A battery early life prediction method and system based on convolutional neural network
By using a convolutional neural network-based method, graphical feature extraction and three-dimensional data matrix construction are performed on battery sampling data. The AlexNet network architecture is then used for early battery life prediction, which solves the problem of low accuracy in existing technologies and achieves high accuracy and reliability in battery life prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SUN YAT SEN UNIV
- Filing Date
- 2023-02-10
- Publication Date
- 2026-06-09
AI Technical Summary
Existing early battery life prediction technologies use neural network models with low accuracy and reliability in predicting battery life, especially with single-channel input data.
A convolutional neural network-based approach is adopted to extract graphical features from battery sampling data, construct a three-dimensional data matrix, and use the prediction model of the AlexNet network architecture to extract high-level features from multi-channel input features for battery life prediction.
It improves the accuracy and reliability of early battery life prediction, and enables rapid and reliable prediction based on a small amount of early data, with a root mean square error of 91.51 cycles and a correlation coefficient of 0.91.
Smart Images
Figure CN116068407B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method and system for predicting the early lifespan of batteries based on convolutional neural networks. Background Technology
[0002] Lithium-ion batteries possess advantages such as high energy density, long lifespan, and fast charging, leading to their widespread application in new energy vehicles and grid storage systems. Traditional lifespan prediction methods utilize a large number of battery cycles, often only predicting battery lifespan when capacity shows a significant decline. Early battery lifespan prediction technologies aim to provide rapid and reliable predictions using only limited early-stage data. However, current early battery lifespan prediction technologies primarily employ neural network models based on single-channel input data, resulting in low accuracy and reliability in predicting battery lifespan. Therefore, the technical problems in these technologies urgently need to be addressed. Summary of the Invention
[0003] In view of this, embodiments of the present invention provide a highly accurate and reliable method and system for predicting early battery life based on convolutional neural networks.
[0004] On one hand, this invention provides a method for predicting early battery life based on convolutional neural networks, the method comprising:
[0005] Obtain sampling data of the battery to be predicted;
[0006] The sampled data is processed by graphical feature extraction to obtain a two-dimensional tensor data set;
[0007] The two-dimensional tensor data set is adjusted and stacked to obtain a three-dimensional data matrix;
[0008] A prediction model is constructed based on a convolutional neural network. The prediction model includes a feature extraction module and a regression prediction module. The convolutional neural network adopts the AlexNet network architecture.
[0009] The three-dimensional data matrix is input into the prediction model to predict battery life and obtain the prediction result.
[0010] Optionally, obtaining the sampling data of the battery to be predicted includes:
[0011] Data is acquired and processed during the discharge process of the battery to be predicted to obtain discharge data;
[0012] The discharge data is sampled for discharge capacity and voltage data to obtain sampled data.
[0013] Optionally, the step of performing graphical feature extraction processing on the sampled data to obtain a two-dimensional tensor data set includes:
[0014] The sampled data is extracted and processed to obtain the discharge capacity data and voltage data within the cycle.
[0015] A first function curve is constructed based on the discharge capacity data and the voltage data;
[0016] The voltage data is differentiated based on the discharge capacity data to obtain the second function curve;
[0017] The third function curve is obtained based on the difference in discharge capacity data within each cycle and the voltage data;
[0018] Voltage values are sampled and processed in parallel on the first function curve, the second function curve, and the third function curve respectively to obtain a two-dimensional tensor data set, which includes capacity voltage characteristics, capacity increment characteristics, and capacity difference characteristics.
[0019] Optionally, the adjustment and stacking of the two-dimensional tensor data set to obtain a three-dimensional data matrix includes:
[0020] Each data point in the two-dimensional tensor dataset is resized, and the resized data points are stacked in the depth direction to obtain a three-dimensional data matrix.
[0021] Optionally, constructing the prediction model based on the convolutional neural network includes:
[0022] A feature extraction module is constructed based on a convolutional neural network;
[0023] A regression prediction module is constructed, which includes a first fully connected layer, a second fully connected layer, a third fully connected layer and a fourth fully connected layer. The third fully connected layer has 100 neurons and the fourth fully connected layer has 1 neuron.
[0024] A prediction model is constructed based on the feature extraction module and the regression prediction module.
[0025] Optionally, before inputting the three-dimensional data matrix into the prediction model for battery life prediction, the method further includes pre-training the prediction model, specifically including:
[0026] The model parameters of the prediction model are tuned according to the optimization algorithm to obtain a trained prediction model.
[0027] Optionally, the step of inputting the three-dimensional data matrix into the prediction model to predict battery life and obtain the prediction result includes:
[0028] The three-dimensional data matrix is input into the feature extraction module of the prediction model to obtain a feature map;
[0029] The regression prediction module of the prediction model performs a mapping process from the feature space to the sample label space on the feature map to obtain the prediction result.
[0030] On the other hand, embodiments of the present invention also provide a battery early life prediction system based on a convolutional neural network, comprising:
[0031] The first module is used to acquire sampling data of the battery to be predicted;
[0032] The second module is used to perform graphical feature extraction processing on the sampled data to obtain a two-dimensional tensor data set, which includes capacity voltage features, capacity increment features, and capacity difference features.
[0033] The third module is used to adjust and stack the two-dimensional tensor data set to obtain a three-dimensional data matrix;
[0034] The fourth module is used to construct a prediction model based on a convolutional neural network. The prediction model includes a feature extraction module and a regression prediction module. The convolutional neural network adopts the AlexNet network architecture.
[0035] The fifth module is used to input the three-dimensional data matrix into the prediction model to predict battery life and obtain the prediction result.
[0036] On the other hand, embodiments of the present invention also disclose an electronic device, including a processor and a memory;
[0037] The memory is used to store programs;
[0038] The processor executes the program to implement the method described above.
[0039] On the other hand, embodiments of the present invention also disclose a computer-readable storage medium storing a program that is executed by a processor to implement the methods described above.
[0040] On the other hand, embodiments of the present invention also disclose a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the aforementioned method.
[0041] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects: The embodiments of the present invention extract graphical features from the sampled data of the battery to be predicted, and combine multiple sets of graphical features to form multi-channel input data that is equivalent to the image input of the convolutional neural network. By applying the classic convolutional neural network, which performs well in the field of image recognition, to the early life prediction of the battery, the reliability of the predicted battery life is improved by utilizing the good feature generation and recognition capabilities of the convolutional neural network. Attached Figure Description
[0042] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0043] Figure 1 This is a flowchart of a battery early life prediction method based on a convolutional neural network provided in an embodiment of this application;
[0044] Figure 2 This is a network architecture diagram of a prediction model provided in an embodiment of this application;
[0045] Figure 3 This is a comparison chart of model predictions and actual values provided in an embodiment of this application. Detailed Implementation
[0046] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0047] First, let's analyze some of the terms used in this application:
[0048] Early battery life prediction: Utilizing limited data from the early stages of battery life, combined with machine learning, deep learning, and other methods, this technology enables rapid prediction of battery life before significant capacity degradation is observed in the early stages of battery degradation.
[0049] Deep learning is a method in machine learning based on learning representations of data. It uses multiple processing layers containing complex structures or multiple nonlinear transformations to abstract data.
[0050] Convolutional Neural Networks (CNNs) are a type of feedforward neural network that can take graphical data, such as image pixels, as input and automatically extract features from the data layer by layer. They are highly invariant to deformations (translation, tilt, etc.) of graphical data. CNNs can learn representations from grid-like data in vision-related applications and have good recognition capabilities.
[0051] As the number of charge-discharge cycles increases, the performance of lithium-ion batteries gradually degrades, and their capacity decreases. This degradation is not linear. When the battery capacity drops below 80% of its initial capacity, the capacity decline accelerates. Traditional battery life prediction methods use a large number of battery cycles, often only predicting battery life when the battery capacity shows a significant decrease. Early battery life prediction technology aims to make rapid and reliable predictions of battery life using only a limited amount of data from the early stages of battery life.
[0052] Early battery life prediction utilizes limited early-stage battery observation data to build models based on physics, mathematics, machine learning, or deep learning to quickly and directly predict battery life. Compared to remaining battery life prediction, early battery life prediction requires even less data, and capacity degradation is not significant in the early stages of battery use, making prediction more challenging. Current mainstream methods for early battery life prediction primarily employ machine learning, selecting capacity-voltage features as single-channel inputs to the network and capturing gradual changes in local areas using non-overlapping kernels to achieve early life prediction. However, the neural network models used in these technologies apply to single-channel input data and do not extract and analyze multiple features, resulting in low accuracy in early battery life prediction. Therefore, this invention aims to improve the accuracy and reliability of early battery life prediction.
[0053] Reference Figure 1 This invention provides a method for predicting early battery life based on a convolutional neural network, comprising:
[0054] S101. Obtain the sampling data of the battery to be predicted;
[0055] S102. Perform graphical feature extraction processing on the sampled data to obtain a two-dimensional tensor data set;
[0056] S103. The two-dimensional tensor data set is adjusted and stacked to obtain a three-dimensional data matrix;
[0057] S104. Construct a prediction model based on a convolutional neural network. The prediction model includes a feature extraction module and a regression prediction module. The convolutional neural network adopts the AlexNet network architecture.
[0058] S105. Input the three-dimensional data matrix into the prediction model to predict battery life and obtain the prediction result.
[0059] In this embodiment of the invention, multiple sets of correlated graphical features are extracted from the sampled data of the battery to be predicted. These extracted features are then stacked in the depth direction to form a three-dimensional matrix, establishing a multi-channel input equivalent to the image input of a convolutional neural network. A prediction model built based on the convolutional neural network is then used to further extract high-level features from the multi-channel input features and map them onto the battery life, thereby achieving reliable prediction of the early battery life and improving prediction accuracy. This embodiment of the invention performs graphical feature extraction processing on the sampled data of the battery to be predicted, extracting a two-dimensional tensor dataset. This two-dimensional tensor dataset includes capacity-voltage features, capacity increment features, and capacity difference features, which exhibit clear correlations, enabling the prediction model to better extract and analyze features, thus improving prediction accuracy. Furthermore, the prediction model built based on the convolutional neural network in this embodiment of the invention can automatically extract features in multiple dimensions, such as time and space, using its own network architecture. This efficiently utilizes the continuous data of each charge-discharge cycle in the early battery life prediction, providing richer and more intelligent features for the early life prediction of lithium-ion batteries.
[0060] As a further preferred embodiment, the step of acquiring the sampling data of the battery to be predicted includes:
[0061] Data is acquired and processed during the discharge process of the battery to be predicted to obtain discharge data;
[0062] The discharge data is sampled for discharge capacity and voltage data to obtain sampled data.
[0063] In this embodiment of the invention, the prediction of early battery life is achieved by building a machine learning model using limited early-stage observational data of the battery to directly and quickly predict its lifespan. Therefore, it is necessary to acquire sampling data of the battery to be predicted. During the discharge process of the battery, various discharge data are collected. During the data collection process, a battery capacity tester can be used to detect discharge data such as overcurrent, voltage, capacity, internal resistance, charging and discharging temperature, and battery cycle life. Then, discharge capacity and voltage data are sampled from the discharge data to obtain the sampled data.
[0064] As a further preferred embodiment, the step of performing graphical feature extraction processing on the sampled data to obtain a two-dimensional tensor data set includes:
[0065] The sampled data is extracted and processed to obtain the discharge capacity data and voltage data within the cycle.
[0066] A first function curve is constructed based on the discharge capacity data and the voltage data;
[0067] The voltage data is differentiated based on the discharge capacity data to obtain the second function curve;
[0068] The third function curve is obtained based on the difference in discharge capacity data within each cycle and the voltage data;
[0069] Voltage values are sampled and processed in parallel on the first function curve, the second function curve, and the third function curve respectively to obtain a two-dimensional tensor data set, which includes capacity voltage characteristics, capacity increment characteristics, and capacity difference characteristics.
[0070] In this embodiment of the invention, the discharge capacity and voltage data of the battery within one cycle are extracted from the sampling data of the battery to be predicted, and respectively represented by Q. d Let v represent the values of Q1 and V. The main purpose of this invention is to directly predict battery life before significant degradation trends are observed. Therefore, in this embodiment, only the data from the first 100 discharge cycles of the battery are used for one cycle. Within one cycle of the battery to be predicted, the discharge capacity is considered a function of the discharge voltage, with the discharge voltage as the horizontal axis. For ease of description, Q1(v) represents the time-series data of the discharge capacity-discharge voltage curve within one cycle, resulting in the first function curve. Next, the first derivative of the discharge capacity with respect to voltage within one discharge cycle is calculated, yielding curve Q1(v). d The rate of change of v, i.e., the second function curve, is denoted as Q2(v). The formula for calculating the second function curve is shown below:
[0071]
[0072] In the formula, j represents the j-th sampling point, Q d (i) represents the discharge capacity data of the i-th sampling point, and v(i) represents the voltage data of the i-th sampling point.
[0073] Furthermore, in terms of cycle time, the discharge capacity data is interpolated. The time-series discharge capacity data from one cycle and the second cycle within the predicted battery life are subtracted to obtain the third function curve, denoted as Q3(v). The formula for calculating the third function curve is shown below:
[0074] Q3 = Q di -Q d2 ;
[0075] Among them, Q di and Q d2 These represent the discharge capacity curve data for the i-th and 2nd cycles, respectively.
[0076] For the three voltage function curves Q1(v), Q2(v), and Q3(v) mentioned above, 100 equidistant voltage sampling points Vs∈R are set within each of the first 100 discharge cycles of the battery, between the upper and lower limits of the battery discharge voltage (2.0V~3.5V). 1×100 Using the voltage values of the sampling points as indices, the corresponding curve data is obtained, resulting in 1*100 one-dimensional time-series data. The one-dimensional time-series data from the first 100 battery life cycles are then concatenated in parallel to form a 100*100 two-dimensional tensor, yielding the capacity-voltage feature F1∈R. 100×100 Capacity increment feature F2∈R 100×100 The capacity difference feature F3∈R 100×100 .
[0077] As a further preferred embodiment, the adjustment and stacking of the two-dimensional tensor data set to obtain a three-dimensional data matrix includes:
[0078] Each data point in the two-dimensional tensor dataset is resized, and the resized data points are stacked in the depth direction to obtain a three-dimensional data matrix.
[0079] In this embodiment of the invention, the two-dimensional tensor data set includes capacity voltage features, capacity increment features, and capacity difference features. The size of the above three two-dimensional graphical features (100*100) is adjusted to 227*227, and they are stacked in the depth direction to obtain a three-dimensional data matrix, thus constructing a three-channel input that is equivalent to the input of a convolutional neural network.
[0080] As a further preferred embodiment, the step of constructing a prediction model based on a convolutional neural network includes:
[0081] A feature extraction module is constructed based on a convolutional neural network;
[0082] A regression prediction module is constructed, which includes a first fully connected layer, a second fully connected layer, a third fully connected layer and a fourth fully connected layer. The third fully connected layer has 100 neurons and the fourth fully connected layer has 1 neuron.
[0083] A prediction model is constructed based on the feature extraction module and the regression prediction module.
[0084] In this embodiment of the invention, a classic convolutional neural network is used to further extract high-level features from the features and map them onto battery life, thereby achieving reliable prediction of early battery life. The early battery life prediction method based on convolutional neural networks can automatically extract features in multiple dimensions such as time and space using its own network architecture, efficiently utilizing continuous data from the early discharge process of battery degradation, and providing richer and more intelligent features for early lithium-ion battery life prediction. A convolutional neural network is a feedforward multi-layer network, which can be subdivided into multiple learning processes such as convolutional layers, activation functions, and fully connected layers. Convolutional neural networks with different topologies composed of various network layers have varying performance. In this embodiment of the invention, the AlexNet network from convolutional neural networks is used as the network architecture of the prediction model. The fully connected layers of the network architecture are modified, with a total of four fully connected layers. The number of neurons in the fully connected layers is modified to 4096, 4096, 100, and 1, respectively. A dropout layer is connected after the first two fully connected layers. The feature extraction part of the prediction model is constructed using multiple convolutional layers, ReLU activation functions, and pooling layers cascaded together. A fully connected layer then maps the feature space to the sample label space and outputs the model's predicted values. The network structure of the prediction model is as follows: Figure 2 As shown.
[0085] This invention constructs a prediction model based on a convolutional neural network, starting with a feature extraction module. The first convolutional layer in this module has 96 kernels (11*11*3) and a stride of 4. A ReLU activation function is then applied, resulting in a feature map of size 55*55*96. This is followed by a local response normalization layer to accelerate training. Next, a max pooling layer is applied. This pooling layer has a 3*3 pooling window and a stride of 2, resulting in an output size of 27*27*96. The 96 feature maps are divided into two groups, each with a size of 27*27*48. A second convolutional layer performs convolution operations within each group. Each group is convolved with 128 kernels of size 5*5*48 with a stride of 1. This is followed by a ReLU activation function and a local response normalization layer. The resulting feature map in each group has a size of 27*27*128, and there are two groups in total. Finally, a max pooling layer is applied. The pooling layer has a 3x3 pooling window and a stride of 2. After individual pooling operations, each group outputs a size of 13x13x128. The third convolutional layer performs global convolution. It receives two concatenated data sets with a size of 13x13x256 as input. This layer has 384 convolutional kernels (3x3x256 each) with a stride of 1. A ReLU activation function is then applied, resulting in a feature map of size 13x13x384. These 384 feature maps are further divided into two groups, each with a size of 13x13x192. The fourth convolutional layer performs intra-group convolution. Each group is convolved with 192 3x3x192 kernels with a stride of 1. A ReLU activation function is then applied. Each group produces a feature map of size 13x13x192, and there are two groups in total. The fifth convolutional layer performs intra-group convolution. Each data set is convolved with 128 3*3*192 convolutional kernels with a stride of 1. This is followed by a ReLU activation function. The resulting feature map in each set is 13*13*128 pixels, and there are two sets in total. Next, a max pooling layer is applied. This pooling layer has a 3*3 pooling window and a stride of 2. After each set undergoes individual pooling, they are joined together to obtain a feature map of size 6*6*256. A regression prediction module is then constructed, consisting of a first fully connected layer, a second fully connected layer, a third fully connected layer, and a fourth fully connected layer. The feature maps extracted by the feature extraction module are fed into the first fully connected layer, which contains 4096 neurons. This is followed by a ReLU activation layer and a dropout layer with a dropout probability of 0.5. Then, a second fully connected layer, also containing 4096 neurons, is applied. This is again followed by a ReLU activation layer and a dropout layer with a dropout probability of 0.5. Finally, the brain passes through the third and fourth fully connected layers, which contain 100 and 1 neuron nodes, respectively.A prediction model is constructed based on the feature extraction module and the regression prediction module. In this embodiment of the invention, the prediction model uses the non-linear, non-saturated ReLU function instead of the non-linear, saturated Sigmoid function as the activation function, improving the gradient decay rate during training. Furthermore, unlike other convolutional neural networks where the pooling window of the pooling layer equals the stride, this embodiment sets a smaller pooling stride, resulting in pooling overlap during pooling. This expands the pooling layer output into multiple smaller features, and sparse coding is used to fuse these multi-level features, reducing the feature dimensionality of the pooling layer output.
[0086] As a further preferred embodiment, before inputting the three-dimensional data matrix into the prediction model for battery life prediction, the method further includes pre-training the prediction model, specifically including:
[0087] The model parameters of the prediction model are tuned according to the optimization algorithm to obtain a trained prediction model.
[0088] In this embodiment of the invention, before the prediction model performs early life prediction on the battery to be predicted, it needs to be trained and tested. The Adam algorithm is chosen as the optimization algorithm for training the prediction model. Compared to the stochastic gradient descent algorithm, the Adam algorithm adds first and second moments and sets specific adaptive learning rates for different parameters. The initial learning rate is set to 0.0005. For the pre-trained prediction model, the number of training iterations is set to 400. Then, experiments are conducted on a battery degradation dataset to test the model. The experiments show that the prediction model has reliable prediction accuracy in early life prediction, and the prediction results are as follows: Figure 3 As shown. After training and testing the prediction model as described above, a deep learning model based on graphical features can be built, which can directly predict battery life using a small amount of early-stage battery data.
[0089] As a further preferred embodiment, the step of inputting the three-dimensional data matrix into the prediction model to predict battery life and obtain the prediction result includes:
[0090] The three-dimensional data matrix is input into the feature extraction module of the prediction model to obtain a feature map;
[0091] The regression prediction module of the prediction model performs a mapping process from the feature space to the sample label space on the feature map to obtain the prediction result.
[0092] In this embodiment of the invention, a three-dimensional data matrix obtained by combining three sets of graphical features is used as the three-channel input data of the model by a prediction model based on a convolutional neural network. After the feature extraction module of the prediction model automatically extracts and learns the features, a feature map is obtained. Then, the regression prediction module of the prediction model performs mapping processing from the feature space to the sample label space on the feature map, and finally outputs the predicted value of battery life.
[0093] Combined with appendix Figure 1 The process of this invention specifically includes: performing graphical feature extraction processing on the sampled data of the battery to be predicted; obtaining a three-dimensional data matrix through stacking processing; and using this matrix as input to a prediction model constructed based on a convolutional neural network to predict battery life, thereby obtaining a prediction result for the early life of the battery. This embodiment of the invention applies the classic convolutional neural network used for image classification problems to the early life prediction of batteries. Multiple sets of graphical features are cascaded to form a three-dimensional matrix equivalent to the image data, which is used as input to the network. A prediction model is built based on the convolutional neural network, enabling it to handle regression prediction problems and make reliable predictions about battery life.
[0094] and Figure 1 Corresponding to the method described above, this embodiment of the invention also provides an electronic device, including a processor and a memory; the memory is used to store a program; the processor executes the program to implement the method described above.
[0095] and Figure 1 Corresponding to the method described above, embodiments of the present invention also provide a computer-readable storage medium storing a program that is executed by a processor to implement the method described above.
[0096] This invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computer device to perform... Figure 1 The method shown.
[0097] In summary, the embodiments of the present invention have the following advantages:
[0098] This invention applies the classic convolutional neural network, which excels in image recognition and classification, to the early life prediction of batteries, extracting more potential features from a small amount of data in the early stages of battery degradation and achieving rapid prediction of early battery life.
[0099] In addition, the embodiments of the present invention combine three sets of features to construct three-dimensional features equivalent to convolutional neural network image data as the multi-channel input of the network, which improves the shortcomings of existing research that mainly uses single-channel two-dimensional data or stacked two-dimensional data of multiple parameters to form a three-dimensional matrix.
[0100] Furthermore, this embodiment of the invention utilizes data from the first 100 cycles of the battery's early lifespan to achieve a prediction effect of 91.51 cycles with a root mean square error and a correlation coefficient of 0.91 on the improved prediction model, thereby improving the accuracy of early battery lifespan prediction.
[0101] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this invention are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is altered and sub-operations described as part of a larger operation are executed independently.
[0102] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.
[0103] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0104] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.
[0105] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.
[0106] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.
[0107] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0108] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.
[0109] The above is a detailed description of the preferred embodiments of the present invention, but the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.
Claims
1. A method for predicting early battery life based on convolutional neural networks, characterized in that, The method includes: Obtain sampling data of the battery to be predicted; The sampled data is processed by graphical feature extraction to obtain a two-dimensional tensor data set; The two-dimensional tensor data set is adjusted and stacked to obtain a three-dimensional data matrix; A prediction model is constructed based on a convolutional neural network. The prediction model includes a feature extraction module and a regression prediction module. The convolutional neural network adopts the AlexNet network architecture. The three-dimensional data matrix is input into the prediction model to predict battery life, and the prediction result is obtained. The step of performing graphical feature extraction processing on the sampled data to obtain a two-dimensional tensor data set includes: The sampled data is extracted and processed to obtain discharge capacity data and voltage data in each cycle. The sampled data includes discharge data collected in multiple cycles during the discharge process of the battery to be predicted. A first function curve is constructed based on the discharge capacity data and the voltage data; the first function curve represents the time-series data of the battery's discharge capacity-discharge voltage curve within one cycle; The voltage data is differentiated based on the discharge capacity data to obtain the second function curve; The formula for calculating the second function curve is as follows: ； In the formula, j Indicates the first j One sampling point, Indicates the first Discharge capacity data at each sampling point Indicates the first Voltage data at each sampling point; The third function curve is obtained based on the difference in discharge capacity data within each cycle and the voltage data; The formula for calculating the third function curve is as follows: ； in, and They represent the first i Discharge capacity curve data for the first and second cycles; Voltage values are sampled and processed in parallel on the first function curve, the second function curve, and the third function curve respectively to obtain a two-dimensional tensor data set, which includes capacity voltage characteristics, capacity increment characteristics, and capacity difference characteristics.
2. The method according to claim 1, characterized in that, The acquisition of sampling data for the battery to be predicted includes: Data is acquired and processed during the discharge process of the battery to be predicted to obtain discharge data; The discharge data is sampled for discharge capacity and voltage data to obtain sampled data.
3. The method according to claim 1, characterized in that, The process of adjusting and stacking the two-dimensional tensor data set to obtain a three-dimensional data matrix includes: Each data point in the two-dimensional tensor dataset is resized, and the resized data points are stacked in the depth direction to obtain a three-dimensional data matrix.
4. The method according to claim 1, characterized in that, The construction of the prediction model based on the convolutional neural network includes: A feature extraction module is constructed based on a convolutional neural network. The feature extraction module consists of multiple convolutional layers, ReLU activation functions, and pooling layers in a cascaded manner. A regression prediction module is constructed, comprising a first fully connected layer, a second fully connected layer, a third fully connected layer, and a fourth fully connected layer. The third fully connected layer has 100 neurons, and the fourth fully connected layer has 1 neuron. A prediction model is constructed based on the feature extraction module and the regression prediction module.
5. The method according to claim 1, characterized in that, Before inputting the three-dimensional data matrix into the prediction model for battery life prediction, the method further includes pre-training the prediction model, specifically including: The model parameters of the prediction model are tuned using an optimization algorithm to obtain a trained prediction model.
6. The method according to claim 1, characterized in that, The step of inputting the three-dimensional data matrix into the prediction model to predict battery life and obtain the prediction result includes: The three-dimensional data matrix is input into the feature extraction module of the prediction model to obtain a feature map; The regression prediction module of the prediction model performs a mapping process from the feature space to the sample label space on the feature map to obtain the prediction result.
7. A battery early life prediction system based on convolutional neural networks, characterized in that, The system is applied to a battery early life prediction method based on a convolutional neural network as described in claim 1, the system comprising: The first module is used to acquire sampling data of the battery to be predicted; The second module is used to perform graphical feature extraction processing on the sampled data to obtain a two-dimensional tensor data set; The third module is used to adjust and stack the two-dimensional tensor data set to obtain a three-dimensional data matrix; The fourth module is used to construct a prediction model based on a convolutional neural network. The prediction model includes a feature extraction module and a regression prediction module. The convolutional neural network adopts the AlexNet network architecture. The fifth module is used to input the three-dimensional data matrix into the prediction model to predict battery life and obtain the prediction result.
8. An electronic device, characterized in that, The electronic device includes a memory and a processor; The memory is used to store programs; The processor executes the program to implement the method of any one of claims 1 to 6.
9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 6.