Lithium ion battery static impedance prediction method and system based on dynamic electrochemical impedance spectroscopy
By constructing the BatImp-ResSeq network, the static electrochemical impedance spectrum is directly predicted from the dynamic electrochemical impedance spectrum data, which solves the problems of long measurement time and low data interpretation in traditional methods, and achieves fast and accurate impedance prediction and reduces hardware costs.
Patent Information
- Application Number
- CN202510597891.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-09
- Publication Date
- 2025-08-08
AI Technical Summary
Traditional electrochemical impedance spectrometry measurement requires the battery to reach a steady state, resulting in a long measurement time and interruption of the charging and discharge process, and the dynamic electrochemical impedance spectrometry data is low.
The static impedance prediction method of lithium-ion batteries based on dynamic electrochemical impedance spectrum is adopted, and the BatImp-ResSeq network is constructed using residual neural network and long and short-term memory network to predict the static electrochemical impedance spectrum through dynamic electrochemical impedance spectrum data, reducing measurement time and improving data interpretability.
It realizes rapid prediction of static electrochemical impedance spectra without the need for steady state of the battery, improves measurement efficiency and accuracy, reduces hardware costs, and is suitable for impedance prediction under different operating conditions.
Smart Images
Figure CN120446787A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of lithium-ion batteries, and in particular relates to a method and system for predicting the static impedance of lithium-ion batteries based on dynamic electrochemical impedance spectroscopy. Background Art
[0002] With the continued depletion of non-renewable energy sources and the urgent need to address global warming, a shift in energy mix is becoming imperative, with a focus on sustainable solutions such as lithium-ion batteries. The widespread use of lithium-ion batteries in portable electric vehicles and large-scale energy storage systems has heightened the importance of accurate health monitoring. As batteries age, their efficiency and reliability change, posing potential safety risks.
[0003] Electrochemical impedance spectroscopy has been demonstrated to be an effective non-invasive battery characterization technique. By measuring the battery's impedance at different frequencies, it provides information about the battery's internal properties, such as electrode-electrolyte interface behavior, charge transfer kinetics, and ion transport characteristics.
[0004] Traditional electrochemical impedance spectroscopy measurements must be performed after the battery reaches steady state, which requires a long rest period and interrupts the battery's charge and discharge processes. Dynamic electrochemical impedance spectroscopy can be measured during battery operation, but is affected by dynamic conditions and its data interpretation is less accurate. Summary of the Invention
[0005] The purpose of the present invention is to provide a method and system for predicting the static impedance of lithium-ion batteries based on dynamic electrochemical impedance spectroscopy, which can directly predict the static impedance spectrum from the dynamic impedance spectrum data without requiring the battery to reach a steady state, thereby overcoming the limitations of traditional electrochemical impedance spectroscopy measurements.
[0006] In order to achieve the above object, the present invention adopts the following technical solutions:
[0007] The static impedance prediction method of lithium-ion batteries based on dynamic electrochemical impedance spectroscopy includes:
[0008] Obtain static electrochemical impedance spectroscopy, dynamic electrochemical impedance spectroscopy data, battery state and temperature and obtain statistical features from the dynamic electrochemical impedance spectroscopy;
[0009] Build a sequence-to-sequence deep learning model that includes residual neural networks and long short-term memory network units, defined as BatImp-ResSeq network;
[0010] The BatImp-ResSeq network is trained using dynamic electrochemical impedance spectroscopy data as input and its corresponding battery state, temperature, and statistical features as labels. The output is the predicted static electrochemical impedance spectrum, which enables it to learn the relationship between dynamic and static electrochemical impedance spectra.
[0011] The collected dynamic electrochemical impedance spectroscopy data and its corresponding battery state, temperature and statistical characteristics are input into the trained BatImp-ResSeq network to predict the corresponding static electrochemical impedance spectrum.
[0012] A further improvement of the present invention is to obtain static electrochemical impedance spectroscopy, dynamic electrochemical impedance spectroscopy data, battery state and temperature and obtain statistical features from the dynamic electrochemical impedance spectroscopy, including: maximum value, minimum value, average value, variance, covariance and ratio; the calculation formula is
[0013] Z max =max(Z i ),Z min =min(Z i )
[0014]
[0015] where Z i Indicates that in i th The impedance spectrum value at the frequency point, N is the total frequency point of the data, μ is the average value, Z max is the maximum value, Z min is the minimum value, σ 2 is the variance, cov is the covariance, and R is the ratio of the real part Z′ to the imaginary part Z″.
[0016] A further improvement of the present invention is that the BatImp-ResSeq network includes a ResNet module and a Seq2Seq structure, and the Seq2Seq structure includes two parts: an encoder and a decoder;
[0017] The ResNet module serves as the embedding layer for data processing, feeding the processed data into the encoder and decoder, each of which consists of two LSTM layers. The ResNet module contains multiple residual blocks, each of which consists of one or more convolutional layers, a batch normalization layer, and a ReLU activation function.
[0018] The encoder is used to extract the features of the dynamic electrochemical impedance spectroscopy data and encode them into hidden states; the decoder generates predicted static data based on the encoder's hidden states and the input dynamic electrochemical impedance spectroscopy data; the decoder first receives the hidden states output by the encoder, and then generates predicted static electrochemical impedance spectroscopy data one by one based on the input dynamic electrochemical impedance spectroscopy data and hidden states; the residual block alleviates the problem of gradient vanishing or exploding during deep neural network training by introducing residual connections; the LSTM layer can capture long-term dependencies in time series data.
[0019] A further improvement of the present invention is that training the BatImp-ResSeq network includes:
[0020] The dynamic electrochemical impedance spectroscopy data is input into the embedding layer 1 for data processing, and its corresponding battery state, temperature and statistical characteristics are input into the embedding layer 2 as labels for data processing;
[0021] The processed labels are input into the encoder, and the output of the encoder and the processed dynamic electrochemical impedance spectroscopy data are input into the decoder. The output of the decoder is the final output of the BatImp-ResSeq network, which is the predicted static electrochemical impedance spectroscopy data.
[0022] A further improvement of the present invention is that the training process of the BatImp-ResSeq network includes:
[0023] The mean square error (MSE) is calculated based on the predicted and actual static electrochemical impedance spectroscopy data as the network's loss function. The root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) are used as evaluation metrics. The adaptive moment estimation optimizer is used to update the model parameters. An early stopping mechanism is implemented to prevent model overfitting.
[0024] A further improvement of the present invention is that the ResNet module is used as an embedding layer to perform data processing, including:
[0025] The structure of the embedding layer 1 is three residual blocks, and the structure of the embedding layer 2 is two linear layers, one ReLU activation function layer and three residual blocks; data processing is performed on labels and inputs respectively.
[0026] A further improvement of the present invention is that the residual block includes:
[0027] The first convolution layer is used to extract features from the input channel and contains a 3×1 convolution kernel and a sliding window with a stride of 1; the first batch of normalization layers is used to stabilize the training process and accelerate convergence; the ReLU activation function is used to introduce nonlinear characteristics; the second convolution layer is used to further extract features and contains a 3×1 convolution kernel and a sliding window with a stride of 1; the second batch of normalization layers is used to normalize the features again; an optional 1×1 convolution layer for implementing jump connections is enabled when the number of input channels does not match the number of output channels; the addition operation is used to add the input features to the convolved features to form a residual connection.
[0028] The static impedance prediction system for lithium-ion batteries based on dynamic electrochemical impedance spectroscopy includes:
[0029] A data acquisition unit, configured to acquire static electrochemical impedance spectroscopy, dynamic electrochemical impedance spectroscopy data, battery state and temperature, and obtain statistical features from the dynamic electrochemical impedance spectroscopy;
[0030] The model building unit is used to build a sequence-to-sequence deep learning model, which contains residual neural network and long short-term memory network units, defined as BatImp-ResSeq network;
[0031] The model training unit is used to train the BatImp-ResSeq network using dynamic electrochemical impedance spectroscopy data as input and its corresponding battery state, temperature, and statistical features as labels. The output is the predicted static electrochemical impedance spectrum, which enables it to learn the relationship between dynamic and static electrochemical impedance spectra.
[0032] The impedance prediction unit is used to input the collected dynamic electrochemical impedance spectroscopy data and its corresponding battery state, temperature and statistical characteristics into the trained BatImp-ResSeq network to predict the corresponding static electrochemical impedance spectrum.
[0033] A further improvement of the present invention is that, in the data acquisition unit, static electrochemical impedance spectroscopy, dynamic electrochemical impedance spectroscopy data, battery state and temperature are obtained and statistical features are obtained from the dynamic electrochemical impedance spectroscopy, including: maximum value, minimum value, average value, variance, covariance and ratio; the calculation formula is
[0034] Z max =max(Z i ),Z min =min(Z i )
[0035]
[0036]
[0037]
[0038]
[0039] where Z i Indicates that in i th The impedance spectrum value at the frequency point, N is the total frequency point of the data, μ is the average value, Z max is the maximum value, Z min is the minimum value, σ 2 is the variance, cov is the covariance, and R is the ratio of the real part Z′ to the imaginary part Z″.
[0040] A further improvement of the present invention is that, in the model construction unit, the BatImp-ResSeq network includes a ResNet module and a Seq2Seq structure, and the Seq2Seq structure includes two parts: an encoder and a decoder;
[0041] The ResNet module serves as the embedding layer for data processing, feeding the processed data into the encoder and decoder, each of which consists of two LSTM layers. The ResNet module contains multiple residual blocks, each of which consists of one or more convolutional layers, a batch normalization layer, and a ReLU activation function.
[0042] The encoder is used to extract the features of the dynamic electrochemical impedance spectroscopy data and encode them into hidden states; the decoder generates predicted static data based on the encoder's hidden states and the input dynamic electrochemical impedance spectroscopy data; the decoder first receives the hidden states output by the encoder, and then generates predicted static electrochemical impedance spectroscopy data one by one based on the input dynamic electrochemical impedance spectroscopy data and hidden states; the residual block alleviates the problem of gradient vanishing or exploding during deep neural network training by introducing residual connections; the LSTM layer can capture long-term dependencies in time series data.
[0043] Compared with the prior art, the present invention has at least the following beneficial technical effects:
[0044] The static impedance prediction method and system for lithium-ion batteries based on dynamic electrochemical impedance spectroscopy provided by the present invention can reduce the time required for static electrochemical impedance spectroscopy measurement. Compared with traditional electrochemical impedance spectroscopy methods, static electrochemical impedance spectrum can be predicted directly from dynamic electrochemical impedance spectroscopy data without long rest periods, thereby improving measurement efficiency.
[0045] This paper uses a deep learning model that combines a residual network with a long short-term memory network, significantly improving generalization capabilities and making it suitable for impedance prediction under different operating conditions. Furthermore, end-to-end learning reduces reliance on electrochemical impedance spectroscopy equipment, thereby reducing hardware costs.
[0046] Compared with traditional electrochemical impedance spectroscopy measurement methods, the present invention not only achieves faster impedance measurement, but also improves the accuracy and robustness of prediction, providing an efficient and feasible solution for battery health assessment. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0048] Figure 1 This is the BatImp-ResSeq network structure diagram.
[0049] Figure 2 This is the specific structure diagram of the residual block.
[0050] Figure 3 The figure is a structural block diagram of the lithium-ion battery static impedance prediction system based on dynamic electrochemical impedance spectroscopy of the present invention. DETAILED DESCRIPTION
[0051] Hereinafter, only certain exemplary embodiments are briefly described. As will be appreciated by those skilled in the art, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and description are to be considered as illustrative in nature and not restrictive.
[0052] It will be understood that when used in this specification and the appended claims, the terms “comprises” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.
[0053] It should also be understood that the terms used in the present specification are only for the purpose of describing particular embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.
[0054] It should be further understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.
[0055] The accompanying drawings illustrate various schematic diagrams of structures according to embodiments disclosed herein. These figures are not drawn to scale; for clarity, some details are exaggerated and some details may be omitted. The shapes of the various regions and layers shown in the figures, as well as their relative sizes and positional relationships, are merely exemplary and may deviate in practice due to manufacturing tolerances or technical limitations. Those skilled in the art may design regions / layers with different shapes, sizes, and relative positions as needed.
[0056] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0057] Example 1
[0058] The static impedance prediction method of lithium-ion batteries based on dynamic electrochemical impedance spectroscopy includes:
[0059] Obtain static electrochemical impedance spectroscopy, dynamic electrochemical impedance spectroscopy data, battery state and temperature and obtain statistical features from the dynamic electrochemical impedance spectroscopy;
[0060] Build a sequence-to-sequence deep learning model that includes residual neural networks and long short-term memory network units, defined as BatImp-ResSeq network;
[0061] The BatImp-ResSeq network is trained using dynamic electrochemical impedance spectroscopy data as input and its corresponding battery state, temperature, and statistical features as labels. The output is the predicted static electrochemical impedance spectrum, which enables it to learn the relationship between dynamic and static electrochemical impedance spectra.
[0062] The collected dynamic electrochemical impedance spectroscopy data and its corresponding battery state, temperature and statistical characteristics are input into the trained BatImp-ResSeq network to predict the corresponding static electrochemical impedance spectrum.
[0063] In this embodiment, static electrochemical impedance spectroscopy, dynamic electrochemical impedance spectroscopy data, battery state and temperature are obtained, and statistical features are obtained from the dynamic electrochemical impedance spectroscopy, including maximum value, minimum value, average value, variance, covariance and ratio; the calculation formula is:
[0064] Z max =max(Z i ),Z min =min(Z i )
[0065]
[0066]
[0067]
[0068]
[0069] where Z i Indicates that in i th The impedance spectrum value at the frequency point, N is the total frequency point of the data, μ is the average value, Z max is the maximum value, Z min is the minimum value, σ 2 is the variance, cov is the covariance, and R is the ratio of the real part Z′ to the imaginary part Z″.
[0070] In this embodiment, the BatImp-ResSeq network includes a ResNet module and a Seq2Seq structure, and the Seq2Seq structure includes two parts: an encoder and a decoder;
[0071] The ResNet module serves as the embedding layer for data processing, feeding the processed data into the encoder and decoder, each of which consists of two LSTM layers. The ResNet module contains multiple residual blocks, each of which consists of one or more convolutional layers, a batch normalization layer, and a ReLU activation function.
[0072] The encoder is used to extract the features of the dynamic electrochemical impedance spectroscopy data and encode them into hidden states; the decoder generates predicted static data based on the encoder's hidden states and the input dynamic electrochemical impedance spectroscopy data; the decoder first receives the hidden states output by the encoder, and then generates predicted static electrochemical impedance spectroscopy data one by one based on the input dynamic electrochemical impedance spectroscopy data and hidden states; the residual block alleviates the problem of gradient vanishing or exploding during deep neural network training by introducing residual connections; the LSTM layer can capture long-term dependencies in time series data.
[0073] In this embodiment, training the BatImp-ResSeq network includes:
[0074] The dynamic electrochemical impedance spectroscopy data is input into the embedding layer 1 for data processing, and its corresponding battery state, temperature and statistical characteristics are input into the embedding layer 2 as labels for data processing;
[0075] The processed labels are input into the encoder, and the output of the encoder and the processed dynamic electrochemical impedance spectroscopy data are input into the decoder. The output of the decoder is the final output of the BatImp-ResSeq network, which is the predicted static electrochemical impedance spectroscopy data.
[0076] In this embodiment, the training process of the BatImp-ResSeq network includes:
[0077] The mean square error (MSE) is calculated based on the predicted and actual static electrochemical impedance spectroscopy data as the network's loss function. The root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) are used as evaluation metrics. The adaptive moment estimation optimizer is used to update the model parameters. An early stopping mechanism is implemented to prevent model overfitting.
[0078] In this embodiment, the ResNet module acts as an embedding layer to perform data processing, including:
[0079] The structure of the embedding layer 1 is three residual blocks, and the structure of the embedding layer 2 is two linear layers, one ReLU activation function layer and three residual blocks; data processing is performed on labels and inputs respectively.
[0080] In this embodiment, the residual block includes:
[0081] The first convolution layer is used to extract features from the input channel and contains a 3×1 convolution kernel and a sliding window with a stride of 1; the first batch of normalization layers is used to stabilize the training process and accelerate convergence; the ReLU activation function is used to introduce nonlinear characteristics; the second convolution layer is used to further extract features and contains a 3×1 convolution kernel and a sliding window with a stride of 1; the second batch of normalization layers is used to normalize the features again; an optional 1×1 convolution layer for implementing jump connections is enabled when the number of input channels does not match the number of output channels; the addition operation is used to add the input features to the convolved features to form a residual connection.
[0082] Example 2
[0083] The present invention provides a method for predicting the static impedance of a lithium-ion battery based on dynamic electrochemical impedance spectroscopy, comprising:
[0084] The method first extracts the battery state, temperature and statistical features of the impedance spectrum, such as maximum, minimum, average, variance, covariance and ratio, from the dynamic electrochemical impedance spectroscopy data.
[0085] Then, a deep learning model based on sequence to sequence (Seq2Seq) was constructed, which contains residual neural network (ResNet) and long short-term memory network (LSTM) units, named BatImp-ResSeq network. The specific network structure is as follows Figure 1 shown.
[0086] The BatImp-ResSeq model adopts a Seq2Seq structure and consists of two parts: an encoder and a decoder. The encoder is used to extract the features of the dynamic electrochemical impedance spectroscopy data and encode it into hidden states. The decoder generates predicted static data based on the hidden states of the encoder and the input dynamic electrochemical impedance spectroscopy data. The residual block alleviates the problem of gradient vanishing or exploding during deep neural network training by introducing residual connections, thereby enhancing the learning and generalization capabilities of the model. The LSTM layer can capture long-term dependencies in time series data. The decoder first receives the hidden state output by the encoder, and then generates predicted static electrochemical impedance spectroscopy data one by one based on the input dynamic electrochemical impedance spectroscopy data and hidden states. The ResNet module contains multiple residual blocks, each of which consists of one or more convolutional layers, batch normalization layers, and ReLU activation functions. Figure 2 This is the specific structure diagram of the residual block.
[0087] Finally, we used dynamic EIS data and its corresponding static EIS data to train our model, enabling it to learn the complex relationship between dynamic and static EIS spectra. By feeding new dynamic EIS data into the trained model, we can predict the corresponding static EIS spectra.
[0088] Example 3
[0089] Eunicell lithium-ion battery dataset:
[0090] The dataset includes electrochemical impedance spectroscopy (EIS) measurements of over 20,000 lithium-ion cells, acquired at various battery SOHs, battery states, and temperatures used in this study. The EIS range extends from 0.02 Hz to 20 kHz. The EIS data pertains to twelve 45 mAh Eunicell LR2032 coin cells. These twelve cells are labeled 25C01 to 25C08, indicating cycling at 25°C. 35C01 to 35C02 and 45C01 to 45C02 represent cells cycled at 35°C and 45°C, respectively. EIS measurements were systematically performed at nine states at each even-numbered cycle, encompassing both static and dynamic EIS. We selected cells 25C01-04, 25C07-08, 35C01, and 45C01 as the training set, and the remaining cells as the test set.
[0091] The evaluation indicators are MAE, MSR, MAPE and R 2 .
[0092] Table 1 Comparison of static electrochemical impedance spectroscopy prediction effects of various models
[0093]
[0094]
[0095] The results showed that there were significant differences in the prediction ability of static electrochemical impedance spectroscopy among the compared models. BatImp-ResSeq had a greater advantage. The MAE, MSE, and MAPE of BatImp-ResSeq were the smallest. 2 The value is close to 1, indicating that BatImp-ResSeq not only has high prediction accuracy but also stable fitting performance. Although the prediction of the imaginary part is more complex, BatImp-ResSeq can still outperform other models in all four evaluation metrics, demonstrating that BatImp-ResSeq has better ability to handle complex data than other comparison models.
[0096] Example 4
[0097] like Figure 3 As shown, the lithium-ion battery static impedance prediction system based on dynamic electrochemical impedance spectroscopy provided by the present invention includes:
[0098] A data acquisition unit, configured to acquire static electrochemical impedance spectroscopy, dynamic electrochemical impedance spectroscopy data, battery state and temperature, and obtain statistical features from the dynamic electrochemical impedance spectroscopy;
[0099] The model building unit is used to build a sequence-to-sequence deep learning model, which contains residual neural network and long short-term memory network units, defined as BatImp-ResSeq network;
[0100] The model training unit is used to train the BatImp-ResSeq network using dynamic electrochemical impedance spectroscopy data as input and its corresponding battery state, temperature, and statistical features as labels. The output is the predicted static electrochemical impedance spectrum, which enables it to learn the relationship between dynamic and static electrochemical impedance spectra.
[0101] The impedance prediction unit is used to input the collected dynamic electrochemical impedance spectroscopy data and its corresponding battery state, temperature and statistical characteristics into the trained BatImp-ResSeq network to predict the corresponding static electrochemical impedance spectrum.
[0102] In this embodiment, the data acquisition unit obtains static electrochemical impedance spectroscopy, dynamic electrochemical impedance spectroscopy data, battery state and temperature, and obtains statistical features from the dynamic electrochemical impedance spectroscopy, including: maximum value, minimum value, average value, variance, covariance and ratio; the calculation formula is:
[0103] Z max =max(Z i ),Z min =min(Z i )
[0104]
[0105]
[0106]
[0107]
[0108] where Z i Indicates that in i th The impedance spectrum value at the frequency point, N is the total frequency point of the data, μ is the average value, Z max is the maximum value, Z min is the minimum value, σ 2 is the variance, cov is the covariance, and R is the ratio of the real part Z′ to the imaginary part Z″.
[0109] In this embodiment, in the model construction unit, the BatImp-ResSeq network includes a ResNet module and a Seq2Seq structure, and the Seq2Seq structure includes two parts: an encoder and a decoder;
[0110] The ResNet module serves as the embedding layer for data processing, feeding the processed data into the encoder and decoder, each of which consists of two LSTM layers. The ResNet module contains multiple residual blocks, each of which consists of one or more convolutional layers, a batch normalization layer, and a ReLU activation function.
[0111] The encoder is used to extract the features of the dynamic electrochemical impedance spectroscopy data and encode them into hidden states; the decoder generates predicted static data based on the encoder's hidden states and the input dynamic electrochemical impedance spectroscopy data; the decoder first receives the hidden states output by the encoder, and then generates predicted static electrochemical impedance spectroscopy data one by one based on the input dynamic electrochemical impedance spectroscopy data and hidden states; the residual block alleviates the problem of gradient vanishing or exploding during deep neural network training by introducing residual connections; the LSTM layer can capture long-term dependencies in time series data.
[0112] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, from all points of view, the embodiments should be regarded as illustrative and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes that fall within the meaning and range of equivalents of the claims are included in the present invention. Any reference signs in the claims should not be construed as limiting the claim to which they relate.
[0113] In addition, it should be understood that although this specification describes the embodiments, not every embodiment contains only one independent technical solution. This description is for clarity only. Those skilled in the art should consider the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art. The above content is only for the purpose of illustrating the technical concept of the present invention and cannot be used to limit the scope of protection of the present invention. Any changes made based on the technical solution in accordance with the technical concept proposed by the present invention fall within the scope of protection of the claims of the present invention.
Claims
1. A method for predicting static impedance of lithium-ion batteries based on dynamic electrochemical impedance spectroscopy, characterized in that: include: Obtain static electrochemical impedance spectroscopy, dynamic electrochemical impedance spectroscopy data, battery state and temperature and obtain statistical features from the dynamic electrochemical impedance spectroscopy; Build a sequence-to-sequence deep learning model that includes residual neural networks and long short-term memory network units, defined as BatImp-ResSeq network; The BatImp-ResSeq network is trained using dynamic electrochemical impedance spectroscopy data as input and its corresponding battery state, temperature, and statistical features as labels. The output is the predicted static electrochemical impedance spectrum, which enables it to learn the relationship between dynamic and static electrochemical impedance spectra. The collected dynamic electrochemical impedance spectroscopy data and its corresponding battery state, temperature and statistical characteristics are input into the trained BatImp-ResSeq network to predict the corresponding static electrochemical impedance spectrum.
2. The method for predicting static impedance of a lithium-ion battery based on dynamic electrochemical impedance spectroscopy according to claim 1, wherein: Obtain static electrochemical impedance spectroscopy, dynamic electrochemical impedance spectroscopy data, battery state and temperature and obtain statistical features from the dynamic electrochemical impedance spectroscopy, including: maximum, minimum, average, variance, covariance and ratio; the calculation formula is Z max =max(Z i ),Z min =min(Z i ) where Z i Indicates that in i th The impedance spectrum value at the frequency point, N is the total frequency point of the data, μ is the average value, Z max is the maximum value, Z min is the minimum value, σ 2 is the variance, cov is the covariance, and R is the ratio of the real part Z′ to the imaginary part Z″.
3. The method for predicting static impedance of a lithium-ion battery based on dynamic electrochemical impedance spectroscopy according to claim 1, wherein: The BatImp-ResSeq network includes the ResNet module and the Seq2Seq structure, which consists of two parts: the encoder and the decoder; The ResNet module serves as the embedding layer for data processing, feeding the processed data into the encoder and decoder, each of which consists of two LSTM layers. The ResNet module contains multiple residual blocks, each of which consists of one or more convolutional layers, a batch normalization layer, and a ReLU activation function. The encoder is used to extract the features of dynamic electrochemical impedance spectroscopy data and encode them into hidden states; The decoder generates predicted static data based on the encoder's hidden state and the input dynamic electrochemical impedance spectroscopy data; The decoder first receives the hidden state output by the encoder, and then generates predicted static electrochemical impedance spectroscopy data one by one based on the input dynamic electrochemical impedance spectroscopy data and hidden state; the residual block alleviates the problem of gradient disappearance or explosion during deep neural network training by introducing residual connections; the LSTM layer can capture long-term dependencies in time series data.
4. The method for predicting static impedance of a lithium-ion battery based on dynamic electrochemical impedance spectroscopy according to claim 3, wherein: Training the BatImp-ResSeq network involves: The dynamic electrochemical impedance spectroscopy data is input into the embedding layer 1 for data processing, and its corresponding battery state, temperature and statistical characteristics are input into the embedding layer 2 as labels for data processing; The processed labels are input into the encoder, and the output of the encoder and the processed dynamic electrochemical impedance spectroscopy data are input into the decoder. The output of the decoder is the final output of the BatImp-ResSeq network, which is the predicted static electrochemical impedance spectroscopy data.
5. The method for predicting static impedance of a lithium-ion battery based on dynamic electrochemical impedance spectroscopy according to claim 4, characterized in that: The training process of the BatImp-ResSeq network includes: The mean square error (MSE) is calculated based on the predicted and actual static electrochemical impedance spectroscopy data as the network's loss function. The root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) are used as evaluation metrics. The adaptive moment estimation optimizer is used to update the model parameters. An early stopping mechanism is implemented to prevent model overfitting.
6. The method for predicting static impedance of lithium-ion batteries based on dynamic electrochemical impedance spectroscopy according to claim 3, characterized in that: The ResNet module acts as an embedding layer for data processing, including: The structure of the embedding layer 1 is three residual blocks, and the structure of the embedding layer 2 is two linear layers, one ReLU activation function layer and three residual blocks; data processing is performed on labels and inputs respectively.
7. The method for predicting static impedance of a lithium-ion battery based on dynamic electrochemical impedance spectroscopy according to claim 3, wherein: The residual block includes: The first convolution layer is used to extract features from the input channel and contains a 3×1 convolution kernel and a sliding window with a stride of 1; the first batch of normalization layers is used to stabilize the training process and accelerate convergence; the ReLU activation function is used to introduce nonlinear characteristics; the second convolution layer is used to further extract features and contains a 3×1 convolution kernel and a sliding window with a stride of 1; the second batch of normalization layers is used to normalize the features again; an optional 1×1 convolution layer for implementing jump connections is enabled when the number of input channels does not match the number of output channels; the addition operation is used to add the input features to the convolved features to form a residual connection.
8. A lithium-ion battery static impedance prediction system based on dynamic electrochemical impedance spectroscopy, characterized in that: include: A data acquisition unit, configured to acquire static electrochemical impedance spectroscopy, dynamic electrochemical impedance spectroscopy data, battery state and temperature, and obtain statistical features from the dynamic electrochemical impedance spectroscopy; The model building unit is used to build a sequence-to-sequence deep learning model, which contains residual neural network and long short-term memory network units, defined as BatImp-ResSeq network; The model training unit is used to train the BatImp-ResSeq network using dynamic electrochemical impedance spectroscopy data as input and its corresponding battery state, temperature, and statistical features as labels. The output is the predicted static electrochemical impedance spectrum, which enables it to learn the relationship between dynamic and static electrochemical impedance spectra. The impedance prediction unit is used to input the collected dynamic electrochemical impedance spectroscopy data and its corresponding battery state, temperature and statistical characteristics into the trained BatImp-ResSeq network to predict the corresponding static electrochemical impedance spectrum.
9. The lithium-ion battery static impedance prediction system based on dynamic electrochemical impedance spectroscopy according to claim 8, characterized in that: In the data acquisition unit, static electrochemical impedance spectroscopy, dynamic electrochemical impedance spectroscopy data, battery state and temperature are obtained, and statistical features are obtained from the dynamic electrochemical impedance spectroscopy, including maximum value, minimum value, average value, variance, covariance and ratio; the calculation formula is Z max =max(Z i ),Z min =min(Z i ) where Z i Indicates that in i th The impedance spectrum value at the frequency point, N is the total frequency point of the data, μ is the average value, Z max is the maximum value, Z min is the minimum value, σ 2 is the variance, cov is the covariance, and R is the ratio of the real part Z′ to the imaginary part Z″.
10. The lithium-ion battery static impedance prediction system based on dynamic electrochemical impedance spectroscopy according to claim 8, characterized in that: In the model construction unit, the BatImp-ResSeq network includes the ResNet module and the Seq2Seq structure, which consists of two parts: the encoder and the decoder; The ResNet module serves as the embedding layer for data processing, feeding the processed data into the encoder and decoder, each of which consists of two LSTM layers. The ResNet module contains multiple residual blocks, each of which consists of one or more convolutional layers, a batch normalization layer, and a ReLU activation function. The encoder is used to extract the features of dynamic electrochemical impedance spectroscopy data and encode them into hidden states; The decoder generates predicted static data based on the encoder's hidden state and the input dynamic electrochemical impedance spectroscopy data; The decoder first receives the hidden state output by the encoder, and then generates predicted static electrochemical impedance spectroscopy data one by one based on the input dynamic electrochemical impedance spectroscopy data and hidden state; the residual block alleviates the problem of gradient disappearance or explosion during deep neural network training by introducing residual connections; the LSTM layer can capture long-term dependencies in time series data.