Lithium battery EIS prediction method and device for electric energy storage

Abstract

The invention provides a lithium battery EIS prediction method for electric power energy storage, and belongs to the field of lithium battery impedance spectroscopy prediction.The lithium battery EIS prediction method comprises the steps that preprocessed battery sample data is divided into a test set, a first data set and a second data set, and geometric parameters of a battery in different states are calculated according to impedance spectroscopy; training an MLP network by taking the data set I as input and the geometric parameters as output, and inputting the data set II into the trained MLP network to obtain geometric features; inputting the second data set into an Encoder network to obtain a feature vector, fusing the feature vector and a geometric feature, inputting the fused feature vector and geometric feature into a Decoder network, training an Encoder-Decoder network based on LSTM by taking the impedance spectrum as output, and constructing an impedance spectrum prediction model based on the trained MLP network and the trained Encoder-Decoder network; predicting an impedance spectrum; the credibility of the impedance spectrum is verified based on a Lin-KK method, and a lithium battery EIS prediction device for electric energy storage is also provided; the problems of insufficient physical interpretation, indefinite feature construction and low reliability of impedance spectrum prediction are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium battery impedance spectrum prediction, and in particular to an EIS prediction method and device for lithium batteries for power energy storage. Background Art

[0002] Impedance spectrum (EIS) data refers to a series of values ​​about battery impedance obtained by the battery impedance spectrum test, which is used to study the electrochemical reactions inside the battery. It obtains the impedance spectrum of the battery by applying small-amplitude AC sinusoidal potential waves of different frequencies to the battery and measuring the ratio of the AC potential to the current signal (i.e., the impedance of the system) as the frequency of the sinusoidal wave changes. In actual research, the obtained battery impedance spectrum data is often used to study the electrochemical performance of the battery, and even to evaluate the battery life. The impedance spectrum data of the battery is mainly obtained by testing with professional impedance spectrum testing instruments, but the impedance spectrum testing instruments are relatively expensive, which makes the cost of obtaining impedance spectrum data high. Therefore, how to obtain impedance spectrum data at low cost has become a problem that needs to be solved urgently.

[0003] At the same time, with the rapid development of renewable energy, power storage technology plays an increasingly critical role in regulating the fluctuations between power supply and demand, improving grid stability, and improving energy efficiency. Lithium batteries, with their high energy density, long cycle life, and lightweight characteristics, have become one of the main battery technologies in power storage systems, and are widely used in power system peak regulation, frequency regulation, off-grid systems, and electric vehicle charging stations. However, with the continuous expansion of the application scope of lithium batteries, the requirements for their performance monitoring, optimization, and life prediction are also increasing. In particular, battery internal resistance and impedance characteristics are important parameters for evaluating the health and life of lithium batteries. By monitoring the impedance spectrum, information such as the electrochemical reaction, electron and ion transmission inside the battery can be obtained, which is of great significance for improving the intelligence level of battery management systems and extending battery life. Therefore, how to obtain the impedance spectrum data of lithium batteries at low cost and high efficiency, and accurately analyze their performance, has become an important research direction in the development and application of power storage systems.

[0004] In this regard, the commonly used solution is to establish a mapping relationship between the battery's current data, voltage data, and temperature data and the impedance spectrum data, and build a corresponding impedance spectrum prediction model. You only need to input the tested current data, voltage data, and temperature data into the impedance spectrum prediction model to predict the corresponding impedance spectrum data. Although this method can use charging and discharging data to predict the impedance spectrum, it often has problems such as insufficient physical explanation, unclear feature construction, and poor generalization ability. In addition, the current data-driven impedance spectrum prediction method does not introduce any mechanism to verify the reliability of impedance spectrum prediction. If the target impedance spectrum in the training set of the data-driven model only barely passes the KK test, then the model may introduce new uncertainties when generalizing the test set, resulting in fluctuations in the reliability of the predicted impedance spectrum. Summary of the invention

[0005] The technical problem to be solved by the present invention is how to solve the problems of insufficient physical explanation, unclear characteristic structure and low reliability in the current lithium battery impedance spectrum prediction.

[0006] The present invention solves the above technical problems through the following technical solutions: an EIS prediction method for electric energy storage lithium battery, the method comprising: The preprocessed battery sample data is divided into a test set, a data set 1, and a data set 2, and the geometric parameters of the battery under different states are calculated according to the impedance spectrum; Taking data set 1 as input and geometric parameters as output, the MLP network is trained to obtain the trained MLP network. Data set 2 is input into the trained MLP network to obtain geometric features. The second dataset is input into the Encoder network to obtain the feature vector, which is then fused with the geometric features and input into the Decoder network. The impedance spectrum is used as the output to train the LSTM-based Encoder-Decoder network to obtain the trained Encoder-Decoder network. The impedance spectrum prediction model is constructed based on the trained MLP network and the trained Encoder-Decoder network. The test set is input into the impedance spectrum prediction model to predict the impedance spectrum; The credibility of the impedance spectrum is verified based on the Lin-KK method, and a reliable impedance spectrum is obtained.

[0007] Beneficial effects: The present invention adopts a two-stage network structure to construct an impedance spectrum prediction model. In the first stage, the mapping relationship between battery sample data and geometric features is established by training the MLP network to ensure that the geometric features output by the trained MLP network match the geometric morphology of the actual impedance spectrum. The geometric features closely related to the impedance spectrum morphology in the battery sample data are extracted by the trained MLP network, and the battery charging and discharging behavior is closely linked to the impedance spectrum morphology, so that the prediction result has a clear physical meaning. In the second stage, the predicted physical geometric features and the battery sample data are input into the Encoder-Decoder network through deep mapping, and the time domain of the battery sample data is extracted by the Encoder network. The time series feature vector is obtained by fusion of the time series feature vector and geometric features, and then input into the Decoder network. The Encoder-Decoder network based on LSTM is trained. The impedance spectrum prediction model is constructed based on the trained MLP network and the trained Encoder-Decoder network. This model not only inherits the adaptability of the data-driven method to the working condition data, but also establishes an interpretable association with the electrochemical mechanism through the explicit physical feature layer. Through the organic fusion of geometric features and time series features, the prediction results have clear physical interpretation and high prediction accuracy, and can achieve the goal of directly mapping the battery sample data collected during the charging and discharging process to the full-frequency domain impedance spectrum of the battery. Finally, the credibility of the predicted impedance spectrum is verified based on the outlier removal algorithm verified by linear Kronig-Kramers to improve the reliability of impedance spectrum prediction.

[0008] Preferably, the battery sample data includes current, voltage, and temperature, and the preprocessing process includes: Performing filtering processing on the collected battery sample data to obtain battery sample data with noise removed; Normalizing the battery sample data after noise removal to obtain normalized battery sample data; The normalized battery sample data is corrected to obtain preprocessed battery sample data.

[0009] Beneficial effects: The present invention removes interference signals by performing noise filtering on the original signals of current, voltage and temperature, eliminates the influence of different measurement ranges on model training by normalizing the data, and ensures the consistency and accuracy of the data at each sampling point by correcting the data.

[0010] Preferably, the process of calculating the geometric parameters of the battery in different states according to the impedance spectrum includes: The impedance spectrum is filtered and fitted to obtain a continuous EIS curve; Based on the first-order derivative and second-order derivative of the EIS curve, the boundary from high frequency to medium frequency and the boundary from medium frequency to low frequency are obtained respectively; Select the horizontal coordinate of the intersection of the high frequency area and the real axis as the parameter , fitting the semicircle equation corresponding to the mid-frequency region, and obtaining the radius of the circle as the parameter , fit the straight line equation in the low frequency region, the slope of the straight line equation is the parameter .

[0011] Beneficial effect: The present invention determines the division position of low-frequency, medium-frequency and high-frequency regions by analyzing the change trend of the first-order and second-order derivatives of the EIS curve. When the first-order derivative drops sharply, it indicates that the EIS transitions from the high-frequency straight line to the medium-frequency semicircle. In the medium-frequency semicircle area, the first-order derivative changes slowly. When the second-order derivative changes suddenly, the corresponding first-order derivative changes from negative to positive, indicating the transition from the medium-frequency area to the low-frequency area. After dividing the frequency range, fit the medium-frequency semicircle to obtain the parameters , and extract the ohmic impedance corresponding to the low-frequency straight line and its slope angle , geometric parameters , , It is a key parameter to describe the EIS geometry.

[0012] Preferably, the process of training the MLP network includes: Input the data into the input layer of the MLP network and flatten it to obtain a one-dimensional input vector; Input the input vector to the first hidden layer of the MLP network and map it to a high-dimensional feature space; The output of the first hidden layer is input into the second hidden layer for further mapping to keep the output high dimensionality; The output of the second hidden layer is input into the output layer to map the geometric parameters; The mean square error between the predicted value and the calculated value of the geometric parameters is calculated. When the mean square error is minimized, the trained MLP network is obtained.

[0013] Beneficial effect: The present invention can ensure the physical meaning and stability of the extracted geometric features by training the MLP network. During the training process, the impedance spectrum data obtained by actual measurement is used to optimize the network parameters through loss functions such as minimum mean square error (MSE) to ensure the output g , r , The geometric features are matched with the actual impedance spectrum. After training, the parameters of the geometric feature extraction module are fixed so as to directly provide stable and physically interpretable auxiliary information in the subsequent stage. The present invention realizes the conversion from conventional BMS sensor data (voltage, current, temperature) to EIS geometric features ( g , r , ). Unlike the traditional method that relies on a dedicated impedance spectrometer to extract these features, the present invention directly predicts the equivalent g , r , parameters, which preserves the physical correlation between EIS characteristics and battery aging mechanisms (such as g Reflects ohmic impedance, r Corresponding to the charge transfer process), and getting rid of the dependence on high-frequency impedance detection equipment.

[0014] Preferably, the first hidden layer uses a nonlinear activation function , the expression is:

[0015] in, is the input vector, express The value of the function.

[0016] Beneficial effects: Using the nonlinear activation function Swish can provide a smoother gradient flow in the network, thereby accelerating the training process and avoiding the gradient vanishing problem.

[0017] Preferably, the process of training the LSTM-based Encoder-Decoder network includes: Input the trained MLP network into the second dataset to obtain geometric features; Input data set 2 into the Encoder network to generate feature vector ; The feature vector Combined with geometric features to obtain comprehensive feature vector ; The comprehensive feature vector Input two decoder networks respectively, and one decoder network outputs the real part of the impedance value corresponding to each frequency point , another Decoder network outputs the imaginary part of the impedance value at each frequency point , according to the real part and the imaginary part Get the predicted impedance spectrum; The root mean square error between the predicted impedance spectrum and the actual impedance spectrum is calculated. When the root mean square error is the smallest, the trained Encoder-Decoder network is obtained.

[0018] Preferably, the process of verifying the credibility of the impedance spectrum based on the Lin-KK method includes: 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 under different orders, and find the optimal order based on the fitting results; The impedance spectrum is fitted based on the Lin-KK model with the optimal order of RC elements to obtain the Lin-KK impedance of the impedance spectrum; The residual of each frequency point in the impedance spectrum is calculated based on Lin-KK impedance, and the abnormal points with residuals greater than the set threshold are eliminated to obtain a reliable impedance spectrum.

[0019] Beneficial effect: When verifying the credibility of the predicted impedance spectrum, 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 predicted impedance spectrum, and the remaining model parameters are finely fitted. By judging the residual of each frequency point in the predicted impedance spectrum, only when the residual is lower than the set threshold, the data is considered to be highly consistent with the actual dynamic behavior of the electrochemical system at the local parameter level. Through the comprehensive verification of the two major aspects of the physical intrinsic consistency of the predicted impedance spectrum and the model topology and parameter consistency, not only the physical authenticity of the impedance spectrum is ensured, but also the stable consistency of the impedance spectrum in the global model structure is guaranteed, providing a solid data foundation for subsequent battery state evaluation and accurate modeling.

[0020] Preferably, the fitting results of the experimental EIS data at different orders are calculated, and the process of finding the optimal order based on the fitting results is: Setting DC resistance ,resistance , time constant The initial value of the resistor is adjusted at each order. and time constant , DC resistance The value of is fixed or fine-tuned within a small range, and the iterative calculation is performed until the preset convergence condition is met to obtain the best fitting parameters under this order, and the best fitting parameters under all orders are obtained by analogy; Based on the best fitting parameters at each order, the experimental EIS data are calculated at each frequency point The real residual and the imaginary residual :

[0021]

[0022] Calculate each order based on the real and imaginary residuals M The average residual under :

[0023] The order corresponding to the minimum average residual value is taken as the optimal order :

[0024] in, , The experimental EIS data are Mid-frequency point The corresponding real and imaginary parts are , are the real and imaginary parts of the impedance obtained by fitting the Lin-KK model, N is the number of frequency points.

[0025] Preferably, the Lin-KK impedance of the impedance spectrum for:

[0026] in, , , The order of RC components is the optimal order. When the DC resistance and the k The resistance of the first RC element, k The time constant of an RC element; The residual at each frequency point in the impedance spectrum includes the real part residual , imaginary residual :

[0027]

[0028] in, , Generate EIS data for Mid-frequency point The corresponding real and imaginary parts are , They are respectively the real part and imaginary part of the impedance obtained by fitting the Lin-KK model when the order of the RC element is the optimal order; When the real residual or the imaginary residual When it is greater than the set threshold, the data corresponding to the frequency point is determined as an abnormal point, and the abnormal point is removed from the impedance spectrum to obtain a reliable impedance spectrum.

[0029] The present invention also provides a lithium battery EIS prediction device for power energy storage, the device comprising: A data processing module is used to divide the preprocessed battery sample data into a test set, a data set 1, and a data set 2, and calculate the geometric parameters of the battery under different states according to the impedance spectrum; The geometric feature extraction module is used to train the MLP network with the data set 1 as input and the geometric parameters as output to obtain the trained MLP network, and input the data set 2 into the trained MLP network to obtain the geometric features; The joint modeling module is used to input the second data set into the Encoder network to obtain a feature vector, which is then fused with the geometric features and input into the Decoder network. The impedance spectrum is used as the output to train the LSTM-based Encoder-Decoder network to obtain a trained Encoder-Decoder network. The impedance spectrum prediction model is constructed based on the trained MLP network and the trained Encoder-Decoder network. A prediction module is used to input the test set into the impedance spectrum prediction model to predict the impedance spectrum; The verification module is used to verify the credibility of the impedance spectrum based on the Lin-KK method to obtain a credible impedance spectrum.

[0030] The present invention also provides the following advantages: (1) The impedance spectrum prediction method of the present invention can predict the full-frequency impedance spectrum online based on the collected battery charging and discharging data, without the need for traditional impedance spectrum test instruments and external disturbance signals. It can solve the problems existing in the prior art, such as expensive test equipment, long test cycle, and insufficient physical interpretation of data.

[0031] (2) The present invention is based on a physical guided neural network (PGNN) with lower computational latency. Compared with complex models (such as the neural network in the figure), the PGNN network has a lower risk of overfitting small sample data, is more adaptable to the long-tail distribution characteristics of battery aging data, and is more robust. The two-stage network structure combines the temporal dynamic characteristics and geometric morphological characteristics. The model inherits the adaptability of the data-driven method to the working condition data, and also uses an explicit physical feature layer (parameters ,parameter ,parameter ) established an interpretable association with the electrochemical mechanism, and further improved the robustness and accuracy of the predictions through cross-validation and model fusion.

[0032] (3) The present invention establishes a physical connection between generated data and experimental data by reusing the Lin-KK verification method. First, the RC order is initialized and set within a wide range, which basically covers the RC orders that may appear in the lithium battery equivalent circuit model. The experimental EIS data is fitted and the real residual and imaginary residual under each order are calculated. The average residual is calculated based on the real residual and the imaginary residual. The order corresponding to the minimum value of the average residual is determined as the optimal order. The optimal fitting order has the function of representing the basic information of the lithium battery equivalent circuit. When the quality of the predicted impedance spectrum is tested, 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 fitting, and then fits the remaining three parameters through the Lin-KK model, so that the evaluation benchmark of the impedance spectrum is consistent with the evaluation benchmark of the experimental EIS data, and successfully links the experimental EIS data with the predicted impedance spectrum data. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A schematic diagram of geometric features in a method for predicting impedance spectra of a lithium battery for electric energy storage provided by an embodiment of the present invention; Figure 2 A schematic diagram of a method for predicting impedance spectrum of a lithium battery for power storage provided by an embodiment of the present invention; Figure 3 A flowchart of a method for predicting impedance spectrum of a lithium battery for power energy storage provided by an embodiment of the present invention; Figure 4 A sub-flow chart of a method for predicting impedance spectrum of a lithium battery for electric energy storage provided by an embodiment of the present invention; Figure 5 A schematic diagram of an impedance spectrum prediction device for a power energy storage lithium battery provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the technical solution of the present invention is clearly and completely described below in combination with specific embodiments and with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] Example 1 See also Figure 2 and Figure 3 , this embodiment provides an EIS prediction method for a lithium battery for power storage, comprising the following steps: S101. During the charging and discharging process of lithium-ion batteries, multiple groups of battery sample data are collected, and the battery sample data are preprocessed to obtain preprocessed battery sample data. The preprocessed battery sample data is divided into a test set, a data set 1, and a data set 2. For example, the preprocessed battery sample data is divided into 10 parts, 1 of which is used as a test set, and the remaining 9 are used as data sets for training and verifying 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 10Hz, thereby forming a time series data matrix of size (3, T), where "3" represents the three channels of voltage, current, and temperature, and T is the number of time steps within the sampling time. In the process of obtaining each group of battery sample data, the impedance value corresponding to the sample battery at a certain frequency is tested synchronously by an impedance spectrum tester. The impedance spectrum data includes impedance values ​​at different frequencies, and the geometric parameters of the battery in different states are calculated based on the impedance spectrum.

[0036] The process of preprocessing the collected battery sample data includes: Filtering the battery sample data to remove interference signals and obtain the battery sample data with noise removed; The battery sample data after noise removal is normalized to obtain normalized battery sample data, so as to eliminate the influence of different measurement ranges on model training; The normalized battery sample data is corrected to obtain the preprocessed battery sample data to ensure the consistency and accuracy of the data at each sampling point. By preprocessing the battery sample data, the high quality and stability of the subsequent network model input data can be guaranteed.

[0037] The process of calculating the geometric parameters of the battery under different states based on the impedance spectrum includes: Input impedance spectrum (real resistance value arranged from high to low frequency) and the imaginary resistance ), using Savitzky-Golay filter to filter and fit to obtain a continuous EIS curve; The changing trends of the first-order derivative and the second-order derivative of the EIS curve are analyzed to obtain the division positions of the low-frequency, medium-frequency and high-frequency regions; among them, the first-order derivative reflects the change in the curvature of the curve, and the second-order derivative is used to capture the inflection point. The first-order derivative drops rapidly from a maximum positive value to close to zero (corresponding to the vertical line → semicircle transition), which represents the physical meaning of ohmic impedance dominance → charge transfer impedance dominance. The second-order derivative has a local maximum (corresponding to the end of the semicircle arc and the beginning of Warburg impedance), which represents the physical meaning of charge transfer impedance dominance → diffusion process dominance. The boundary from high frequency to medium frequency is obtained based on the first-order derivative, and the boundary from medium frequency to low frequency is obtained based on the second-order derivative.

[0038] The horizontal coordinate value of the intersection of the high frequency region of the EIS curve and the real axis is selected as the parameter , use the least squares method to fit the semicircle equation corresponding to the intermediate frequency region, and get the radius of the circle as the parameter , using the real and imaginary data in the low-frequency region, the least squares method is used to fit the linear equation in the low-frequency region. The slope value of the linear equation is the parameter .

[0039] See also Figure 1 , geometric parameters , , These three parameters correspond to the key geometric features of the impedance spectrum, among which the parameters is the voltage value corresponding to the intersection of the impedance spectrum with the real axis in the Nyquist diagram, indicating the ohmic resistance of the battery. is the fitting radius of the semicircle in the intermediate frequency region of the Nyquist diagram, reflecting the interface reaction and related electrochemical processes. It is the slope angle of the straight line in the low-frequency region of the Nyquist diagram, reflecting the diffusion process and transmission limitation effect.

[0040] S102, using the data set 1 as input and the geometric parameters as output, training the MLP network, and when the loss function is minimized, obtaining the trained MLP network, inputting the data set 2 into the trained MLP network to obtain the geometric features.

[0041] The data set used for training and verification in S101 can be divided into 9 parts, and 2 parts are selected as data set 1 for MLP network training. Data set 1 is used as input and geometric parameters are used as output. The MLP network is trained and the mean square error (MSE) is used as the loss function. When the loss function is minimized, the trained MLP network is obtained. The trained MLP network can extract geometric features that are highly consistent with the actual impedance spectrum data morphology from the input current, voltage, and temperature data. After the training is completed, the MLP network parameters are frozen for geometric feature generation in the second training stage.

[0042] The process of training an MLP network includes: Input the battery sample data (two-dimensional data of size 3×T) into the input layer of the preset MLP network and flatten it to obtain a one-dimensional input vector; The input vector is input into the first hidden layer of the preset MLP network and mapped to a high-dimensional feature space, such as a 256-dimensional feature space. The first hidden layer uses a nonlinear activation function (such as or ) Enhance the ability to express features and extract the electrochemical features and impedance spectrum geometry implicit in the charge and discharge data. Nonlinear activation function The mathematical expression is:

[0043] in, is the input vector, express The value of the function.

[0044] Non-linear activation function It can provide a smoother gradient flow in the network, thereby speeding up the training process and avoiding the gradient vanishing problem.

[0045] The output of the first hidden layer is input into the second hidden layer for further mapping, so as to keep the output high-dimensional, and the dimension of the present invention is 256; The output of the second hidden layer is input into the output layer, and the geometric parameters (parameters ,parameter ,parameter ), the equivalent parameters are predicted directly from the operating data through a deep learning network ,parameter ,parameter , which preserves the physical correlation between EIS characteristics and battery aging mechanisms (such as parameters Reflects ohmic impedance, parameters Corresponding to the charge transfer process), and getting rid of the dependence on high-frequency impedance detection equipment.

[0046] S103, inputting the second data set into the Encoder network to obtain a feature vector, fusing the feature vector with the geometric feature and inputting the feature vector into the Decoder network, taking the impedance spectrum as the output, training the LSTM-based Encoder-Decoder network, and obtaining the trained Encoder-Decoder network when the loss function is minimized. Based on the trained MLP network and the trained Encoder-Decoder network, an impedance spectrum prediction model is constructed.

[0047] The remaining 7 data sets in S101 used for training and verification are used as data set 2 to train the Encoder-Decoder network. The battery sample data of data set 2 is input into the trained MLP network to obtain geometric features. The battery sample data is input into the Encoder network to generate feature vectors. , the feature vector generated by the Encoder network It is a characteristic vector that represents the dynamic characteristics of current, voltage, and temperature time series data and can describe the dynamic changes of the battery charging and discharging process. and geometric features (parameters ,parameter ,parameter ) to obtain the comprehensive feature vector , comprehensive feature vector Input two decoder networks respectively, and one decoder network outputs the real part of the impedance value corresponding to each frequency point , another Decoder network outputs the imaginary part of the impedance value at each frequency point , according to the real part and the imaginary part The impedance spectrum EIS can be obtained.

[0048] The Encoder-Decoder network adopts an encoder-decoder structure based on an improved long short-term memory network to realize the prediction of the full-frequency domain impedance spectrum. In order to improve the prediction performance of the model, the present invention is divided into two stages when training the LSTM-based Encoder-Decoder network. In the first stage, the Encoder network and the two Decoder networks are jointly trained to obtain the Encoder-Decoder network after preliminary training, and the ability of the Decoder network to extract features from the battery sample data is enhanced. In the second stage, the parameters of the Encoder network are frozen, and then the two Decoder networks are independently trained. Based on the feature vector extracted by the encoder, the encoder extracts the feature vector. , improve the decoder's mapping ability to the real and imaginary parts, 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.

[0049] When training the MLP network and the LSTM-based Encoder-Decoder network, cross-validation and model fusion strategies can be used for training, which can effectively alleviate the additional prediction errors that may be caused by sample distribution deviation in a single model, and improve the prediction accuracy and robustness of the model. For example, when training the MLP network, the data set 1 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 the model parameters are modified during the training process. The validation set is used to validate the trained model and adjust the hyperparameters of the model. In the data set 1, one group is selected as the validation set, and the rest are selected as the training set, to obtain multiple groups of data sets, each of which includes a validation set and multiple training sets. Using each data set to train and validate the MLP network, multiple independent trained MLP networks can be obtained. Similarly, when training the Encoder-Decoder network, the second data set is divided into a training set and a validation set according to a set ratio. One set is selected in turn from the second data set as a validation set, and the rest are selected as training sets to obtain multiple sets of data sets. Each set of data sets includes a validation set and multiple training sets. The Encoder-Decoder network is trained and verified with each set of data sets, so that multiple independent trained Encoder-Decoder networks can be obtained.

[0050] S104, input the test set into the impedance spectrum prediction model to predict the impedance spectrum. When the test set is input 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, and the multiple groups of output results are weighted and fused to obtain geometric features. The test set is input into the encoder network of multiple independent trained encoder-decoder networks to obtain feature vectors, and the feature vectors are spliced ​​with the geometric features and 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 root mean square error The calculation formula is:

[0051] in, and Respectively represent i The real and imaginary parts of the actual EIS at each frequency point, and are the real and imaginary parts of the prediction results respectively.

[0052] The Temporal Attention Mechanism is introduced into the Encoder-Decoder network. This mechanism dynamically assigns different attention weights to each time step during the charging and discharging process of the lithium battery, allowing the model to focus on the most critical time period in the charging and discharging process, so as to better pay attention to the battery behavior in the critical time period and more accurately predict the impedance spectrum.

[0053] Attention weights at each time step The calculation formula is:

[0054] in, For each time step t The larger the attention score, the more attention is paid to the time step. and input features The weighted calculation is:

[0055] in, Usually for hidden states or other hidden layer outputs to linearly transform the weight matrix used to map the hidden vector to the same value as the input On the same or additively compatible dimensions, b is the corresponding bias term, which is consistent with other bias usages in neural networks and is used to provide a learnable translation for linear transformations, i.e. and b They are the weight matrix and bias term used to linearly map the hidden state or input vector.

[0056] The present invention adopts a two-stage network structure to construct an impedance spectrum prediction model. In the first stage, the mapping relationship between battery sample data and geometric features is established by training the MLP network to ensure that the geometric features output by the trained MLP network match the geometric morphology of the actual impedance spectrum. The geometric features closely related to the impedance spectrum morphology in the battery sample data are extracted by the trained MLP network, and the battery charging and discharging behavior is closely linked to the impedance spectrum morphology, so that the prediction results have clear physical meanings. In the second stage, the predicted physical geometric features and the battery sample data are input into the Encoder-Decoder network through deep mapping, and the time series features of the battery sample data are extracted through the Encoder network. Vector, and the time series feature vector is fused with the geometric features and then input into the Decoder network, the Encoder-Decoder network based on LSTM is trained, and the 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 the working 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 time series 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 charging and discharging process to the full-frequency domain impedance spectrum of the battery.

[0057] The impedance spectrum prediction method of the present invention can predict the full-frequency impedance spectrum online based on the collected battery charge and discharge data, without the need for traditional impedance spectrum test instruments and external disturbance signals, and can solve the problems of expensive test equipment, long test cycle, and insufficient physical interpretation of data in the prior art. Compared with the traditional EIS prediction method, there is no need to interrupt the battery operation for impedance scanning, and real-time online monitoring is achieved; compared with the pure data-driven model, the present invention uses parameters to predict the impedance spectrum of the battery. , , The physical meaning of the battery state change mode is explained (such as the parameters A sudden increase in the value indicates a loss of active substances), which helps improve the prediction accuracy. Compared with the hybrid model, the present invention uses the geometric characteristics of the impedance spectrum as an intermediate variable rather than the final output for the first time, allowing the network to learn physical laws and data distribution at the same time.

[0058] The PGNN network structure (with more than 80% fewer parameters than Transformer) has lower computational latency. Compared with complex models (such as the neural network in the figure), the PGNN network has a lower risk of overfitting small sample data, is more adaptable to the long-tail distribution characteristics of battery aging data, and is more robust. The two-stage network structure combines the time series dynamic characteristics and geometric morphological characteristics. The model inherits the adaptability of the data-driven method to the working condition data, and through the explicit physical feature layer (parameters ,parameter ,parameter ) established an interpretable association with the electrochemical mechanism, and further improved the robustness and accuracy of the prediction 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., and can realize real-time online monitoring of battery impedance spectra, thus facilitating the construction of precise battery management systems.

[0059] S105. Verify the credibility of the impedance spectrum based on the Lin-KK method to obtain a reliable impedance spectrum. Figure 4 ,The specific process of credibility verification includes: S1051. Obtain the experimental EIS data of the battery. The experimental EIS data is obtained by testing the battery under the test standard using a high-precision electrochemical workstation. The experimental EIS data includes frequency data. , real data , imaginary data , where the frequency data Includes a series of positive frequency points, real data For each positive frequency point corresponding to the real part, imaginary part data is the imaginary part corresponding to each positive frequency point.

[0060] The maximum and minimum orders of RC elements in the Lin-KK model are set, the fitting results of the experimental EIS data under different orders are calculated, and the optimal order is found based on the fitting results.

[0061] Among them, the impedance of the Lin-KK model is for:

[0062] in, is the DC resistance in the lithium battery equivalent circuit, The equivalent circuit of lithium battery k The resistance of the RC element, The equivalent circuit of lithium battery k The time constant of an RC element is M is the order of RC elements in the lithium battery equivalent circuit.

[0063] The present invention sets the maximum order of RC elements in the Lin-KK model is 50, the minimum order is 1. In other embodiments, the model order can be simplified. For example, for some data, if the contribution of high-order RC components is small, the calculation efficiency can be improved by simplifying the model order (selecting a more appropriate maximum order and minimum order). The least squares method is used to fit the experimental EIS data at different orders. The fitting process includes: Setting DC resistance ,resistance , time constant For example, the average value of the experimental EIS data in the high frequency region can be used as the DC resistance The initial value of resistance is selected as a small positive value. The initial value of the time constant is a fixed value between 1ms and 10ms. The initial value of can also be determined by analyzing the data characteristics.

[0064] Degree M The value of is taken from the minimum order To the maximum order , the maximum order in this embodiment is 50, the minimum order is 1, for each order M When selected, adjust the resistance at the same time and time constant , DC resistance The value of is fixed or fine-tuned within a small range, and the iterative calculation is performed until the preset convergence condition is met to obtain the best fitting parameters under this order. The preset convergence condition can be 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. The best fitting parameters under all orders are obtained by analogy. The best fitting parameters include DC resistance ,resistance , time constant .

[0065] Based on the best fitting parameters at each order, the experimental EIS data are calculated at each frequency point The real residual and the imaginary residual :

[0066]

[0067] Calculate each order based on the real and imaginary residuals M The average residual under :

[0068] The order corresponding to the minimum average residual value is taken as the optimal order :

[0069] in, is the experimental EIS data, , are the frequency points in the experimental EIS data. The corresponding real and imaginary parts are , are the real and imaginary parts of the impedance obtained by fitting the Lin-KK model, N is the number of frequency points.

[0070] S1052, fitting the impedance spectrum obtained in step S106 based on the Lin-KK model with the optimal order of RC elements to obtain the Lin-KK impedance of the impedance spectrum; the impedance spectrum includes frequency data , real data , imaginary data , where the frequency data Includes a series of positive frequency points, real data The real and imaginary data corresponding to each frequency point is the imaginary part corresponding to each frequency point.

[0071] The RC element order is selected as the optimal order obtained in step S1051 above. The impedance spectrum obtained in step S104 is fitted by the least square method, such as the Levenberg-Marquardt algorithm. The fitting method is similar to the fitting process of step S1051 above, that is, the DC resistance is first set. ,resistance , time constant For example, the average value of the impedance spectrum in the high frequency region can be used as the DC resistance The initial value of resistance is selected as a small positive value. The initial value of the time constant is a fixed value between 1ms and 10ms. The initial value of can also be judged by analyzing the generated impedance spectrum characteristics. During the fitting process, the resistance is adjusted at the same time. and time constant , DC resistance The value of is fixed or fine-tuned in a small range, and the iterative calculation is performed until the preset convergence condition is met to obtain the optimal order. The best fitting parameters under , , .

[0072] Lin-KK impedance of impedance spectrum for:

[0073] in, , , The order of RC components is the optimal order. When the DC resistance and the k The resistance of the first RC element, k The time constant of an RC element.

[0074] The present invention basically covers the possible RC orders of 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 predicted impedance spectrum, 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 through the Lin-KK model, so that the evaluation benchmark of the predicted impedance spectrum is consistent with the evaluation benchmark of the experimental EIS data, and successfully links the experimental EIS data with the data-driven predicted impedance spectrum.

[0075] S1053. Calculate the residual of each frequency point in the impedance spectrum based on the Lin-KK impedance, remove abnormal points whose residual is greater than a set threshold, and obtain a reliable impedance spectrum.

[0076] The residual at each frequency point in the impedance spectrum includes the real part residual , imaginary residual :

[0077]

[0078] in, To generate EIS data, , The frequency points in the generated EIS data are The corresponding real and imaginary parts are , They are respectively the real and imaginary parts of the impedance obtained by fitting the Lin-KK model when the order of the RC element is the optimal order.

[0079] When the real residual or the imaginary residual Greater than the set threshold In this embodiment, the threshold is set is 0.01 or 0.001, the data corresponding to the frequency point is determined as an abnormal point, and the abnormal point is removed from the impedance spectrum to obtain a reliable impedance spectrum, that is, only when the real residual And the imaginary residual In other embodiments, the threshold value can be set according to different experimental conditions and data quality. Make adaptive adjustments to further improve inspection accuracy.

[0080] After screening the predicted impedance spectrum, the screened EIS data is obtained as a reliable impedance spectrum output. The screened EIS data removes unqualified frequency points, meets the standards of the Lin-KK model, and can be used for further analysis or modeling.

[0081] The present invention verifies the credibility of the impedance spectrum based on the Lin-KK method. First, the experimental EIS data from a high-precision electrochemical workstation is used 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 predicted impedance spectrum, and the remaining model parameters are finely fitted. By judging the residual of each frequency point in the predicted impedance spectrum, only when the residual is lower than the set threshold, the data is considered to be highly consistent with the real dynamic behavior of the electrochemical system at the local parameter level. Through the comprehensive verification of the physical intrinsic consistency of the predicted impedance spectrum and the model topology and parameter consistency, not only the physical authenticity of the impedance spectrum is ensured, but also the stable consistency of the impedance spectrum in the global model structure is guaranteed, which provides a solid data foundation for subsequent battery state evaluation and precise modeling.

[0082] Example 2 See also Figure 5 This embodiment provides an EIS prediction device for lithium batteries for power storage, the device comprising: The data processing module is used to divide the preprocessed battery sample data into a test set, a data set 1, and a data set 2, and calculate the geometric parameters of the battery under different states according to the impedance spectrum; the battery sample data includes current, voltage, and temperature. The process of preprocessing the collected battery sample data includes: Performing filtering processing on the collected battery sample data to obtain battery sample data with noise removed; Normalizing the battery sample data after noise removal to obtain normalized battery sample data; The normalized battery sample data is corrected to obtain preprocessed battery sample data.

[0083] The process of calculating the geometric parameters of the battery under different states based on the impedance spectrum includes: The impedance spectrum is filtered and fitted to obtain a continuous EIS curve; Based on the first-order derivative and second-order derivative of the EIS curve, the boundary from high frequency to medium frequency and the boundary from medium frequency to low frequency are obtained respectively; Select the horizontal coordinate of the intersection of the high frequency area and the real axis as the parameter , fitting the semicircle equation corresponding to the mid-frequency region, and obtaining the radius of the circle as the parameter , fit the straight line equation in the low frequency region, the slope of the straight line equation is the parameter .

[0084] The geometric feature extraction module is used to train the MLP network with the data set 1 as input and the geometric parameters as output. When the loss function is minimized, the trained MLP network is obtained, and the data set 2 is input into the trained MLP network to obtain the geometric features. The process of training the MLP network includes: Input the data into the input layer of the MLP network and flatten it to obtain a one-dimensional input vector; Input the input vector to the first hidden layer of the MLP network and map it to a high-dimensional feature space; The output of the first hidden layer is input into the second hidden layer for further mapping to keep the output high dimensionality; The output of the second hidden layer is input into the output layer to map the geometric parameters; The mean square error between the predicted value and the calculated value of the geometric parameters is calculated. When the mean square error is minimized, the trained MLP network is obtained.

[0085] The first hidden layer uses the nonlinear activation function Swish, which is expressed as: .

[0086] The joint modeling module is used to input the second data set into the Encoder network to obtain a feature vector, which is then fused with the geometric features and input into the Decoder network. The impedance spectrum is used as the output to train the LSTM-based Encoder-Decoder network. When the loss function is minimized, the trained Encoder-Decoder network is obtained. The impedance spectrum prediction model is constructed based on the trained MLP network and the trained Encoder-Decoder network. The process of training the LSTM-based Encoder-Decoder network includes: Input the trained MLP network into the second dataset to obtain geometric features; Input data set 2 into the Encoder network to generate feature vector ; The feature vector Combined with geometric features to obtain comprehensive feature vector ; The comprehensive feature vector Input two decoder networks respectively, and one decoder network outputs the real part of the impedance value corresponding to each frequency point , another Decoder network outputs the imaginary part of the impedance value at each frequency point , according to the real part and the imaginary part Get the predicted impedance spectrum; The root mean square error between the predicted impedance spectrum and the actual impedance spectrum is calculated. When the root mean square error is the smallest, the trained Encoder-Decoder network is obtained.

[0087] A prediction module is used to input the test set into the impedance spectrum prediction model to predict the impedance spectrum; The verification module is used to verify the credibility of the impedance spectrum based on the Lin-KK method to obtain a credible impedance spectrum.

[0088] The process of verifying the credibility of impedance spectrum based on Lin-KK method includes: 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 under different orders, and find the optimal order based on the fitting results; The impedance spectrum is fitted based on the Lin-KK model with the optimal order of RC elements to obtain the Lin-KK impedance of the impedance spectrum; The residual of each frequency point in the impedance spectrum is calculated based on Lin-KK impedance, and the abnormal points with residuals greater than the set threshold are eliminated to obtain a reliable impedance spectrum.

[0089] Among them, the fitting results of experimental EIS data under different orders are calculated, and the process of finding the optimal order based on the fitting results is: Setting DC resistance ,resistance , time constant The initial value of the resistor is adjusted at each order. and time constant , DC resistance The value of is fixed or fine-tuned within a small range, and the iterative calculation is performed until the preset convergence condition is met to obtain the best fitting parameters under this order, and the best fitting parameters under all orders are obtained by analogy; Based on the best fitting parameters at each order, the experimental EIS data are calculated at each frequency point The real residual and the imaginary residual :

[0090]

[0091] Calculate each order based on the real and imaginary residuals M The average residual under :

[0092] The order corresponding to the minimum average residual value is taken as the optimal order :

[0093] in, , The experimental EIS data are Mid-frequency point The corresponding real and imaginary parts are , are the real and imaginary parts of the impedance obtained by fitting the Lin-KK model, N is the number of frequency points.

[0094] Lin-KK impedance of impedance spectrum for:

[0095] in, , , The order of RC components is the optimal order. When the DC resistance and the k The resistance of the first RC element, k The time constant of an RC element; The residual at each frequency point in the impedance spectrum includes the real part residual , imaginary residual :

[0096]

[0097] in, , Generate EIS data for Mid-frequency point The corresponding real and imaginary parts are , They are respectively the real part and imaginary part of the impedance obtained by fitting the Lin-KK model when the order of the RC element is the optimal order; When the real residual or the imaginary residual When it is greater than the set threshold, the data corresponding to the frequency point is determined as an abnormal point, and the abnormal point is removed from the impedance spectrum to obtain a reliable impedance spectrum.

[0098] Example 3 This embodiment takes a data set containing charge and discharge and EIS data of 11 new cylindrical lithium iron phosphate batteries as the research object.

[0099] Each battery was subjected to constant current (CC) discharge test under different conditions (SOC from 5% to 100%, step size of 5%), the test conditions included discharge current of 0.12 A, cut-off voltage of 2 V and ambient temperature of 20°C, and the capacity of each battery was between 633 mAh and 683 mAh. For each battery, electrochemical impedance spectroscopy (EIS) test was performed under different SOC conditions, the test frequency range was 0.01 Hz to 1000 Hz, and 28 sets of real and imaginary data were obtained for each test. For each EIS test, the first 100 current, voltage, and temperature data points were intercepted from the discharge process under the corresponding SOC as input data, and a total of 220 sets of data samples were finally formed, of which battery B07 was randomly selected as the test sample, and the remaining data were used for model training and verification.

[0100] To achieve online prediction, this embodiment first preprocesses the original current, voltage, and temperature time series data collected during the charging and discharging process. The preprocessing steps include: (1) Use digital filtering algorithms to suppress noise in raw data and remove high-frequency interference; (2) Normalize the current, voltage, and temperature data according to their respective measurement ranges; (3) Correct the time series data during the acquisition process to ensure the consistency of the data in the time domain.

[0101] The preprocessed data forms a matrix of size (3, 100), where "3" represents the three channels of voltage, current, and temperature, providing high-quality input for subsequent feature extraction and prediction.

[0102] Next, the preprocessed data is input into the two-stage physics-guided neural network (PGNN) model of the present invention for full-frequency EIS prediction. The first stage uses a pre-trained geometric feature extraction module based on a multi-layer perceptron (MLP) network structure and uses a Swish activation function to enhance feature expression capabilities. The flattened (3, 100) data is converted into key parameters describing the EIS geometry. ,parameter ,parameter The specific implementation steps are: (1) Flatten the (3, 100) data into a one-dimensional vector; (2) After two hidden layers (each with 256 neurons and using the Swish activation function), three scalars are generated through the output layer: parameters ,parameter ,parameter , where the parameter Reflects the ohmic internal resistance corresponding to the intersection of EIS and the real axis in the Nyquist diagram, parameter The radius of the semicircle fitting in the mid-frequency region represents the electrochemical reaction at the interface. It reflects the slope of the straight line in the low-frequency region and characterizes the diffusion and transmission limitation effects.

[0103] In this stage, the geometric parameters at each SOC are pre-calculated using the actual measured EIS data, and the MLP is trained using the mean square error (MSE) as the loss function until the extracted geometric features are highly consistent with the actual EIS morphology. After the training is completed, the module parameters are frozen and used as physical constraints for the second stage prediction.

[0104] The second stage adopts an encoder-decoder structure based on an improved long short-term memory network (LSTM), in which a temporal attention mechanism is introduced. This mechanism calculates the attention weight for each time step, allowing the model to focus on the characteristics of the key time period in the charging and discharging process, thereby improving the prediction accuracy: (1) Input the preprocessed voltage and current time series data into a single-layer LSTM encoder to extract its time series dynamic features and generate a feature vector ; (2) The feature vector output by the encoder Compared with the geometric features obtained in the first stage ( , , ) to form a comprehensive feature vector ; (3) Use two independent LSTM decoders to decode the comprehensive features One decoder is responsible for predicting the real part of EIS and the other decoder is responsible for predicting the imaginary part; (4) The two decoders each map the output to the impedance value at each frequency point through a fully connected layer, and merge the real and imaginary results to obtain the complete predicted EIS.

[0105] During the training process, the AdamW optimizer was used (the initial learning rate was set to 0.001, combined with a dynamic learning rate decay strategy), and a gradient clipping strategy was introduced to jointly train the encoder and decoder parameters. The goal was to minimize the RMSE between the predicted EIS and the actual measured EIS at each frequency point.

[0106] In order to further improve the robustness and generalization ability of the model, this embodiment adopts cross-validation and model fusion strategy. The specific implementation method is: (1) Among the 11 battery samples, battery B07 is fixed as the test set, and one of the remaining 10 battery samples is selected as the validation set, and the rest are used as the training set, forming multiple data partitioning schemes; (2) Train the model for each data partitioning scheme separately to obtain multiple independent training models; (3) The EIS prediction results of each model under the same SOC conditions in the test set are integrated using a weighted fusion method to form the final prediction output. This strategy effectively alleviates the prediction bias of a single model that may be caused by differences in sample distribution and significantly improves the overall prediction accuracy and stability.

[0107] For the evaluation of prediction performance, this embodiment uses the root mean square error (RMSE) as the main evaluation indicator. After multiple experimental statistics, under all SOC conditions, the overall EIS prediction RMSE of the test set is lower than 1 mΩ, and the prediction error of each frequency band is maintained at a low level, which proves the high accuracy and stability of the method of the present invention.

[0108] The experimental results show that under the conditions of SOC of 30% and 55%, the predicted EIS curve is highly consistent with the actual measured data at most frequency points; among them, the real part prediction accuracy is high, and the imaginary part has a slight deviation in the low-frequency area, but the overall error is controlled within the allowable range. Furthermore, by statistically analyzing the RMSE distribution under different SOCs, the results show that the median RMSE under each SOC condition is less than 1 mΩ, and the maximum error does not exceed 0.1Ω, which fully verifies the effectiveness of this method under multiple working conditions.

[0109] This example uses experimental verification of 11 battery samples to prove that the full-frequency domain EIS online prediction method based on physical guided neural network can effectively use the voltage and current data collected during the charging and discharging process to achieve high-precision prediction of the impedance spectrum of lithium-ion batteries. This method is online and real-time, low-cost, simple in structure, and has good physical interpretation and robustness. It is suitable for practical applications in battery status monitoring and health management in electric vehicles, energy storage systems, and battery management systems.

[0110] 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 prediction method for lithium batteries for power storage, characterized by: Methods include: The preprocessed battery sample data is divided into a test set, a data set 1, and a data set 2, and the geometric parameters of the battery under different states are calculated according to the impedance spectrum; Taking data set 1 as input and geometric parameters as output, the MLP network is trained to obtain the trained MLP network. Data set 2 is input into the trained MLP network to obtain geometric features. The second dataset is input into the Encoder network to obtain the feature vector, which is then fused with the geometric features and input into the Decoder network. The impedance spectrum is used as the output to train the LSTM-based Encoder-Decoder network to obtain the trained Encoder-Decoder network. The impedance spectrum prediction model is constructed based on the trained MLP network and the trained Encoder-Decoder network. The test set is input into the impedance spectrum prediction model to predict the impedance spectrum; The credibility of the impedance spectrum is verified based on the Lin-KK method, and a reliable impedance spectrum is obtained.

2. The EIS prediction method for lithium battery for power energy storage according to claim 1, characterized in that: Battery sample data includes current, voltage, and temperature. The preprocessing process includes: Performing filtering processing on the collected battery sample data to obtain battery sample data with noise removed; Normalizing the battery sample data after noise removal to obtain normalized battery sample data; The normalized battery sample data is corrected to obtain preprocessed battery sample data.

3. The EIS prediction method for lithium battery for power energy storage according to claim 1, characterized in that: The process of calculating the geometric parameters of the battery under different states based on the impedance spectrum includes: The impedance spectrum is filtered and fitted to obtain a continuous EIS curve; Based on the first-order derivative and second-order derivative of the EIS curve, the boundary from high frequency to medium frequency and the boundary from medium frequency to low frequency are obtained respectively; Select the horizontal coordinate of the intersection of the high frequency area and the real axis as the parameter , fitting the semicircle equation corresponding to the mid-frequency region, and obtaining the radius of the circle as the parameter , fit the straight line equation in the low frequency region, the slope of the straight line equation is the parameter .

4. The EIS prediction method for lithium battery for power energy storage according to claim 1, characterized in that: The process of training an MLP network includes: Input the data into the input layer of the MLP network and flatten it to obtain a one-dimensional input vector; Input the input vector to the first hidden layer of the MLP network and map it to a high-dimensional feature space; The output of the first hidden layer is input into the second hidden layer for further mapping to keep the output high dimensionality; The output of the second hidden layer is input into the output layer to map the geometric parameters; The mean square error between the predicted value and the calculated value of the geometric parameters is calculated. When the mean square error is minimized, the trained MLP network is obtained.

5. The EIS prediction method for lithium battery for power energy storage according to claim 4, characterized in that: The first hidden layer uses a nonlinear activation function , the expression is: in, is the input vector, express The value of the function.

6. The EIS prediction method for lithium battery for power energy storage according to claim 1, characterized in that: The process of training an LSTM-based Encoder-Decoder network includes: Input the trained MLP network into the second dataset to obtain geometric features; Input data set 2 into the Encoder network to generate feature vector ; The feature vector Combined with geometric features to obtain comprehensive feature vector ; The comprehensive feature vector Input two decoder networks respectively, and one decoder network outputs the real part of the impedance value corresponding to each frequency point , another Decoder network outputs the imaginary part of the impedance value at each frequency point , according to the real part and the imaginary part Get the predicted impedance spectrum; The root mean square error between the predicted impedance spectrum and the actual impedance spectrum is calculated. When the root mean square error is the smallest, the trained Encoder-Decoder network is obtained.

7. The EIS prediction method for lithium battery for power energy storage according to claim 1, characterized in that: The process of verifying the credibility of impedance spectrum based on Lin-KK method includes: 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 under different orders, and find the optimal order based on the fitting results; The impedance spectrum is fitted based on the Lin-KK model with the optimal order of RC elements to obtain the Lin-KK impedance of the impedance spectrum; The residual of each frequency point in the impedance spectrum is calculated based on Lin-KK impedance, and the abnormal points with residuals greater than the set threshold are eliminated to obtain a reliable impedance spectrum.

8. The EIS prediction method for lithium battery for power energy storage according to claim 7, characterized in that: The fitting results of experimental EIS data under different orders are calculated. The process of finding the optimal order based on the fitting results is as follows: Setting DC resistance ,resistance , time constant The initial value of the resistor is adjusted at each order. and time constant , DC resistance The value of is fixed or fine-tuned within a small range, and the iterative calculation is performed until the preset convergence condition is met to obtain the best fitting parameters under this order, and the best fitting parameters under all orders are obtained by analogy; Based on the best fitting parameters at each order, the experimental EIS data are calculated at each frequency point The real residual and the imaginary residual : Calculate each order based on the real and imaginary residuals M The average residual under : The order corresponding to the minimum average residual value is taken as the optimal order : in, , The experimental EIS data are Mid-frequency point The corresponding real and imaginary parts are , are the real and imaginary parts of the impedance obtained by fitting the Lin-KK model, N is the number of frequency points.

9. The EIS prediction method for lithium battery for power energy storage according to claim 7, characterized in that: Lin-KK impedance of impedance spectrum for: in, , , The order of RC components is the optimal order. When the DC resistance and the k The resistance of the first RC element, k The time constant of an RC element; The residual at each frequency point in the impedance spectrum includes the real part residual , imaginary residual : in, , Generate EIS data for Mid-frequency point The corresponding real and imaginary parts are , They are respectively the real part and imaginary part of the impedance obtained by fitting the Lin-KK model when the order of the RC element is the optimal order; When the real residual or the imaginary residual When it is greater than the set threshold, the data corresponding to the frequency point is determined as an abnormal point, and the abnormal point is removed from the impedance spectrum to obtain a reliable impedance spectrum.

10. An EIS prediction device for lithium batteries for power storage, characterized in that: The device includes: A data processing module is used to divide the preprocessed battery sample data into a test set, a data set 1, and a data set 2, and calculate the geometric parameters of the battery under different states according to the impedance spectrum; The geometric feature extraction module is used to train the MLP network with the data set 1 as input and the geometric parameters as output to obtain the trained MLP network, and input the data set 2 into the trained MLP network to obtain the geometric features; The joint modeling module is used to input the second data set into the Encoder network to obtain a feature vector, which is then fused with the geometric features and input into the Decoder network. The impedance spectrum is used as the output to train the LSTM-based Encoder-Decoder network to obtain a trained Encoder-Decoder network. The impedance spectrum prediction model is constructed based on the trained MLP network and the trained Encoder-Decoder network. A prediction module is used to input the test set into the impedance spectrum prediction model to predict the impedance spectrum; The verification module is used to verify the credibility of the impedance spectrum based on the Lin-KK method to obtain a credible impedance spectrum.

Images