Early prediction method and system for cycle life of lithium ion battery based on NSGAII-CNN-LSTM

Abstract

The invention discloses a lithium ion battery cycle life early prediction method and system based on NSGAII-CNN-LSTM, and belongs to the field of lithium ion battery fault prediction and health management. The method comprises the following steps: acquiring working data of a to-be-tested lithium ion battery; inputting the obtained working data of the lithium ion battery to be tested into the trained NSGAII-CNN-LSTM model, and outputting an early prediction result of the cycle life of the lithium ion battery; wherein the training of the NSGAII-CNN-LSTM model comprises the following steps: determining high-correlation working data of the cycle life of the lithium ion battery, and constructing a data set of the working data; and establishing an NSGAII-CNN-LSTM model, and training the established NSGAII-CNN-LSTM model by using the established data set, so as to obtain a trained NSGAII-CNN-LSTM model. According to the method, the NSGAII, the CNN and the LSTM are fused, the NSGAII-CNN-LSTM model is proposed to process the early cycle life prediction task of the lithium ion battery, experimental evaluation shows that the performance on different tasks and data sets is superior to that of CNN and LSTM neural networks, and the method can be used for quickly verifying the service life of a new material of the lithium ion battery and quickly evaluating the advantages and disadvantages of a new manufacturing process of the lithium ion battery.

Description

Technical Field

[0001] The present invention belongs to the field of lithium-ion battery fault prediction and health management, and specifically relates to an NSGAⅡ-CNN-LSTM-based lithium-ion battery cycle life early prediction method and system. Background Art

[0002] Lithium batteries have the advantages of high output voltage, high energy density, low self-discharge rate, long cycle life and high reliability. They are widely used in power storage systems such as hydropower, thermal power, wind power and solar power stations, as well as important fields such as transportation, military equipment, aerospace, etc. If a battery failure occurs during use, it is likely to cause the performance of the corresponding power equipment or system to degrade or fail, thereby increasing costs and even causing accidents such as fire and explosion. If the cycle life of lithium batteries can be accurately predicted early in the life cycle of lithium batteries and used to guide the use of lithium batteries, such accidents will be eliminated. In addition, the early prediction of the cycle life of lithium batteries can be used to verify the shortening of life verification time and speed up the development cycle during the research and development of battery materials. It can also quickly verify the new manufacturing process of batteries, and can also classify / grade new batteries according to their expected life. Therefore, accurate early prediction of the cycle life of batteries has great practical significance.

[0003] There are three typical methods for predicting the cycle life of lithium-ion batteries: the first is an experience-based method, the second is a model-based method, and the third is a data-driven method. The experience-based method is to complete the battery cycle life prediction through the rules of historical data. The experience-based method is simple and fast, but it is only a rough estimate of the cycle life and does not meet the current accuracy requirements. The model-based method completes the cycle life prediction by analyzing the operating environment and internal material properties of the battery, but it takes a lot of time and is cumbersome to establish an accurate model. The data-driven method is based on statistical analysis and realizes cycle life prediction by mining the intrinsic relationship between input and response output. The data-driven method does not require specific knowledge of material properties, structure or failure mechanism. Although it is more convenient in modeling, the specific aging mechanism of the battery cannot be clearly expressed. In recent years, deep neural networks, as a type of machine learning algorithm, have been widely used for their powerful function mapping capabilities when facing highly nonlinear and complex multidimensional systems. Especially in time series processing problems, recurrent neural networks, long short-term memory networks, gated unit networks, etc. are widely used due to their powerful feature mining capabilities. However, recurrent neural networks and their variants have limited ability to extract data features, and their hyperparameters cannot be adaptively adjusted according to application scenarios, making it difficult for them to achieve ideal prediction accuracy.

[0004] Therefore, it is necessary to design and develop an early prediction method and system for the cycle life of lithium-ion batteries based on NSGAⅡ-CNN-LSTM for the above-mentioned problems. Summary of the invention

[0005] The purpose of the present invention is to provide an early prediction method for the cycle life of a lithium-ion battery based on NSGAⅡ-CNN-LSTM in view of the problems existing in the prior art. By integrating a non-dominated sorting genetic algorithm (NSGAⅡ), a convolutional neural network (CNN) and a long short-term memory network (LSTM), an NSGAⅡ-CNN-LSTM model is proposed to handle the task of early cycle life prediction of lithium-ion batteries. Not only is the prediction accuracy significantly improved, but it also helps to quickly verify the service life of new battery materials and quickly evaluate the advantages and disadvantages of new battery manufacturing processes.

[0006] An early prediction method for lithium-ion battery cycle life based on NSGAⅡ-CNN-LSTM, comprising:

[0007] Obtain the working data of the lithium-ion battery to be tested;

[0008] The obtained working data of the lithium-ion battery to be tested is input into the trained NSGAⅡ-CNN-LSTM model to output the early prediction result of the cycle life of the lithium-ion battery; wherein the training of the NSGAⅡ-CNN-LSTM model includes:

[0009] Identify the operating data with high relevance to the cycle life of lithium-ion batteries and construct a dataset of the operating data;

[0010] Build an NSGAⅡ-CNN-LSTM model, including a 3D CNN module and an LSTM module. The 3D CNN module and the LSTM module are connected through a tile layer. The 3D CNN module is used to extract features from input working data, and the LSTM module is used to process the data after feature extraction and enhance the feature representation of the data.

[0011] The constructed data set is used to train the constructed NSGAⅡ-CNN-LSTM model to obtain the trained NSGAⅡ-CNN-LSTM model.

[0012] Furthermore, the working data with high relevance to the cycle life of lithium-ion batteries are determined, including:

[0013] Acquire various working data of the lithium-ion battery, and respectively calculate the level difference and the square of the level difference between each working data and the cycle life of the lithium-ion battery;

[0014] The Spearman correlation coefficient formula is calculated based on the calculated rank difference and rank difference square, and the value result between [-1, 1] is obtained. The value close to 1 indicates a strong positive correlation between the two variables, and the value close to -1 indicates a strong negative correlation.

[0015] Based on the calculation results of the Spearman correlation coefficient of each working data, the working data with high correlation to the cycle life of the lithium-ion battery is determined.

[0016] Furthermore, the 3D CNN module includes:

[0017] 3D convolutional layer, which slides over the input data and performs convolution operations;

[0018] Normalization layer, used to normalize the input of each layer;

[0019] ReLU layer, used to introduce nonlinear factors and alleviate the gradient vanishing problem;

[0020] The 3D pooling layer is used to downsample the feature map output by the convolutional layer.

[0021] Furthermore, the LSTM module includes:

[0022] LSTM layer, used to selectively retain or forget information;

[0023] A self-attention mechanism that computes a relevance score between each position in the sequence and every other position;

[0024] Fully connected layer, used for features extracted by LSTM layer and self-attention mechanism layer.

[0025] Furthermore, the training of the NSGAⅡ-CNN-LSTM model also includes:

[0026] Based on the early prediction results of lithium-ion battery cycle life, the root mean square error, mean absolute percentage error and mean absolute error are calculated, and the hyperparameters in the 3D CNN module are optimized using a non-dominated sorting genetic algorithm based on the error data.

[0027] Furthermore, the 10-fold cross-validation method is used to optimize the number of layers of the 3D CNN module.

[0028] An early prediction system for lithium-ion battery cycle life based on NSGAⅡ-CNN-LSTM, comprising:

[0029] A data acquisition module is used to acquire the working data of the lithium-ion battery to be tested;

[0030] The battery life prediction module inputs the obtained working data of the lithium-ion battery to be tested into the trained NSGAⅡ-CNN-LSTM model, and outputs the early prediction result of the cycle life of the lithium-ion battery; wherein the training of the NSGAⅡ-CNN-LSTM model includes:

[0031] Identify the operating data with high relevance to the cycle life of lithium-ion batteries and construct a dataset of the operating data;

[0032] Build an NSGAⅡ-CNN-LSTM model, including a 3D CNN module and an LSTM module. The 3D CNN module and the LSTM module are connected through a tile layer. The 3D CNN module is used to extract features from input working data, and the LSTM module is used to process the data after feature extraction and enhance the feature representation of the data.

[0033] The constructed data set is used to train the constructed NSGAⅡ-CNN-LSTM model to obtain the trained NSGAⅡ-CNN-LSTM model.

[0034] According to one aspect of the present invention, there is provided an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the target detection method based on infrared and visible light image fusion when executing the computer program.

[0035] According to one aspect of the present invention, a computer-readable storage medium is provided, which stores a computer program, and when the computer program is executed by a processor, the steps of the target detection method based on infrared and visible light image fusion are implemented.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. The present invention proposes an early prediction method and system for the cycle life of lithium-ion batteries based on NSGAⅡ-CNN-LSTM. By establishing an NSGAⅡ-CNN-LSTM model to predict the cycle life of lithium-ion batteries, the service life of new battery materials can be quickly verified during the development of new battery materials, and the advantages and disadvantages of new battery manufacturing processes can be quickly evaluated.

[0038] 2. The present invention proposes an early prediction method and system for the cycle life of lithium-ion batteries based on NSGAⅡ-CNN-LSTM, which uses a non-dominated sorting genetic algorithm to optimize hyperparameters and a 10-fold cross-validation method to optimize the number of layers of the 3D CNN module, thereby improving the prediction accuracy of the NSGAⅡ-CNN-LSTM model, thereby improving the classification of newly manufactured batteries, helping to guide end users to use batteries reasonably, and preventing safety accidents such as battery combustion and explosion. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] 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, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 A schematic diagram of the NSGAⅡ-CNN-LSTM model structure provided in an embodiment of the present invention;

[0041] Figure 2 A schematic diagram of correlation analysis between early working data of a lithium-ion battery and battery cycle life according to an embodiment of the present invention;

[0042] Figure 3 A schematic diagram of the NSGAⅡ-CNN-LSTM model training process provided in an embodiment of the present invention;

[0043] Figure 4 A schematic diagram of the prediction results of the NSGAⅡ-CNN-LSTM model provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] The present invention provides a structural schematic diagram of an early prediction method for the cycle life of a lithium ion battery based on NSGAⅡ-CNN-LSTM, comprising: selecting highly correlated working data of a lithium ion battery as a data set, and dividing the data into a training data set, a test data set and a verification data set; dividing the training data into vectors and scalars, inputting the training data into a 3D CNN module for convolution pooling processing, and inputting the training data into an LSTM module for sequence processing, and connecting the 3D CNN module and the LSTM module through a tiled layer; and finally predicting the cycle life of the lithium ion battery based on the verification data set and the NSGAⅡ-CNN-LSTM model to obtain a prediction result.

[0046] Specifically, the lithium-ion battery working data includes charging capacity (Qc), charging voltage (CV), charging temperature (Tc), discharging temperature (Td), charging time (CT), discharging time (DT) and initial capacity (IC). The Spearman correlation coefficient is used to perform correlation analysis on these lithium-ion battery working data. The results are as follows: Figure 2 The specific steps are as follows:

[0047] Acquire various working data of the lithium-ion battery, and respectively calculate the level difference and the square of the level difference between each working data and the cycle life of the lithium-ion battery;

[0048] The Spearman correlation coefficient formula is calculated based on the calculated rank difference and rank difference square, and the value result between [-1, 1] is obtained. The value close to 1 indicates a strong positive correlation between the two variables, and the value close to -1 indicates a strong negative correlation. The Spearman correlation coefficient calculation formula is as follows:

[0049]

[0050] Among them, x represents the working data of lithium-ion battery, y represents the cycle life of lithium-ion battery, and i is the number of samples;

[0051] Based on the calculation results of the Spearman correlation coefficient of each working data, the working data with high correlation to the cycle life of the lithium-ion battery is determined.

[0052] Specifically, lithium-ion battery working data is selected as a data set and divided into a training data set, a test data set and a verification data set. In an embodiment of the present invention, the three are divided according to 70%:15%:15%, and the training data set is input into the input layer, wherein the discharge voltage, discharge capacity and dQ / dV data are constructed into vector data, and the cycle life data is constructed into scalar data.

[0053] Specifically, the 3D CNN module includes: a 3D convolution layer for sliding on the input data and performing convolution operations; a normalization layer for normalizing the input of each layer; a ReLU layer for introducing nonlinear factors to alleviate the gradient vanishing problem; and a 3D pooling layer for downsampling the feature map output by the convolution layer. The LSTM module includes: an LSTM layer for selectively retaining or forgetting information; a self-attention mechanism for calculating the correlation score between each position in the sequence and other positions; and a fully connected layer for features extracted by the LSTM layer and the self-attention mechanism layer. In addition, the tiling layer is used to integrate the multi-dimensional feature tensors output from the 3DCNN module, reduce the relevant high-dimensional data to one-dimensional vector data, and input it into the LSTM module.

[0054] Specifically, based on the early prediction results of lithium-ion battery cycle life, the root mean square error (RMSE), mean absolute percentage error (MAPE) and mean absolute error (MAE) are calculated, and the non-dominated sorting genetic algorithm is used to optimize the hyperparameters in the 3D CNN module, where the hyperparameters include filter size (FS), number of filters (NF), batch size (BS), learning rate (LR), and L2 regularization coefficient (L2R); in addition, the 10-fold cross-validation method is used to optimize the number of layers of the 3D CNN module to obtain the optimal network structure, such as Figure 3 As shown in the figure, the verification error is the smallest when a three-layer CNN is used. As the number of layers increases, overfitting occurs and the error increases.

[0055] Specifically, based on the validation data set and the NSGAⅡ-CNN-LSTM model, the cycle life of lithium-ion batteries is predicted, and the prediction results are obtained, such as Figure 4 As shown in Table 1, the prediction accuracy of the NSGAⅡ-CNN-LSTM neural network on the test set data is very high, and the predicted cycle life is very close to the actual cycle life. As shown in Table 1, the average absolute percentage error of the predicted cycle life of lithium-ion batteries is only 7.1%.

[0056]

[0057] The embodiment of the present invention also provides a lithium ion battery cycle life early prediction system based on NSGAⅡ-CNN-LSTM, and the specific implementation process thereof includes: adopting a data acquisition module to collect the working data of the lithium electronic battery, including charging capacity (Qc), charging voltage (CV), charging temperature (Tc), discharging temperature (Td), charging time (CT), discharging time (DT) and initial capacity (IC); adopting a battery life prediction module to input the obtained data set into the NSGAⅡ-CNN-LSTM model, and output the early prediction result of the lithium ion battery cycle life, wherein the training of the NSGAⅡ-CNN-LSTM model includes dividing the data set into a training data set, a test data set and a verification data set according to a proportion; building the NSGAⅡ-CNN-LSTM model, including a 3D CNN module and an LSTM module, wherein the 3D CNN module and the LSTM module are connected by a tiling layer, the hyperparameters of the 3D CNN module are optimized by the NSGAⅡ algorithm, the test data set is used for performance evaluation, the model parameters are adjusted, and the validity of the model is verified by the verification data set. Finally, for the lithium-ion battery to be tested, the working data is input into the trained NSGAⅡ-CNN-LSTM model to predict the cycle life of the lithium-ion battery and obtain the corresponding prediction results.

[0058] Based on the same inventive concept as the aforementioned embodiment, an embodiment of the present invention further provides an electronic device, including a memory and a processor, the memory being used to store computer-executable instructions, and the processor being used to execute computer-executable instructions, to implement the early prediction method for the cycle life of a lithium-ion battery based on NSGAⅡ-CNN-LSTM as proposed in the aforementioned embodiment.

[0059] This embodiment also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by the processor, it is used to quickly verify the service life of new lithium-ion battery materials during the development of new lithium-ion battery materials, and to quickly evaluate the advantages and disadvantages of new lithium-ion battery manufacturing processes. The storage medium can be any non-volatile storage device such as a hard disk, a solid-state hard disk, a flash drive, an optical disk, etc., which is used to store computer program codes and necessary data files. The stored computer program includes: a data acquisition module, a battery life prediction module; wherein the battery life prediction module includes: a data set establishment unit, a network model building unit and a life prediction unit.

[0060] Finally, it should be pointed out that the above specific embodiments are only representative examples of the present invention. Obviously, the present invention is not limited to the above specific embodiments, and there are many variations. Any simple modification, equivalent changes and modifications made to the above specific embodiments based on the technical essence of the present invention should be considered to belong to the protection scope of the present invention.

Claims

1. An early prediction method for lithium-ion battery cycle life based on NSGAⅡ-CNN-LSTM, comprising: Obtain the working data of the lithium-ion battery to be tested; The obtained working data of the lithium-ion battery to be tested is input into the trained NSGAⅡ-CNN-LSTM model to output the early prediction result of the cycle life of the lithium-ion battery; wherein the training of the NSGAⅡ-CNN-LSTM model includes: Identify the operating data with high relevance to the cycle life of lithium-ion batteries and construct a dataset of the operating data; Build an NSGAⅡ-CNN-LSTM model, including a 3D CNN module and an LSTM module. The 3D CNN module and the LSTM module are connected through a tile layer. The 3D CNN module is used to extract features from input working data, and the LSTM module is used to process the data after feature extraction and enhance the feature representation of the data. The constructed data set is used to train the constructed NSGAⅡ-CNN-LSTM model to obtain the trained NSGAⅡ-CNN-LSTM model.

2. The early prediction method for lithium-ion battery cycle life based on NSGAⅡ-CNN-LSTM according to claim 1 is characterized in that: Determine highly relevant working data for the cycle life of lithium-ion batteries, including: Acquire various working data of the lithium-ion battery, and respectively calculate the level difference and the square of the level difference between each working data and the cycle life of the lithium-ion battery; The Spearman correlation coefficient formula is calculated based on the calculated rank difference and rank difference square, and the value result between [-1, 1] is obtained. The value close to 1 indicates a strong positive correlation between the two variables, and the value close to -1 indicates a strong negative correlation. Based on the calculation results of the Spearman correlation coefficient of each working data, the working data with high correlation to the cycle life of the lithium-ion battery is determined.

3. The early prediction method for lithium-ion battery cycle life based on NSGAⅡ-CNN-LSTM according to claim 1 is characterized in that: The 3D CNN module includes: 3D convolutional layer, which slides over the input data and performs convolution operations; Normalization layer, used to normalize the input of each layer; ReLU layer, used to introduce nonlinear factors and alleviate the gradient vanishing problem; The 3D pooling layer is used to downsample the feature map output by the convolutional layer.

4. The early prediction method for lithium-ion battery cycle life based on NSGAⅡ-CNN-LSTM according to claim 1 is characterized in that: The LSTM module includes: LSTM layer, used to selectively retain or forget information; A self-attention mechanism that computes a relevance score between each position in the sequence and every other position; Fully connected layer, used for features extracted by LSTM layer and self-attention mechanism layer.

5. The method for early prediction of lithium-ion battery cycle life based on NSGAⅡ-CNN-LSTM according to claim 1, characterized in that: The training of the NSGAⅡ-CNN-LSTM model also includes: Based on the early prediction results of lithium-ion battery cycle life, the root mean square error, mean absolute percentage error and mean absolute error are calculated, and the hyperparameters in the 3D CNN module are optimized using a non-dominated sorting genetic algorithm based on the error data.

6. The method for early prediction of lithium-ion battery cycle life based on NSGAⅡ-CNN-LSTM according to claim 3, characterized in that: The 10-fold cross-validation method is used to optimize the number of layers of the 3D CNN module.

7. An early prediction system for lithium-ion battery cycle life based on NSGAⅡ-CNN-LSTM, comprising: A data acquisition module is used to acquire the working data of the lithium-ion battery to be tested; The battery life prediction module inputs the obtained working data of the lithium-ion battery to be tested into the trained NSGAⅡ-CNN-LSTM model, and outputs the early prediction result of the cycle life of the lithium-ion battery; wherein the training of the NSGAⅡ-CNN-LSTM model includes: Identify the operating data with high relevance to the cycle life of lithium-ion batteries and construct a dataset of the operating data; Build an NSGAⅡ-CNN-LSTM model, including a 3D CNN module and an LSTM module. The 3D CNN module and the LSTM module are connected through a tile layer. The 3D CNN module is used to extract features from input working data, and the LSTM module is used to process the data after feature extraction and enhance the feature representation of the data. The constructed data set is used to train the constructed NSGAⅡ-CNN-LSTM model to obtain the trained NSGAⅡ-CNN-LSTM model.

8. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the early prediction method of lithium-ion battery cycle life based on NSGAⅡ-CNN-LSTM according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the early prediction method of lithium-ion battery cycle life based on NSGAⅡ-CNN-LSTM according to any one of claims 1 to 6 are implemented.

Images