Battery remaining life prediction method and device, electronic equipment and storage medium

Abstract

The invention provides a battery remaining life prediction method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining battery charging data, inputting the battery charging data as three channels into a convolutional neural network for feature extraction, and obtaining feature matrixes corresponding to the three channels, obtaining a corresponding adjacency matrix based on the feature matrix and the corresponding transpose matrix, splicing the adjacency matrix to obtain a fusion adjacency matrix, splicing the feature matrix to obtain a fusion feature matrix, obtaining a multi-channel fusion feature based on the fusion adjacency matrix and the fusion feature matrix, and inputting the feature matrix into a channel interactive attention mechanism. According to the method, a channel interaction attention matrix is obtained, a fusion adjacency matrix is enhanced based on the channel interaction attention matrix, a target prediction feature is obtained based on the enhanced fusion adjacency matrix and the multi-channel fusion feature, and the remaining life of the battery is predicted based on the target prediction feature, so that the accuracy of lithium battery life prediction is improved through the method.

Description

Technical Field

[0001] The present application relates to the technical field of image recognition, and particularly relates to a method, device, electronic device and storage medium for predicting the remaining battery life. Background Art

[0002] With the rapid development of the new energy vehicle industry, new energy vehicles are gradually becoming the focus of market attention. One of its core components - the power battery not only undertakes the important task of storing energy, but also directly affects the vehicle's performance, cruising range and safety. Power batteries, especially lithium batteries, have been widely used in the field of electric vehicles due to their high energy density, long cycle life and environmental protection characteristics.

[0003] However, the performance of lithium batteries will inevitably decline during use, which is mainly due to the complex electrochemical reaction process inside and the influence of external environmental factors. As the number of charge and discharge cycles increases, the health state of the battery gradually decreases until it reaches the retirement standard. Therefore, accurately estimating and predicting the remaining service life of the battery is of great significance for optimizing the battery management strategy, improving the vehicle operation strategy, extending the battery service life and preventing potential safety risks.

[0004] In the related art, battery life prediction based on models or through data-driven methods are two relatively common ways. However, battery life prediction based on models is to predict the battery performance using a mathematical model. However, due to the many non-linear characteristics of the service life of lithium batteries during the attenuation process, it is difficult to predict, so it is very difficult to accurately predict the battery life through a mathematical model. The data-driven method predicts the battery life through a large amount of data. Although good prediction results can be obtained through data-driven methods when the data is sufficient, when facing the non-linear and dynamically changing data generated by the complex electrochemical system of lithium batteries, traditional data-driven prediction still has limitations in extracting internal relationships, resulting in low prediction accuracy. Summary of the Invention

[0005] In view of the above-mentioned disadvantages of the related art, the present application provides a method, device, electronic device and storage medium for predicting the remaining battery life to solve the technical problem of low accuracy in predicting the remaining battery life.

[0006] The present application provides a method for predicting the remaining life of a battery. The method for predicting the remaining life of the battery includes: obtaining battery charging data, where the battery charging data includes constant current charging data, constant voltage charging data, and internal resistance cycle data; using the battery charging data as three channels to input into a convolutional neural network for feature extraction to obtain corresponding feature matrices for the three channels, where the feature matrices include a constant current feature matrix, a constant voltage feature matrix, and an internal resistance cycle feature matrix; obtaining adjacency matrices corresponding to the three channels based on the feature matrices of the three channels and their corresponding transposed matrices, where the adjacency matrices include a constant current adjacency matrix, a constant voltage adjacency matrix, and an internal resistance cycle adjacency matrix; splicing the constant current adjacency matrix, the constant voltage adjacency matrix, and the internal resistance cycle adjacency matrix to obtain a fused adjacency matrix, splicing the constant current feature matrix, the constant voltage feature matrix, and the internal resistance cycle feature matrix to obtain a fused feature matrix, and obtaining multi-channel fused features based on the fused adjacency matrix and the fused feature matrix; inputting the feature matrices of the three channels into a channel interaction attention mechanism to obtain a channel interaction attention matrix; enhancing the fused adjacency matrix based on the channel interaction attention matrix to obtain an enhanced fused adjacency matrix, and performing an aggregation operation on the enhanced fused adjacency matrix and the multi-channel fused features to obtain target prediction features; inputting the target prediction features into a fully connected layer to obtain a prediction result for the remaining life of the battery.

[0007] In an embodiment of the present application, splicing the constant current adjacency matrix, the constant voltage adjacency matrix, and the internal resistance cycle adjacency matrix to obtain a fused adjacency matrix includes: using the constant current adjacency matrix as the first adjacency node, using the constant voltage adjacency matrix as the second adjacency node, and using the internal resistance cycle adjacency matrix as the third adjacency node; connecting the first adjacency node, the second adjacency node, and the third adjacency node pairwise to obtain the fused adjacency matrix.

[0008] In an embodiment of the present application, splicing the constant current adjacency matrix, the constant voltage adjacency matrix, and the internal resistance cycle adjacency matrix to obtain a fused adjacency matrix includes: using the constant current adjacency matrix as the first adjacency node, using the constant voltage adjacency matrix as the second adjacency node, and using the internal resistance cycle adjacency matrix as the third adjacency node; connecting the first adjacency node, the second adjacency node, and the third adjacency node pairwise to obtain the fused adjacency matrix.

[0009] In an embodiment of the present application, obtaining the adjacency matrices corresponding to three channels based on the feature matrices of the three channels and the corresponding transposed matrices includes: multiplying the constant current feature matrix by the transposed matrix corresponding to the constant current feature matrix to obtain the constant current adjacency matrix; multiplying the constant voltage feature matrix by the transposed matrix corresponding to the constant voltage feature matrix to obtain the constant voltage adjacency matrix; multiplying the internal resistance cycle feature matrix by the transposed matrix corresponding to the internal resistance cycle feature matrix to obtain the internal resistance cycle adjacency matrix.

[0010] In an embodiment of the present application, obtaining the multi-channel fusion features based on the fused adjacency matrix and the fused feature matrix includes: performing a symmetric normalization operation on the fused adjacency matrix, and obtaining the pre-activated multi-channel fusion features based on the symmetrically normalized fused adjacency matrix, the fused feature matrix, and a preset weight matrix; activating the pre-activated multi-channel fusion features based on a preset activation function to obtain the multi-channel fusion features.

[0011] In an embodiment of the present application, aggregating the enhanced fused adjacency matrix and the multi-channel fusion features to obtain the target prediction features includes:

[0012]

[0013] where, H out is the target prediction feature, ReLU is the preset activation function, A h is the enhanced fused adjacency matrix, is the normalization process on the enhanced fused adjacency matrix, H d is the multi-channel fusion feature, and W is the preset weight matrix.

[0014] In an embodiment of the present application, inputting the feature matrices of the three channels into the channel interaction attention mechanism to obtain the channel interaction attention matrix includes: performing a symmetric normalization operation on the adjacency matrix of each channel, obtaining the aggregated features of each channel based on the feature matrix of each channel and the symmetrically normalized adjacency matrix, where the aggregated features include the aggregated constant current feature, the aggregated constant voltage feature, and the aggregated internal resistance cycle feature; performing pairwise product interaction on the aggregated constant current feature, the aggregated constant voltage feature, and the aggregated internal resistance cycle feature to obtain a plurality of product interaction features; processing each product interaction feature through a hyperbolic tangent function and a normalized exponential function to obtain the attention coefficient vector corresponding to each product interaction feature; obtaining the channel interaction attention matrix based on the attention coefficient vector corresponding to each product interaction feature and the plurality of product interaction features.

[0015] In an embodiment of the present application, before inputting the target prediction feature into the fully connected layer to obtain the battery remaining life prediction result, it includes: acquiring historical charging data and the remaining life value corresponding to the historical charging data, inputting the historical charging data into the convolutional neural network for feature extraction, and obtaining historical prediction features based on the extracted feature matrix;

[0016] Using the historical prediction features as the training data set and the remaining life value corresponding to the historical prediction features as the training label;

[0017] Training the fully connected layer based on the training data set and the training label, so that the fully connected layer obtains the corresponding battery remaining life prediction result based on the target prediction feature.

[0018] An embodiment of the present application further provides a battery remaining life prediction device, which includes: a charging data input module for acquiring battery charging data, where the battery charging data includes constant current charging data, constant voltage charging data, and internal resistance cycle data; a data feature extraction module for inputting the battery charging data as three channels into a convolutional neural network for feature extraction to obtain feature matrices corresponding to the three channels, where the feature matrices include a constant current feature matrix, a constant voltage feature matrix, and an internal resistance cycle feature matrix; a calculation adjacency matrix module for obtaining adjacency matrices corresponding to the three channels based on the feature matrices of the three channels and their corresponding transposed matrices, where the adjacency matrices include a constant current adjacency matrix, a constant voltage adjacency matrix, and an internal resistance cycle adjacency matrix; a first feature fusion module for splicing the constant current adjacency matrix, the constant voltage adjacency matrix, and the internal resistance cycle adjacency matrix to obtain a fused adjacency matrix, splicing the constant current feature matrix, the constant voltage feature matrix, and the internal resistance cycle feature matrix to obtain a fused feature matrix, and obtaining multi-channel fused features based on the fused adjacency matrix and the fused feature matrix; a calculation attention matrix module for inputting the feature matrices of the three channels into a channel interaction attention mechanism to obtain a channel interaction attention matrix; a second feature fusion module for enhancing the fused adjacency matrix based on the channel interaction attention matrix to obtain an enhanced fused adjacency matrix, and performing an aggregation operation on the enhanced fused adjacency matrix and the multi-channel fused features to obtain a target prediction feature; a battery life prediction module for inputting the target prediction feature into a fully connected layer to obtain a battery remaining life prediction result.

[0019] An embodiment of the present application further provides an electronic device, which includes: one or more processors; a storage device for storing one or more programs, and when the one or more programs are executed by the one or more processors, the electronic device realizes the battery remaining life prediction method as described in any one of the above embodiments.

[0020] An embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor of a computer, the computer is enabled to execute the battery remaining life prediction method as described in any one of the above embodiments.

[0021] Advantages of the present application: An embodiment of the present application provides a battery remaining life prediction method, device, electronic device and storage medium. The method includes obtaining battery charging data, and using the battery charging data as three channels to input into a convolutional neural network for feature extraction to obtain feature matrices corresponding to the three channels. Based on the feature matrices and their corresponding transposed matrices, corresponding adjacency matrices are obtained. The adjacency matrices are concatenated to obtain a fused adjacency matrix, the feature matrices are concatenated to obtain a fused feature matrix, and multi-channel fused features are obtained based on the fused adjacency matrix and the fused feature matrix. The feature matrices are input into a channel interaction attention mechanism to obtain a channel interaction attention matrix, and the fused adjacency matrix is enhanced based on the channel interaction attention matrix. Based on the enhanced fused adjacency matrix and the multi-channel fused features, target prediction features are obtained, and the battery remaining life is predicted based on the target prediction features. By performing feature extraction on the battery charging data and performing feature fusion within multiple channels and interactive fusion of multiple channels, the relationship between different channels and the importance of features can be better utilized, providing more comprehensive and richer battery charging data, thereby improving the expression ability and generalization ability of the model.

[0022] It should be understood that the above general description and subsequent detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a schematic diagram of an implementation environment of a battery remaining life prediction method shown in an exemplary embodiment of the present application;

[0024] Figure 2 is a flowchart of a battery remaining life prediction method shown in an exemplary embodiment of the present application;

[0025] Figure 3 is a schematic diagram of the construction process of a fused adjacency matrix shown in an exemplary embodiment of the present application;

[0026] Figure 4 is a schematic diagram of a channel interaction attention mechanism shown in an exemplary embodiment of the present application;

[0027] Figure 5 is a schematic diagram of the aggregation process of the enhanced fused adjacency matrix shown in an exemplary embodiment of the present application;

[0028] Figure 6It is a block diagram of a battery remaining life prediction device shown in an exemplary embodiment of the present application;

[0029] Figure 7 It is a schematic structural diagram of an electronic device shown in an exemplary embodiment of the present application. Specific embodiments

[0030] The following uses specific specific examples to illustrate the implementation manners of the present application. Those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in this specification. The present application can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0031] It should be noted that the drawings provided in the following embodiments only illustrate the basic concept of the present application in a schematic manner. Therefore, only the components related to the present application are shown in the drawings, rather than being drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and ratio of each component in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0032] It should be noted that in the present application, "first", "second", etc. are only used to distinguish similar objects, and are not used to limit the order or sequence of similar objects. The described "including", "having", etc. are deformed, indicating that the scope covered by the subject of this word is not exclusive except for the examples shown by this word.

[0033] It can be understood that the various numerical numbers, step numbers, etc. recorded in the present application are for the convenience of description and are not used to limit the scope of the present application. The size of the labels in the present application does not mean the sequence of execution. The execution sequence of each process should be determined by its function and internal logic.

[0034] In the following description, a large number of details are discussed to provide a more thorough explanation of the embodiments of the present application. However, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present application difficult to understand.

[0035] Embodiments of the present application respectively propose a battery remaining life prediction method, a battery remaining life prediction device, an electronic device, a computer-readable storage medium, and a computer program product. The following will describe these embodiments in detail.

[0036] Please refer toFigure 1 , Figure 1 is a schematic diagram of the implementation environment of a battery remaining life prediction method shown in an exemplary embodiment of the present application.

[0037] As Figure 1 shown, the implementation environment may include a data acquisition device 101 and a computer device 102. The computer device 102 may be at least one of a microcomputer, an embedded computer, a neural network computer, etc. The data acquisition device 101 is used to collect charging data during battery charging, including constant current charging data, constant voltage charging data, and internal resistance cycle data. The computer device 102 is used to perform feature extraction and feature aggregation on the collected battery charging data to obtain target prediction features, and perform battery remaining life prediction based on the target prediction features to obtain a battery remaining life prediction result.

[0038] Please refer to Figure 2 , Figure 2 is a flowchart of a battery remaining life prediction method shown in an exemplary embodiment of the present application. This method can be applied to the Figure 1 implementation environment shown, and this method can also be applicable to other exemplary implementation environments and be specifically executed by devices in other implementation environments. This embodiment does not limit the implementation environment applicable to this method.

[0039] As Figure 2 shown, in an exemplary embodiment, the battery remaining life prediction method at least includes steps S210 to S270, which are introduced in detail as follows:

[0040] Step S210, obtain battery charging data, where the battery charging data includes constant current charging data, constant voltage charging data, and internal resistance cycle data.

[0041] In an embodiment of the present application, the constant current charging data includes the value of the constant current charging time of the battery under test changing with the number of cycles during multiple chargings of the battery under test. The constant voltage charging time includes the value of the constant voltage charging time of the battery under test changing with the number of cycles during multiple chargings of the battery under test. The internal resistance cycle data includes the value of the internal resistance of the battery under test changing with the number of cycles during multiple chargings of the battery under test.

[0042] Step S220, input the battery charging data as three channels into a convolutional neural network for feature extraction to obtain feature matrices corresponding to the three channels. The feature matrices include a constant current feature matrix, a constant voltage feature matrix, and an internal resistance cycle feature matrix.

[0043] In an embodiment of the present application, input the battery charging data as three channels and perform feature extraction on each channel based on the convolutional neural network, that is, Hi = CNN(X i ), where i = 1, 2, 3. When i = 1, it represents the channel corresponding to the constant current charging data. When i = 2, it represents the channel corresponding to the constant voltage charging data. When i = 3, it represents the channel corresponding to the internal resistance cycle data. H i represents the feature matrices corresponding to the three channels. In this embodiment, the feature size of H i is 20, and X i represents the data corresponding to the three channels. For example, X 1 represents the constant current charging data, and X 1 can be [100, 98, 97,...], which is the value of the constant current charging time varying with the number of charging cycles.

[0044] Step S230: Obtain the adjacency matrices corresponding to the three channels based on the transposed matrices of the feature matrices of the three channels. The adjacency matrices include the constant current adjacency matrix, the constant voltage adjacency matrix, and the internal resistance cycle adjacency matrix.

[0045] In an embodiment of the present application, obtaining the adjacency matrices corresponding to the three channels based on the feature matrices of the three channels and their corresponding transposed matrices includes: multiplying the constant current feature matrix by the transposed matrix corresponding to the constant current feature matrix to obtain the constant current adjacency matrix; multiplying the constant voltage feature matrix by the transposed matrix corresponding to the constant voltage feature matrix to obtain the constant voltage adjacency matrix; multiplying the internal resistance cycle feature matrix by the transposed matrix corresponding to the internal resistance cycle feature matrix to obtain the internal resistance cycle adjacency matrix.

[0046] In an embodiment of the present application, multiplying the feature matrix of each channel by its corresponding transposed matrix to obtain the adjacency matrix corresponding to the channel, where A i is the adjacency matrix, and the size of the adjacency matrix can be 20×20, H i is the feature matrix with a size of 20, is the transposed matrix corresponding to H i . After obtaining the adjacency matrices and feature matrices corresponding to the three channels respectively, graph construction can be performed on the channels based on the adjacency matrix and feature matrix of each channel. Through G i =(H i , A i ) represents the three channels, where i = 1, 2, 3, H i represents the feature matrix, and A i represents the adjacency matrix. When i = 1, it represents the channel corresponding to the constant current charging data. When i = 2, it represents the channel corresponding to the constant voltage charging data. When i = 3, it represents the channel corresponding to the internal resistance cycle data.

[0047] Step S240: Concatenate the constant-current adjacency matrix, the constant-voltage adjacency matrix, and the internal-resistance cyclic adjacency matrix to obtain a fused adjacency matrix. Concatenate the constant-current feature matrix, the constant-voltage feature matrix, and the internal-resistance cyclic feature matrix to obtain a fused feature matrix. And obtain the multi-channel fused features based on the fused adjacency matrix and the fused feature matrix.

[0048] In one embodiment of the present application, concatenating the constant-current adjacency matrix, the constant-voltage adjacency matrix, and the internal-resistance cyclic adjacency matrix to obtain a fused adjacency matrix includes: taking the constant-current adjacency matrix as the first adjacency node, taking the constant-voltage adjacency matrix as the second adjacency node, and taking the internal-resistance cyclic adjacency matrix as the third adjacency node; connecting the first adjacency node, the second adjacency node, and the third adjacency node pairwise to obtain the fused adjacency matrix.

[0049] Please refer to Figure 3 , Figure 3 is a schematic diagram of the construction process of the fused adjacency matrix shown in an exemplary embodiment of the present application. In one embodiment of the present application, combine the adjacency matrices of three channels, regard each adjacency matrix as a node, and regard the adjacency matrices of three channels as three nodes. By connecting these three nodes pairwise, a 3×3 adjacency matrix can be obtained. Since each node (the adjacency matrix of each channel) is a 20×20 adjacency matrix, a fused adjacency matrix A with a size of 60×60 can be obtained through such a combination method. r Then, concatenate the feature matrices of the three channels based on this method to obtain a fused feature matrix H with a size of 60. r .

[0050] In one embodiment of the present application, obtaining the multi-channel fused features based on the fused adjacency matrix and the fused feature matrix includes: performing a symmetric normalization operation on the fused adjacency matrix, and obtaining the pre-activated multi-channel fused features based on the symmetrically normalized fused adjacency matrix, the fused feature matrix, and a preset weight matrix; activating the pre-activated multi-channel fused features based on a preset activation function to obtain the multi-channel fused features.

[0051] In one embodiment of the present application, obtaining the multi-channel fused features based on the fused adjacency matrix and the fused feature matrix includes:

[0052]

[0053] In Equation (1), H d is the multi-channel fused features, A r is the fused adjacency matrix, H r is the fused feature matrix, is the symmetric normalization operation on the fused adjacency matrix, W is the preset first weight matrix, and ReLU is the preset activation function.

[0054] In one embodiment of the present application, the weight matrix can be random, and the network model training will automatically optimize the weight matrix.

[0055] In one embodiment of the present application, through the above operations, the connection patterns and signal information of the feature nodes in three channels can be combined, and the information in different channels is fused together, so as to provide more comprehensive and richer battery life information, which helps the model better understand and represent the characteristics of battery life, and improves the prediction ability of battery life.

[0056] Step S250: Input the feature matrices of three channels into the channel interaction attention mechanism to obtain a channel interaction attention matrix.

[0057] In one embodiment of the present application, inputting the feature matrices of three channels into the channel interaction attention mechanism to obtain a channel interaction attention matrix includes: performing symmetric normalization on the adjacency matrix of each channel, obtaining the aggregated features of each channel based on the feature matrix of each channel and the symmetrically normalized adjacency matrix, and the aggregated features include the aggregated features after constant current, the aggregated features after constant voltage, and the aggregated features after internal resistance cycle; performing pairwise product interaction on the aggregated features after constant current, the aggregated features after constant voltage, and the aggregated features after internal resistance cycle to obtain a plurality of product interaction features; processing each product interaction feature through a hyperbolic tangent function and a normalized exponential function to obtain an attention coefficient vector corresponding to each product interaction feature; and obtaining a channel interaction attention matrix based on the attention coefficient vector corresponding to each product interaction feature and the plurality of product interaction features.

[0058] Please refer to Figure 4 , Figure 4 which is a schematic diagram of the channel interaction attention mechanism shown in an exemplary embodiment of the present application. In one embodiment of the present application, through the product operation of multiple channels, the interaction of important features can be amplified to emphasize their roles. By calculating the products between the features of multiple channels, the feature interactions with important influences in multiple channels can be highlighted, thereby improving the criticality of the key interactions of the model. Input the graph G i =(H i , A i ) of three channels into the channel interaction attention mechanism. First, perform graph convolution calculation on each channel to obtain the hidden spatial feature relationships within each channel, including:

[0059]

[0060] In Equation (2), i = 1, 2, 3, H ti is H t1 , H t2 and H t3 , which respectively represent the aggregated features of three channels, Ht1 is the characteristic after constant current polymerization, H t2 is the characteristic after constant voltage polymerization, H t3 is the characteristic after internal resistance cycle polymerization, ReLU is a preset activation function is the operation of symmetrically normalizing the adjacency matrix, H i is the feature matrix corresponding to each channel, W is a preset second weight matrix

[0061] In an embodiment of the present application, then pairwise product interactions are performed on the three generated aggregated features to respectively obtain the product interaction feature H of channel 1,1 11 , the product interaction feature H of channel 1,2 12 , the product interaction feature H of channel 1,3 13 , the product interaction feature H of channel 2,1 21 , the product interaction feature H of channel 2,2 22 , the product interaction feature H of channel 2,3 23 , the product interaction feature H of channel 3,1 31 , the product interaction feature H of channel 3,2 32 , the product interaction feature H of channel 3,3 33 . Among them, H 11 , H 22 , H 33 respectively represent product interactions with their own channels

[0062] In an embodiment of the present application, each product interaction feature is successively subjected to tanh (hyperbolic tangent function) and softmax (normalized exponential function) operations to obtain an attention coefficient vector (size 20) corresponding to each product interaction feature, and then each attention coefficient vector is multiplied by its transpose matrix to obtain a product interaction matrix corresponding to each product interaction feature. The product interaction matrix is of size 20×20

[0063] In an embodiment of the present application, each channel is regarded as a node, and the three nodes are connected pairwise. For example: the feature corresponding to the first channel node and the second channel node is H 12 , and in this way, a node matrix of size 3×3 can be obtained. Through the node matrix and the corresponding product interaction matrix, the channel interaction attention matrix A a is obtained

[0064] Step S260, based on the channel interaction attention matrix, enhance the fused adjacency matrix to obtain an enhanced fused adjacency matrix, and perform an aggregation operation on the enhanced fused adjacency matrix and the multi-channel fused features to obtain the target prediction feature

[0065] In an embodiment of the present application, the channel interaction attention matrix Aa Multiply with the fused adjacency matrix A r to obtain the enhanced fused adjacency matrix A h .

[0066] In an embodiment of the present application, the enhanced fused adjacency matrix and the multi-channel fused features are subjected to an aggregation operation to obtain the target prediction features, including:

[0067]

[0068] In Equation (3), H out is the target prediction feature, ReLU is a preset activation function, A h is the enhanced fused adjacency matrix, is the normalization process of the enhanced fused adjacency matrix, H d is the multi-channel fused feature, and W is a preset weight matrix.

[0069] Please refer to Figure 5 , Figure 5 which is a schematic diagram of the aggregation process of the enhanced fused adjacency matrix shown in an exemplary embodiment of the present application. In an embodiment of the present application, based on the fused adjacency matrix and the fused feature matrix, a multi-channel fused feature is obtained through graph convolution. The fused feature matrix is multiplied with the channel interaction attention matrix to obtain the enhanced fused adjacency matrix. The enhanced fused adjacency matrix and the multi-channel fused feature are again subjected to an aggregation operation through graph convolution to obtain the final output feature, that is, the target prediction feature. Through the process shown in Figure 5 , the relationship between different channels and the feature importance can be better utilized to improve the expression ability and generalization ability of the model.

[0070] Step S270, input the target prediction feature into the fully connected layer to obtain the battery remaining life prediction result.

[0071] In an embodiment of the present application, before inputting the target prediction feature into the fully connected layer to obtain the battery remaining life prediction result, it includes: obtaining historical charging data and the corresponding remaining life values of the historical charging data, inputting the historical charging data into a convolutional neural network for feature extraction, and obtaining historical prediction features based on the extracted feature matrix; using the historical prediction features as the training data set and the remaining life values corresponding to the historical prediction features as the training labels; training the fully connected layer based on the training data set and the training labels so that the fully connected layer obtains the corresponding battery remaining life prediction result based on the target prediction feature.

[0072] In an embodiment of the present application, when the target prediction feature is input into the trained fully connected layer, the fully connected layer will output the corresponding battery remaining life prediction result based on the target prediction feature.

[0073] Please refer to Figure 6 , Figure 6 which is a block diagram of a battery remaining life prediction device shown in an exemplary embodiment of the present application. This device can be applied to Figure 1 the implementation environment shown in

[0074] As Figure 6 shown, this exemplary battery remaining life prediction device includes: a charging data input module 601, a data feature extraction module 602, a calculation adjacency matrix module 603, a first feature fusion module 604, a calculation attention matrix module 605, a second feature fusion module 606, and a battery life prediction module 607.

[0075] The charging data input module 601 is used to obtain battery charging data, and the battery charging data includes constant current charging data, constant voltage charging data, and internal resistance cycle data;

[0076] The data feature extraction module 602 is used to input the battery charging data as three channels into a convolutional neural network for feature extraction to obtain corresponding feature matrices for the three channels, and the feature matrices include a constant current feature matrix, a constant voltage feature matrix, and an internal resistance cycle feature matrix;

[0077] The calculation adjacency matrix module 603 is used to obtain the corresponding adjacency matrices for the three channels based on the feature matrices of the three channels and their corresponding transposed matrices, and the adjacency matrices include a constant current adjacency matrix, a constant voltage adjacency matrix, and an internal resistance cycle adjacency matrix;

[0078] The first feature fusion module 604 is used to splice the constant current adjacency matrix, the constant voltage adjacency matrix, and the internal resistance cycle adjacency matrix to obtain a fused adjacency matrix, splice the constant current feature matrix, the constant voltage feature matrix, and the internal resistance cycle feature matrix to obtain a fused feature matrix, and obtain multi-channel fused features based on the fused adjacency matrix and the fused feature matrix;

[0079] The calculation attention matrix module 605 is used to input the feature matrices of the three channels into a channel interaction attention mechanism to obtain a channel interaction attention matrix;

[0080] The second feature fusion module 606 is used to enhance the fused adjacency matrix based on the channel interaction attention matrix to obtain an enhanced fused adjacency matrix, and perform an aggregation operation on the enhanced fused adjacency matrix and the multi-channel fused features to obtain target prediction features;

[0081] The battery life prediction module 607 is used to input the target prediction features into a fully connected layer to obtain a battery remaining life prediction result.

[0082] Figure 7 The figure shows a schematic structural diagram of a computer system of an electronic device suitable for implementing the embodiments of the present application. It should be noted that Figure 7 The shown computer system 700 of the electronic device is only an example, and should not impose any limitations on the functions and usage scope of the embodiments of the present application.

[0083] As Figure 7 shown, the computer system 700 includes a central processing unit (CPU) 701, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 702 or the program loaded from the storage section 708 into the random access memory (RAM) 703, such as executing the method described in the above embodiments. In the RAM 703, various programs and data required for system operation are also stored. The CPU 701, ROM 702, and RAM 703 are connected to each other via a bus 704. The input / output (I / O) interface 705 is also connected to the bus 704.

[0084] The following components are connected to the I / O interface 705: an input section 1206 including a keyboard, a mouse, etc.; an output section 707 including, for example, a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage section 708 including a hard disk, etc.; and a communication section 709 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the I / O interface 705 as needed. A removable medium 711, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 710 as needed so that a computer program read from it can be installed into the storage section 708 as needed.

[0085] In particular, according to an embodiment of the present application, the processes described above with reference to the flowchart can be implemented as computer software programs. For example, an embodiment of the present application includes a computer program product that includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication section 709 and / or installed from the removable medium 711. When the computer program is executed by the central processing unit (CPU) 701, various functions defined in the system of the present application are executed.

[0086] It should be noted that the computer-readable medium shown in the embodiments of the present application can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of the computer-readable storage medium may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable computer program. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted by any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0087] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present application. Among them, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the above-mentioned module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, as well as the combination of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0088] The units involved in the embodiments described in the present application can be implemented in software or in hardware, and the described units can also be provided in a processor. Among them, the names of these units do not, in some cases, constitute a limitation on the unit itself.

[0089] Another aspect of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor of a computer, the computer is caused to execute the battery remaining life prediction method as described above. The computer-readable storage medium may be included in the electronic device described in the above embodiments, or may exist separately and not be assembled into the electronic device.

[0090] Another aspect of the present application also provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computer device to execute the battery remaining life prediction method provided in the above various embodiments.

[0091] The above embodiments are only illustrative of the principles and effects of the present application, and are not used to limit the present application. Any person familiar with this technology can make modifications or changes to the above embodiments without departing from the spirit and scope of the present application. Therefore, all equivalent modifications or changes completed by those with ordinary knowledge in the technical field without departing from the spirit and technical ideas disclosed in the present application should still be covered by the claims of the present application.

Claims

1. A method for predicting remaining battery life, characterized in that: The battery remaining life prediction method comprises: Acquiring battery charging data, wherein the battery charging data includes constant current charging data, constant voltage charging data and internal resistance cycle data; Inputting the battery charging data as three channels into a convolutional neural network for feature extraction, obtaining feature matrices corresponding to the three channels, wherein the feature matrices include a constant current feature matrix, a constant voltage feature matrix, and an internal resistance cycle feature matrix; Based on the characteristic matrices of the three channels and the corresponding transposed matrices, the adjacency matrices corresponding to the three channels are obtained, wherein the adjacency matrices include a constant current adjacency matrix, a constant voltage adjacency matrix and an internal resistance cycle adjacency matrix; The constant current adjacency matrix, the constant voltage adjacency matrix and the internal resistance cycle adjacency matrix are spliced ​​to obtain a fused adjacency matrix, the constant current feature matrix, the constant voltage feature matrix and the internal resistance cycle feature matrix are spliced ​​to obtain a fused feature matrix, and multi-channel fusion features are obtained based on the fused adjacency matrix and the fused feature matrix; Input the feature matrices of the three channels into the channel interaction attention mechanism to obtain a channel interaction attention matrix; Based on the channel interaction attention matrix, the fused adjacency matrix is ​​enhanced to obtain an enhanced fused adjacency matrix, and the enhanced fused adjacency matrix and the multi-channel fusion feature are aggregated to obtain a target prediction feature; The target prediction features are input into the fully connected layer to obtain the battery remaining life prediction result.

2. The method for predicting remaining battery life according to claim 1, characterized in that: The fused adjacency matrix obtained by splicing the constant current adjacency matrix, the constant voltage adjacency matrix and the internal resistance cycle adjacency matrix comprises: The constant current adjacency matrix is ​​used as the first adjacency node, the constant voltage adjacency matrix is ​​used as the second adjacency node, and the internal resistance cycle adjacency matrix is ​​used as the third adjacency node; The first adjacent nodes, the second adjacent nodes, and the third adjacent nodes are connected in pairs to obtain the fused adjacency matrix.

3. The method for predicting remaining battery life according to claim 1, characterized in that: Based on the feature matrices of the three channels and the corresponding transposed matrices, the adjacency matrices corresponding to the three channels include: Multiplying the constant current characteristic matrix by the transposed matrix corresponding to the constant current characteristic matrix to obtain the constant current adjacency matrix; Multiplying the constant-pressure characteristic matrix by a transposed matrix corresponding to the constant-pressure characteristic matrix to obtain the constant-pressure adjacency matrix; The internal resistance cycle feature matrix is ​​multiplied by a transposed matrix corresponding to the internal resistance cycle feature matrix to obtain the internal resistance cycle adjacency matrix.

4. The method for predicting remaining battery life according to claim 1, characterized in that: Obtaining multi-channel fusion features based on the fusion adjacency matrix and the fusion feature matrix includes: Performing a symmetric normalization operation on the fused adjacency matrix, and obtaining multi-channel fusion features before activation based on the fused adjacency matrix after the symmetric normalization operation, the fusion feature matrix and a preset weight matrix; The multi-channel fusion feature before activation is activated based on a preset activation function to obtain the multi-channel fusion feature.

5. The method for predicting remaining battery life according to claim 1, characterized in that: Aggregating the enhanced fusion adjacency matrix and the multi-channel fusion features to obtain target prediction features includes: Among them, H out is the target prediction feature, ReLU is the preset activation function, A h is the enhanced fusion adjacency matrix, In order to normalize the enhanced fusion connection matrix, H d is the multi-channel fusion feature, and W is the preset weight matrix.

6. The method for predicting remaining battery life according to any one of claims 1 to 5, characterized in that: The feature matrices of the three channels are input into the channel interaction attention mechanism, and the channel interaction attention matrix is ​​obtained, including: Performing a symmetric normalization operation on the adjacency matrix of each channel, and obtaining the aggregated features of each channel based on the feature matrix of each channel and the adjacency matrix after the symmetric normalization operation, wherein the aggregated features include constant current aggregated features, constant voltage aggregated features, and internal resistance cycle aggregated features; Perform product interaction on the constant current post-polymerization characteristics, the constant voltage post-polymerization characteristics and the internal resistance cycle post-polymerization characteristics in pairs to obtain a plurality of product interaction characteristics; Processing each of the product interaction features through a hyperbolic tangent function and a normalized exponential function to obtain an attention coefficient vector corresponding to each of the product interaction features; The channel interaction attention matrix is ​​obtained based on the attention coefficient vector corresponding to each of the product interaction features and the multiple product interaction features.

7. The method for predicting remaining battery life according to any one of claims 1 to 5, characterized in that: The target prediction feature is input into the fully connected layer to obtain the battery remaining life prediction result, including: Acquire historical charging data and remaining life values ​​corresponding to the historical charging data, input the historical charging data into the convolutional neural network for feature extraction, and obtain historical prediction features based on the extracted feature matrix; Using the historical prediction features as a training data set, and using the remaining life values ​​corresponding to the historical prediction features as training labels; The fully connected layer is trained based on the training data set and the training labels so that the fully connected layer obtains the corresponding battery remaining life prediction result based on the target prediction feature.

8. A battery remaining life prediction device, characterized in that: The battery remaining life prediction device comprises: A charging data input module, used to obtain battery charging data, wherein the battery charging data includes constant current charging data, constant voltage charging data and internal resistance cycle data; A data feature extraction module, used to input the battery charging data as three channels into a convolutional neural network for feature extraction, to obtain feature matrices corresponding to the three channels, wherein the feature matrices include a constant current feature matrix, a constant voltage feature matrix, and an internal resistance cycle feature matrix; A module for calculating an adjacency matrix is ​​used to obtain adjacency matrices corresponding to the three channels based on the characteristic matrices of the three channels and the corresponding transposed matrices, wherein the adjacency matrices include a constant current adjacency matrix, a constant voltage adjacency matrix and an internal resistance cycle adjacency matrix; A first feature fusion module, used for splicing the constant current adjacency matrix, the constant voltage adjacency matrix and the internal resistance cycle adjacency matrix to obtain a fused adjacency matrix, splicing the constant current feature matrix, the constant voltage feature matrix and the internal resistance cycle feature matrix to obtain a fused feature matrix, and obtaining multi-channel fusion features based on the fused adjacency matrix and the fused feature matrix; An attention matrix calculation module is used to input the feature matrices of the three channels into the channel interaction attention mechanism to obtain a channel interaction attention matrix; A second feature fusion module is used to enhance the fused adjacency matrix based on the channel interaction attention matrix to obtain an enhanced fused adjacency matrix, and to aggregate the enhanced fused adjacency matrix and the multi-channel fusion features to obtain a target prediction feature; The battery life prediction module is used to input the target prediction features into the fully connected layer to obtain the battery remaining life prediction result.

9. An electronic device, characterized in that: The electronic device comprises: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the electronic device to implement the battery remaining life prediction method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor of a computer, the computer is caused to execute the battery remaining life prediction method as described in any one of claims 1 to 7.

Images