Battery life prediction method based on imaging single-cycle sampling amount and related device

Abstract

The invention relates to a battery life prediction method based on imaging single-cycle sampling quantity and a related device. The method comprises the following steps: sampling a battery sampling amount during a single charge-discharge cycle of a battery to obtain a plurality of single-cycle sampling amounts of the battery to form a single-cycle sampling amount sequence; segmenting the single-cycle sampling amount sequence into a plurality of sub-sequences with equal length; calculating the distance between the subsequences, and generating a sampling amount image corresponding to the single-cycle sampling amount sequence; and inputting the sampling quantity image into a battery life prediction model to obtain the expected cycle life of the battery. The method can accurately predict the remaining service life of the battery in the ultra-early stage of battery use.

Description

Technical Field

[0001] The present invention relates to the field of battery technology, and in particular to a battery life prediction method based on imaged single-cycle sampling and a related device. Background Art

[0002] With the increasing popularity of electric vehicles and renewable energy, lithium-ion batteries have gradually become the first choice for electrified transportation and energy storage systems due to their advantages such as high energy density, low self-discharge rate and long cycle life. Among them, the types of lithium-ion batteries include but are not limited to lithium iron phosphate batteries, ternary lithium batteries, lithium titanate batteries, etc. In addition, in order to meet the diverse needs of different application scenarios, other types of battery technologies such as sodium ion batteries, lithium metal batteries, semi-solid batteries, solid-state batteries, etc. are also constantly developing. However, since battery performance will gradually decline with the increase of usage time, it will lead to a series of safety hazards and economic losses. Therefore, how to accurately estimate the battery's state of health (State of Health, SOH) and predict its remaining useful life (Remaining Useful Life, RUL) has become a key technical issue to ensure operating efficiency and safety.

[0003] At present, traditional battery life prediction methods mainly include methods based on physical models and / or data-driven methods. Among them, physical models usually rely on a deep understanding of the battery degradation mechanism, but due to their large computational requirements, they cannot be efficiently applied to actual battery life predictions. Data-driven methods, especially machine learning and deep learning algorithms, require the analysis of a large amount of historical data to obtain more accurate prediction results. Once the amount of data is insufficient, the accuracy of the prediction results will be seriously affected. Therefore, it is necessary to provide a battery life prediction method and related devices based on the imaged single-cycle sampling amount. Summary of the invention

[0004] The present invention provides a battery life prediction method and related device based on image single-cycle sampling, which improves the problem that the prior art cannot accurately and quickly predict the remaining service life of the battery due to the need for a large amount of historical data or complex physical models to predict the battery life.

[0005] The present invention provides a battery life prediction method based on imaged single-cycle sampling, the method comprising: sampling the battery sampling amount during a single charge and discharge cycle of the battery to obtain multiple single-cycle sampling amounts of the battery to form a single-cycle sampling amount sequence; dividing the single-cycle sampling amount sequence into multiple subsequences of equal length; calculating the distance between each subsequence to generate a sampling amount image corresponding to the single-cycle sampling amount sequence; and inputting the sampling amount image into a battery life prediction model to obtain the expected cycle life of the battery.

[0006] In one embodiment of the present invention, the single-cycle sampling sequence is divided into multiple subsequences of equal length, including: size normalizing the length of the single-cycle sampling sequence according to a preset sequence length; performing data normalization on the size-normalized single-cycle sampling sequence; and performing sliding sampling on the data-normalized single-cycle sampling sequence according to a preset sampling length to obtain multiple subsequences of equal length.

[0007] In one embodiment of the present invention, the length of the single-cycle sampling sequence is size normalized according to a preset sequence length, including: counting the length of the single-cycle sampling sequence: if the length of the single-cycle sampling sequence is greater than the preset sequence length, extracting multiple single-cycle sampling quantities from the single-cycle sampling sequence based on a downsampling method to form a single-cycle sampling sequence of the sequence length; if the length of the single-cycle sampling sequence is less than the preset sequence length, expanding the single-cycle sampling sequence based on an interpolation method to form a single-cycle sampling sequence of the sequence length.

[0008] In one embodiment of the present invention, the single-cycle sampling amount is one, and the distance between each subsequence is calculated to generate a sampling amount image corresponding to the single-cycle sampling amount sequence, including: calculating the distance between each subsequence and generating a distance matrix; mapping each distance value in the distance matrix to a grayscale value to generate a sampling amount image corresponding to the single-cycle sampling amount sequence.

[0009] In one embodiment of the present invention, there are multiple types of single-cycle sampling amounts, and the distance between each subsequence is calculated to generate a sampling amount image corresponding to the single-cycle sampling amount sequence, including: for each single-cycle sampling amount: calculating the distance between each subsequence and generating a distance matrix; mapping each distance value in the distance matrix to a grayscale value to generate a sampling amount image corresponding to the single-cycle sampling amount sequence; stacking the sampling amount images of all single-cycle sampling amount sequences according to the channel dimension to form a multi-channel sampling amount image and using it as the sampling amount image of all single-cycle sampling amount sequences.

[0010] In one embodiment of the present invention, the battery life prediction model is AlexNet, and the battery life prediction model is one. The step of inputting the sampled image into the battery life prediction model to obtain the expected cycle life of the battery includes: inputting the sampled image into the convolutional network of the battery life prediction model, and extracting image features of the sampled image layer by layer; wherein the convolutional network includes multiple cascaded convolutional layers; and inputting the image features generated by the last convolutional layer of the convolutional network into the fully connected network of the battery life prediction model to obtain the expected cycle life of the battery.

[0011] In one embodiment of the present invention, the battery life prediction model is AlexNet, and there are multiple battery life prediction models, and the network architectures of the multiple battery life prediction models are the same but the parameters are different. The sampling image is input into the battery life prediction model to obtain the expected cycle life of the battery, including: for each battery life prediction model: inputting the sampling image into the convolutional network of the battery life prediction model, and extracting the image features of the sampling image layer by layer; wherein the convolutional network includes multiple cascaded convolutional layers; inputting the image features generated by the last convolutional layer of the convolutional network into the fully connected network of the battery life prediction model, and mapping the image features into a one-dimensional feature vector; flattening and splicing each feature vector, and inputting the spliced ​​feature vector into a fully connected layer to obtain the expected cycle life of the battery.

[0012] In one embodiment of the present invention, a battery life prediction device based on imaged single-cycle sampling is also provided, the device comprising: a data acquisition module, used to sample the battery sampling amount during a single charge and discharge cycle of the battery, obtain multiple single-cycle sampling amounts of the battery, and form a single-cycle sampling amount sequence; a sequence segmentation module, used to segment the single-cycle sampling amount sequence into multiple sub-sequences of equal length; an imaging module, used to calculate the distance between each sub-sequence, and generate a sampling amount image corresponding to the single-cycle sampling amount sequence; a battery life prediction module, used to input the sampling amount image into a battery life prediction model to obtain the expected cycle life of the battery.

[0013] In one embodiment of the present invention, an electronic device is also provided, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the electronic device implements any of the above-mentioned battery life prediction methods based on visualized single-cycle sampling.

[0014] In one embodiment of the present invention, a computer-readable storage medium is further provided, on which a computer program is stored. When the computer program is executed by a processor of a computer, the computer executes any of the above-mentioned battery life prediction methods based on graphical single-cycle sampling.

[0015] As described above, the present invention proposes a battery life prediction method and related device based on imaged single-cycle sampling, which has the following beneficial effects: the battery sampling amount is collected during a single charge and discharge cycle of the battery, a single-cycle sampling amount sequence is constructed, and it is divided into multiple subsequences of equal length. By calculating the distance between each subsequence, a sampling amount image is generated, and the battery life prediction model is called to realize the prediction of the remaining cycle life of the battery. The present invention does not need to rely on complex physical models or a large amount of historical data. It only needs to use the battery sampling amount during a single charge and discharge cycle to achieve accurate prediction of the remaining service life of the battery in any period, and can even accurately predict the remaining service life of the battery in the ultra-early stage of battery use, effectively improving the accuracy and efficiency of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic flow chart of a battery life prediction method based on imaged single-cycle sampling provided by an embodiment of the present invention;

[0017] Figure 2 Displays a current, voltage and capacity curve graph provided by an embodiment of the present invention;

[0018] Figure 3a A visualization image showing a distance matrix of voltage provided by an embodiment of the present invention;

[0019] Figure 3b Displaying a visualization image of a distance matrix of current provided by an embodiment of the present invention;

[0020] Figure 3c A visualization image showing a distance matrix of a capacity provided by an embodiment of the present invention;

[0021] Figure 4 Shows an overall flow chart provided for an embodiment of the present invention;

[0022] Figure 5 Shown is a structural block diagram of a battery life prediction device based on imaged single-cycle sampling provided by an embodiment of the present invention;

[0023] Figure 6 Shown is a structural schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0025] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0026] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0027] The present invention provides a battery life prediction method based on imaged single-cycle sampling. The battery sampling amount is collected during a single charge and discharge cycle of the battery, a single-cycle sampling amount sequence is constructed, and the sequence is divided into multiple subsequences of equal length. By calculating the distance between each subsequence, a sampling amount image is generated, and a battery life prediction model is called to predict the remaining cycle life of the battery. The present invention does not need to rely on complex physical models or a large amount of historical data. It only needs to use the battery sampling amount during a single charge and discharge cycle to accurately predict the remaining service life of the battery in the ultra-early stage of battery use, effectively improving the accuracy and efficiency of the prediction.

[0028] See also Figure 1 The battery life prediction method based on the image single cycle sampling includes the following steps:

[0029] S1. Sampling a battery sampling amount during a single charge and discharge cycle of the battery to obtain a plurality of single-cycle sampling amounts of the battery to form a single-cycle sampling amount sequence.

[0030] A single charge and discharge cycle refers to a complete cycle process in which the battery goes from a fully discharged state (i.e., the charge is 0%) to a fully charged state (i.e., the charge is 100%) through a charging process, and then discharges again to a fully discharged state. Correspondingly, a single-cycle sampling quantity refers to the battery sampling quantity of the battery during a single charge and discharge cycle. The battery sampling quantity includes but is not limited to current, voltage, and capacity, wherein the capacity refers to the cumulative charge and discharge capacity, and the capacity gradually increases in the charging state and gradually decreases in the discharging state. During a single charge and discharge cycle, the battery sampling quantity of the battery is recorded by a sampling device according to a preset sampling interval, and the battery sampling quantity is arranged in sequence according to the sampling time to generate a time-series single-cycle sampling quantity sequence. As an example, if the battery sampling quantity is current, the single-cycle current sequence may be I={I1,I2,I3,…,I n}, where I n is the current at the nth sampling point. It is understandable that the start and end time points of a single charge and discharge cycle and the corresponding sampling values ​​can be determined based on the battery sampling amount. For example, taking the battery sampling amount as voltage, the charging start point is the minimum voltage value, and the discharge end point is when the voltage drops back to the minimum voltage again. The time period between two adjacent voltage minimum values ​​and the corresponding sampling amount data are used as the sampling amount recording range of a single charge and discharge cycle. As an example, Figure 2 As shown in the figure, the starting point of a single charge and discharge cycle is time 0 and the end point is the 50th minute. During this period, charging is first performed, and the voltage gradually rises from the minimum value of 2V to the peak, and then discharging is performed, and the voltage gradually decreases from the peak to the minimum value of 2V. The change trend of current and capacity is similar to that of voltage, so it will not be described in detail here.

[0031] It should be noted that the battery types described in this application include but are not limited to lithium-ion batteries, sodium-ion batteries, lithium metal batteries, semi-solid batteries, solid-state batteries, etc., and the types of lithium-ion batteries include but are not limited to lithium iron phosphate batteries, ternary lithium batteries, lithium titanate batteries, etc. Preferably, the present application uses lithium-ion batteries. It should be further noted that, considering that there may be differences in the sampling points of different sampling amounts during the sampling period, in order to facilitate subsequent processing, after forming a single-cycle sampling amount sequence, it also includes: aligning the sampling times of different types of single-cycle sampling amounts.

[0032] S2. Divide the single-cycle sampling sequence into multiple subsequences of equal length.

[0033] For each single-cycle sampling amount, starting from the start position of the sequence, a sliding window of a preset length W is moved on the sequence according to a preset step length S to divide it into multiple subsequences of length W, as shown in formula (1):

[0034]

[0035] Where N is the total number of subsequences, L is the length of the single-cycle sampling amount sequence, W is the size of the sliding window (i.e., the number of data points included in each subsequence), and S is the step size (i.e., the size by which the sliding window moves backward each time). denotes rounding down. It should be noted that for the same single-cycle sampling amount sequence, the division method of the subsequences of the present invention can be any one of the following: 1) there is no overlap in the sampling amount between the subsequences; 2) there is an interval in the sampling amount between the subsequences; 3) there is an overlap in the sampling amount between the subsequences. Those skilled in the art can select the specific division method adaptively based on the actual performance of the battery, and it is not limited here.

[0036] Specifically, for the first division method: when S = W, that is, when the step size is the same as the size of the sliding window, the subsequences are adjacent to each other, and there is no overlap or interval in the sampling amount between them. For example, if the length L of the single-cycle sampling amount sequence is 9, the size W of the sliding window and the step size S are both 3, then the divided subsequences are [a1, a2, a3], [a4, a5, a6], [a7, a8, a9], where a i is the i-th single-cycle sampling amount.

[0037] For the second division method: when S > W, that is, when the step size is greater than the size of the sliding window, there is an interval between the subsequences. For example, if the length L of the single-cycle sampling amount sequence is 11, the size W of the sliding window and the step size S are 3 and 4 respectively, then the divided subsequences are [a1, a2, a3], [a5, a6, a7], [a9, a 10 , a 11 , where a i is the i-th single-cycle sampling amount.

[0038] For the third division method: when S < W, that is, when the step size is less than the size of the sliding window, there is an overlap in the sampling amount between the subsequences, that is, part of the sampling amount will exist in two adjacent subsequences at the same time. For example, if the length L of the single-cycle sampling amount sequence is 7, the size W of the sliding window and the step size S are 3 and 2 respectively, then the divided subsequences are [a1, a2, a3], [a3, a4, a5], [a5, a6, a7], where a i is the i-th single-cycle sampling amount.

[0039] In an embodiment of the present invention, the dividing the single-cycle sampling amount sequence into multiple equal-length subsequences includes:

[0040] Performing size normalization processing on the length of the single-cycle sampling amount sequence according to a preset sequence length;

[0041] Performing data normalization processing on the single-cycle sampling amount sequence after size normalization;

[0042] According to the preset sampling length, the single-cycle sampling sequence after data normalization is subjected to sliding sampling to obtain multiple subsequences of equal length.

[0043] Considering different types of single-cycle sampling, due to different sampling points, the length of the generated single-cycle sampling sequence may be different. In order to facilitate unified processing, each single-cycle sampling sequence needs to be size normalized to adjust each single-cycle sampling sequence to the same sequence length L. For each size-normalized single-cycle sampling sequence: normalize each single-cycle sampling sequence to a preset range, thereby eliminating dimensional differences to improve the accuracy of subsequent model processing. It is understandable that the data normalization method includes but is not limited to standardization, minimum-maximum normalization, etc., which are not limited here. The single-cycle sampling sequence after data normalization is subjected to sliding sampling according to a preset sliding window length and a preset step size, thereby dividing a single-cycle sampling sequence into several subsequences. It should be noted that when there is only one type of single-cycle sampling, the above-mentioned size normalization process is also required to make the corresponding single-cycle sampling sequence suitable for subsequent model processing.

[0044] In one embodiment of the present invention, the size normalization process is performed on the length of the single cycle sampling sequence according to the preset sequence length, including:

[0045] Count the length of a single-cycle sampling sequence:

[0046] If the length of the single-cycle sampling amount sequence is greater than a preset sequence length, extracting multiple single-cycle sampling amounts from the single-cycle sampling amount sequence based on a downsampling method to form a single-cycle sampling amount sequence of the sequence length;

[0047] If the length of the single-cycle sampling sequence is less than the preset sequence length, the single-cycle sampling sequence is expanded based on the interpolation method to form a single-cycle sampling sequence of the sequence length.

[0048] When the single-cycle sampling sequence is size-normalized, the corresponding processing methods are different due to different sizes, so it is necessary to first count the length of the single-cycle sampling sequence, where the length of the single-cycle sampling sequence refers to the total amount of single-cycle sampling contained in the sequence. If the length is exactly equal to the preset sequence length, there is no need to perform size normalization on it. If the length is greater than the preset sequence length, it is necessary to crop the sequence by downsampling, and by reducing the sequence length, the cropped sequence length is equal to the preset sequence length. If the length is less than the preset sequence length, it is necessary to add new data to the original single-cycle sampling sequence by interpolation to make it conform to the sequence length. Among them, the interpolation method includes but is not limited to bilinear interpolation, Lagrange interpolation, etc.

[0049] S3. Calculate the distance between each subsequence and generate a sampling volume image corresponding to the single-cycle sampling volume sequence.

[0050] For each single-cycle sampling amount, each subsequence is regarded as a vector, and the distance between each subsequence is calculated to generate a distance matrix, and the distance matrix is ​​regarded as the sampling amount image of the single-cycle sampling amount sequence, and the distance matrix is ​​input into the battery life prediction model to predict the remaining cycle life of the battery. It can be understood that in addition to the above method, Gram angle field, Markov transfer field and other methods can also be used to image the time series data, which is not limited here. Figure 3a , Figure 3b and Figure 3c As shown in Figure 1, it shows an example of distance matrix visualization. In the above three figures, the darker the color, the greater the distance between the corresponding two subsequences. Figure 3a It is the sampling image of voltage. The horizontal and vertical coordinates are the coordinate positions in the distance matrix respectively. The color bar on the right is the distance value of the corresponding position in the voltage distance matrix. The distance matrix is ​​intuitively displayed in the form of an image to clearly reflect the similarity between voltage subsequences. Figure 3b and Figure 3c They are the sampling images of current and capacity respectively. The specific information is similar to that of the sampling image of voltage, so they will not be described in detail here.

[0051] Specifically, in one embodiment of the present invention, the single-cycle sampling amount is one type, and the step of calculating the distance between each subsequence to generate a sampling amount image corresponding to the single-cycle sampling amount sequence includes:

[0052] Calculate the distance between each subsequence and generate a distance matrix;

[0053] Each distance value in the distance matrix is ​​mapped to a grayscale value to generate a sampling volume image corresponding to a single-cycle sampling volume sequence.

[0054] For each single-cycle sampling sequence: first, divide it into multiple subsequences according to the above method, regard each subsequence as a W-dimensional vector (W is the total number of samples contained in the subsequence), and calculate the distance between each subsequence and the remaining subsequences, as shown in formula (2):

[0055] F i,j =‖p i -p j ‖ (2)

[0056] Among them, p i and p j are the ith and jth subsequences respectively, F i,j is the distance between the ith subsequence and the jth subsequence. After calculating the distances between all subsequences in the above manner, a distance matrix is ​​formed. Among them, the rows and columns of the distance matrix correspond to each subsequence, and the elements in the matrix represent the distances between the subsequences of the corresponding rows and columns. It should be noted that since the distance metric has symmetry, the distance matrix is ​​a symmetric matrix, and the elements on the diagonal of the matrix represent the distance between the subsequence and itself, so the value is zero. Each distance value in the distance matrix is ​​normalized, so as to be mapped to the range of grayscale values ​​(i.e., 0 to 255). The normalized distance matrix is ​​used as the pixel matrix of the image, and each normalized distance value is a pixel point in the image, so that the distance matrix can be regarded as a grayscale sample image. As one example, the distance value can be mapped to the grayscale value in the following manner: the minimum distance value and the maximum distance value in the distance matrix are respectively corresponding to the grayscale value 0 and the grayscale value 255, and then the other distance values ​​are linearly mapped to obtain the corresponding grayscale value.

[0057] It is understandable that in the present invention, the distance calculation method between two subsequences can use Euclidean distance, or other distance calculation methods such as Manhattan distance, Chebyshev distance, Minkowski distance, etc. Those skilled in the art can select them based on the characteristics of the battery sampling amount, and there is no limitation here.

[0058] In one embodiment of the present invention, there are multiple single-cycle sampling amounts, and the step of calculating the distances between the subsequences to generate a sampling amount image corresponding to the single-cycle sampling amount sequence includes:

[0059] For each single cycle sampling volume:

[0060] Calculate the distance between each subsequence and generate a distance matrix;

[0061] Mapping each distance value in the distance matrix into a grayscale value to generate a sampling volume image corresponding to a single-cycle sampling volume sequence;

[0062] The sampling volume images of all single-cycle sampling volume sequences are stacked according to the channel dimension to form a multi-channel sampling volume image and used as the sampling volume image of all single-cycle sampling volume sequences.

[0063] If there are multiple types of single-cycle sampling, for each type of single-cycle sampling, after generating the corresponding sampling image according to the above method, based on the pre-set rules, each sampling image corresponds to a channel, and multiple single-channel grayscale images are stacked together to form a multi-channel comprehensive sampling image. Through this multi-channel approach, the battery life prediction model can learn the relationship between different types of sampling, thereby improving the accuracy of subsequent life prediction.

[0064] S4. Inputting the sample amount image into a battery life prediction model to obtain the expected cycle life of the battery.

[0065] The sampling image is input into the battery life prediction model, and features in the sampling image with a high correlation with the cycle life of the battery are identified by extracting features from the sampling image, thereby obtaining the remaining cycle life of the battery. The battery life prediction model is pre-trained and is a CNN architecture, including but not limited to AlexNet, VGG, ResNet, etc.

[0066] In one embodiment of the present invention, the battery life prediction model is AlexNet, the battery life prediction model is one, and the step of inputting the sample amount image into the battery life prediction model to obtain the expected cycle life of the battery includes:

[0067] Inputting the sample amount image into the convolutional network of the battery life prediction model, and extracting image features of the sample amount image layer by layer; wherein the convolutional network includes a plurality of cascaded convolutional layers;

[0068] The image features generated by the last convolutional layer of the convolutional network are input into the fully connected network of the battery life prediction model to obtain the expected cycle life of the battery.

[0069] The sample amount image generated above is input into the AlexNet model. AlexNet has five cascaded convolutional layers and three fully connected layers, and the five cascaded convolutional layers constitute a convolutional network. Specifically, the sample amount image is input into the convolutional network, and the shallow image features are extracted in the first convolutional layer. The subsequent convolutional layers integrate the features extracted by the previous convolutional layer and extract the deep image features by reducing the size of the convolution kernel. The deep image features generated by the last convolutional layer of the convolutional network are input into the fully connected layer, and the deep image features are weighted combined and nonlinearly mapped, so as to map them into a one-dimensional feature vector, and through the output layer, the expected cycle life of the battery is obtained. It can be understood that the sample amount image described in this embodiment can be either a single-channel sample amount image or a multi-channel sample amount image, which is not limited here.

[0070] In one embodiment of the present invention, the battery life prediction model is AlexNet, there are multiple battery life prediction models, the network architectures of the multiple battery life prediction models are the same but the parameters are different, and the inputting of the sample amount image into the battery life prediction model to obtain the expected cycle life of the battery includes:

[0071] For each battery life prediction model:

[0072] Inputting the sample amount image into the convolutional network of the battery life prediction model, and extracting image features of the sample amount image layer by layer; wherein the convolutional network includes a plurality of cascaded convolutional layers;

[0073] Inputting the image features generated by the last convolutional layer of the convolutional network into the fully connected network of the battery life prediction model, and mapping the image features into a one-dimensional feature vector;

[0074] Each feature vector is flattened and concatenated, and the concatenated feature vector is input into a fully connected layer to obtain the expected cycle life of the battery.

[0075] In order to improve the accuracy of the prediction, in this embodiment, by using multiple AlexNet battery life prediction models with the same network architecture but different sampling amounts, the sampling amount image is processed to obtain the expected cycle life of the battery. Specifically, the same sampling amount image is input into each battery life prediction model. For each battery life prediction model: through a convolutional network composed of multiple cascaded convolutional layers, the features of the sampling amount image are extracted layer by layer, and the image features generated by the last convolutional layer are input into the fully connected network of the model to generate a one-dimensional feature vector. After all battery life prediction models are processed in the above manner, the feature vectors output by all battery life prediction models are flattened and spliced ​​together to form a comprehensive feature vector, and the spliced ​​comprehensive feature vector is input into a fully connected layer to obtain the expected cycle life of the battery. Since this spliced ​​feature vector contains feature information extracted by multiple models at the same time, it can more comprehensively and accurately reflect the relationship between the sampling amount image and the battery life.

[0076] As a specific example, Figure 4As shown, the categories of single-cycle sampling are current, voltage and capacity, so the sampling image input to the model is a three-channel image, and the battery life prediction model is AlexNet. After collecting the current, voltage and capacity during a single charge and discharge from the battery, constructing the single-cycle voltage sequence, single-cycle current sequence and single-cycle capacity sequence, the preprocessed single-cycle voltage sequence, single-cycle current sequence and single-cycle capacity sequence are obtained through preprocessing processes such as size normalization and data standardization. According to the preset sliding window size, each time series sequence is sampled according to the preset step size, and each sequence is divided into multiple subsequences of equal length. For each single-cycle sampling amount: by calculating the distance between each pair of subsequences, the corresponding N*N distance matrix (N is the total number of subsequences in the single-cycle sampling amount sequence) is generated, and the distance matrix is ​​regarded as the corresponding sampling amount image. For current, voltage and capacity, the current image, voltage image and capacity image are generated accordingly. The three images are stacked to the corresponding channels to form a three-channel image and adjusted to a uniform size (such as 224). The adjusted three-channel images are respectively input into three trained battery life prediction models (such as AlexNet), and a feature vector of a preset scale (such as 1000) is generated by the first battery life prediction model AlexNet, the second battery life prediction model AlexNet2 and the third battery life prediction model AlexNet3. These three feature vectors are spliced ​​together to form a comprehensive feature vector with a scale of 3000, and the remaining cycle life of the battery is obtained accordingly. Experiments show that when only a single cycle data is used, the average absolute percentage error of the cycle life prediction of the present invention for the public battery data set used for model evaluation is within 7.1%. It can be seen that the use of this method of the present invention can show excellent results in the prediction of the ultra-early stage of battery life, not only can high-precision prediction of battery life be achieved, but also significantly reduce the dependence on large-scale historical data and complex physical models, so this method of the present invention has strong promotion and practicality.

[0077] See also Figure 5 The battery life prediction device 100 based on the imaged single-cycle sampling amount includes: a data acquisition module 110, a sequence segmentation module 120, an imaging module 130 and a battery life prediction module 140. The above-mentioned data acquisition module 110 is used to sample the battery sampling amount during a single charge and discharge cycle of the battery, obtain multiple single-cycle sampling amounts of the battery, and form a single-cycle sampling amount sequence. The sequence segmentation module 120 is used to divide the single-cycle sampling amount sequence into multiple subsequences of equal length. The imaging module 130 is used to calculate the distance between each subsequence and generate a sampling amount image corresponding to the single-cycle sampling amount sequence. The battery life prediction module 140 is used to input the sampling amount image into the battery life prediction model to obtain the expected cycle life of the battery.

[0078] For the specific limitations of the battery life prediction device based on the imaged single-cycle sampling amount, please refer to the limitations of the battery life prediction method based on the imaged single-cycle sampling amount above, which will not be repeated here. Each module in the above-mentioned battery life prediction device based on the imaged single-cycle sampling amount can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in hardware format, or can be stored in the memory of the computer device in software format, so that the processor can call the operations corresponding to the above modules.

[0079] It should be noted that, in order to highlight the innovative part of the present invention, the present embodiment does not introduce modules that are not closely related to solving the technical problem proposed by the present invention, but this does not mean that there are no other modules in the present embodiment.

[0080] See also Figure 6 The electronic device 1 may include a memory 12, a processor 13 and a bus, and may also include a computer program stored in the memory 12 and executable on the processor 13, such as a battery life prediction program based on graphical single-cycle sampling.

[0081] Among them, the memory 12 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 12 can be an internal storage unit of the electronic device 1, such as a mobile hard disk of the electronic device 1. In other embodiments, the memory 12 can also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 12 can also include both an internal storage unit of the electronic device 1 and an external storage device. The memory 12 can not only be used to store application software and various types of data installed in the electronic device 1, such as the code for battery life prediction based on the imaged single-cycle sampling amount, but also can be used to temporarily store data that has been output or is to be output.

[0082] In some embodiments, the processor 13 may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips. The processor 13 is the control core (Control Unit) of the electronic device 1, and uses various interfaces and lines to connect various components of the entire electronic device 1, and executes or runs programs or modules stored in the memory 12 (for example, a battery life prediction program based on an imaged single-cycle sampling amount, etc.), and calls data stored in the memory 12 to execute various functions of the electronic device 1 and process data.

[0083] The processor 13 executes the operating system and various installed applications of the electronic device 1. The processor 13 executes the applications to implement the steps in the above-mentioned battery life prediction method based on the imaged single-cycle sampling amount.

[0084] Exemplarily, the computer program may be divided into one or more modules, which are stored in the memory 12 and executed by the processor 13 to complete the present application. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the electronic device 1. For example, the computer program may be divided into a data acquisition module 110, a sequence segmentation module 120, an imaging module 130, and a battery life prediction module 140.

[0085] The above-mentioned integrated unit implemented in the form of a software function module can be stored in a computer-readable storage medium, and the computer-readable storage medium can be non-volatile or volatile. The above-mentioned software function module is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a computer device, or a network device, etc.) or a processor to perform part of the functions of the battery life prediction method based on the imaged single-cycle sampling amount described in each embodiment of the present application.

[0086] In summary, the present invention discloses a battery life prediction method and related device based on imaged single-cycle sampling, which collects battery sampling during a single charge and discharge cycle of the battery, constructs a single-cycle sampling sequence, and divides it into multiple equal-length subsequences. By calculating the distance between each subsequence, a sampling image is generated, and the battery life prediction model is called to predict the remaining cycle life of the battery. The present invention does not need to rely on complex physical models or a large amount of historical data. It only needs to use the battery sampling during a single charge and discharge cycle to accurately predict the remaining service life of the battery in the ultra-early stage of battery use, effectively improving the accuracy and efficiency of the prediction. Compared with the traditional method, the present invention greatly simplifies the data collection and processing process, and cleverly establishes a connection between the battery sampling and the image, analyzes the imaged battery sampling through a deep learning model, and realizes an accurate prediction of the expected cycle life of the battery. Therefore, the present invention effectively overcomes the various shortcomings of the prior art and has a high industrial utilization value.

[0087] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A battery life prediction method based on image-based single-cycle sampling, characterized in that: The method comprises: Sampling a battery sampling amount during a single charge and discharge cycle of the battery to obtain a plurality of single-cycle sampling amounts of the battery to form a single-cycle sampling amount sequence; Split the single-cycle sampling sequence into multiple subsequences of equal length; The distances between the subsequences are calculated to generate a sampling image corresponding to the single-cycle sampling sequence; and the sampling image is input into a battery life prediction model to obtain the expected cycle life of the battery.

2. The battery life prediction method based on imaged single-cycle sampling according to claim 1, characterized in that: The step of dividing the single cycle sampling sequence into a plurality of subsequences of equal length includes: According to the preset sequence length, the length of the single-cycle sampling sequence is normalized; Perform data normalization on the size-normalized single-cycle sampling sequence; According to the preset sampling length, the single-cycle sampling sequence after data normalization is subjected to sliding sampling to obtain multiple subsequences of equal length.

3. The battery life prediction method based on imaged single-cycle sampling according to claim 2, characterized in that: The process of normalizing the length of the single-cycle sampling sequence according to the preset sequence length includes: Count the length of a single-cycle sampling sequence: If the length of the single-cycle sampling sequence is greater than a preset sequence length, extracting multiple single-cycle sampling amounts from the single-cycle sampling sequence based on a downsampling method to form a single-cycle sampling sequence of the sequence length; If the length of the single-cycle sampling sequence is less than the preset sequence length, the single-cycle sampling sequence is expanded based on the interpolation method to form a single-cycle sampling sequence of the sequence length.

4. The battery life prediction method based on imaged single-cycle sampling according to claim 1, characterized in that: The single-cycle sampling amount is one type, and the distance between each subsequence is calculated to generate a sampling amount image corresponding to the single-cycle sampling amount sequence, including: Calculate the distance between each subsequence and generate a distance matrix; Each distance value in the distance matrix is ​​mapped to a grayscale value to generate a sampling volume image corresponding to a single-cycle sampling volume sequence.

5. The battery life prediction method based on imaged single-cycle sampling according to claim 1, characterized in that: There are multiple single-cycle sampling amounts. The distance between each subsequence is calculated to generate a sampling amount image corresponding to the single-cycle sampling amount sequence, including: For each single cycle sampling volume Calculate the distance between each subsequence and generate a distance matrix; Mapping each distance value in the distance matrix into a grayscale value to generate a sampling image corresponding to a single-cycle sampling sequence; The sampling volume images of all single-cycle sampling volume sequences are stacked according to the channel dimension to form a multi-channel sampling volume image and used as the sampling volume image of all single-cycle sampling volume sequences.

6. The battery life prediction method based on imaged single-cycle sampling according to claim 1, characterized in that: The battery life prediction model is AlexNet, and the battery life prediction model is one. Inputting the sample amount image into the battery life prediction model to obtain the expected cycle life of the battery includes: Inputting the sample image into the convolutional network of the battery life prediction model to extract image features of the sample image layer by layer; wherein the convolutional network includes a plurality of cascaded convolutional layers; The image features generated by the last convolutional layer of the convolutional network are input into the fully connected network of the battery life prediction model to obtain the expected cycle life of the battery.

7. The battery life prediction method based on imaged single-cycle sampling according to claim 1, characterized in that: The battery life prediction model is AlexNet, and there are multiple battery life prediction models, each having the same network architecture but different parameters. Inputting the sample image into the battery life prediction model to obtain the expected cycle life of the battery includes: For each battery life prediction model Inputting the sample image into the convolutional network of the battery life prediction model to extract image features of the sample image layer by layer; wherein the convolutional network includes a plurality of cascaded convolutional layers; Inputting the image features extracted by the last convolutional layer of the convolutional network into the fully connected network of the battery life prediction model, and mapping the image features into a one-dimensional feature vector; Each feature vector is flattened and concatenated, and the concatenated feature vector is input into a fully connected layer to obtain the expected cycle life of the battery.

8. A battery life prediction device based on image-based single-cycle sampling, characterized in that: The device comprises: a data acquisition module, configured to sample a battery sample amount during a single charge and discharge cycle of the battery, obtain a plurality of single-cycle sample amounts of the battery, and form a single-cycle sample amount sequence; A sequence segmentation module is used to segment a single cycle sampling sequence into multiple subsequences of equal length; An imaging module is used to calculate the distance between each subsequence and generate a sampling image corresponding to a single-cycle sampling sequence; A battery life prediction module is used to input the sample amount image into a battery life prediction model to obtain the expected cycle life of the battery.

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 life prediction method based on the graphical single-cycle sampling amount as described in any one of claims 1 to 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 life prediction method based on the graphical single-cycle sampling amount as claimed in any one of claims 1 to 7.

Images