SOH prediction method and system based on CNN-LSTM-Attention model

Abstract

The invention discloses an SOH prediction method and system based on a CNN-LSTM-Attention model, and a readable storage medium. The method comprises the following steps: testing and collecting battery EIS data and battery impedance capacity data; the method comprises the following steps: constructing and training a CNN-LSTM-Attention model; and inputting the EIS data into the CNN-LSTM-Attention model to carry out SOH prediction, and outputting a prediction result. According to the method, the CNN model, the LSTM model and the Attention model are combined, and the SOH is jointly predicted based on the EIS data. EIS is regarded as a characteristic, and a real part and an imaginary part of impedance data corresponding to each frequency are regarded as characteristic values. A CNN model is used for extracting spatial features of an impedance spectrum, an LSTM model is used for extracting a time sequence change rule of the impedance spectrum, Attention is used for further selecting important parts in spatial and temporal features of EIS data, and SOH is jointly predicted. The pertinence of the model on data processing is improved, and the SOH prediction efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of electrochemical technology, and more specifically, to a SOH prediction method, system and readable storage medium based on a CNN-LSTM-Attention model. Background Art

[0002] Electrochemical Impedance Spectroscopy (EIS) is a test method used to characterize the electrochemical process inside a battery. EIS can be used to test the impedance spectrum of the battery, including real impedance and imaginary impedance. The attenuation of the battery's state of health (SOH) will lead to an increase in the battery's internal resistance and AC impedance. Therefore, the impedance spectrum obtained by the EIS test can be used to predict the battery's SOH.

[0003] Chinese patent CN114325403A published on April 12, 2022 provides a lithium-ion battery life detection method and system based on electrochemical impedance spectroscopy testing. The method includes an electrochemical impedance spectroscopy testing step, a neural network-based impedance spectrum learning and analysis step, and a life detection step. The electrochemical impedance spectrum sampling rate can be improved through electrochemical impedance spectroscopy testing, and then the electrochemical impedance spectrum is trained using a deep learning image recognition algorithm or prediction algorithm. The trained model analyzes the real-time measured impedance spectrum to identify the SOH state of the lithium-ion battery impedance spectrum, realize non-destructive testing of lithium-ion batteries, and ensure the effectiveness and reliability of the detection. However, the scheme is not very targeted for the use of EIS data, and the prediction efficiency of SOH needs to be improved. Summary of the invention

[0004] In view of the above problems, the object of the present invention is to provide a SOH prediction method, system and readable storage medium based on the CNN-LSTM-Attention model.

[0005] The first aspect of the present invention provides a SOH prediction method based on a CNN-LSTM-Attention model. The method comprises the following steps:

[0006] Test and collect battery EIS data and battery impedance capacity data;

[0007] Build and train the CNN-LSTM-Attention model;

[0008] The EIS data is input into the CNN-LSTM-Attention model for SOH prediction, and the prediction results are output.

[0009] Preferably, the collecting of battery EIS data is specifically:

[0010] Collect battery EIS data at preset frequency intervals;

[0011] When collecting battery EIS data, record the real and imaginary impedance data of the battery EIS data corresponding to each frequency.

[0012] Preferably, the CNN-LSTM-Attention model includes: an input layer, a CNN model, an LSTM model, an Attention model, and a linear regression layer;

[0013] The input layer is used to input EIS data into the CNN model;

[0014] The CNN model extracts EIS spatial characteristics from the principal component data to obtain EIS spatial features, and outputs the EIS spatial features to the LSTM model;

[0015] The LSTM model extracts the time series change pattern of EIS based on the spatial features of EIS, and then outputs the time series change pattern of EIS to the Attention model;

[0016] The Attention model extracts key features of EIS spatial features based on the time series change pattern of EIS;

[0017] The linear regression layer predicts the battery SOH based on the key features in the EIS spatial features.

[0018] Preferably, the CNN model includes a two-layer CNN structure.

[0019] Preferably, the CNN structure includes two convolutional layers and one pooling layer, the activation functions of the convolutional layers are all relu functions, the number of filters is set to 32, the kernel size is 1, and the pooling size of the pooling layer is also 1.

[0020] Preferably, the LSTM model includes an LSTM layer and a fully connected layer, the number of neurons in the LSTM layer is 32, the number of neurons in the fully connected layer is 16, and the activation functions are both relu functions.

[0021] Preferably, the Attention model extracts key features of the EIS spatial features according to the time series variation law of the EIS, specifically:

[0022] Input the EIS time series into the Attention model and learn an attention weight vector;

[0023] Apply the attention weight to the hidden state of each time step in the EIS time series and perform weighted summation to obtain the weighted EIS time series;

[0024] The weighted EIS time series are sorted by importance, and the EIS data with importance greater than the preset importance in the EIS spatial features are extracted as the key features in the EIS spatial features.

[0025] Preferably, the Attention model is a BahdanauAttention structure;

[0026] BahdanauAttention is divided into two steps: Encoder and Decoder. The calculation formulas are:

[0027] (1) Encoder:

[0028] h i =tanh(W[h i-1 ,x i ])

[0029] o i =softmax(Vh i )

[0030] where h i Represents input x i The hidden state vector of i represents the normalized output vector of the hidden state at time i, and W and V represent the weight matrices to be learned.

[0031] (2)Decoder:

[0032] The first step is to generate the semantic vector at time t:

[0033]

[0034]

[0035]

[0036] Among them, c t represents the semantic vector at time t; e ti is the hidden state vector h at the i-th moment in the encoder i For the hidden state s at the decoder at time t t The degree of influence; α tk It means that e is transformed by the softmax function ti The normalized value, W a represents the weight matrix to be learned, v a represents the weight vector to be learned, and N represents the total time;

[0037] The second step is to pass the hidden layer information and predict:

[0038] s t =tanh(W[s t-1 ,y t-1 ,c t ])

[0039] o t =softmax(Vs t )

[0040] Among them, s t represents the hidden state of the decoder at time t, y t-1 represents t-1 prediction targets, o t represents the prediction distribution.

[0041] Preferably, the activation function of the linear regression layer is a linear function;

[0042] The loss function of the CNN-LSTM-Attention model is mean square error, the optimizer is rmsprop, and the evaluation metric is mean absolute error.

[0043] A second aspect of the present invention provides a SOH prediction system based on a CNN-LSTM-Attention model, comprising a memory and a processor, wherein the memory comprises a SOH prediction method program based on the CNN-LSTM-Attention model, and when the SOH prediction method program based on the CNN-LSTM-Attention model is executed by the processor, the steps of the SOH prediction method based on the CNN-LSTM-Attention model are implemented.

[0044] A third aspect of the present invention provides a readable storage medium, which includes a SOH prediction method program based on a CNN-LSTM-Attention model. When the SOH prediction method program based on the CNN-LSTM-Attention model is executed by a processor, the steps of the SOH prediction method based on the CNN-LSTM-Attention model are implemented.

[0045] Compared with the prior art, the beneficial effect of the technical solution of the present invention is as follows: the present invention combines CNN, LSTM and Attention models to jointly predict SOH based on EIS data. EIS is regarded as a feature, and the real and imaginary parts of the impedance data corresponding to each frequency are used as feature values. The CNN model is used to extract the spatial features of the impedance spectrum, the LSTM model is used to extract the time series variation law of the impedance spectrum, and Attention is used to further select the important parts of the spatiotemporal features of the EIS data to jointly predict SOH. This increases the pertinence of the model to data processing and improves the efficiency of SOH prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flow chart of a SOH prediction method based on a CNN-LSTM-Attention model described in Example 1.

[0047] Figure 2 This is the structure diagram of the CNN-LSTM-Attention model.

[0048] Figure 3 This is the BahdanauAttention structure diagram.

[0049] Figure 4 The following is a sample diagram of SOH prediction results based on the CNN-LSTM-Attention model. DETAILED DESCRIPTION

[0050] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0051] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0052] Example 1

[0053] like Figure 1 As shown, this embodiment discloses a SOH prediction method based on a CNN-LSTM-Attention model. The method comprises the following steps:

[0054] Test and collect battery EIS data and battery impedance capacity data;

[0055] Build and train the CNN-LSTM-Attention model;

[0056] The EIS data is input into the CNN-LSTM-Attention model for SOH prediction, and the prediction results are output.

[0057] It should be noted that this application improves CNN-LSTM, adds an attention mechanism, combines CNN, LSTM and Attention models, and jointly predicts SOH based on EIS data. EIS is regarded as a feature, and the real and imaginary parts of the impedance data corresponding to each frequency are used as feature values. The CNN model is used to extract the spatial characteristics of the impedance spectrum, the LSTM model is used to extract the time series variation law of the impedance spectrum, and Attention is used to further select the important parts of the spatiotemporal characteristics of the EIS data to jointly predict SOH. In order to increase the pertinence of the model to data processing, the efficiency of SOH prediction is improved.

[0058] According to an embodiment of the present invention, the collecting of battery EIS data is specifically:

[0059] Collect battery EIS data at preset frequency intervals;

[0060] When collecting battery EIS data, record the real and imaginary impedance data of the battery EIS data corresponding to each frequency.

[0061] It should be noted that testing and collecting battery electrochemical impedance spectrum data requires professional impedance spectrum testing instruments and needs to cover a certain frequency range. The frequency range can be set from 10-2HZ to 105HZ. Multiple frequency values ​​are selected at certain intervals in the middle of the frequency range. The frequency values ​​are generally selected from 5 to 60. In this embodiment, 60 are selected as a sample.

[0062] When collecting data, it is necessary to record the real and imaginary data of the impedance corresponding to each frequency, as well as the target value that needs to be predicted, which is the corresponding battery capacity data.

[0063] According to an embodiment of the present invention, the CNN-LSTM-Attention model includes: an input layer, a CNN model, an LSTM model, an Attention model, and a linear regression layer;

[0064] The input layer is used to input EIS data into the CNN model;

[0065] The CNN model extracts EIS spatial characteristics from the principal component data to obtain EIS spatial features, and outputs the EIS spatial features to the LSTM model;

[0066] The LSTM model extracts the time series change pattern of EIS based on the spatial features of EIS, and then outputs the time series change pattern of EIS to the Attention model;

[0067] The Attention model extracts key features of EIS spatial features based on the time series change pattern of EIS;

[0068] The linear regression layer predicts the battery SOH based on the key features in the EIS spatial features.

[0069] It should be noted that in the model using EIS to predict SOH, the convolution neural network (CNN) can extract the spatial features of the impedance spectrum, and the long short-term memory network (LSTM) can learn the time series variation of the impedance spectrum. The attention mechanism can learn the relative importance of each time step in the sequence, thereby strengthening the model's understanding of the sequence context, and can dynamically select the important parts of the sequence to obtain a more effective sequence expression.

[0070] like Figure 2 As shown, in this embodiment, the CNN-LSTM-Attention model input parts real1 to real60 represent the real data of the impedance data corresponding to high frequency to low frequency, and imaginary1 to imaginary60 represent the imaginary data of the impedance data corresponding to high frequency to low frequency;

[0071] Next is a 2-layer CNN structure with the activation function of relu. After that, the convolution calculation result needs to be subjected to the maximum pooling operation. Then, the maximum pooling calculation result is flattened, that is, flattened. After that, the flattened layer is input into the LSTM structure with the activation function of relu. Then, the output of LSTM is input into an Attention model to further learn the relative importance of each time step in the sequence. Next, it is connected to a fully connected layer. Finally, the SOH is output through the linear regression layer.

[0072] In summary, CNN is used to learn the local spatial features of the sequence; LSTM is used to capture the temporal dependencies in the sequence; the attention mechanism strengthens the model's understanding of the sequence context; the combination of the three can effectively learn sequence data with temporal and local correlation structures.

[0073] According to an embodiment of the present invention, the CNN model includes a two-layer CNN structure.

[0074] According to an embodiment of the present invention, the CNN structure includes two convolutional layers and one pooling layer, the activation functions of the convolutional layers are all relu functions, the number of filters is set to 32, the kernel size is 1, and the pooling size of the pooling layer is also 1.

[0075] It should be noted that CNN can automatically learn the spatial correlation of input data and extract high-level abstract features.

[0076] The CNN structure of this embodiment includes two convolutional layers and one pooling layer. The activation function of the convolutional layer is the relu function, the number of filters is set to 32, the kernel size is 1, and the pooling size of the pooling layer is also 1. Each convolutional layer can extract high-level features of the input data, and the pooling layer implements downsampling to reduce the data dimension. Finally, the three-dimensional features are mapped to a one-dimensional vector through the Flatten layer as the spatial feature expression of the EIS.

[0077] According to an embodiment of the present invention, the LSTM model includes an LSTM layer and a fully connected layer, the number of neurons in the LSTM layer is 32, the number of neurons in the fully connected layer is 16, and the activation functions are both relu functions.

[0078] It should be noted that LSTM is a special RNN that can learn long-term dependencies and capture dynamic changes in time series data. The LSTM model described in this embodiment includes an LSTM layer and a fully connected layer. The number of neurons in the LSTM layer is 32, the number of neurons in the fully connected layer is 16, and the activation function is the relu function. The LSTM layer can learn the long-term dependencies of the input data and identify the dynamic change pattern of the time series. The fully connected layer further extracts the LSTM output to obtain the high-level feature expression of the time series.

[0079] According to an embodiment of the present invention, the Attention model extracts key features of the EIS spatial features according to the time series variation law of the EIS, specifically:

[0080] Input the EIS time series into the Attention model and learn an attention weight vector;

[0081] Apply the attention weight to the hidden state of each time step in the EIS time series and perform weighted summation to obtain the weighted EIS time series;

[0082] The weighted EIS time series are sorted by importance, and the EIS data with importance greater than the preset importance in the EIS spatial features are extracted as the key features in the EIS spatial features.

[0083] It should be noted that the CNN-LSTM model has extracted the spatiotemporal features of the EIS data. The output results of the CNN-LSTM are used as the input of the Attention model to further calculate the relative importance of the features and assign higher weights to relatively important features. In other words, the important parts of the features are extracted, and the important parts of the features are the key features.

[0084] According to an embodiment of the present invention, the Attention model described in this embodiment is a BahdanauAttention structure; specifically, Figure 3 shown.

[0085] BahdanauAttention is divided into two steps: Encoder and Decoder. The calculation formulas are:

[0086] (1) Encoder:

[0087] h i =tanh(W[h i-1 ,x i ])

[0088] o i =softmax(Vh i )

[0089] where h i Represents input x i The hidden state vector of i represents the normalized output vector of the hidden state at time i, and W and V represent the weight matrices to be learned.

[0090] (2)Decoder:

[0091] The first step is to generate the semantic vector at time t:

[0092]

[0093]

[0094]

[0095] Among them, c t represents the semantic vector at time t; e ti is the hidden state vector h at the i-th moment in the encoder i For the hidden state s at the decoder at time t t The degree of influence; α tk It means that e is transformed by the softmax function ti The normalized value, W a represents the weight matrix to be learned, v a represents the weight vector to be learned, and N represents the total time;

[0096] The second step is to pass the hidden layer information and predict:

[0097] s t =tanh(W[s t-1 ,y t-1 ,ct ])

[0098] o t =softmax(Vs t )

[0099] Among them, s t represents the hidden state of the decoder at time t, y t-1 represents t-1 prediction targets, o t represents the prediction distribution.

[0100] According to the embodiment of the present invention, the output layer of the CNN-LSTM-Attention model is a linear regression layer, the predicted value is the battery capacity, i.e., SOH, and the activation function is a linear function. The loss function of the CNN-LSTM-Attention model is the mean square error (MSE); the optimizer is rmsprop; the evaluation metric is the mean absolute error (MAE), i.e., the sum of the absolute values ​​of the difference between the target value and the predicted value.

[0101] As a specific embodiment, the present invention is described below in conjunction with specific data.

[0102] This embodiment tests the effect on a public data set, data source: https: / / github.com / YunweiZhang / ML-identify-battery-degradation, and uses the impedance data therein to predict the battery capacity. The data set is divided into a training set and a test set in a ratio of 8:2, the CNN-LSTM-Attention model is trained on the training set, and the effect is tested on the test set. The test data has a total of 120 features, the feature numbers are 1 to 120, the first 60 features are the real part data of the impedance data corresponding to the frequency from high to low, and the last 60 features are the imaginary part data of the impedance data corresponding to the frequency from high to low.

[0103] Table 1 Battery EIS prediction battery SOH data set sample (partial)

[0104] f1 f2 f3 …… f119 f120 SOH 0.3847 0.39156 0.39684 …… 0.29026 0.32795 37.2027 0.38886 0.39343 0.40066 …… 0.28064 0.323 36.2230 0.39038 0.396 0.4025 …… 0.28773 0.32955 35.5893 0.39194 0.39643 0.40406 …… 0.29768 0.33673 35.1081 0.39225 0.3976 0.4054 …… 0.28396 0.32562 34.7673

[0105] Among them, f1 to f120 represent the EIS frequency characteristics, and the SOH here represents the maximum capacity of the battery and also reflects the health of the battery.

[0106] The mean of the test set is 30.5773, the standard deviation is 4.0661, and the mean of the prediction set is 30.471, the standard deviation is 3.9586, which shows that the data distribution of the prediction results is basically consistent with the data distribution of the real data.

[0107] The root mean square error between the prediction set and the test set is only 0.3652, indicating that the prediction effect of the model is good.

[0108] The prediction results are shown in the following table:

[0109] Table 2 Example of prediction results of SOH using EIS based on CNN-LSTM-Attention model

[0110]

[0111]

[0112] The prediction results are as follows: Figure 4 As shown:

[0113] The solid line above the graph represents the true value, and the dotted line represents the predicted value. The numbers below are the true value, and the numbers above are the predicted value. The dotted line below represents the error value, and the numbers are the error. From the overlap degree of the broken lines and the error value, it can be seen that the prediction effect is good.

[0114] Example 2

[0115] This embodiment discloses a SOH prediction system based on a CNN-LSTM-Attention model, including a memory and a processor, wherein the memory includes a SOH prediction method program based on the CNN-LSTM-Attention model, and when the SOH prediction method program based on the CNN-LSTM-Attention model is executed by the processor, the steps of a SOH prediction method based on a CNN-LSTM-Attention model described in Example 1 are implemented.

[0116] Example 3

[0117] The present embodiment discloses a readable storage medium, which includes a SOH prediction method program based on a CNN-LSTM-Attention model. When the SOH prediction method program based on the CNN-LSTM-Attention model is executed by a processor, the steps of a SOH prediction method based on a CNN-LSTM-Attention model as described in Example 1 are implemented.

[0118] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0119] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0120] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0121] Those skilled in the art can understand that: all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), disks or optical disks, and other media that can store program codes.

[0122] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

Claims

1. A SOH prediction method based on CNN-LSTM-Attention model, characterized in that: The method comprises the following steps: Test and collect battery EIS data and battery impedance capacity data; Build and train the CNN-LSTM-Attention model; The EIS data is input into the CNN-LSTM-Attention model for SOH prediction, and the prediction results are output.

2. According to claim 1, a SOH prediction method based on CNN-LSTM-Attention model is characterized in that: The CNN-LSTM-Attention model includes: an input layer, a CNN model, an LSTM model, an Attention model, and a linear regression layer; The input layer is used to input EIS data into the CNN model; The CNN model extracts EIS spatial characteristics from the principal component data to obtain EIS spatial features, and outputs the EIS spatial features to the LSTM model; The LSTM model extracts the time series change pattern of EIS based on the spatial features of EIS, and then outputs the time series change pattern of EIS to the Attention model; The Attention model extracts key features of EIS spatial features based on the time series change pattern of EIS; The linear regression layer predicts the battery SOH based on the key features in the EIS spatial features.

3. The SOH prediction method based on the CNN-LSTM-Attention model according to claim 2 is characterized in that: The CNN model includes a two-layer CNN structure.

4. The SOH prediction method based on the CNN-LSTM-Attention model according to claim 3 is characterized in that: The CNN structure includes two convolutional layers and one pooling layer. The activation functions of the convolutional layers are all relu functions. The number of filters is set to 32, the kernel size is 1, and the pooling size of the pooling layer is also 1.

5. The SOH prediction method based on the CNN-LSTM-Attention model according to claim 4 is characterized in that: The LSTM model includes an LSTM layer and a fully connected layer. The number of neurons in the LSTM layer is 32, the number of neurons in the fully connected layer is 16, and the activation function is the relu function.

6. The SOH prediction method based on the CNN-LSTM-Attention model according to claim 5 is characterized in that: The Attention model extracts key features of EIS spatial features according to the time series variation law of EIS, specifically: Input the EIS time series into the Attention model and learn an attention weight vector; Apply the attention weight to the hidden state of each time step in the EIS time series and perform weighted summation to obtain the weighted EIS time series; The weighted EIS time series are sorted by importance, and the EIS data with importance greater than the preset importance in the EIS spatial features are extracted as the key features in the EIS spatial features.

7. The SOH prediction method based on the CNN-LSTM-Attention model according to claim 4 is characterized in that: The Attention model is a BahdanauAttention structure; BahdanauAttention is divided into two steps: Encoder and Decoder. The calculation formulas are: (1) Encoder: h i =tanh(W[h i-1 ,x i ]) o i =softmax(Vh i ) where h i Represents input x i The hidden state vector of i represents the normalized output vector of the hidden state at time i, W and V represent the weight matrices to be learned; (2)Decoder: The first step is to generate the semantic vector at time t: Among them, c t represents the semantic vector at time t; e ti is the hidden state vector h at the i-th moment in the encoder i For the hidden state s at the decoder at time t t The degree of influence; α tk It means that e is transformed by the softmax function ti The normalized value, W a represents the weight matrix to be learned, v a represents the weight vector to be learned, and N represents the total time; The second step is to pass the hidden layer information and predict: S t =tanh(W[s t-1 ,y t-1 ,c t ]) o t =softmax(Vs t ) Among them, s t represents the hidden state of the decoder at time t, y t-1 represents t-1 prediction targets, o t represents the prediction distribution.

8. The SOH prediction method based on the CNN-LSTM-Attention model according to claim 7, characterized in that: The activation function of the linear regression layer is a linear function; The loss function of the CNN-LSTM-Attention model is mean square error, the optimizer is rmsprop, and the evaluation metric is mean absolute error.

9. A SOH prediction system based on CNN-LSTM-Attention model, characterized in that: It includes a memory and a processor, wherein the memory includes a SOH prediction method program based on a CNN-LSTM-Attention model, and when the SOH prediction method program based on a CNN-LSTM-Attention model is executed by the processor, the steps of a SOH prediction method based on a CNN-LSTM-Attention model as described in any one of claims 1 to 8 are implemented.

10. A readable storage medium, characterized in that: The readable storage medium includes a SOH prediction method program based on the CNN-LSTM-Attention model. When the SOH prediction method program based on the CNN-LSTM-Attention model is executed by a processor, the steps of a SOH prediction method based on the CNN-LSTM-Attention model as described in any one of claims 1 to 8 are implemented.

Images