SEI is the double-layer constant phase element, α is CPE SEI The fractional order of R CT is the charge transfer polarization resistance, CPE CT is the load-transmitting constant phase element, β is CPE CT The fractional order of R DS is the diffusion polarization resistance, CPE DS is a diffuse constant phase element, γ is CPE DS The fractional order of .

[0035] S3: Using the relaxation voltage data, offline identification of relaxation voltage model parameters is performed to form a parameter set Θ; wherein offline identification of relaxation voltage model parameters is specifically as follows: offline identification or fitting of relaxation voltage model parameters is performed through relaxation voltage data, wherein the offline identification or fitting process is a process in which the root mean square error between the model voltage and the measured voltage is minimized, and a characteristic parameter set is obtained through equation calculation, wherein a parameter in the characteristic parameter set characterizes the time constant variation law of the first-order RC model.

[0036] In this embodiment, the relaxation voltage model structure is as follows: Figure 2 As shown in the figure, the relaxation voltage model parameters are identified offline using the relaxation voltage data, and their mathematical expressions in the time domain are:

[0037]

[0038] Where t is the relaxation time, and the meanings of other parameters are as follows: Figure 2 As shown, OCV is the open circuit voltage of the battery, U ct , U ds and U0 are the voltages of battery charge transfer polarization, concentration polarization and ohmic polarization, respectively. OCV, U ct and U ds The sum is the relaxation voltage (U RLX ), R0 is the ohmic internal resistance, τ ct is the time constant of charge transfer polarization, which is equal to R ct With C ct The product of ds is the time constant of concentration polarization, which is equal to R ds With C ds The product of V ct and V ds is the initial voltage value of charge transfer polarization and concentration polarization.

[0039] S4: Taking the parameter set P and the parameter set θ as input, the neural network model is trained offline to obtain a mapping relationship model between the battery relaxation voltage model and the fractional-order equivalent circuit model;

[0040] S5: Apply the offline trained neural network model online to reconstruct the multi-frequency impedance spectrum curve of the battery. The method of applying the offline trained neural network model online is: collect local relaxation signals online, calculate characteristic parameters by fitting, and input them into the trained neural network model to obtain the fractional-order model parameter set at this moment, and solve the full-frequency impedance spectrum curve of the battery to be tested in the frequency domain with the fractional-order model parameter set, so as to reconstruct the multi-frequency impedance spectrum curve of the battery online.

[0041] The above steps S1-S4 are the offline training part. In the offline training part, steps S2 and S3 are model parameter identification and data processing, among which the relaxation voltage test and full-frequency electrochemical impedance spectroscopy test process need to be carried out under different SOCs and different aging degrees. The data obtained in this way can improve the accuracy and robustness of the training model when the neural network model is trained offline; S4 is neural network training.

[0042] The above step S5 is the online application part, which only requires online collection of short-term relaxation voltage signals, and online fitting of the relaxation voltage model parameters of the battery to calculate and obtain the characteristic parameter set, which is then input into the trained neural network model to obtain the parameters of the fractional-order model at this moment, and online reconstruct the full-frequency electrochemical impedance spectrum curve of the battery.

[0043] Figure 4 This is the process of offline training and online application in the method of this embodiment, which includes four parts: data acquisition, relaxation voltage processing, impedance spectrum processing and model building. Figure 4 As shown, in the data acquisition part, relaxation voltage test and impedance spectrum test are used to construct relaxation voltage and impedance spectrum data covering batteries with different SOCs and different aging degrees; in the impedance spectrum processing and relaxation voltage processing parts, relaxation voltage and impedance spectrum data are respectively used to obtain relaxation model parameters and fractional-order model parameters through offline identification methods; in the model building part, the relaxation model parameters and fractional-order model parameters obtained by identification are input into the neural network for training. The full-frequency electrochemical impedance spectrum curve of the battery is reconstructed by the method of this embodiment. In order to be closer to the actual conditions, the signal collected online is a short-time relaxation signal and a local relaxation signal is used. At the same time, the sampling frequency and signal duration are variable. The impedance spectrum of batteries with different SOCs and aging degrees is reconstructed online. The errors and results are shown in FIG. Figure 5 As shown in the figure, Figure a is the relative error distribution diagram of EIS prediction in the test set. The maximum relative error is only 6.1%, and the relative error of 86% of the samples is within 5%, and 98.4% of the samples are within 6%. Figures b and c respectively show the 8 results with the best and worst prediction effects in the test set. Except for the two groups of EIS in the worst results with serious deviations in the mid-frequency and low-frequency predictions, the remaining groups can basically clearly describe the trajectory of EIS. It can be seen that the proposed method can be used to accurately, quickly and cost-effectively reconstruct the full-frequency impedance spectrum curve of the battery online.

[0044] The above is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any technician in the relevant technical field, without departing from the scope of the technical solution of the present invention, makes any form of equivalent replacement or modification to the technical solution and technical content disclosed in the present invention, which does not depart from the content of the technical solution of the present invention and still falls within the protection scope of the present invention.

Claims

1. A method for online generation of full-frequency electrochemical impedance spectroscopy of a battery, characterized in that: Based on the relaxation voltage, a mapping relationship between a battery relaxation voltage model and a fractional-order equivalent circuit model is established; the steps include: S1: Select a certain number of lithium-ion batteries as test batteries, and after accelerated aging test, perform relaxation voltage test and full-frequency electrochemical impedance spectroscopy test on the test batteries to obtain relaxation voltage data and impedance spectrum data; S2: using the impedance spectrum data, offline identifying the parameters of the fractional-order equivalent circuit model, and sorting the RC component parameters from high to low according to the time constant to form a parameter set P; S3: Using the relaxation voltage data, offline identifying relaxation voltage model parameters and forming a parameter set θ; S4: Taking the parameter set P and the parameter set θ as input, the neural network model is trained offline to obtain a mapping relationship model between the battery relaxation voltage model and the fractional-order equivalent circuit model; S5: Apply the offline trained neural network model online to reconstruct the multi-frequency impedance spectrum curve of the battery.

2. The method for online generation of full-frequency electrochemical impedance spectroscopy of a battery according to claim 1, characterized in that: In step S1, the relaxation voltage test is: applying the same current to the test battery and then immediately letting it stand, and measuring the output voltage response of the battery.

3. The method for online generation of full-frequency electrochemical impedance spectroscopy of a battery according to claim 1, characterized in that: In step S1, the full-frequency electrochemical impedance spectroscopy test is: using an impedance spectroscopy instrument to measure the full-frequency impedance spectrum of the test battery to obtain an impedance spectrum curve.

4. The method for online generation of full-frequency electrochemical impedance spectroscopy of a battery according to claim 1, characterized in that: In step S2, the method for offline identifying the fractional-order equivalent circuit model parameters is: finding the optimal fractional-order equivalent circuit model parameters through an optimization algorithm or a search algorithm to minimize the root mean square error between the model impedance and the measured impedance in the full frequency domain.

5. The method for online generation of full-frequency electrochemical impedance spectroscopy of a battery according to claim 1, characterized in that: In step S3, offline identification of relaxation voltage model parameters is specifically: offline identification or fitting of relaxation voltage model parameters through relaxation voltage data, the offline identification or fitting process is the process of minimizing the root mean square error between the model voltage and the measured voltage, and the characteristic parameter set is obtained by equation calculation.

6. The method for online generation of full-frequency electrochemical impedance spectroscopy of a battery according to claim 5, characterized in that: A parameter in the characteristic parameter set characterizes the variation law of the time constant of the first-order RC model.

7. The method for online generation of full-frequency electrochemical impedance spectroscopy of a battery according to claim 1, characterized in that: In step 5, the method for applying the offline trained neural network model online is: online collection of local relaxation signals, calculation of characteristic parameters by fitting, and input into the trained neural network model to obtain the fractional-order model parameter set at this moment, thereby online reconstructing the multi-frequency impedance spectrum curve of the battery.

8. The method for online generation of full-frequency electrochemical impedance spectrum of a battery according to claim 7, characterized in that: In step 5, the fractional-order model parameter set is solved in the frequency domain to obtain a full-frequency impedance spectrum curve of the battery to be tested, thereby completing the reconstruction process.

Images