Characteristic parameter set test method for power battery health state prediction

Abstract

The invention relates to the technical field of power battery health state rapid detection, and particularly discloses a characteristic parameter set test method for power battery health state prediction, which comprises the following steps of: extracting characteristic parameters and forming a characteristic parameter set; the characteristic parameter extraction comprises the following steps: A01: EIS data processing: performing EIS intermediate frequency interval test on power batteries with different SOH aging degrees to obtain alternating current impedance data of the batteries; a02, DRT analysis is carried out on the EIS atlas, A03, HPPC data processing is carried out, specifically, HPPC testing is carried out on power batteries of different SOHs, and direct current internal resistance data of the batteries are obtained; a04, extracting single pulse data, and calculating characteristic parameters; a05, carrying out EIS and HPPC experiment testing under multiple conditions; forming a characteristic parameter set, which comprises the following steps: B01, identifying and removing an abnormal value by data processing software, and removing a background and a reserved basis of a previous check value; b02, standardization processing, feature selection F1, feature extraction F2 and feature combination F3; and B03, selecting a reference model, selecting verification indexes and performing feature screening verification.

Description

Technical Field

[0001] The present application relates to the technical field of rapid detection of the health status of power batteries, and specifically discloses a characteristic parameter set testing method for predicting the health status of power batteries. Background Art

[0002] At present, the power batteries widely used in electric vehicles are facing the problem of retirement. As the service life increases, corrosion, active material shedding, electrolyte decomposition or volatilization may occur inside the battery, resulting in performance degradation, affecting equipment efficiency and possibly causing safety hazards such as overheating and short circuit. Therefore, the performance evaluation of retired power batteries is crucial to determine whether the battery is suitable for reuse in energy storage systems or needs to be scrapped, which is of great significance to resource utilization and environmental protection.

[0003] At present, battery life detection mainly relies on capacity testing and internal resistance testing. The former is time-consuming, while the latter is quick but not accurate enough. With the development of big data and machine learning technologies, prediction methods based on these technologies are becoming a new trend, which can significantly reduce test time and cost while improving the accuracy of life prediction. By constructing a mathematical model, the battery performance degradation trend can be effectively predicted, providing a scientific basis for battery reuse and safe recycling.

[0004] But at the same time, the training of the model requires a large amount of data and characteristic parameters related to battery performance to optimize the model and improve the accuracy of the model. These characteristic parameters can be basic parameters such as the battery's charge and discharge rate, operating temperature range, charge and discharge depth, and frequency of use. They can also be data such as changes in the voltage curve, changes in internal resistance, temperature gradients, and changes in the chemical composition of the battery. Reasonable selection of characteristic parameters can more comprehensively reflect the actual operating status of the battery, thereby improving the prediction accuracy of the model. In addition, diversified characteristic parameters help the model better adapt to different application scenarios and environmental conditions, enhance the generalization ability of the model, and reveal the deep-seated mechanisms behind battery performance degradation.

[0005] In short, the selection of characteristic parameters is a key step in building an efficient and accurate power battery life prediction model. By scientifically and rationally selecting and processing these parameters, not only can the prediction performance of the model be significantly improved, but it can also provide a more scientific basis for the efficient reuse and safe recycling of batteries.

[0006] In battery performance testing, DC internal resistance and AC impedance are two key parameters. These parameters can provide important information about the health status of the battery, especially for retired batteries. Impedance testing can effectively evaluate the aging of the battery and provide basic data for further building a prediction model for the battery state of health (SOH). In addition, screening and determining representative characteristic parameters from the collected impedance characteristic data is crucial to ensure the accuracy of the model.

[0007] The DC internal resistance is usually calculated through the hybrid pulse power characterization (HPPC) test, while the AC impedance is measured through the electrochemical impedance spectroscopy (EIS) technique.

[0008] As a dynamic test method, HPPC test is mainly used to evaluate the pulse charge and discharge performance of power batteries, especially the power density and energy density of the battery. The test starts from a fully charged state, and the transient response capability of the battery is evaluated by pulse discharge and charging at different depths of discharge (DOD). The test process usually includes rapid discharge and charge pulses, as well as rest time between pulses to ensure that the battery reaches electrochemical and thermal equilibrium. After analysis and calculation, the collected HPPC test data can be used to extract characteristic parameters such as charge / discharge ohmic internal resistance, charge / discharge polarization internal resistance, rebound voltage, and voltage drop. These characteristic parameters are not clearly defined and rely on subjective judgment;

[0009] As a static test method, EIS test detects the AC impedance characteristics of the battery at different frequencies. The results of EIS test are usually expressed on Nyquist diagram or Bode diagram, which can reflect the electrochemical kinetic process of the battery. The spectrum mainly includes three regions. The high-frequency region (>1kHz) mainly reflects the electronic conduction and electrolyte resistance of the battery; the medium-frequency region (1kHz to 1Hz) mainly reflects the charge transfer impedance and SEI film resistance; the low-frequency region (<1Hz) mainly reflects the diffusion process. The spectrum is analyzed by relaxation time distribution (DRT) to obtain the characteristic parameters of AC impedance. DRT analysis converts impedance data from frequency domain to time domain, extracts the relaxation time constant (τ) of each electrochemical process, thereby reflecting the response speed of these processes and providing detailed information about the electrochemical process inside the battery. For example, the time constant (τ) determines the peak position and quantifies the relaxation time of a specific kinetic process inside the battery, that is, the time scale required for the system to recover from a non-equilibrium state to an equilibrium state; the half-peak width reflects the dispersion of the impedance distribution; the peak height reflects the intensity of the impedance distribution; and the peak area reflects the total amount of the impedance distribution.

[0010] The advantage of DC internal resistance testing is that it is simple and fast to implement, and is suitable for field testing and battery management system (BMS) integration. However, it can only provide partial information about the battery and cannot distinguish in detail the different electrochemical processes inside the battery; in addition, the DC internal resistance test results may be affected by temperature and the charge and discharge status of the battery. The AC impedance test provides more detailed internal information of the battery, can distinguish the various electrochemical processes inside the battery, and helps to understand the internal mechanism of the battery. As the battery ages, the internal impedance will change. The EIS test can further reflect the health status of the battery and reveal possible problem areas inside the battery, such as the shedding of active substances and the degradation of electrode materials.

[0011] Existing methods often rely on a single parameter, such as internal resistance and capacity, while ignoring other important parameters that may affect the health of the battery. In addition, these methods usually stay at the measurement of surface parameters, fail to deeply understand and analyze the aging state and mechanism inside the battery, and do not perform comprehensive impedance quantification;

[0012] This method combines the DC internal resistance and AC impedance data, fully considers the impact of the test environment, comprehensively collects relevant data under different temperature conditions and charge states, and extracts a set of key characteristic parameter sets for data modeling. These characteristic parameters can more comprehensively reflect the health status of the battery, providing a more scientific and reliable basis for the performance evaluation, reuse and safe recycling of retired power batteries. Summary of the invention

[0013] This method is based on the electrochemical perspective and solves the limitation of incomplete coverage of characteristic parameters in the past. It integrates the two impedance characteristics, analyzes and quantifies the internal impedance state of the battery, extracts key characteristic parameters, and ensures the validity and reliability of the modeling data.

[0014] In order to achieve the above object, the present invention provides the following basic scheme:

[0015] A characteristic parameter set testing method for predicting the health status of a power battery comprises the following steps:

[0016] Feature parameter extraction and feature parameter set formation;

[0017] Feature parameter extraction includes the following steps:

[0018] A01: EIS data processing: Perform EIS medium frequency range test on the same batch of the same type of power batteries with different SOH aging degrees to obtain the battery AC impedance data;

[0019] A02: Perform DRT analysis on the EIS spectrum and extract peak characteristics as characteristic parameters;

[0020] A03: HPPC data processing: Perform HPPC tests on power batteries with different SOH to obtain the DC internal resistance data of the battery;

[0021] A04: Extract single pulse data and calculate characteristic parameters;

[0022] A05: EIS and HPPC experimental tests under multiple conditions;

[0023] The formation of the characteristic parameter set includes the following steps:

[0024] B01: Data processing software, identification and removal of outliers, background and retention basis for removal of previously checked values;

[0025] B02: standardization, feature selection F1, feature extraction F2 and feature combination F3;

[0026] B03: Select a benchmark model, select validation metrics, and perform feature screening validation.

[0027] The principles and effects of this basic solution are:

[0028] 1. Compared with the existing technology, the characteristic parameters cover a wide range and in-depth. By extracting multiple characteristic parameters, this method provides more comprehensive information on the internal performance status of the battery. These parameters cover multiple aspects such as the impedance characteristics, voltage response and time constant of the battery. By comprehensively analyzing these multi-dimensional characteristic parameters, this method can obtain a high-quality characteristic parameter set for rapid prediction of the battery health status.

[0029] 2. Compared with the existing technology, the test efficiency is improved. EIS focuses on the medium frequency range test, which is different from the previous full frequency acquisition, reducing unnecessary test time and improving test efficiency. The data in the medium frequency area mainly reflects the charge transfer process inside the battery and the characteristics of the electrode / electrolyte interface. Compared with high frequency and low frequency, the medium frequency is less affected by external interference and can more accurately reflect the true state of the battery. At the same time, the medium frequency acquisition step is set to 40, which is different from the conventional 10, improving the accuracy and reliability of the test data.

[0030] Furthermore, the EIS intermediate frequency range is from 1 kHz to 1 Hz.

[0031] Further, the EIS and HPPC experimental tests under multiple conditions include temperature testing and state of charge testing;

[0032] Temperature test: Test at three temperatures: -20°C, 25°C and 55°C to evaluate the performance of the battery in low temperature, room temperature and high temperature environments;

[0033] State of Charge Test: Tests are performed every 10% from 10% to 100% SOC to evaluate the performance of the battery at different states of charge.

[0034] Furthermore, in step B01, all characteristic parameters extracted from the EIS test and the HPPC test are presented in the xlsx format after identifying and removing outliers. The characteristic data are processed into the npz format with high reading efficiency using basic data processing methods based on Python algorithm libraries such as Pandas, Numpy and Scikit-learn.

[0035] Furthermore, in the box plot, a box is constructed based on the quartiles (Q1, Q2, Q3) of the data. The box contains the middle 50% of the data and extends up and down to a certain multiple of the interquartile range (IQR = Q3-Q1). Data points that are less than Q1-1.5*IQR or greater than Q3+1.5*IQR are considered outliers, and these outlier points are directly deleted from the data set. After identifying and removing outliers, if the proportion of outliers in the current column is less than 5%, the mean filling, median filling, forward filling, backward filling or interpolation method are used for appropriate filling. If the proportion is greater than or equal to 5%, samples with a large number of missing values ​​are directly deleted to ensure the integrity of the data. Among them, if the data is normally distributed, the mean filling is mostly used; if the data has more skew, the median filling is mostly used.

[0036] Furthermore, the environmental conditions of the battery when collecting EIS data, such as whether the temperature is abnormal (deviation from the set value> 5°C), whether the equipment is working normally (the equipment is not abnormal and the test process is not interrupted), etc. If there are abnormal values ​​caused by such factors, the data needs to be collected again.

[0037] Furthermore, the Z-Score standardization method was selected for the standardization process, and its calculation formula is Z = (x-μ) / σ, where μ is the mean value of the data set and σ is the standard deviation.

[0038] Furthermore, a benchmark model is selected: a model such as a support vector machine (SVM) is used to perform model prediction based on the final features constructed by feature extraction and feature selection. The rationality of the constructed features is analyzed by comparing the accuracy of the predicted value with the true value.

[0039] Further, select the verification indicator: the evaluation indicator is mainly based on the custom AccRatio, which measures the relative deviation between the predicted value and the true value, ensuring that the model prediction is as close to the actual data as possible. The larger the AccRatio, the smaller the deviation between the predicted value and the true value, that is, the higher the prediction accuracy of the model. Formula:

[0040]

[0041] Among them, yi represents the true value of the i-th sample, represents the predicted value of the i-th sample, and n represents the total number of samples. The range of AccRatio is between (0,1]. The larger the value, the better the prediction effect. When the predicted value is completely accurate, the score tends to 1. When the prediction deviation is large, the score will gradually decrease.

[0042] Furthermore, models such as SVM are used to perform model prediction on the three types of features F1, F2, and F3 respectively. The optimal feature column combination is obtained by comparing the accuracy of the predicted values ​​and the true values. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0044] Figure 1 A schematic diagram of an EIS graph (left) and a DRT analysis (right) in a characteristic parameter set testing method for predicting the health status of a power battery proposed in an embodiment of the present application is shown;

[0045] Figure 2 A schematic diagram of a battery pulse response curve in a characteristic parameter set testing method for predicting the health status of a power battery proposed in an embodiment of the present application is shown;

[0046] Figure 3 A schematic diagram showing the formation of a characteristic parameter set in a characteristic parameter set testing method for predicting the health status of a power battery proposed in an embodiment of the present application is shown;

[0047] Figure 4 A schematic diagram of a characteristic parameter set in a characteristic parameter set testing method for predicting the health status of a power battery proposed in an embodiment of the present application is shown;

[0048] Figure 5 The overall flow chart of a characteristic parameter set testing method for predicting the health status of a power battery proposed in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0049] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments.

[0050] Implementation example Figure 1 , Figure 2 , Figure 3 , Figure 4 and Figure 5 As shown:

[0051] A characteristic parameter set testing method for predicting the health status of a power battery comprises the following steps:

[0052] Feature parameter extraction and feature parameter set formation;

[0053] Feature parameter extraction includes the following steps:

[0054] A01: EIS data processing: Perform EIS medium frequency range test on the same batch of the same type of power batteries with different SOH aging degrees to obtain the battery AC impedance data;

[0055] The EIS intermediate frequency range is between 1kHz and 1Hz.

[0056] Specific:

[0057] Select the same batch of the same type of power batteries with different SOH aging degrees for subsequent testing. In this step, the SOH of power batteries with different aging degrees needs to be tested by yourself.

[0058] The operation process is usually as follows:

[0059] 1. Select batteries of the same type and from the same batch with good appearance, and record basic information, including the manufacturer's rated capacity Q1, cut-off voltage, charge and discharge rate, etc.;

[0060] 2. Use the charge and discharge instrument to perform a complete charge and discharge test to obtain the current capacity Q2 of the battery;

[0061] 3. Calculate the SOH of this battery = (Q2 / Q1) × 100%;

[0062] 4. Select batteries with SOH ≥ 60% and wait for subsequent testing;

[0063] A02: Perform DRT analysis on the EIS spectrum and extract peak characteristics as characteristic parameters;

[0064] A03: HPPC data processing: Perform HPPC tests on power batteries with different SOH to obtain the DC internal resistance data of the battery;

[0065] A04: Extract single pulse data and calculate characteristic parameters;

[0066] A05: EIS and HPPC experimental tests under multiple conditions;

[0067] Specific:

[0068] EIS and HPPC experimental tests under multiple conditions include temperature test and state of charge test;

[0069] Temperature test: Test at three temperatures: -20°C, 25°C and 55°C to evaluate the performance of the battery in low temperature, room temperature and high temperature environments;

[0070] State of Charge Test: Test every 10% from 10% to 100% SOC to evaluate the performance of the battery at different states of charge

[0071] The formation of the characteristic parameter set includes the following steps:

[0072] B01: Data processing software, identifying and removing outliers. All characteristic parameters extracted from EIS and HPPC tests are presented in xlsx format. Basic data processing methods based on Python-based Pandas, Numpy, Scikit-learn and other algorithm libraries are used to process characteristic data into npz format with high reading efficiency.

[0073] In the box plot, boxes are constructed based on the quartiles (Q1, Q2, Q3) of the data. The boxes contain the middle 50% of the data and extend up and down to data points within a certain multiple of the interquartile range (IQR = Q3-Q1). Data points that are less than Q1-1.5*IQR or greater than Q3+1.5*IQR are considered outliers and are directly deleted from the data set.

[0074] After identifying and removing outliers, if the proportion of outliers in the current column is less than 5%, use mean filling, median filling, forward filling, backward filling or interpolation to fill in appropriately. If the proportion is greater than or equal to 5%, directly delete samples with a large number of missing values ​​to ensure data integrity. Among them, if the data is normally distributed, use mean filling more often; if the data is more skewed, use median filling more often.

[0075] ③Remove the background of the previous check value

[0076] The environmental conditions of the battery when collecting EIS data, such as whether the temperature is abnormal (deviation from the set value> 5°C), whether the equipment is working normally (the equipment is not abnormal and the test process is not interrupted), etc. If there are abnormal values ​​caused by such factors, the data needs to be collected again.

[0077] ④ Basis for retention

[0078] First, pay attention to the meaning of this value. If the outlier can be explained based on the physical principles of battery performance, such as short-term fluctuations that may occur in specific battery aging stages or operating conditions, but does not represent battery failure, then it can be retained. Secondly, look at the overall characteristics of the data distribution. If the existence of the outlier deviates from the overall data distribution, but removing it will cause the data distribution to be overly skewed or lose important information, affecting the evaluation of the overall battery performance, it also needs to be considered for retention.

[0079] B02: ① Standardization processing: The Z-Score standardization method is selected for standardization processing. Its calculation formula is Z = (x-μ) / σ, where μ is the mean value of the data set and σ is the standard deviation. This method achieves the dimensional unification of different dimensional features by mapping the data to a distribution with a mean of 0 and a standard deviation of 1, optimizes the efficiency of subsequent model solving, and retains the distance information between samples while processing the sample anomaly.

[0080] ②Feature selection F1:

[0081] Feature selection is a method of selecting part of the column data (with representative features) based on the original data. In order to efficiently screen out linearly related and nonlinearly related features, the Pearson correlation coefficient and mutual information method are used on the standardized data to quickly screen out linearly related features and nonlinearly related features. By calculating the Pearson correlation coefficient between the eigenvalue and SOH, setting the threshold and judging, the threshold is set to 0.3. If the absolute value of the correlation coefficient is less than 0.3, it is considered that the linear correlation between the value and SOH is low;

[0082] At the same time, combined with mutual information, a threshold is set by calculating the mutual information between the eigenvalue and SOH. The threshold is set to 0.1. If the calculated result is also less than the threshold, it indicates that the linear or nonlinear dependence between the eigenvalue and SOH is weak, that is, the correlation is low, and the eigenvalue can be eliminated.

[0083] ③Feature extraction F2:

[0084] Feature extraction is a method of mapping raw data to another vector space through a regular mathematical transformation. To achieve this goal, principal component analysis (PCA) and linear discriminant analysis (LDA) are used to reduce the dimension of feature data based on standardized data, converting the original feature space into a new feature space that is more compact and more representative.

[0085] PCA analysis calculates the covariance matrix between each feature, performs eigendecomposition on the covariance matrix, and obtains the eigenvalues ​​and their corresponding eigenvectors. The eigenvalues ​​are sorted, and the first few eigenvectors with a cumulative explained variance of more than 90% are selected as principal components. The original data set is converted to a new coordinate system defined by the selected principal components to achieve data dimensionality reduction.

[0086] LDA analyzes and calculates the scatter matrix, including the intra-class scatter matrix (SW) and the inter-class scatter matrix (SB), solves the generalized eigenvalue problem of S-1WSB, and finds the optimal linear discriminant. The eigenvectors corresponding to the first d largest eigenvalues ​​are selected as the new feature axis, where d is usually the number of categories minus one (if the data set contains C categories, then d = C-1), and the original data is projected into the new feature space determined by LDA, enhancing the difference between classes while reducing the dimension.

[0087] ④Feature combination F3:

[0088] The two types of data feature columns obtained based on feature extraction (F1) and feature selection (F2) are combined and spliced ​​in the horizontal direction to obtain a new feature combination F3.

[0089] Step B03:

[0090] ①Select the benchmark model:

[0091] Models such as support vector machines (SVM) are used to perform model prediction based on the final features constructed by feature extraction and feature selection. The rationality of constructed features is analyzed by comparing the accuracy of predicted values ​​with the true values.

[0092] ②Select verification indicators:

[0093] The evaluation index is mainly based on the custom AccRatio, which measures the relative deviation between the predicted value and the true value, ensuring that the model prediction is as close to the actual data as possible. The larger the AccRatio, the smaller the deviation between the predicted value and the true value, that is, the higher the prediction accuracy of the model. Formula:

[0094]

[0095] Among them, yi represents the true value of the i-th sample, represents the predicted value of the i-th sample, and n represents the total number of samples. The range of AccRatio is between (0,1]. The larger the value, the better the prediction effect. When the predicted value is completely accurate, the score tends to 1. When the prediction deviation is large, the score will gradually decrease.

[0096] ③Feature screening verification:

[0097] Using models such as SVM, model prediction is performed on the three types of features F1, F2, and F3 respectively. By comparing the accuracy of the predicted values ​​and the true values, the optimal feature column combination (data set) and its corresponding feature processing method are obtained.

[0098] like Figure 1 As shown:

[0099] The EIS spectrum was collected to obtain the AC impedance data, and then the characteristic parameters were obtained. The EIS spectrum was subjected to DRT analysis. τ1 represents the charge transfer resistance, τ2 represents the solid electrolyte interface (SEI) film resistance, and τ3 represents the diffusion resistance. The extracted characteristic parameters are shown in Table 1 below:

[0100] AC impedance characteristic parameters <![CDATA[R ct (t1)]]> <![CDATA[R sei (t2)]]> <![CDATA[R Wb (t3)]]> Time constant (τ) 0.00324 0.0181 0.0389 <![CDATA[Peak area (P area )]]> 0.00174 0.00255 0.00428 <![CDATA[Peak height (P h )]]> 0.00321 0.00224 0.00299 Full width at half maximum (FWHM) 0.53929 1.13 1.47063

[0101] Table 1

[0102] like Figure 2 As shown, the DC internal resistance characteristic parameters are obtained, and the DC internal resistance data is obtained by collecting the pulse curve, and then the characteristic parameters are calculated;

[0103] Among them, the ohmic internal resistance R0 and the polarization internal resistance Rp are uniformly calculated using the charging side data, and the calculation of each characteristic parameter is as shown in Table 2:

[0104] DC internal resistance characteristic parameters Calculation process result <![CDATA[Ohmic internal resistance (R0)]]> ΔUt3 / ΔIt3 10.31mΩ <![CDATA[Polarization internal resistance (R p )]]> (U4-U3) / ΔIt3 4.10mΩ <![CDATA[Platform voltage (V plat )]]> U1 3.28V <![CDATA[Rebound voltage (V R )]]> U3-U2 0.328V <![CDATA[Pressure Drop (V D )]]> U4-U5 0.161V

[0105] Table 2

[0106] Feature parameter set:

[0107] The obtained battery characteristic parameters are summarized and analyzed, such as Figure 3 As shown; Xn represents the characteristic parameter of AC impedance, and Yn represents the characteristic parameter of DC internal resistance, as shown Figure 4 As shown:

[0108] This method is based on the electrochemical perspective and solves the limitation of incomplete coverage of characteristic parameters in the past. It integrates the two impedance characteristics, analyzes and quantifies the internal impedance state of the battery, extracts key characteristic parameters, and ensures the validity and reliability of the modeling data.

[0109] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A characteristic parameter set testing method for predicting the health status of a power battery, characterized in that: The following steps are involved: Feature parameter extraction and feature parameter set formation; Feature parameter extraction includes the following steps: A01: EIS data processing: Perform EIS medium frequency range test on the same batch of the same type of power batteries with different SOH aging degrees to obtain the battery's AC impedance data; A02: Perform DRT analysis on the EIS spectrum and extract peak characteristics as characteristic parameters; A03: HPPC data processing: Perform HPPC tests on power batteries with different SOH to obtain the DC internal resistance data of the battery; A04: Extract single pulse data and calculate characteristic parameters; A05: EIS and HPPC experimental tests under multiple conditions; The formation of the characteristic parameter set includes the following steps: B01: Data processing software, identification and removal of outliers, background and retention basis for removal of previously checked values; B02: standardization, feature selection F1, feature extraction F2 and feature combination F3; B03: Select a benchmark model, select validation metrics, and perform feature screening validation.

2. A characteristic parameter set testing method for power battery health status prediction according to claim 1, characterized in that: The EIS intermediate frequency range is from 1 kHz to 1 Hz.

3. A characteristic parameter set testing method for power battery health status prediction according to claim 2, characterized in that: The EIS and HPPC experimental tests under multiple conditions include temperature testing and state of charge testing; Temperature test: Test at three temperatures: -20°C, 25°C and 55°C to evaluate the performance of the battery in low temperature, room temperature and high temperature environments; State of Charge Test: Tests are performed every 10% from 10% to 100% SOC to evaluate the performance of the battery at different states of charge.

4. A characteristic parameter set testing method for power battery health status prediction according to claim 1, characterized in that: In step B01, all characteristic parameters extracted from the EIS test and HPPC test are presented in xlsx format after outliers are identified and removed. The characteristic data are processed into npz format with high reading efficiency using basic data processing methods based on Python algorithm libraries such as Pandas, Numpy and Scikit-learn.

5. A characteristic parameter set testing method for power battery health status prediction according to claim 4, characterized in that: In the box plot, the box is constructed based on the quartiles (Q1, Q2, Q3) of the data. The box contains the middle 50% of the data and extends up and down to a certain multiple of the interquartile range (IQR = Q3-Q1). Data points less than Q1-1.5*IQR or greater than Q3+1.5*IQR are considered outliers and are directly deleted from the data set. After identifying and removing outliers, if the proportion of outliers in the current column is less than 5%, the mean filling, median filling, forward filling, backward filling or interpolation method are used for appropriate filling. If the proportion is greater than or equal to 5%, samples with a large number of missing values ​​are directly deleted to ensure the integrity of the data. Among them, if the data is normally distributed, the mean filling is often used; if the data has more skew, the median filling is often used.

6. A characteristic parameter set testing method for power battery health status prediction according to claim 5, characterized in that: The environmental conditions of the battery when collecting EIS data, such as whether the temperature is abnormal (deviation from the set value> 5°C), whether the equipment is working normally (the equipment is not abnormal and the test process is not interrupted), etc. If there are abnormal values ​​caused by such factors, the data needs to be collected again.

7. A characteristic parameter set testing method for power battery health status prediction according to claim 6, characterized in that: The Z-Score standardization method was selected for standardization processing, and its calculation formula is Z = (x-μ) / σ, where μ is the mean of the data set and σ is the standard deviation.

8. According to the feature parameter set testing method for power battery health status prediction according to claim 1 or 6, a benchmark model is selected: a model such as a support vector machine (SVM) is used to perform model prediction based on the final features constructed by feature extraction and feature selection. The rationality of constructing the features is analyzed by comparing the accuracy of the predicted value with the true value.

9. According to the characteristic parameter set test method for predicting the health status of power batteries described in claim 8, select the verification index: the evaluation index is mainly based on the customized AccRatio, which measures the relative deviation between the predicted value and the true value, ensuring that the model prediction is as close to the actual data as possible. The larger the AccRatio, the smaller the deviation between the predicted value and the true value, that is, the higher the prediction accuracy of the model. Formula: in, yi represents the true value of the i-th sample, represents the predicted value of the i-th sample, and n represents the total number of samples. The range of AccRatio is between (0,1]. The larger the value, the better the prediction effect. When the predicted value is completely accurate, the score tends to 1. When the prediction deviation is large, the score will gradually decrease.

10. According to the feature parameter set testing method for predicting the health status of a power battery as described in claim 9, a model such as SVM is used to perform model prediction on the three types of features F1, F2, and F3 respectively, and the optimal feature column combination is obtained by performing precision comparison analysis between the predicted values ​​and the true values.

Images