EIS data quality inspection method and device based on Lin-KK verification

Abstract

The invention provides an EIS data quality inspection method based on Lin-KK verification, and belongs to the field of EIS data qualification inspection.The method comprises the steps that experimental EIS data of a battery is obtained, the maximum order and the minimum order of an RC element in a Lin-KK model are set, fitting results of the experimental EIS data under different orders are calculated, and the optimal order is found based on the fitting results; fitting is carried out on the generated EIS data based on a Lin-KK model with the RC element order being the optimal order, and Lin-KK impedance of the generated EIS data is obtained; the residual error of each frequency point in the EIS data is generated based on Lin-KK impedance calculation, abnormal points with the residual errors larger than a set threshold value are removed, and qualified data are obtained; the invention further provides an EIS data quality inspection device based on Lin-KK verification. The physical authenticity of the generated EIS data is ensured, and the stability and consistency of the generated EIS data on a global model structure are also ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of EIS data qualification inspection, and particularly to an EIS data quality inspection method and device based on Lin-KK verification. Background Art

[0002] Electrochemical Impedance Spectroscopy (EIS), as an important tool for characterizing the state of lithium batteries, can obtain multiple electrochemical characteristics of the system through frequency response. However, with the increasing diversification of EIS data acquisition methods, especially in the case of generating EIS data using non-traditional methods (such as artificial intelligence), how to ensure data quality has become an urgent problem to be solved.

[0003] Traditional EIS data quality inspection methods mainly target data collected by high-precision electrochemical workstations, and such data usually has high quality. In contrast, some EIS data generated based on artificial intelligence often rely on data collected by electrochemical workstations as the training basis. For example, the Chinese patent application for invention "Battery impedance spectrum prediction method, device, medium and equipment based on MAE model" with the publication number CN118191613A discloses a battery impedance spectrum prediction method based on the MAE model, including: obtaining the actual battery state data of the target battery, where at least one of the actual current data, actual voltage data, and actual temperature data in the actual battery state data is missing; inputting the actual battery state data into the trained encoder in the trained MAE model to obtain the latent vector corresponding to the actual battery state data, and the trained MAE model is a trained MAE model that can restore the missing data in the actual current data, actual voltage data, and actual temperature data; inputting the latent vector into the impedance spectrum prediction model to obtain the impedance spectrum data of the target battery. However, it only realizes the prediction of impedance spectrum data, but does not verify the quality of the prediction results. Due to the inherent black-box characteristics of artificial intelligence methods, the generated data may not be fully physically constrained, and it is impossible to ensure that the generated data strictly follows basic principles such as causality, linearity, and the Kramers-Kronig relationship, so the data quality is difficult to guarantee. Summary of the Invention

[0004] The technical problem to be solved by the present invention is: how to inspect the qualification of EIS data generated by non-traditional methods and ensure that the generated EIS data has inherent physical authenticity and conforms to the physical characteristics inherent in the electrochemical system.

[0005] The present invention solves the above technical problem through the following technical solutions: an EIS data quality inspection method based on Lin-KK verification, the method includes:

[0006] S1. Obtain the experimental EIS data of the battery, set the maximum and minimum orders of the RC elements in the Lin-KK model, calculate the fitting results of the experimental EIS data at different orders, and find the optimal order based on the fitting results;

[0007] S2. Fit the generated EIS data based on the Lin-KK model with the optimal order of the RC elements to obtain the Lin-KK impedance of the generated EIS data;

[0008] S3. Calculate the residuals of each frequency point in the generated EIS data based on the Lin-KK impedance, and remove the abnormal points with residuals greater than the set threshold to obtain qualified data.

[0009] First, the present invention uses the experimental EIS data to determine the optimal equivalent circuit RC element order of the current battery through the Lin-KK method, and this order represents the global topological characteristics of the overall impedance spectrum of the battery; then, keep the same order on the generated EIS data, and finely fit the remaining model parameters. By judging the residuals of each frequency point in the generated EIS data, only when the residuals are lower than the set threshold, the data is considered to highly coincide with the true dynamic behavior of the electrochemical system at the local parameter level. Through the comprehensive verification of the two aspects of the physical internal consistency of the generated EIS data and the model topology and parameter consistency, it not only ensures the physical authenticity of the generated EIS data, but also guarantees the stable consistency of the generated EIS data in the global model structure, providing a solid data basis for subsequent battery state evaluation and accurate modeling.

[0010] Preferably, the experimental EIS data of the battery in S1 is obtained by testing the battery with an electrochemical workstation under test standards. The experimental EIS data includes multiple positive frequency points and the real part data and imaginary part data corresponding to each frequency point.

[0011] Preferably, the impedance Z exp,LinKK (ω) of the Lin-KK model in S1 is:

[0012]

[0013] Among them, is the DC resistance in the equivalent circuit of the lithium battery, is the resistance of the kth RC element in the equivalent circuit of the lithium battery, is the time constant of the kth RC element in the equivalent circuit of the lithium battery, and M is the order of the RC elements in the equivalent circuit of the lithium battery.

[0014] Preferably, the calculation process of the fitting results of the experimental EIS data at different orders in S1 is as follows:

[0015] Set the DC resistance Resistance Time constant The initial value of, at each order, simultaneously adjusts the resistance and the time constant DC resistance The value of is fixed or slightly adjusted within a small range, and iterative calculations are performed until the preset convergence condition is met to obtain the best-fit parameters at this order. By analogy, the best-fit parameters at all orders are obtained.

[0016] Preferably, the calculation process for finding the optimal order based on the fitting result in S1 is as follows:

[0017] Based on the best-fit parameters at each order, calculate the real part residual ΔRe i of the experimental EIS data at each frequency point ω exp (ω i ) and the imaginary part residual ΔIm exp (ω i ):

[0018]

[0019]

[0020] Calculate the average residual Avg_Residual (M,exp) at each order M based on the real part residual and the imaginary part residual:

[0021]

[0022] Take the order corresponding to the minimum average residual value as the optimal order M optimaI :

[0023]

[0024] where Z exp,Re (ω i ), Z exp,Im (ω i ) are the real part and the imaginary part of the experimental EIS data Z exp (ω i ) at the frequency point ω i respectively, and Z exp,LinKK,Re (ω i ), Z exp,LinKK,Im (ω i ) are the real part and the imaginary part of the impedance obtained by fitting through the Lin-KK model respectively, and N is the number of frequency points.

[0025] Preferably, the EIS data generated in S2 is generated by harmonic injection or generated by artificial intelligence.

[0026] Preferably, the Lin-KK impedance Z of the generated EIS datagen,LinKK (ω) is as follows:

[0027]

[0028] Wherein, R gen,ohm , R gen,k , τ gen,ohm are respectively the DC resistance, the resistance of the k-th RC element, and the time constant of the k-th RC element in the equivalent circuit after fitting when the order of the RC element is the optimal order M optimal .

[0029] Preferably, the residual of each frequency point in the generated EIS data in S3 includes the real part residual ΔRe gen (ω i ), and the imaginary part residual ΔIm gen (ω i ):

[0030]

[0031] Wherein, Z gen,Re (ω i ), Z gen,Im (ω i ) are respectively the real part and the imaginary part corresponding to the frequency point ω gen (ω i ) in the generated EIS data Z i ; Z gen,LinKK,Re (ω i ), Z gen,LinKK,Im (ω i ) are respectively the real part and the imaginary part of the impedance obtained by fitting with the Lin-KK model when the order of the RC element is the optimal order;

[0032] When the real part residual ΔRe gen (ω i ) or the imaginary part residual ΔIm gen (ω i ) is greater than the set threshold, the data corresponding to this frequency point is determined as an abnormal point, and the abnormal point is removed from the generated EIS data to obtain qualified data.

[0033] Preferably, the set threshold is 0.01 or 0.001.

[0034] The present invention also provides an EIS data quality inspection device based on Lin-KK verification. The device includes:

[0035] A model structure optimization module, configured to obtain the experimental EIS data of the battery, set the maximum order and the minimum order of the RC element in the Lin-KK model, calculate the fitting results of the experimental EIS data at different orders, and find the optimal order based on the fitting results;

[0036] A generated EIS data fitting module is used to fit the generated EIS data based on the Lin-KK model with the optimal order of RC elements to obtain the Lin-KK impedance of the generated EIS data.

[0037] A data quality inspection module is used to calculate the residuals of each frequency point in the generated EIS data based on the Lin-KK impedance, eliminate the abnormal points with residuals greater than the set threshold, and obtain qualified data.

[0038] The advantages provided by the present invention are as follows:

[0039] (1) Firstly, the present invention uses the experimental EIS data from a high-precision electrochemical workstation to determine the optimal equivalent circuit RC element order of the current battery through the Lin-KK method, and this order represents the global topological characteristics of the overall impedance spectrum of the battery; then, the same order is maintained for the generated EIS data, and the remaining model parameters are finely fitted. By judging the residuals of each frequency point in the generated EIS data, only when the residuals are lower than the set threshold, the data is considered to be highly consistent with the true dynamic behavior of the electrochemical system at the local parameter level. Through the comprehensive verification of the physical internal consistency of the generated EIS data and the consistency of the model topology and parameters, it not only ensures the physical authenticity of the generated EIS data, but also guarantees the stable consistency of the generated EIS data in the global model structure, which is applicable to the qualification inspection of EIS data generated by non-traditional methods and provides a solid data basis for subsequent battery state evaluation and accurate modeling.

[0040] (2) The present invention establishes the physical connection between the generated data and the experimental data by reusing the Lin-KK verification method. Firstly, the RC order is initialized and set within a relatively wide range, which basically covers the possible RC orders of the lithium battery equivalent circuit model. By fitting the experimental EIS data and calculating the real part residuals and imaginary part residuals at each order, the average residual is calculated based on the real part residuals and imaginary part residuals, and the order corresponding to the minimum value of the average residual is determined as the optimal order. This optimal fitting order has the function of representing the basic information of the lithium battery equivalent circuit. When inspecting the quality of the generated EIS data, in order to ensure the consistency of the model under different working conditions and facilitate comparison, the present invention fixes the RC element order of the model as the optimal order obtained by the above fitting, and then fits the remaining three parameters of the generated EIS data through the Lin-KK model, making the evaluation benchmark of the generated EIS data consistent with that of the experimental EIS data, and successfully connecting the experimental EIS data with the generated EIS data.

[0041] (3) The present invention adopts an automated residual calculation and data screening process, avoiding manual intervention, improving the inspection efficiency and accuracy, and being able to effectively enhance the reliability of EIS data inspection. It can be widely applied in fields such as electrochemical research, lithium battery analysis, and material research, and has strong practical application value. Description of the Drawings

[0042] Figure 1 It is a flowchart of the EIS data quality inspection method based on Lin-KK verification provided by an embodiment of the present invention;

[0043] Figure 2 It is a schematic diagram of the EIS data quality inspection device based on Lin-KK verification provided by an embodiment of the present invention. Detailed Embodiments

[0044] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the following describes the technical solutions of the present invention clearly and completely in conjunction with specific embodiments and with reference to the accompanying drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0045] Embodiment 1

[0046] As Figure 1 shown, this embodiment provides an EIS data quality inspection method based on Lin-KK verification, including the following steps:

[0047] Step 1: Obtain the experimental EIS data of the battery. The experimental EIS data is obtained by testing the battery with a high-precision electrochemical workstation under test standards. The experimental EIS data includes frequency data ω i , real part data Z exp,Re (ω i ), and imaginary part data Z exp,Im (ω i ). Among them, the frequency data ω i includes a series of positive frequency points, the real part data Z exp,Re (ω i ) is the real part corresponding to each positive frequency point, and the imaginary part data Z exp,Im (ω i ) is the imaginary part corresponding to each positive frequency point.

[0048] Set the maximum and minimum orders of the RC components in the Lin-KK model, calculate the fitting results of the experimental EIS data at different orders, and find the optimal order based on the fitting results.

[0049] Among them, the impedance Z of the Lin-KK modelexp,LinKK ω is:

[0050]

[0051] wherein, is the DC resistance in the equivalent circuit of the lithium battery, is the resistance of the k-th RC element in the equivalent circuit of the lithium battery, is the time constant of the k-th RC element in the equivalent circuit of the lithium battery, and M is the order of the RC elements in the equivalent circuit of the lithium battery.

[0052] The present invention sets the maximum order M of the RC elements in the Lin-KK model max to be 50, and the minimum order M min to be 1. In other embodiments, the model order can be simplified. In practical applications, when there are already equivalent circuit parameters with high precision, the contribution of each order of RC elements to the overall response can be further analyzed for certain batteries. Specifically, by performing least squares fitting on the experimental EIS data and observing the change in the fitting error at different orders, if it is found that increasing the order has very limited improvement on the fitting accuracy, it indicates that the contribution of the high-order RC elements is small. Thus, based on the original set maximum order (e.g., 50) and minimum order (e.g., 1), the order range can be appropriately reduced according to the characteristics of the data, such as selecting a more appropriate order within 10 orders above and below the original parameters. This method can not only simplify the model structure and improve the calculation efficiency, but also make the model more in line with the requirements of practical applications while ensuring the fitting accuracy. And the least squares method is used to fit the experimental EIS data at different orders. The fitting process includes:

[0053] Set the DC resistance resistance time constant initial values. For example, the average value of the experimental EIS data in the high-frequency region can be used as the initial value of the DC resistance initial value, and a small positive value can be selected as the initial value of the resistance initial value, and a fixed value between 1 ms and 10 ms can be selected as the initial value of the time constant , or a judgment can be made by analyzing the data characteristics.

[0054] The value of the order M is between M min and M max . In this embodiment, the maximum order M max is 50, and the minimum order M mun is 1. For each selected order M, the resistance and the time constant DC resistance The value is fixed or slightly adjusted within a small range, and iterative calculations are performed until a preset convergence condition is met to obtain the best fitting parameters at this order. The preset convergence condition can be that the change in the model parameters to be fitted (DC resistance resistance time constant ) is less than the set threshold, or the number of iterations reaches the upper limit value. And so on, the best fitting parameters at all orders are obtained. The best fitting parameters include DC resistance resistance time constant

[0055] Based on the best fitting parameters at each order, calculate the real part residual ΔRe u and the imaginary part residual ΔIm exp of the experimental EIS data at each frequency point ω i : exp (ω u ) are calculated as follows:

[0056]

[0057] Calculate the average residual Avg_Residual (M,exp) at each order M according to the real part residual and the imaginary part residual:

[0058]

[0059] Take the order corresponding to the minimum average residual value as the optimal order M optimal :

[0060]

[0061] where Z exp (ω i ) is the experimental EIS data, Z exp,Re (ω i ), Z exp,Im (ω i ) are the real part and the imaginary part of the impedance corresponding to the frequency point ω i in the experimental EIS data respectively, Z exp,LinKK,Re (ω i ), Z exp,LinKK,Im (ω i ) are the real part and the imaginary part of the impedance obtained by fitting through the Lin-KK model respectively, and N is the number of frequency points.

[0062] Step 2: Fit the generated EIS data based on the Lin-KK model with the optimal order of RC elements to obtain the Lin-KK impedance of the generated EIS data; the generated EIS data is generated by harmonic injection or by artificial intelligence, and the generated EIS data includes frequency data ω i 、real part data Z gen,Re (ω i ), imaginary part data Z gen,Im (ω i ), where the frequency data ω i includes a series of positive frequency points, and the real part data Z gen,Re (ω i ) is the real part corresponding to each frequency point, and the imaginary part data Z gen,Im (ω i ) is the imaginary part corresponding to each frequency point.

[0063] The generated EIS data of the present invention is obtained by non-traditional methods (such as artificial intelligence). The following introduces a method for generating EIS data based on a physics-guided neural network.

[0064] S201: During the charging and discharging process of the lithium-ion battery, collect multiple groups of battery sample data, preprocess the battery sample data to obtain the preprocessed battery sample data, and divide the preprocessed battery sample data into a test set, a data set one, and a data set two. For example, divide the preprocessed battery sample data into 10 parts, 1 part as the test set, and the remaining 9 parts as the data set for training and validating the model. The battery sample data includes current data, voltage data, and temperature data, and the temperature data refers to the battery surface temperature signal. The data sampling rate is generally set to about 10 Hz, thus forming a time series data matrix of size (3, T), where "3" represents three channels of voltage, current, and temperature, and T is the number of time steps within the sampling duration. During the process of obtaining each group of battery sample data, simultaneously test the impedance value corresponding to the sample battery at a certain frequency through an impedance spectroscopy test instrument. The impedance spectroscopy data includes impedance values at different frequencies, and calculate the geometric parameters of the battery in different states according to the impedance spectroscopy.

[0065] The process of preprocessing the collected battery sample data includes:

[0066] Filter the battery sample data to remove interference signals and obtain the battery sample data with noise removed;

[0067] Normalize the battery sample data with noise removed to obtain the normalized battery sample data to eliminate the influence of different measurement ranges on model training;

[0068] Calibrate the normalized battery sample data to obtain the preprocessed battery sample data, ensuring the consistency and accuracy of the data at each sampling point. Through the preprocessing of the battery sample data, the high quality and stability of the input data for the subsequent network model can be guaranteed.

[0069] The process of calculating the geometric parameters of the battery in different states based on the impedance spectrum includes:

[0070] Input the impedance spectrum (the real part resistance value Z′(f i ) and the imaginary part resistance value Z″(f i )) arranged from high to low in frequency, filter and fit using the Savitzky-Golay filter to obtain a continuous EIS curve;

[0071] Analyze the change trends of the first derivative and second derivative of the EIS curve to obtain the division positions of the low-frequency, mid-frequency, and high-frequency regions; among them, the first derivative reflects the change in the curve curvature, and the second derivative is used to capture the inflection point. The first derivative rapidly drops from a large positive value to close to zero (corresponding to the transition from a vertical line to a semicircle), representing the physical meaning that the ohmic impedance dominates → the charge transfer impedance dominates. The second derivative has a local maximum value (corresponding to the end of the semicircular arc and the start of the Warburg impedance), representing the physical meaning that the charge transfer impedance dominates → the diffusion process dominates. Based on the first derivative, the boundary from high frequency to mid frequency is obtained, and based on the second derivative, the boundary from mid frequency to low frequency is obtained.

[0072] Select the abscissa value of the intersection point of the high-frequency region of the EIS curve and the real axis as the parameter g, use the least squares method to fit the semicircle equation corresponding to the mid-frequency region to obtain the radius value of the circle as the parameter r, and use the real part and imaginary part data in the low-frequency region to fit the straight line equation in the low-frequency region using the least squares method. The slope value of the straight line equation is the parameter θ.

[0073] See Figure 1 , the three parameters of geometric parameters g, r, and θ respectively correspond to the key geometric features of the impedance spectrum. Among them, the parameter g is the voltage value corresponding to the intersection point of the impedance spectrum and the real axis in the Nyquist diagram, indicating the ohmic resistance of the battery. The parameter r is the fitting radius of the semicircle in the mid-frequency region of the Nyquist diagram, reflecting the interfacial reaction and related electrochemical processes. The parameter θ is the slope angle of the straight line in the low-frequency region of the Nyquist diagram, reflecting the diffusion process and transport limitation effect.

[0074] S202: Use Dataset 1 as the input and the geometric parameters as the output to train the MLP network. When the loss function is minimized, obtain the trained MLP network. Input Dataset 2 into the trained MLP network to obtain the geometric features.

[0075] The dataset used for training and validation in S201 can be divided into nine parts, and two parts are selected as the first dataset for training the MLP network. The first dataset is used as the input, and the geometric parameters are used as the output. The MLP network is trained using the mean square error (MSE) as the loss function. When the loss function is minimized, the trained MLP network is obtained. The trained MLP network can extract geometric features from the input current, voltage, and temperature data that highly match the actual impedance spectrum data. After training, the parameters of the MLP network are frozen and used for geometric feature generation in the second training stage.

[0076] The process of training the MLP network includes:

[0077] Input the battery sample data (two-dimensional data with a size of 3×T) into the input layer of the preset MLP network, and flatten it to obtain a one-dimensional input vector;

[0078] Input the input vector into the first hidden layer of the preset MLP network to map it to a high-dimensional feature space, such as a 256-dimensional feature space. The first hidden layer uses a non-linear activation function (such as Swish or Mish) to enhance the feature expression ability and realize the extraction of the electrochemical features and impedance spectrum geometric morphology hidden in the charge and discharge data. The mathematical expression of the non-linear activation function Swish is:

[0079]

[0080] The non-linear activation function Swish can provide a smoother gradient flow in the network, thereby accelerating the training process and avoiding the problem of gradient disappearance.

[0081] Input the output of the first hidden layer into the second hidden layer for further mapping, and keep the output high-dimensional. The high dimension in the present invention is 256;

[0082] Input the output of the second hidden layer into the output layer, and map to obtain geometric parameters (parameter g, parameter r, parameter θ). The equivalent parameters g, r, and θ are directly predicted from the operating condition data through the deep learning network, which not only retains the physical correlation between the EIS characteristics and the battery aging mechanism (such as parameter g reflecting the ohmic impedance and parameter r corresponding to the charge transfer process), but also gets rid of the dependence on high-frequency impedance detection equipment.

[0083] S203. Input the second dataset into the Encoder network to obtain a feature vector. After fusing the feature vector with the geometric features, input it into the Decoder network with the impedance spectrum as the output. Train the LSTM-based Encoder-Decoder network. When the loss function is minimized, the trained Encoder-Decoder network is obtained. An impedance spectrum prediction model is constructed based on the trained MLP network and the trained Encoder-Decoder network.

[0084] Use the remaining 7 samples in the dataset for training and validation as Dataset 2 to train the Encoder-Decoder network. Input the battery sample data of Dataset 2 into the trained MLP network to obtain geometric features. Input the battery sample data into the Encoder network to generate the feature vector f. * The feature vector f generated by the Encoder network * represents the dynamic features of current, voltage, and temperature time series data and can describe the dynamic changes in the battery charging and discharging process. The feature vector f * is concatenated with the geometric features (parameter g, parameter r, parameter θ) to obtain the comprehensive feature vector F. The comprehensive feature vector F is input into two Decoder networks respectively. One Decoder network outputs the real part Re corresponding to the impedance value at each frequency point, and the other Decoder network outputs the imaginary part Im corresponding to the impedance value at each frequency point. The impedance spectrum EIS can be obtained based on the real part Re and the imaginary part Im.

[0085] The Encoder-Decoder network adopts an encoder-decoder structure based on the improved long short-term memory network to realize the prediction of the full-frequency impedance spectrum. To improve the prediction performance of the model, in the present invention, when training the Encoder-Decoder network based on LSTM, it is divided into two stages. In the first stage, the Encoder network and the two Decoder networks are jointly trained to obtain the preliminarily trained Encoder-Decoder network, enhancing the ability of the Decoder network to extract features from the battery sample data. In the second stage, the parameters of the Encoder network are frozen, and then the two Decoder networks are independently trained respectively. Based on the feature vector f extracted by the encoder * to improve the mapping ability of the decoder for the real part and the imaginary part, and obtain the trained Encoder-Decoder network. The root mean square error (RMSE) is used as the loss function during the training process. When the loss function is minimized, the trained Encoder-Decoder network is obtained.

[0086] When training the MLP network and the LSTM-based Encoder-Decoder network, cross-validation and model fusion strategies can be adopted for training, which can effectively alleviate the additional prediction errors that may be caused by the sample distribution deviation of a single model and improve the prediction accuracy and robustness of the model. For example, when training the MLP network, the first dataset is divided into a training set and a validation set according to a set ratio. The training set is used to train the model and modify the model parameters during the training process. The validation set is used to validate the trained model and adjust the hyperparameters of the model. In the first dataset, a group is sequentially selected as the validation set, and the rest are used as the training set to obtain multiple groups of datasets. Each group of datasets includes a group of validation sets and multiple groups of training sets. Training and validating the MLP network with each group of datasets can obtain multiple independent trained MLP networks. Similarly, when training the Encoder-Decoder network, the second dataset is divided into a training set and a validation set according to a set ratio. In the second dataset, a group is sequentially selected as the validation set, and the rest are used as the training set to obtain multiple groups of datasets. Each group of datasets includes a group of validation sets and multiple groups of training sets. Training and validating the Encoder-Decoder network with each group of datasets can obtain multiple independent trained Encoder-Decoder networks.

[0087] S204. Input the test set into the impedance spectrum prediction model to predict the impedance spectrum. When inputting the test set into the impedance spectrum prediction model, the current, voltage, and temperature data are input into multiple trained MLP networks to obtain multiple groups of output results. The multiple groups of output results are weighted and fused to obtain geometric features. The test set is input into the Encoder network in multiple independent trained Encoder-Decoder networks to obtain feature vectors. The feature vectors and geometric features are concatenated and then input into the Decoder network to obtain multiple groups of output results. The multiple groups of output results are weighted and fused to obtain the predicted impedance spectrum, and the prediction performance of the model is evaluated based on the root mean square error. The calculation formula for the root mean square error RMSE is:

[0088]

[0089] where Re(Z i ) and Im(Z i ) respectively represent the real part and the imaginary part of the actual EIS at the i-th frequency point, and Re(Z i ′) and Im(Z i ′) are the prediction results.

[0090] The attention mechanism (Temporal Attention Mechanism) is introduced into the Encoder-Decoder network. This mechanism dynamically assigns different attention weights to each time step during the charge and discharge process of lithium batteries, enabling the model to focus on the most critical time periods during the charge and discharge process, so as to better pay attention to the battery behavior in critical time periods and more accurately predict the impedance spectrum.

[0091] The attention weight α for each time step t is calculated by the following formula:

[0092]

[0093] where, e t is the attention score for each time step t. The larger it is, the more attention is paid to that time step. It is obtained by weighted calculation of the hidden state h t of the LSTM and the input feature x t as follows:

[0094] e t = tanh(W h h t + W h x t + b)

[0095] where, W h is usually the weight matrix for linearly transforming the hidden state h t-1 or the output of other hidden layers, used to map the hidden vector to the same or additively compatible dimension as the input x t . b is the corresponding bias term, which is consistent with the usage of other biases in the neural network and is used to provide a learnable translation amount for the linear transformation. That is, W h and b are the weight matrix and bias term used for linear mapping of the hidden state or input vector respectively.

[0096] The present invention constructs an impedance spectrum prediction model using a two-stage network structure. In the first stage, by training an MLP network, a mapping relationship between battery sample data and geometric features is established to ensure that the geometric features output by the trained MLP network match the geometric form of the actual impedance spectrum. The geometric features closely related to the impedance spectrum form in the battery sample data are extracted through the trained MLP network, linking the battery charge and discharge behavior tightly with the impedance spectrum form, making the prediction results have clear physical meanings. In the second stage, through deep mapping, the predicted physical geometric features and the battery sample data are jointly input into an Encoder-Decoder network. The temporal feature vector of the battery sample data is extracted through the Encoder network, and after fusing the temporal feature vector with the geometric features, it is input into the Decoder network. The Encoder-Decoder network based on LSTM is trained, and an impedance spectrum prediction model is constructed according to the trained MLP network and the trained Encoder-Decoder network. This model not only inherits the adaptability of the data-driven method to operating condition data but also establishes an interpretable association with the electrochemical mechanism through an explicit physical feature layer. Through the organic fusion of geometric features and temporal features, the prediction results have clear physical interpretability and high prediction accuracy, and can achieve the goal of directly mapping the battery sample data collected during the charge and discharge process to the full-frequency domain impedance spectrum of the battery.

[0097] The impedance spectrum prediction method of the present invention can online predict the full-frequency domain impedance spectrum based on the collected battery charge and discharge data, without the need for traditional impedance spectrum testing instruments and external perturbation signals, and can solve the problems existing in the prior art such as expensive testing equipment, long testing cycle, and insufficient physical interpretation of data. Compared with the traditional EIS prediction method, it does not require interrupting the battery operation for impedance scanning and realizes real-time online monitoring; compared with the pure data-driven model, the present invention provides an explanation of the battery state change mode through the physical meanings of parameters g, r, θ (such as a sudden increase in the value of parameter r indicating the loss of active substances), helping to improve the prediction accuracy; compared with the hybrid model, the present invention first uses the impedance spectrum geometric features as intermediate variables rather than the final output, enabling the network to learn physical laws and data distributions simultaneously.

[0098] The PGNN network structure (with a parameter quantity more than 80% less than that of Transformer) has a lower computational latency. Compared with complex models (such as graph neural networks), the PGNN network has a lower overfitting risk for small-sample data, is more adaptable to the long-tail distribution characteristics of battery aging data, and has stronger robustness. The two-stage network structure combines temporal dynamic features and geometric morphological features. The model not only inherits the adaptability of data-driven methods to operating condition data but also establishes an interpretable association with the electrochemical mechanism through an explicit physical feature layer (parameter g, parameter r, parameter θ). The prediction robustness and accuracy are further improved through cross-validation and model fusion. The present invention is applicable to the prediction of impedance spectra of various types of lithium-ion batteries. Especially in the fields of electric vehicles, energy storage systems, etc., it can realize real-time online monitoring of battery impedance spectra and help build a precise battery management system.

[0099] Select the order of the RC element as the optimal order M obtained in the above step 1 optimal , and through the least squares method, such as the Levenberg-Marquardt algorithm, fit the generated EIS data. The fitting process is similar to that in the above step 1, that is, first set the DC resistance resistance time constant initial values. For example, the average value of the generated EIS data in the high-frequency region can be used as the DC resistance initial value, and a small positive value can be selected as the initial value of the resistance . A fixed value between 1 ms and 10 ms can be selected as the initial value of the time constant , or a judgment can be made by analyzing the characteristics of the generated data. During the fitting process, adjust the resistance and the time constant while keeping the DC resistance value fixed or slightly adjusted within a small range, and perform iterative calculations until the preset convergence condition is met to obtain the best fitting parameters R optimal , R gen,ohm , R gen,k , τ gen,ohm at the optimal order M.

[0100] The Lin-KK impedance Z gen,LinKK (ω) of the generated EIS data is:

[0101]

[0102] where R gen,ohm , R gen,k , τ gen,ohm are the DC resistance, resistance optimalWhen , the DC resistance, the resistance of the kth RC element, and the time constant of the kth RC element in the fitted equivalent circuit are given.

[0103] The present invention basically covers the RC orders that may appear in the lithium battery equivalent circuit model within a wide range (1-50), and determines the optimal equivalent circuit RC element order of the current battery through the Lin-KK method. This order represents the global topological characteristics of the overall impedance spectrum of the battery, that is, the optimal fitting order has the function of representing the basic information of the lithium battery equivalent circuit. When testing the quality of the generated EIS data, in order to ensure the consistency of the model under different working conditions and facilitate comparison, the present invention fixes the RC element order of the model to the optimal order obtained by the above fitting, and then fits the remaining three parameters of the generated EIS data through the Lin-KK model, so that the evaluation benchmark of the generated EIS data is consistent with the evaluation benchmark of the experimental EIS data, and successfully links the experimental EIS data with the generated EIS data.

[0104] Step 3: Generate the residual of each frequency point in the EIS data based on the Lin-KK impedance calculation, remove the abnormal points whose residual is greater than the set threshold, and obtain qualified data.

[0105] The residuals for each frequency point in the generated EIS data include the real residual ΔRe gen (ω i ), imaginary residual ΔIm gen (ω i ):

[0106]

[0107] Among them, Z gen (ω i ) is used to generate EIS data, Z gen,Re (ω i ), Z gen,Im (ω i ) are the frequency points ω in generating EIS data. i The corresponding real and imaginary parts, Z gen,LinKK,Re (ω i ), Z gen,LinKK,Im (ω i ) are the real and imaginary parts of the impedance respectively obtained by fitting the Lin-KK model when the order of RC elements is the optimal order.

[0108] When the real residual ΔRe gen (ω i ) or the imaginary residual ΔIm gen (ω i)When it is greater than the set threshold ∈, in this embodiment, the set threshold ∈ is 0.01 or 0.001. The data corresponding to this frequency point is determined as an abnormal point, and the abnormal point is removed from the generated EIS data to obtain qualified data. That is, only when the real part residual ΔRe gen (ω i ) ≤ ∈ and the imaginary part residual ΔIm gen (ω i ) ≤ ∈, the data at this frequency point is considered qualified. In other embodiments, the set threshold ∈ can be adaptively adjusted according to different experimental conditions and data quality to further improve the inspection accuracy.

[0109] After screening the generated EIS data through the above steps, the screened EIS data is output as qualified data. The screened EIS data removes unqualified frequency points, conforms to the standard of the Lin-KK model, and can be used for further analysis or modeling.

[0110] The EIS data quality inspection method based on Lin-KK verification of the present invention is applicable to the qualification inspection of EIS data generated by non-traditional methods. Although non-traditional methods such as artificial intelligence have significant advantages in data generation and pattern recognition, due to the opacity of their internal operation mechanisms, that is, the "black box" problem of data generated by artificial intelligence, such as deep neural networks. When the target EIS in the training set of the deep neural network barely passes the KK test, new uncertainties may be introduced when the network generalizes the test set, resulting in fluctuations in the reliability of the generated EIS data. The generated data may ignore or misunderstand some basic physical laws, resulting in limitations in the physical meaning of the output results.

[0111] The present invention first uses the experimental EIS data from a high-precision electrochemical workstation to determine the optimal equivalent circuit RC element order of the current battery through the Lin-KK method. This order represents the global topological characteristics of the overall impedance spectrum of the battery. Then, the same order is maintained on the generated EIS data, and the remaining model parameters are finely fitted. By judging the residuals of each frequency point in the generated EIS data, only when the residuals are lower than the set threshold, the data is considered to highly match the true dynamic behavior of the electrochemical system at the local parameter level. Through the comprehensive verification of the two major aspects of the physical internal consistency and the model topology and parameter consistency of the generated EIS data, not only the physical authenticity of the generated EIS data is ensured, but also the stable consistency of the generated EIS data in the global model structure is guaranteed, providing a solid data foundation for subsequent battery state assessment and accurate modeling.

[0112] Example 2

[0113] See Figure 2, an EIS data quality inspection device based on Lin-KK verification, the device includes:

[0114] A model structure optimization module, which is used to obtain the experimental EIS data of the battery, set the maximum and minimum orders of the RC elements in the Lin-KK model, calculate the fitting results of the experimental EIS data at different orders, and find the optimal order based on the fitting results; the experimental EIS data of the battery is obtained by testing the battery with an electrochemical workstation under test standards, and the experimental EIS data includes multiple positive frequency points and the real part data and imaginary part data corresponding to each frequency point. The impedance Z exp,LinKK (ω) of the Lin-KK model is:

[0115]

[0116] Among them, is the DC resistance in the equivalent circuit of the lithium battery, is the resistance of the kth RC element in the equivalent circuit of the lithium battery, is the time constant of the kth RC element in the equivalent circuit of the lithium battery, and M is the order of the RC elements in the equivalent circuit of the lithium battery.

[0117] The calculation process of the fitting results of the experimental EIS data at different orders is as follows:

[0118] Set the initial values of the DC resistance resistance time constant . At each order, while adjusting the resistance and the time constant the value of the DC resistance is fixed or slightly adjusted within a small range, and iterative calculation is performed until the preset convergence condition is met to obtain the best fitting parameters at this order. By analogy, the best fitting parameters at all orders are obtained.

[0119] The calculation process of finding the optimal order based on the fitting results is as follows:

[0120] Based on the best fitting parameters at each order, calculate the real part residual ΔRe i (ω exp ) and the imaginary part residual ΔIm i (ω exp ) of the experimental EIS data at each frequency point ω i :

[0121]

[0122]

[0123] Calculate the average residual Avg_Residual at each order M based on the real - part residual and the imaginary - part residual (M,exp) :

[0124]

[0125] Take the order corresponding to the minimum average residual value as the optimal order M optimaI :

[0126]

[0127] Among them, Z exp,Re (ω i ), Z exp,Im (ω i ) are the real part and the imaginary part corresponding to the frequency point ω exp (ω i ) in the experimental EIS data Z i respectively. Z exp,LinKK,Re (ω i ), Z exp,LinKK,Im (ω i ) are the real part and the imaginary part of the impedance obtained by fitting through the Lin - KK model respectively. N is the number of frequency points.

[0128] Generate an EIS data fitting module, which is used to fit the generated EIS data based on the Lin - KK model with the optimal order of RC elements, and obtain the Lin - KK impedance of the generated EIS data. The generated EIS data is generated by harmonic injection or generated by artificial intelligence. The Lin - KK impedance Z gen,LinKK (ω) is:

[0129]

[0130] Among them, R gen,ohm , R gen,k , τ gen,ohm are the DC resistance, the resistance of the k - th RC element, and the time constant of the k - th RC element in the equivalent circuit after fitting when the order of the RC element is the optimal order M optimal .

[0131] A data quality inspection module, which is used to calculate the residual of each frequency point in the generated EIS data based on the Lin - KK impedance, remove the abnormal points with residuals greater than the set threshold, and obtain qualified data. The residual of each frequency point in the generated EIS data includes the real - part residual ΔRe gen (ω i ), the imaginary - part residual ΔIm gen (ω i ):

[0132]

[0133] Among them, Z gen,Re (ω i ) and Z gen,Im (ω i ) are respectively the real part and the imaginary part of the frequency point ω gen (ω i ) corresponding to the generated EIS data Z i ; Z gen,LinKK,Re (ω i ) and Z gen,LinKK,Im (ω i ) are respectively the real part and the imaginary part of the impedance obtained by fitting the Lin-KK model when the order of the RC element is the optimal order;

[0134] When the real part residual ΔRe gen (ω i ) or the imaginary part residual ΔIm gen (ω i ) is greater than the set threshold, the data corresponding to this frequency point is determined as an abnormal point, and the abnormal point is removed from the generated EIS data to obtain qualified data. The set threshold of the present invention is 0.01 or 0.001.

[0135] Example 3

[0136] This example takes the EIS data to be tested generated by deep learning as an example to introduce the EIS data quality inspection method based on Lin-KK verification of the present invention.

[0137] Step 1: First, test the EIS data of the lithium battery through an electrochemical workstation on the premise of ensuring compliance with the test standards, and input the experimental data into the algorithm: Set the initial parameters: the maximum order Mmax = 50, and the minimum order Mmin = 1. According to the Lin-KK model, fit the experimental data by the least square method, calculate the fitting results at different orders M, and select the M value with the smallest average residual as the optimal order. After calculation, when M = 12, the obtained residuals are the smallest, representing the optimal order M optimal = 12 of the EIS fitting of this lithium battery in the current state.

[0138] Generate a set of EIS data of a lithium battery through deep learning, as shown in Table 1. The data includes the real part and the imaginary part at multiple frequency points, and the resistance value is named Z gen (ω i ), the real part is Z gen,Re (ω i ), and the imaginary part is Z gen,Im (ω i ).

[0139] Table 1 Generated EIS quantity

[0140]

[0141]

[0142] Step 2: Set the RC element order of the Lin-KK model to 12, and fit the generated EIS data using the least squares method to determine the model parameter R gen,ohm , R gen,k and τ gen,ohm , and obtain the Lin-KK impedance for generating EIS data.

[0143] Step 3: Calculate and generate the real residual ΔRe at each frequency point in the EIS data cen (ω i ) and the imaginary residual ΔIm gen (ω ω ), the set threshold value ∈ of the residual is 0.01, and the outliers are eliminated according to the set threshold value. For each frequency point, if its real residual or imaginary residual is greater than the threshold value, the data is considered to be an outlier. In this embodiment, the impedances corresponding to frequencies of 0.01, 0.08 and 0.2 are outliers. The impedances corresponding to frequencies of 0.01, 0.08 and 0.2 are eliminated from the generated EIS data, and finally qualified data is obtained and output, as shown in Table 2.

[0144] Table 2 Qualified data

[0145] Frequency (Hz) Real part resistance Imaginary part resistance 0.02 0.117617 0.046533 0.03 0.108673 0.038067 0.05 0.09803 0.028694 0.1 0.085391 0.020198 0.3 0.076002 0.012679 0.5 0.074076 0.009792 0.8 0.072663 0.008398 1 0.071708 0.007613 2 0.07098 0.006233 3 0.06955 0.005809 5 0.06874 0.00555 8 0.066375 0.005576 10 0.06568 0.005676 11 0.065661 0.005559 21 0.06346 0.005957 31 0.061879 0.006895 61 0.05899 0.007119 81 0.057658 0.007187 100 0.056372 0.006874 110 0.056085 0.006666 210 0.052784 0.006629 310 0.05097 0.006136 510 0.048762 0.004934 810 0.046775 0.003128 1000 0.045726 0.001993

[0146] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. EIS Data Quality Inspection Method Based on Lin-KK Verification Characterized in that: The method includes: S1. Obtain the experimental EIS data of the battery, set the maximum and minimum orders of the RC components in the Lin-KK model, calculate the fitting results of the experimental EIS data at different orders, and find the optimal order based on the fitting results; S2. Fit the generated EIS data based on the Lin-KK model with the optimal order of the RC component to obtain the Lin-KK impedance of the generated EIS data; S3. Calculate the residuals of each frequency point in the generated EIS data based on the Lin-KK impedance, and remove the abnormal points with residuals greater than the set threshold to obtain qualified data.

2. The EIS data quality inspection method based on Lin-KK verification according to claim 1 Characterized in that: In S1, the experimental EIS data of the battery is obtained by testing the battery with an electrochemical workstation under test standards. The experimental EIS data includes multiple positive frequency points and the real part data and imaginary part data corresponding to each frequency point.

3. The EIS data quality inspection method based on Lin-KK verification according to claim 1 Characterized in that: The impedance Z of the Lin-KK model in S1 exp,LinKK (ω) is as follows: Among them, is the DC resistance in the equivalent circuit of the lithium battery, is the resistance of the k-th RC element in the equivalent circuit of the lithium battery, is the time constant of the k-th RC element in the equivalent circuit of the lithium battery, and M is the order of the RC elements in the equivalent circuit of the lithium battery.

4. The EIS data quality inspection method based on Lin-KK verification according to claim 1 Characterized in that: The calculation process of the fitting results of the experimental EIS data at different orders in S1 is: Set the DC resistance Resistance Time constant The initial value of, at each order, simultaneously adjust the resistance And the time constant DC resistance The value is fixed or slightly adjusted within a small range, and iterative calculations are performed until the preset convergence condition is met to obtain the best fitting parameters at this order, and so on to obtain the best fitting parameters at all orders.

5. The EIS data quality inspection method based on Lin-KK verification according to claim 4 Characterized in that: The calculation process of finding the optimal order based on the fitting results in S1 is: Based on the best - fit parameters at each order, calculate the real - part residuals ΔRe i of the experimental EIS data at each frequency point ω exp (ω i ) and the imaginary - part residuals ΔIm exp (ω i ) as follows: Calculate the average residual Avg_Residual at each order M based on the real part residual and the imaginary part residual (M,exp) : Take the order corresponding to the minimum average residual value as the optimal order M optimal : Among them, Z exp,Re (ω i ) and Z exp,Im (ω i ) are the real part and the imaginary part of the impedance corresponding to the frequency point ω i in the experimental EIS data Z exp (ω i ), respectively. Z exp,LinKK,Re (ω i ) and Z exp,LinKK,Im (ω i ) are the real part and the imaginary part of the impedance obtained by fitting through the Lin-KK model, respectively. N is the number of frequency points.

6. The EIS data quality inspection method based on Lin-KK verification according to claim 1 Characterized in that: In S2, the generated EIS data is generated by harmonic injection or generated by artificial intelligence.

7. The EIS data quality inspection method based on Lin-KK verification according to claim 1 Characterized in that: Lin-KK impedance Z for generating EIS data gen,LinKK (ω) is as follows: where R gen,ohm , R hen,k , τ gen,ohm are respectively the DC resistance, the resistance of the k-th RC element, and the time constant of the k-th RC element in the equivalent circuit after fitting when the order of the RC element is the optimal order M optimal .

8. The EIS data quality inspection method based on Lin-KK verification according to claim 1 Characterized in that: The residuals of each frequency point in the EIS data generated in S3 include the real part residual ΔRe gen (ω i ), the imaginary part residual ΔIm gen (ω i ): Among them, Z gen,Re (ω i ), Z gen,Im (ω i ) are the real part and the imaginary part of the frequency point ω gen (ω i ) corresponding to in the generated EIS data Z i ; Z gen,LinKK,Re (ω i ), Z gen,LinKK,Im (ω i ) are the real part and the imaginary part of the impedance obtained by fitting the Lin-KK model when the order of the RC element is the optimal order, respectively; When the real part residual ΔRe gen (ω i ) or the imaginary part residual ΔIm gen (ω i ) is greater than the set threshold, the data corresponding to this frequency point is determined as an abnormal point, and the abnormal point is removed from the generated EIS data to obtain qualified data.

9. The EIS data quality inspection method based on Lin-KK verification according to claim 8 Characterized in that: The set threshold is 0.01 or 0.

001.

10. EIS Data Quality Inspection Device Based on Lin-KK Verification Characterized in that: The device includes: A model structure optimization module, used to obtain the experimental EIS data of the battery, set the maximum and minimum orders of the RC components in the Lin-KK model, calculate the fitting results of the experimental EIS data at different orders, and find the optimal order based on the fitting results; A generated EIS data fitting module, used to fit the generated EIS data based on the Lin-KK model with the optimal order of the RC component to obtain the Lin-KK impedance of the generated EIS data; A data quality inspection module, used to calculate the residuals of each frequency point in the generated EIS data based on the Lin-KK impedance, and remove the abnormal points with residuals greater than the set threshold to obtain qualified data.

Images