VMD-CNN-LSTM-based lithium battery remaining service life prediction method, apparatus and device, and medium

Abstract

The invention belongs to the field of lithium battery service life prediction, and particularly relates to a VMD-CNN-LSTM-based lithium battery remaining service life prediction method and device, equipment and a medium. The method comprises the following steps: extracting discharge average voltage, discharge voltage lowest point time and equal voltage drop discharge time from a battery discharge voltage curve as health factors of a lithium battery, decomposing the health factors into different frequency components and residual errors through VMD decomposition, and inputting the residual errors and each frequency component into a model for independent prediction. And reconstructing the prediction result to obtain a battery capacity prediction value. Compared with an LSTM model and a CNN-LSTM model, the VMD-CNN-LSTM prediction model provided by the invention has the advantages that the prediction error is minimum, and the prediction result is more accurate.

Description

Technical Field

[0001] The present invention belongs to the field of lithium battery service life prediction, and in particular relates to a method, device, equipment and medium for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM. Background Art

[0002] With the rapid development of renewable energy and the growth of the electric vehicle market, the demand for high-performance and high-safety lithium batteries continues to increase. By predicting the life of lithium batteries, the service life of lithium batteries can be effectively extended, the frequency of lithium battery replacement can be reduced, energy efficiency can be improved, and the operating cost of energy systems can be reduced. Accurate prediction of lithium battery life can help to timely detect signs of battery aging and performance degradation, reduce the risk of sudden failure of lithium batteries, and thus improve the safety of lithium battery systems. By predicting the life of lithium batteries, the service life of lithium batteries can be effectively extended, the number of waste lithium batteries can be reduced, and it is beneficial to environmental protection and resource recycling. Therefore, predicting the service life of lithium batteries can help optimize the design and manufacture of lithium batteries to meet the growing demand for energy storage. In general, the research background of lithium battery service life prediction mainly comes from the demand for clean energy and high-performance energy storage systems, and the research significance lies in improving energy efficiency, promoting lithium battery safety, environmental protection and resource conservation, and promoting the development of battery management technology.

[0003] The life of lithium batteries is affected by many factors, including the number of charge and discharge cycles, temperature, current charge and discharge rate, chemical movement, etc. These factors interact with each other, making prediction complicated. The attenuation of lithium battery life usually shows nonlinear characteristics. As the use time increases, the performance of lithium batteries may decline rapidly. This nonlinear relationship increases the difficulty of life prediction. It is a challenge to obtain a large amount of lithium battery life data in real environments, and it is also necessary to overcome many difficulties to establish an accurate mathematical model to describe the lithium battery life attenuation process.

[0004] Regarding the prediction of RUL (remaining useful life) of lithium batteries, domestic and foreign scholars have mainly carried out research work from the following three aspects: model-driven methods, data-driven methods, and methods based on the fusion of the two (model-driven, data-driven). Among them, the model-driven method is divided into electrochemical model, equivalent circuit model and empirical model-driven method, etc. This method requires researchers to have rich development experience. The data-driven method has benefited from the continuous progress of artificial intelligence (AI) and machine learning in recent years. Combined with the large amount of data generated in the production and manufacturing of lithium batteries and the experimental process, it meets the requirements of machine learning methods for large data volume. The method based on the fusion of the two needs to give full play to the advantages of the two driving methods to improve the accuracy of the machine learning prediction model, which has become the main research direction of domestic and foreign scholars at this stage. Due to the lack of data research on the input model, the life of lithium batteries gradually decays with the increasing number of charge and discharge cycles, and there are safety hazards such as battery short circuit and explosion. The existing prediction method is affected by the noise of lithium battery experimental data and the defects of the single algorithm model itself, and there will be phenomena such as gradient disappearance and gradient explosion, resulting in low accuracy of the prediction method and inaccurate prediction results. Therefore, it is urgent to study and develop a prediction method based on deep learning to improve the accuracy of the prediction of the remaining useful life of lithium batteries. Summary of the invention

[0005] In order to improve the accuracy of remaining service life prediction of lithium batteries, the present invention provides a method, device, equipment and medium for predicting remaining service life of lithium batteries based on VMD-CNN-LSTM.

[0006] The present invention is achieved through the following technical solutions:

[0007] In a first aspect, a method for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM is provided, comprising the following steps:

[0008] S101: Using NASA's lithium battery dataset, divide the dataset into a test set and a training set;

[0009] S102: extracting the average discharge voltage, the lowest discharge voltage time and the equal voltage drop discharge time of the lithium battery in the training set during the discharge process as the health factor of the lithium battery;

[0010] S103: decomposing the original data of lithium battery discharge in the training set into modal components of different frequencies by using a VMD decomposition method, and determining the residuals therein;

[0011] S104: Use convolutional neural network (CNN) to extract features from raw data of lithium battery discharge, and combine it with long short-term memory network (LSTM) to solve the long dependency problem of time series data;

[0012] S105: Construct VMD-CNN-LSTM model to predict lithium battery capacity;

[0013] S106: Input the test set into the VMD-CNN-LSTM model to further obtain the predicted capacity of the lithium battery;

[0014] S107: Calculate the remaining service life of the lithium battery based on the predicted capacity of the lithium battery and the failure threshold of the lithium battery.

[0015] As a further improvement of the technical solution of the method of the present invention, in step S101, the NASA lithium battery data set is a data set of 18650 lithium batteries disclosed by the NASA PCoE laboratory, and the data set includes four lithium batteries B0005, B0006, B0007, and B0018.

[0016] As a further improvement of the technical solution of the method of the present invention, in step S101, before the data set is divided into a test set and a training set, the original data of the NASA lithium battery data set is preprocessed, and the preprocessing includes: removing outliers in the original data and standardizing the original data.

[0017] As a further improvement of the technical solution of the method of the present invention, step S103 specifically includes: obtaining the single-sided spectrum of each modal component through Hibert transform, then adding an adjustment item to modulate the baseband bandwidth of each spectrum, and finally processing the demodulated signal to calculate the bandwidth of each modal component.

[0018] As a further improvement of the technical solution of the method of the present invention, in step S104, 1D-CNN in CNN is used as a feature extraction module, and the convolution kernel of CNN is applied to extract features of the original data of lithium battery discharge, and the obtained data is used as the input of LSTM through pooling and activation operations.

[0019] As a further improvement of the technical solution of the method of the present invention, in step S106, the test set is converted into a matrix of the dimension required by the VMD-CNN-LSTM model, the potential relationship between the test set parameters is extracted as a feature in the form of convolution, the obtained feature is input into the pooling layer, the result obtained by the pooling layer is input into the LSTM, the cell state is updated through the forget gate, the input gate, and the output gate, and a prediction is made based on the data relationship between the previous moment and the subsequent moment, and the predicted capacity of the lithium battery is output through the fully connected layer.

[0020] In the second aspect, a device for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM is provided, including a data preprocessing module, a data decomposition module, a model training module, and a data reconstruction module;

[0021] Data preprocessing module: collects the raw data of lithium battery discharge from NASA, preprocesses the raw data of lithium battery discharge, including outlier removal and data standardization, and uses the preprocessed raw data of lithium battery discharge as a data set, and divides the data set into a training set and a test set;

[0022] Data decomposition module: decompose the original data of lithium battery discharge in the training set into modal components of different frequencies through VMD decomposition method, and determine the residuals therein;

[0023] Model training module: construct a VMD-CNN-LSTM model, use the training set in the data set to train the VMD-CNN-LSTM model, and input the test set into the trained VMD-CNN-LSTM model to obtain the predicted capacity of the lithium battery;

[0024] Data reconstruction module: Calculate the predicted capacity of the lithium battery and the failure threshold of the lithium battery to obtain the remaining service life of the lithium battery.

[0025] In a third aspect, a computer device is provided, comprising a memory, a processor and a transceiver which are communicatively connected in sequence, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the above-mentioned method for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM.

[0026] In a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores instructions.

[0027] When the instructions are executed on a computer, the above-mentioned method for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM is executed.

[0028] The present invention provides a method, device, equipment and medium for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM, which has the following advantages over the prior art:

[0029] (1) The method of the present invention improves the RUL prediction accuracy of lithium batteries: the remaining service life of lithium batteries is predicted by adopting the VMD-CNN-LSTM model. Compared with the traditional LSTM and CNN-LSTM models, the present invention can provide more accurate prediction results. The method can better capture the evolution law of lithium battery performance, achieve more accurate estimation of the remaining service life of lithium batteries, and provide accurate decision-making basis for lithium battery management systems.

[0030] (2) The present invention improves the operation level of lithium battery management system: accurate prediction of the remaining service life of lithium batteries is crucial to the normal operation of lithium battery management system. The method provided by the present invention can accurately predict the remaining service life of lithium batteries and provide accurate remaining available time for lithium battery management system, which will optimize the charging and discharging strategy of lithium batteries, extend battery life, improve system reliability, and achieve more effective energy management in the fields of electric vehicles, renewable energy, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0033] Figure 1 This is a flow chart of the remaining service life prediction method of lithium batteries based on VMD-CNN-LSTM.

[0034] Figure 2 This is the basic unit structure diagram of the CNN network.

[0035] Figure 3 This is the basic unit structure diagram of the LSTM network.

[0036] Figure 4 This is the prediction result chart of B0005.

[0037] Figure 5 This is the prediction result chart of B0006.

[0038] Figure 6 This is the prediction result chart of B0007.

[0039] Figure 7 This is the prediction result chart of B0018. DETAILED DESCRIPTION

[0040] In order to more clearly understand the above-mentioned objectives, features and advantages of the present invention, the scheme of the present invention will be further described below. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0041] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present invention, rather than all of the embodiments.

[0042] The specific embodiments of the present invention are described in detail below.

[0043] The present invention provides a specific embodiment of a method for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM, comprising the following steps:

[0044] S101: Using NASA's lithium battery dataset, divide the dataset into a test set and a training set;

[0045] S102: extracting the average discharge voltage, the lowest discharge voltage time and the equal voltage drop discharge time of the lithium battery in the training set during the discharge process as the health factor of the lithium battery;

[0046] S103: decomposing the original data of lithium battery discharge in the training set into modal components of different frequencies by using a VMD decomposition method, and determining the residuals therein;

[0047] S104: Use convolutional neural network (CNN) to extract features from raw data of lithium battery discharge, and combine it with long short-term memory network (LSTM) to solve the long dependency problem of time series data;

[0048] S105: Construct VMD-CNN-LSTM model to predict lithium battery capacity;

[0049] S106: Input the test set into the VMD-CNN-LSTM model to further obtain the predicted capacity of the lithium battery;

[0050] S107: Calculate the remaining service life of the lithium battery based on the predicted capacity of the lithium battery and the failure threshold of the lithium battery.

[0051] Step S101 specifically includes the following steps:

[0052] Step 1.1: Use the 18650 lithium battery dataset with a rated capacity of 2Ah published by NASA to extract valid information such as voltage, current, and temperature during all charge and discharge cycles of B0005, B0006, B0007, and B0018 lithium batteries.

[0053] Step 1.2, data preprocessing, includes the following steps:

[0054] a. Remove outliers: Draw a curve of the raw data of lithium battery discharge and identify outliers by drawing the curve. By observing the distribution of data points on the curve, data points that deviate far from the normal trend are considered outliers. Outliers usually represent erroneous data and will affect the accuracy of subsequent analysis. Therefore, outliers need to be removed to ensure the accuracy and reliability of the data.

[0055] b. Data standardization: The Z-score method is used to transform the original data without outliers so that it falls within the interval of [0,1].

[0056] c. Training set and test set division: The dataset is divided into training set and test set in a ratio of 7:3. The training set is used to train the subsequent VMD-CNN-LSTM model. After the model training is completed, the test set is used to evaluate the model performance.

[0057] Step S102: extract the average discharge voltage, the time of the lowest discharge voltage, and the discharge time of equal voltage drop in the lithium battery of the training set during the discharge process as the health factor of the lithium battery. The specific definition of each health factor is as follows:

[0058] The average discharge voltage is defined as the average value of the discharge voltage of the lithium battery during different charge and discharge cycles. The calculation method is:

[0059]

[0060] Where: Indicates the charge and discharge cycle of lithium battery; For lithium battery The average value of discharge voltage during the charge and discharge cycle; Indicates the number of discharge voltage time sampling points in a charge and discharge cycle of a lithium battery; For lithium battery Charge and discharge cycles, time The discharge voltage value of lithium battery.

[0061] The time of the lowest discharge voltage point is defined as the time corresponding to when the lithium battery voltage drops to the lowest point during the discharge process.

[0062] The constant voltage drop discharge time is defined as the time required for the lithium battery voltage to drop from 3.7V to 3.5V during the discharge process. The calculation method is:

[0063]

[0064] Where: For lithium battery Discharge time with constant voltage drop during the first charge and discharge cycle; For lithium battery The moment when the discharge voltage of the first charge-discharge cycle is 3.7V; For lithium battery The moment when the discharge voltage of the first charge and discharge cycle is 3.5V.

[0065] During the discharge process of lithium batteries, the battery capacity shows an overall attenuation trend but there are local fluctuations. This is due to the influence of noise and measurement errors, coupled with the capacity re-increase effect caused by the complex side reactions inside the lithium battery. If the lithium battery life is predicted directly, the error will be large. Therefore, the original data of lithium battery discharge in the training set is decomposed into different frequency modal components through the VMD decomposition method. Each modal component is independently input into the model prediction to effectively reduce the influence of noise and errors and improve the accuracy of the prediction. Specifically, the following steps are included:

[0066] The VMD decomposition process is essentially a process of solving the variational problem. The purpose is to minimize the sum of the bandwidths of all modal components. First, the single-sided spectrum of each modal component is obtained through Hibert transform, and then the adjustment term is added to modulate the baseband bandwidth of each spectrum. Finally, the demodulated signal is processed to calculate the bandwidth of each modal component. The constrained variational model derived from this is:

[0067] { min { u k } , { w k } { ∑ k = 1 K || ∂ t [( δ ( t ) + j π t )* u k ( t )] e − j ω k t || 2 2 } s . t . ∑ k u k = f

[0068] Where: is the formula operator, the purpose is to minimize the sum of the bandwidths of each modal component and at the same time satisfy The following formula (constraint condition) means that the sum of the modal component bandwidths is equal to the original signal ; is the sampling time; is the number of modal components; and They represent the first modal components and corresponding center frequencies; represents the convolution operator; is the Dirac function; is the gradient operator; represents an imaginary unit; Indicates that the spectrum is shifted to the baseband. It is the abbreviation of "subject to", which means "restricted to" in Chinese. The following formula is an independent condition.

[0069] Solve the above equation and introduce the quadratic penalty term and Lagrange multipliers , the solution formula of the mode is:

[0070] L [ u k , w k , λ ] = α ∑ k || ∂ t [ δ ( t ) + j π t * u k ( t ) ] e − j w k t | | 2 2 + | | f ( t ) − ∑ k u k ( t ) | | 2 2 + < λ ( t ) , f ( t ) − ∑ k u k ( t ) >

[0071] Optimization based on alternating direction multiplier iteration algorithm and :

[0072]

[0073]

[0074] Where: For the Iteration No. modal components; For the modal components; is the center frequency; For the Iteration No. The center frequency of a component.

[0075] The above two formulas represent the search for the optimal solution of modal decomposition and center frequency. Its essence is an iterative process, and the optimal solution is found through continuous iteration. The modal component with the lowest center frequency among all modal components is the residual.

[0076] Step S104 specifically includes the following steps:

[0077] like Figure 2 As shown in the figure, 1D-CNN is used as the feature extraction module, and the convolution kernel of the convolutional neural network (CNN) is used to extract features from the raw data of lithium battery discharge, and the obtained data is used as the input of the subsequent long short-term memory network (LSTM) through pooling and activation operations. Specifically:

[0078] Step 4.1, the convolution layer performs convolution operation on the input data (raw data of lithium battery discharge) and the convolution kernel to extract the potential features of the input data. The fixed-size convolution kernel senses and scans the entire data domain like the human eye. Multiple convolution kernels with different weights evaluate and extract different features of various aspects of the input data through convolution operation. The specific operation of the convolution layer is as follows:

[0079]

[0080] Where: For the Tier convolution kernel weight matrix; For the Layer output; For the The output of the layer Features is the bias term.

[0081] In this embodiment, the rectified linear unit (ReLU) is selected as the activation function of the convolutional layer, and its expression is:

[0082]

[0083] Where: Represents the domain, which is equivalent to , which is usually used in the ReLU activation function express.

[0084] The high-dimensional image data in image processing contains a larger amount of information, and a multi-dimensional convolutional neural network is preferably used. However, the raw data of lithium battery discharge is one-dimensional time series data, so this embodiment uses 1D-CNN to extract features from the raw data of lithium battery discharge.

[0085] Step 4.2: After the original data of lithium battery discharge is extracted by the convolution layer, multiple feature matrices are obtained according to different convolution kernels. In order to extract enough hidden information, the dimension of the output feature of the convolution layer is generally large. The function of the pooling layer is to downsample the input features, filter and select many features at the same time, and strengthen some significant features, which is equivalent to the effective information filtering of human vision for observing things. The operator of the pooling layer is called the pooling kernel, which scans the feature matrix output by the convolution layer by traversal. The specific formula is as follows:

[0086]

[0087] Where: For the Tier The feature matrix elements, Indicates Tier The maximum value of the elements in the feature matrix, After pooling Tier The feature matrix elements; for Pooling kernels.

[0088] Step S105 specifically includes the following steps:

[0089] Step 5.1: Figure 3 In a single LSTM neuron, there are three gate structures: forget gate, input gate and output gate. The forget gate determines which information is discarded from the previous cell state. The input gate determines which new input information is used to update the previous cell state and keep it in the current cell state. The output gate determines whether the current cell state is output to the hidden layer.

[0090] use As step length The input at Indicates step length The cell state, Indicates step length The output and step length of the hidden layer The input of the hidden layer, Indicates step length The cell state, Indicates step length The output of the hidden layer at Indicates step length The information entering the memory cell, Indicates step length Candidate values ​​for memory information.

[0091] set up , , and They represent different deviations. , , , , , They represent the weights of different matrices respectively. Represents the sigmoid function, which is the same as the tanh function, both of which are activation functions.

[0092] Step 5.2, the output value of the forget gate is , is a value greater than 0 and less than 1, and is defined as follows:

[0093]

[0094] The input gate consists of two parts. The first part outputs how much information is updated, and the second part outputs the candidate cell state at the current moment. The two outputs are expressed as follows:

[0095]

[0096]

[0097] Combine the forget gate and the input gate to get the current cell state , which is expressed as follows:

[0098]

[0099] Output gate, combined and , activated by the sigmoid function, we get , which is expressed as follows:

[0100]

[0101] Finally, the step length The output of the hidden layer , which is expressed as follows:

[0102]

[0103] Steps S106 and S107 specifically include the following steps:

[0104] Step 6.1: Convert the test set into a matrix of the dimension required by the network and input it into the VMD-CNN-LSTM model.

[0105] Step 6.2: Extract the potential relationship between parameters as features in the form of convolution.

[0106] Step 6.3: Input the obtained features into the pooling layer, and reduce the model parameters and enhance the features through pooling calculation.

[0107] Step 6.4: Input the results obtained from the pooling layer into the long short-term memory network, update the cell state through the forget gate, input gate, and output gate, and make predictions based on the data relationship between the previous moment and the subsequent moment, and output the predicted capacity of the lithium battery through the fully connected layer.

[0108] Step 6.5: Calculate the remaining service life of the lithium battery by combining the current predicted capacity with the lithium battery failure threshold.

[0109] In order to further evaluate the accuracy of the prediction results of the remaining service life of lithium batteries by the VMD-CNN-LSTM model constructed by the present invention, the test set was input into the LSTM model, CNN-LSTM model and VMD-CNN-LSTM model respectively to compare the prediction results. Figure 4 , Figure 5 , Figure 6 and Figure 7 As shown in the result graph, the prediction curves of the LSTM model and the CNN-LSTM model are always higher than the true value curves among the three models, and the errors are obvious. The prediction result curve of the VMD-CNN-LSTM model has the highest consistency with the actual data curve.

[0110] The embodiment of the present invention uses the root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) to calculate the error between the predicted value and the true value. The specific steps are as follows:

[0111] Step 7.1, calculate the root mean square error. The root mean square error is the deviation between the predicted value and the true value, which can be used as an accuracy assessment of the comprehensive effect.

[0112]

[0113] Step 7.2, calculate the mean absolute error. The mean absolute error is the average of the absolute values ​​of the deviations between the predicted value and the true value, which can better reflect the actual situation of the error.

[0114]

[0115] Step 7.3, calculate the mean absolute percentage error. The mean absolute percentage error is the average of the absolute percentage of the deviation between the predicted value and the true value, which can reflect the size of the relative error of the prediction.

[0116]

[0117] In the above three formulas, Indicates the charge and discharge cycle of lithium batteries. is the actual service life of lithium battery; is the predicted remaining service life of the lithium battery; is the total number of charge and discharge cycles.

[0118] Table 1 is the RMSE, MAE, and MAPE results of the three models. It can be seen from the table that in the data prediction of the three models, the LSTM model has the largest RMSE, MAE, and MAPE values, followed by the CNN-LSTM model, and the VMD-CNN-LSTM model has the smallest value. This shows that the prediction performance has been significantly improved through VMD decomposition.

[0119] Table 1 RMSE, MAE, and MAPE of the three models

[0120] This embodiment further provides a device for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM, including a data preprocessing module, a data decomposition module, a model training module, and a data reconstruction module; Data preprocessing module: collects the raw data of lithium battery discharge from NASA, preprocesses the raw data of lithium battery discharge, including outlier removal and data standardization, and uses the preprocessed data as a data set, and divides the data set into a training set and a test set; Data decomposition module: decompose the original data of lithium battery discharge in the training set into modal components of different frequencies through VMD decomposition method, and determine the residuals therein; Model training module: construct a VMD-CNN-LSTM model, use the training set in the data set to train the VMD-CNN-LSTM model, and input the test set into the trained VMD-CNN-LSTM model to obtain the predicted capacity of the lithium battery; Data reconstruction module: Calculate the predicted capacity of the lithium battery and the failure threshold of the lithium battery to obtain the remaining service life of the lithium battery.

[0121] Furthermore, this embodiment provides a computer device, comprising a memory, a processor and a transceiver that are communicatively connected in sequence, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute a method for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM as described above.

[0122] Furthermore, this embodiment provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions. When the instructions are executed on a computer, the method for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM as described above is executed.

[0123] The above is only a specific implementation of the present invention, which enables those skilled in the art to understand or implement the present invention. Although detailed descriptions are given with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments, and they should all be covered by the protection scope of the claims.

Claims

1. A method for predicting the remaining service life of lithium batteries based on VMD-CNN-LSTM, characterized in that: The following steps are involved: S101: Using NASA's lithium battery dataset, divide the dataset into a test set and a training set; S102: extracting the average discharge voltage, the time of the lowest discharge voltage, and the equal voltage drop discharge time during the discharge process of the lithium battery in the training set as health factors of the lithium battery; S103: decomposing the original data of lithium battery discharge in the training set into modal components of different frequencies by using a VMD decomposition method, and determining the residuals therein; S104: Use CNN to extract features from the raw data of lithium battery discharge, and combine it with LSTM to solve the long dependency problem of time series data; S105: Construct VMD-CNN-LSTM model to predict lithium battery capacity; S106: Input the test set into the VMD-CNN-LSTM model to further obtain the predicted capacity of the lithium battery; S107: Calculate the remaining service life of the lithium battery based on the predicted capacity of the lithium battery and the failure threshold of the lithium battery.

2. According to claim 1, a method for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM is characterized in that: In step S101, the NASA lithium battery data set is a data set of 18650 lithium batteries disclosed by the NASA PCoE laboratory, and the data set includes four types of lithium batteries: B0005, B0006, B0007, and B0018.

3. A method for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM according to claim 2, characterized in that: In step S101, before the data set is divided into a test set and a training set, the original data of the NASA lithium battery data set is preprocessed, and the preprocessing includes: removing outliers in the original data and standardizing the original data.

4. The method for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM according to claim 1, characterized in that: Step S103 specifically includes: obtaining a single-sided spectrum of each modal component through Hibert transform, then adding an adjustment item to modulate the baseband bandwidth of each spectrum, and finally processing the demodulated signal to calculate the bandwidth of each modal component.

5. The method for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM according to claim 1, characterized in that: In step S104, 1D-CNN in CNN is used as a feature extraction module, and the convolution kernel of CNN is applied to extract features from the original data of lithium battery discharge, and the obtained data is used as the input of LSTM through pooling and activation operations.

6. The method for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM according to claim 1, characterized in that: In step S106, the test set is converted into a matrix of the required dimension of the VMD-CNN-LSTM model, the potential relationship between the test set parameters is extracted as a feature in the form of convolution, the obtained features are input into the pooling layer, the results obtained from the pooling layer are input into the LSTM, the cell state is updated through the forget gate, input gate, and output gate, and prediction is made based on the data relationship between the previous moment and the subsequent moment, and the predicted capacity of the lithium battery is output through the fully connected layer.

7. A device for predicting the remaining service life of lithium batteries based on VMD-CNN-LSTM, characterized in that: It includes data preprocessing module, data decomposition module, model training module and data reconstruction module; Data preprocessing module: collects the raw data of lithium battery discharge from NASA, preprocesses the raw data of lithium battery discharge, including outlier removal and data standardization, and uses the preprocessed raw data of lithium battery discharge as a data set, and divides the data set into a training set and a test set; Data decomposition module: decompose the preprocessed raw data of lithium battery discharge in the training set into modal components of different frequencies through VMD decomposition method, and determine the residuals therein; Model training module: construct a VMD-CNN-LSTM model, use the training set in the data set to train the VMD-CNN-LSTM model, and input the test set into the trained VMD-CNN-LSTM model to obtain the predicted capacity of the lithium battery; Data reconstruction module: Calculate the predicted capacity of the lithium battery and the failure threshold of the lithium battery to obtain the remaining service life of the lithium battery.

8. A computer device, characterized in that: It comprises a memory, a processor and a transceiver which are communicatively connected in sequence, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute a method for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer readable storage medium has instructions stored thereon. When the instruction is executed on a computer, a method for predicting the remaining service life of a lithium battery based on VMD-CNN-LSTM as described in any one of claims 1 to 6 is executed.

Images