Lithium ion battery health state evaluation method and device based on image recognition

Abstract

The invention discloses a lithium ion battery health state evaluation method and device based on image recognition, and the method comprises the steps: obtaining a lithium ion battery data set through a set database, and carrying out the preliminary arrangement, and obtaining an arranged data set; importing the sorted data set into MATLAB for analysis and processing through a set function to obtain a sample data set; determining a voltage interval of constant-current charging voltage data sampling of the lithium battery through a capacity increment analysis strategy; sampling is carried out from the sample data set according to the voltage interval, and processing is carried out through a Gramb angle field strategy to generate a Gramb angle field graph; constructing a battery health state evaluation model based on an improved ResNet34 network, and training to obtain a trained battery health state evaluation model; and inputting the target battery data into the trained battery health state evaluation model for evaluation, and outputting a target battery health state evaluation result. The accuracy of battery health state evaluation can be improved, and the calculation cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of battery health status assessment, and in particular to a lithium-ion battery health status assessment method and device based on image recognition. Background Art

[0002] With the advancement of industrialization and urbanization, my country's demand for energy continues to grow. In this process, the rapid development of fields such as electric vehicles and energy storage systems continues to drive the demand for efficient and reliable energy storage solutions, especially the sharp increase in demand for lithium-ion batteries. In response to this demand, the health status estimation technology of lithium batteries has become extremely important because it is directly related to the operating efficiency and life safety of the battery system.

[0003] At the technical level, with the integration of information technology and new energy technology, the use of advanced methods such as deep learning and big data analysis to monitor and manage the health status of batteries has become mainstream. These technologies can not only improve the accuracy of battery status monitoring and prediction, but also provide support for battery performance optimization and life extension, so as to better meet the needs of industry and the market. At present, the methods for estimating the health status of lithium-ion batteries mainly include direct measurement, model-based methods, and data-driven methods. However, these methods all have some disadvantages: the disadvantage of the direct measurement method is that it may be necessary to interrupt the normal use of the battery for measurement, and the measurement results are easily affected by external factors such as battery temperature and usage status; the model-based method usually requires complex parameter calibration and high computational costs; the performance of the data-driven method strongly depends on the quality and quantity of the data, and the transparency and interpretability of the model are low. In previous studies, a research team used a one-dimensional convolutional neural network to process the charging voltage time series data of lithium-ion batteries to estimate the health status of lithium batteries. Although the effectiveness of one-dimensional convolutional neural networks in processing such time series data has been verified to a certain extent, its feature extraction ability is still limited. In addition, another research team tried to convert one-dimensional data into two-dimensional images for health status estimation, but the results were not ideal because the data processing method was too simple.

[0004] Therefore, how to invent a lithium-ion battery health status assessment method based on image recognition to improve the accuracy of battery health status assessment and reduce computing costs has become an urgent problem to be solved. Summary of the invention

[0005] To this end, the present invention provides a lithium-ion battery health status assessment method and device based on image recognition, which selects the key voltage segment in the battery charging process based on incremental capacity analysis (ICA) and Gram Angle Field (GASF) technology, and converts the time series data into a two-dimensional image. Combined with the image recognition deep learning algorithm, the health status of the lithium-ion battery can be accurately estimated.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a lithium-ion battery health status assessment method based on image recognition, comprising:

[0007] Acquire a lithium-ion battery data set by setting a database; preliminarily organize the lithium-ion battery data set to obtain an organized data set; import the organized data set into MATLAB by setting a function for analysis and processing to obtain a sample data set;

[0008] Determine the voltage range for sampling the constant current charging voltage data of the lithium battery through the capacity increment analysis strategy;

[0009] Sampling from the sample data set according to the voltage interval to obtain sampled data; processing the sampled data using a Gram angle field strategy to generate a Gram angle field map;

[0010] Constructing a battery health status assessment model based on an improved ResNet34 network; training the battery health status assessment model by setting a loss function and setting an optimizer to obtain the trained battery health status assessment model;

[0011] The target battery data is input into the trained battery health status assessment model for assessment, and the target battery health status assessment result is output.

[0012] As a preferred solution of the lithium-ion battery health status assessment method based on image recognition, in the process of importing the sorted data set into MATLAB for analysis and processing through the setting function to obtain the sample data set, the sorted data set is imported into MATLAB through the readtable or readmatrix function; the sorted data set is filtered and denoised to obtain a high-quality data set; and the high-quality data set is processed through MATLAB to obtain the sample data set.

[0013] As a preferred solution of the lithium-ion battery health status assessment method based on image recognition, in the process of determining the voltage interval of the constant current charging voltage data sampling of the lithium battery through the capacity increment analysis strategy, the battery terminal voltage is differentiated relative to the capacity through the capacity increment analysis strategy to obtain the voltage interval of the constant current charging voltage data sampling of the lithium battery; the differential processing calculation formula is:

[0014]

[0015] Where Q represents the battery charge; V represents the battery voltage; k is the sampling time; and L is the step size of the difference.

[0016] As a preferred solution of the lithium-ion battery health status assessment method based on image recognition, in the process of processing the sampled data by the Gram angular field strategy to generate the Gram angular field map, the sampled data is normalized to obtain normalized data; the normalized data is converted by an inverse cosine function to obtain angled data; the angled data is calculated by trigonometric functions and angle formulas to quantify the correlation between each pair of time points to generate the Gram angular field map;

[0017] The mathematical expression of the normalization process is:

[0018]

[0019] In the formula, is the value obtained after normalization of the i-th sampling data; x i is the i-th sampling voltage data; x max is the maximum value in the sampled data sequence; x min is the minimum value in the sampled data sequence;

[0020] The conversion expression of the arccosine function is:

[0021]

[0022] In the formula, r i is the normalized time index; φ i is the angle value of the i-th sampled data after arccosine mapping; N is the length of the sampled data sequence;

[0023] The formula for calculating the angled data by trigonometric functions and angle formulas is:

[0024]

[0025] In the formula, φ i is the angle value of the i-th sampling data after arc cosine mapping.

[0026] As a preferred solution of the lithium-ion battery health status assessment method based on image recognition, in the process of training the battery health status assessment model through the set loss function and the set optimizer, the set loss function is a mean square error loss function; the set optimizer is a Rectified Adam optimizer.

[0027] The present invention also provides a lithium-ion battery health status assessment device based on image recognition, based on the above lithium-ion battery health status assessment method based on image recognition, comprising:

[0028] The sample data set acquisition module is used to acquire a lithium-ion battery data set by setting a database; preliminarily organize the lithium-ion battery data set to obtain an organized data set; and import the organized data set into MATLAB for analysis and processing by setting a function to obtain a sample data set;

[0029] A sampling voltage interval determination module is used to determine the voltage interval for sampling the constant current charging voltage data of the lithium battery through a capacity increment analysis strategy;

[0030] A Gram angle field diagram generating module is used to sample from the sample data set according to the voltage interval to obtain sampled data; and to process the sampled data using a Gram angle field strategy to generate a Gram angle field diagram;

[0031] A battery health status assessment model construction and training module is used to construct a battery health status assessment model based on an improved ResNet34 network; the battery health status assessment model is trained by setting a loss function and setting an optimizer to obtain the trained battery health status assessment model;

[0032] The battery health status assessment model processing module is used to input the target battery data into the trained battery health status assessment model for assessment and output the target battery health status assessment result.

[0033] As a preferred solution of the lithium-ion battery health status assessment device based on image recognition, in the sample data set acquisition module, in the process of importing the sorted data set into MATLAB for analysis and processing through the setting function to obtain the sample data set, the sorted data set is imported into MATLAB through the readtable or readmatrix function; the sorted data set is filtered and denoised to obtain a high-quality data set; and the high-quality data set is processed through MATLAB to obtain the sample data set.

[0034] As a preferred solution of the lithium-ion battery health status assessment device based on image recognition, in the sampling voltage interval determination module, in the process of determining the voltage interval of the constant current charging voltage data sampling of the lithium battery through the capacity increment analysis strategy, the battery terminal voltage is differentiated relative to the capacity through the capacity increment analysis strategy to obtain the voltage interval of the constant current charging voltage data sampling of the lithium battery; the differential processing calculation formula is:

[0035]

[0036] Where Q represents the battery charge; V represents the battery voltage; k is the sampling time; and L is the step size of the difference.

[0037] As a preferred solution of the lithium-ion battery health status assessment device based on image recognition, in the Gram angular field map generation module, in the process of processing the sampled data by the Gram angular field strategy to generate the Gram angular field map, the sampled data is normalized to obtain normalized data; the normalized data is converted by an inverse cosine function to obtain angled data; the angled data is calculated by trigonometric functions and angle formulas to quantify the correlation between each pair of time points to generate the Gram angular field map;

[0038] The mathematical expression of the normalization process is:

[0039]

[0040] In the formula, is the value obtained after normalization of the i-th sampling data; x i is the i-th sampling voltage data; x max is the maximum value in the sampled data sequence; x min is the minimum value in the sampled data sequence;

[0041] The conversion expression of the arccosine function is:

[0042]

[0043] In the formula, r i is the normalized time index; φ i is the angle value of the i-th sampled data after arccosine mapping; N is the length of the sampled data sequence;

[0044] The formula for calculating the angled data by trigonometric functions and angle formulas is:

[0045]

[0046] In the formula, φ i is the angle value of the i-th sampling data after arc cosine mapping.

[0047] As a preferred solution of the lithium-ion battery health status assessment device based on image recognition, in the battery health status assessment model construction and training module, in the process of training the battery health status assessment model through the set loss function and the set optimizer, the set loss function is a mean square error loss function; the set optimizer is a RectifiedAdam optimizer.

[0048] The present invention has the following advantages: the present invention obtains a lithium-ion battery data set by setting a database; the lithium-ion battery data set is preliminarily sorted to obtain a sorted data set; the sorted data set is imported into MATLAB by setting a function for analysis and processing to obtain a sample data set; the voltage interval for sampling the constant current charging voltage data of the lithium battery is determined by a capacity increment analysis strategy; sampling is performed from the sample data set according to the voltage interval to obtain sampling data; the sampling data is processed by a Gram angle field strategy to generate a Gram angle field diagram; a battery health state assessment model is constructed based on an improved ResNet34 network; the battery health state assessment model is trained by setting a loss function and setting an optimizer to obtain a trained battery health state assessment model; the target battery data is input into the trained battery health state assessment model for evaluation, and the target battery health state assessment result is output. The present invention utilizes the two-dimensional image conversion of the GASF technology to enhance the spatial characteristic expression of the data, so that the ResNet34 model can more accurately capture the subtle changes in the battery aging process, and the estimation result is highly consistent with the actual battery performance degradation. The present invention combines the voltage segment division of ICA technology and the image recognition capability of ResNet34, which greatly improves the accuracy of SOH estimation. Compared with traditional methods, it shows obvious performance advantages and is suitable for complex and dynamically changing practical application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0050] The structures, proportions, sizes, etc. illustrated in this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with the technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantial technical significance. Any structural modification, change in proportion or adjustment of size shall still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0051] Figure 1 This is a schematic diagram of the flow of the lithium-ion battery health status assessment method based on image recognition provided in Example 1 of the present invention;

[0052] Figure 2A schematic diagram of the charging voltage curve of CS2_35 in all cycles in the lithium-ion battery health status assessment method based on image recognition provided in Example 1 of the present invention;

[0053] Figure 3 This is a schematic diagram of the ICA curve of CS2_35 in all cycles in the lithium-ion battery health status assessment method based on image recognition provided in Example 1 of the present invention;

[0054] Figure 4 A schematic diagram of a GASF image generated by CS2_35 at charging voltage segments (1, 173, 427, 641) in different cycles in the lithium-ion battery health status assessment method based on image recognition provided in Example 1 of the present invention;

[0055] Figure 5 This is a schematic diagram of the modified ResNet34 structure in the lithium-ion battery health status assessment method based on image recognition provided in Example 1 of the present invention;

[0056] Figure 6 A schematic diagram comparing the accuracy of health status estimation using and not using GASF two-dimensional representation in a possible embodiment provided in Embodiment 1 of the present invention;

[0057] Figure 7 A schematic diagram showing a comparison of absolute errors of health status estimation results using and not using GASF two-dimensional representation in a possible embodiment provided in Embodiment 1 of the present invention;

[0058] Figure 8 A schematic diagram showing a comparison of the health status estimation accuracy of the present invention and the existing mainstream method in a possible embodiment provided in Embodiment 1 of the present invention;

[0059] Fig. 9 A schematic diagram showing a comparison of absolute errors of health status estimation results of the present invention and the existing mainstream method in a possible embodiment provided in Embodiment 1 of the present invention;

[0060] Fig.10 This is a schematic diagram of the architecture of a lithium-ion battery health status assessment device based on image recognition provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0061] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] Example 1

[0063] See also Figure 1 Embodiment 1 of the present invention provides a method for evaluating the health status of a lithium-ion battery based on image recognition, comprising the following steps:

[0064] S1. Acquire a lithium-ion battery data set by setting a database; preliminarily organize the lithium-ion battery data set to obtain an organized data set; import the organized data set into MATLAB by setting a function for analysis and processing to obtain a sample data set;

[0065] S2. Determine the voltage range for sampling the constant current charging voltage data of the lithium battery through the capacity increment analysis strategy;

[0066] S3, sampling from the sample data set according to the voltage interval to obtain sampled data; processing the sampled data by using the Gram angle field strategy to generate a Gram angle field map;

[0067] S4. Construct a battery health status assessment model based on the improved ResNet34 network; train the battery health status assessment model by setting a loss function and setting an optimizer to obtain the trained battery health status assessment model;

[0068] S5. Input the target battery data into the trained battery health status assessment model for assessment, and output the target battery health status assessment result.

[0069] In this embodiment, in step S1, a lithium-ion battery data set is obtained by setting a database; the lithium-ion battery data set is preliminarily sorted to obtain a sorted data set; the sorted data set is imported into MATLAB by setting a function for analysis and processing to obtain a sample data set;

[0070] Specifically, as of now, many laboratories in the world have provided authoritative data sets for the entire life cycle of lithium-ion batteries. The example data set used in this embodiment adopts the battery degradation data set of the University of Maryland provided by the CALCE (Center for Advanced Life Cycle Engineering) of the university. It mainly conducts system testing on two lithium-ion batteries - CS2 (rated capacity 1.10Ah) and CX2 (rated capacity 1.35Ah) - at room temperature, including information such as charge and discharge behavior and internal resistance, and regards the battery capacity dropping to about 80% of the initial capacity as the end of life.

[0071] After data acquisition, the first thing to do is preliminary data processing, including downloading the data file and opening and preliminary inspection using data processing software (such as Excel or Python). The purpose of this stage is to identify and remove possible outliers or missing data to ensure the accuracy of subsequent analysis. After dealing with the basic issues, the next task is data organization, including extracting key parameters such as voltage, current, and temperature during battery charging and discharging. At the same time, it is also necessary to select representative data periods for analysis in order to better observe and understand the attenuation trend of battery performance over time.

[0072] Data analysis in the MATLAB environment is the core of this process. First, use the readtable or readmatrix function to import the processed data into MATLAB. Next, the data is further preprocessed, including filtering and denoising, to improve data quality. After that, using MATLAB's powerful data processing and visualization capabilities, the battery charging curve is plotted to analyze and observe the changes in the time required for the voltage to rise to 4.2V during different charge and discharge cycles. Figure 2 As shown in the figure, the constant current charging curve of CS2_35 in the data set is shown throughout its life cycle. The vertical axis is voltage and the horizontal axis is time. The color of the curve changes from bright to dark to indicate the increase in the number of cycles. As the number of cycles increases, the charging voltage curve shifts to the upper left as a whole, and the time required for the voltage to rise to 4.2V is also shortened.

[0073] Ultimately, these detailed analysis results will be used to evaluate the health status of the battery. These analyses not only provide in-depth insights into the battery degradation mechanism, but also provide scientific decision support for battery management and maintenance. Throughout the analysis process, MATLAB graphics and visualization tools are widely used to generate intuitive charts and images. These visual outputs not only enhance the expression of technical reports, but also effectively support project presentations and results discussions.

[0074] In this embodiment, in step S2, the voltage interval for sampling the constant current charging voltage data of the lithium battery is determined by a capacity increment analysis strategy;

[0075] Specifically, in order to find out the charging segment that is most significantly affected by battery aging, in order to determine the sampling scheme for the charging curve, the battery terminal voltage is differentiated relative to the capacity (dQ / dV) using the incremental capacity analysis (ICA) strategy to determine the voltage range for sampling the constant current charging voltage data of the lithium battery; the differential processing calculation formula is:

[0076]

[0077] Where Q represents the battery charge; V represents the battery voltage; k is the sampling time; and L is the step size of the difference.

[0078] In this embodiment, in the Oxford dataset, Figure 3 As shown in the figure, in the voltage range of [3.8-4.0] V, the differential of capacity to voltage has an obvious peak. This interval is a key indicator of battery aging characteristics. In this interval, 64 points of data are sampled every 15 seconds as health characteristics. This precise sampling strategy can effectively capture the behavior of the battery at the critical voltage stage and provide important data for evaluating the health status of the battery.

[0079] Data noise is a common challenge when applying ICA. To overcome this problem, Kalman filtering technology is used to process data in the analysis, which effectively reduces the impact of noise and improves the accuracy of data analysis. Through this method, the aging characteristics of the battery in a specific voltage range can be observed, and then the relationship between the battery state of health (SOH) can be quantitatively analyzed.

[0080] This analysis not only provides in-depth insights into the battery degradation mechanism, but also provides a scientific basis for the formulation of battery management and maintenance strategies. The application of Kalman filtering improves the accuracy of data processing and can accurately analyze the health status of the battery even when the data quality is disturbed. The implementation of this method significantly enhances the scientificity and practicality of battery performance monitoring.

[0081] In this embodiment, in step S3, sampling is performed from the sample data set according to the voltage interval to obtain sampled data; the sampled data is processed by the Gram angle field strategy to generate a Gram angle field map;

[0082] Wherein, in the process of processing the sampled data by the Gram angular field strategy to generate the Gram angular field map, the sampled data is normalized to obtain normalized data; the normalized data is converted by an inverse cosine function to obtain angled data; the angled data is calculated by a trigonometric function and an angle formula to quantify the correlation between each pair of time points to generate the Gram angular field map;

[0083] Specifically, to convert a time series consisting of n values ​​into a two-dimensional image, the original time series data must first be normalized to the range of [-1,1]. This step is necessary because it ensures the uniformity of the data in subsequent processing and the effectiveness of comparison. The mathematical expression of the normalization process is:

[0084]

[0085] In the formula, is the value obtained after normalization of the i-th sampling data; x i is the i-th sampling voltage data; x max is the maximum value in the sampled data sequence; x min is the minimum value in the sampled data sequence;

[0086] After the normalization process, the data values ​​are transformed by the arccosine function, encoding the value of each time series point as an angle in the polar coordinate system, and the timestamps corresponding to these values ​​are expressed as radii in polar coordinates. The arccosine function conversion expression is:

[0087]

[0088] In the formula, r i is the normalized time index; φ i is the angle value of the i-th sampled data after arccosine mapping; N is the length of the sampled data sequence;

[0089] This transformation of representing the time series as polar coordinates has two significant advantages. First, this transformation has a one-to-one bijective property, which establishes a direct correspondence between the original time series data points and their representation in polar coordinates. Second, this transformation effectively preserves the temporal information of the time series, because the radial coordinates in the polar coordinates are directly associated with the timestamps of the time series, ensuring that the continuity and order of time are not lost during the transformation.

[0090] In the final step of the transformation process, the correlation between each pair of time points can be quantified by subtracting the angles between different time points represented in polar coordinates using trigonometric functions and the angular formula:

[0091]

[0092] In the formula, φ i is the angle value of the i-th sampling data after arc cosine mapping.

[0093] This calculation reflects the dynamic relationships and interactions between time series data points, providing a rich source of information for further data analysis and pattern recognition.

[0094] In this embodiment, in Cell1 of the Oxford battery data set, the Gram Angle Field (GASF) is used to show the changes under different charge and discharge cycles. Figure 4As shown in the GASF chart, as the number of charge and discharge cycles increases, the GASF representation shows changes, and the time pattern of the original sequence is effectively preserved along the main diagonal. This shows that although the battery has undergone multiple cycles, its basic charge and discharge characteristics and time relationship can still be clearly identified and tracked through GASF, proving the application value and effect of this data conversion technology in battery performance analysis and health monitoring.

[0095] In this embodiment, in step S4, a battery health state assessment model is constructed based on the improved ResNet34 network; the battery health state assessment model is trained by setting a loss function and setting an optimizer to obtain the trained battery health state assessment model;

[0096] Specifically, a battery health status assessment model is built based on an improved ResNet34 network; two-dimensional images generated by the Gram Angular Field (GASF) method are input into the neural network together with their corresponding SOH labels. The GASF method effectively converts time series data into an image format that can capture changes in battery performance over time and provide rich features for deep learning models to learn.

[0097] Among them, convolutional neural network (CNN) is a key category in deep learning, especially in the field of image processing. The architecture of CNN is based on the principles of local receptive field, weight sharing and pooling, aiming to optimize computational efficiency by effectively reducing the number of network parameters and cleverly managing the translation invariance of images. A standard CNN mainly consists of convolutional layers, activation layers, pooling layers and fully connected layers.

[0098] ResNet34 is a relatively complex variant of the deep residual network in the CNN model family, which solves a key problem in deep learning networks - network degradation. This is achieved through its unique feature - "residual blocks". These blocks contain a shortcut mechanism that allows input data to bypass one or more layers and then be added to the subsequent results, thus avoiding the problem of gradient vanishing and network degradation. The performance of ResNet34 is characterized by its depth of 34 layers (excluding input and output layers), and it performs well in a variety of applications.

[0099] In a standard ResNet-34 architecture, the dimension of the last fully connected layer is usually the same as the total number of target classification categories, and this configuration is mainly used for image classification tasks. However, given that our research goal is to infer the state of health (SOH) from 2D graphical representations - these representations are obtained by applying the Gram Angle Field (GASF) transformation to the selected charging voltage sequence. Obviously, the task we are facing is regression rather than classification. Therefore, as Figure 5As shown, the present invention strategically modifies the architecture of the fully connected layer and reduces its output dimension to a single unit, so that the model can better handle the prediction of continuous values.

[0100] In this embodiment, the mean square error (MSE) is used as the loss function. MSE is a commonly used loss function that calculates the average of the sum of squares of the differences between the model's predicted values ​​and the actual values. This method is particularly effective in regression problems because it can amplify larger errors and force the model to prioritize reducing these larger prediction errors. At the same time, the training of the battery health status assessment model uses the RectifiedAdam (RAdam) optimizer. RAdam is a variant of the Adam optimizer that integrates the advantages of adaptive learning rate adjustment and the stability of stochastic gradient descent (SGD). This optimizer is particularly suitable for dealing with unstable training processes because it can reduce the sensitivity to learning rate selection in the early stages of training while avoiding the model from falling into local optimal solutions during training.

[0101] In this embodiment, the number of iterations of model training is set to 450 and the batch size is set to 16 based on experimental and resource utilization considerations. A higher number of iterations ensures that the model has enough opportunities to learn complex patterns in the data, while a moderate batch size can balance the computational burden of training and the speed of model updates. This configuration makes model training not only efficient, but also relatively stable on different hardware configurations.

[0102] The trained battery health status assessment model can accurately predict the SOH of lithium-ion batteries, thereby providing scientific decision support for battery management, extending the battery life, and optimizing the battery performance.

[0103] In this embodiment, in step S5, the target battery data is input into the trained battery health status assessment model for assessment, and the target battery health status assessment result is output.

[0104] Specifically, the trained model is used to predict the SOH of new battery data. The new battery data is also processed and normalized by GASF to ensure the quality of the input data. The SOH value output by the model provides an intuitive assessment of the current health status of the battery, which is extremely important for formulating battery maintenance and replacement strategies. By analyzing these predicted values, the performance and life of the battery can be effectively monitored, thereby optimizing the overall efficiency of the battery. Over time, the model will be continuously iterated and optimized based on the data and feedback collected in actual use to improve the accuracy of the prediction and the reliability of the model, ensuring the continued adaptability and foresight of the evaluation method.

[0105] In a possible embodiment, in order to evaluate the performance of the SOH estimation method proposed in the present invention, a verification example is provided as follows:

[0106] The present invention uses three indicators, namely, root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE), as evaluation criteria, as shown in the following formula:

[0107]

[0108] In the formula, x i is the true value of the lithium battery SOH; N is the length of the sampling data sequence; is the estimated value of SOH.

[0109] In the University of Maryland dataset, CS2_36 and CS2_37 were selected as training sets, and CS2_35 was selected as the validation set. As shown in Table 1, the training time was 47 minutes.

[0110]

[0111] Table 1 Experimental design

[0112] The subsequent processing and analysis steps are the same as before. Figure 6 and Figure 7 As shown, the comparison between the two-dimensional method and the one-dimensional method in the SOH estimation results for CS2_35 is shown;

[0113] like Figure 8 and Fig. 9 As shown in the figure, the comparison between the present invention and the TCN and GRU methods is given. The results show that the two-dimensional method is significantly better than the one-dimensional method, TCN and GRU in terms of SOH estimation accuracy, and has achieved the lowest error in each round of testing.

[0114]

[0115] Table 2 Experimental results

[0116] Table 2 lists the comprehensive comparison of the three indicators of MAE, RMSE and MAPE. The results also prove that the two-dimensional method is always better than other methods on this dataset.

[0117] In summary, the present invention obtains a lithium-ion battery data set by setting a database; preliminarily organizes the lithium-ion battery data set to obtain an organized data set; imports the organized data set into MATLAB for analysis and processing by setting a function to obtain a sample data set; determines the voltage interval for sampling the constant current charging voltage data of the lithium battery through a capacity increment analysis strategy; samples from the sample data set according to the voltage interval to obtain sampled data; processes the sampled data through a Gram angular field strategy to generate a Gram angular field diagram; constructs a battery health status assessment model based on an improved ResNet34 network; trains the battery health status assessment model by setting a loss function and setting an optimizer to obtain a trained battery health status assessment model; inputs the target battery data into the trained battery health status assessment model for evaluation, and outputs the target battery health status assessment result. The present invention utilizes the two-dimensional image conversion of GASF technology to enhance the spatial characteristic expression of the data, so that the ResNet34 model can more accurately capture the subtle changes in the battery aging process, and the estimation result is highly consistent with the actual battery performance degradation. The present invention combines the voltage segment division of ICA technology and the image recognition capability of ResNet34, which greatly improves the accuracy of SOH estimation. Compared with traditional methods, it shows obvious performance advantages and is suitable for complex and dynamically changing practical application scenarios.

[0118] It should be noted that the method of the embodiment of the present disclosure can be performed by a single device, such as a computer or a server. The method of the present embodiment can also be applied in a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present disclosure, and the multiple devices will interact with each other to complete the described method.

[0119] It should be noted that the above describes some embodiments of the present disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the above embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0120] Example 2

[0121] See also Fig.10 Embodiment 2 of the present invention also provides a lithium-ion battery health status assessment device based on image recognition, comprising:

[0122] The sample data set acquisition module 001 is used to acquire a lithium-ion battery data set by setting a database; preliminarily organize the lithium-ion battery data set to obtain an organized data set; and import the organized data set into MATLAB for analysis and processing by setting a function to obtain a sample data set;

[0123] The sampling voltage interval determination module 002 is used to determine the voltage interval for sampling the constant current charging voltage data of the lithium battery through a capacity increment analysis strategy;

[0124] A Gram angle field diagram generating module 003 is used to sample from the sample data set according to the voltage interval to obtain sampled data; and to process the sampled data using a Gram angle field strategy to generate a Gram angle field diagram;

[0125] The battery health state assessment model construction and training module 004 is used to construct a battery health state assessment model based on the improved ResNet34 network; the battery health state assessment model is trained by setting a loss function and setting an optimizer to obtain the trained battery health state assessment model;

[0126] The battery health status assessment model processing module 005 is used to input the target battery data into the trained battery health status assessment model for assessment, and output the target battery health status assessment result.

[0127] In this embodiment, in the sample data set acquisition module 001, in the process of importing the sorted data set into MATLAB for analysis and processing through the setting function to obtain the sample data set, the sorted data set is imported into MATLAB through the readtable or readmatrix function; the sorted data set is filtered and denoised to obtain a high-quality data set; the high-quality data set is processed through MATLAB to obtain the sample data set.

[0128] In this embodiment, in the sampling voltage interval determination module 002, in the process of determining the voltage interval of the constant current charging voltage data sampling of the lithium battery through the capacity increment analysis strategy, the battery terminal voltage is differentiated relative to the capacity through the capacity increment analysis strategy to obtain the voltage interval of the constant current charging voltage data sampling of the lithium battery; the differential processing calculation formula is:

[0129]

[0130] Where Q represents the battery charge; V represents the battery voltage; k is the sampling time; and L is the step size of the difference.

[0131] In this embodiment, in the Gram angle field map generation module 003, in the process of processing the sampled data by the Gram angle field strategy to generate the Gram angle field map, the sampled data is normalized to obtain normalized data; the normalized data is converted by an inverse cosine function to obtain angled data; the angled data is calculated by a trigonometric function and an angle formula to quantify the correlation between each pair of time points to generate the Gram angle field map;

[0132] The mathematical expression of the normalization process is:

[0133]

[0134] In the formula, is the value obtained after normalization of the i-th sampling data; x i is the i-th sampling voltage data; x max is the maximum value in the sampled data sequence; x min is the minimum value in the sampled data sequence;

[0135] The conversion expression of the arccosine function is:

[0136]

[0137] In the formula, r i is the normalized time index; φ i is the angle value of the i-th sampled data after arccosine mapping; N is the length of the sampled data sequence;

[0138] The formula for calculating the angled data by trigonometric functions and angle formulas is:

[0139]

[0140] In the formula, φ i is the angle value of the i-th sampling data after arc cosine mapping.

[0141] In this embodiment, in the battery health status assessment model construction and training module 004, in the process of training the battery health status assessment model through the setting loss function and the setting optimizer, the setting loss function is a mean square error loss function; the setting optimizer is a RectifiedAdam optimizer.

[0142] It should be noted that the information interaction, execution process and other contents between the modules of the above-mentioned system are based on the same concept as the method embodiment in Example 1 of the present application, and the technical effects they bring are the same as those of the method embodiment of the present application. For specific contents, please refer to the description in the method embodiment shown above in the present application, and will not be repeated here.

[0143] Example 3

[0144] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium, in which program code for a lithium-ion battery health status assessment method based on image recognition is stored, and the program code includes instructions for executing the lithium-ion battery health status assessment method based on image recognition of embodiment 1 or any possible implementation thereof.

[0145] The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more available media. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0146] Example 4

[0147] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;

[0148] The processor and the memory communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the lithium-ion battery health status assessment method based on image recognition of Example 1 or any possible implementation thereof.

[0149] Specifically, the processor can be implemented by hardware or by software. When implemented by hardware, the processor can be a logic circuit, an integrated circuit, etc.; when implemented by software, the processor can be a general-purpose processor implemented by reading software codes stored in a memory. The memory can be integrated in the processor or can be located outside the processor and exist independently.

[0150] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable systems. The computer instructions can be stored in a computer-readable storage medium, or transmitted from a computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center.

[0151] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing system, they can be concentrated on a single computing system, or distributed on a network composed of multiple computing systems, and optionally, they can be implemented by a program code executable by a computing system, so that they can be stored in a storage system and executed by the computing system, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0152] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. A lithium-ion battery health status assessment method based on image recognition, characterized in that: include: Get lithium-ion battery data set by setting database; Preliminarily sorting the lithium-ion battery data set to obtain a sorted data set; The sorted data set is imported into MATLAB for analysis and processing by setting a function to obtain a sample data set; Determine the voltage range for sampling the constant current charging voltage data of the lithium battery through the capacity increment analysis strategy; Sampling from the sample data set according to the voltage interval to obtain sampled data; processing the sampled data using a Gram angle field strategy to generate a Gram angle field map; Constructing a battery health status assessment model based on an improved ResNet34 network; training the battery health status assessment model by setting a loss function and setting an optimizer to obtain the trained battery health status assessment model; The target battery data is input into the trained battery health status assessment model for assessment, and the target battery health status assessment result is output.

2. The method for evaluating the health status of a lithium-ion battery based on image recognition according to claim 1, characterized in that: In the process of importing the sorted data set into MATLAB for analysis and processing through the setting function to obtain the sample data set, the sorted data set is imported into MATLAB through the readtable or readmatrix function; the sorted data set is filtered and denoised to obtain a high-quality data set; The high-quality data set is processed by MATLAB to obtain the sample data set.

3. The method for evaluating the health status of a lithium-ion battery based on image recognition according to claim 2, characterized in that: In the process of determining the voltage interval of the constant current charging voltage data sampling of the lithium battery through the capacity increment analysis strategy, the battery terminal voltage is differentiated relative to the capacity through the capacity increment analysis strategy to obtain the voltage interval of the constant current charging voltage data sampling of the lithium battery; the differential processing calculation formula is: Where Q represents the battery charge; V represents the battery voltage; k is the sampling time; and L is the step size of the difference.

4. The method for evaluating the health status of a lithium-ion battery based on image recognition according to claim 3, characterized in that: In the process of processing the sampled data by using the Gram angular field strategy to generate the Gram angular field map, normalizing the sampled data to obtain normalized data; The normalized data is converted by an inverse cosine function to obtain angled data; the angled data is calculated by a trigonometric function and an angle formula to quantify the correlation between each pair of time points and generate the Gram angle field map; The mathematical expression of the normalization process is: In the formula, is the value obtained after normalization of the i-th sampling data; x i is the i-th sampling voltage data; x max is the maximum value in the sampled data sequence; x min is the minimum value in the sampled data sequence; The conversion expression of the arccosine function is: In the formula, r i is the normalized time index; φ i is the angle value of the i-th sampled data after arccosine mapping; N is the length of the sampled data sequence; The formula for calculating the angled data by trigonometric functions and angle formulas is: In the formula, φ i is the angle value of the i-th sampling data after arc cosine mapping.

5. The method for evaluating the health status of a lithium-ion battery based on image recognition according to claim 4, characterized in that: In the process of training the battery health status assessment model by using the setting loss function and the setting optimizer, the setting loss function is a mean square error loss function; and the setting optimizer is a RectifiedAdam optimizer.

6. A lithium-ion battery health status assessment device based on image recognition, using a lithium-ion battery health status assessment method based on image recognition according to any one of claims 1 to 5, characterized in that: include: A sample data set acquisition module is used to acquire a lithium-ion battery data set by setting a database; Preliminarily sorting the lithium-ion battery data set to obtain a sorted data set; The sorted data set is imported into MATLAB for analysis and processing by setting a function to obtain a sample data set; A sampling voltage interval determination module is used to determine the voltage interval for sampling the constant current charging voltage data of the lithium battery through a capacity increment analysis strategy; A Gram angle field diagram generating module is used to sample from the sample data set according to the voltage interval to obtain sampled data; and to process the sampled data using a Gram angle field strategy to generate a Gram angle field diagram; A battery health status assessment model construction and training module is used to construct a battery health status assessment model based on an improved ResNet34 network; the battery health status assessment model is trained by setting a loss function and setting an optimizer to obtain the trained battery health status assessment model; The battery health status assessment model processing module is used to input the target battery data into the trained battery health status assessment model for assessment and output the target battery health status assessment result.

7. The lithium-ion battery health status assessment device based on image recognition according to claim 6, characterized in that: In the sample data set acquisition module, in the process of importing the sorted data set into MATLAB for analysis and processing through the setting function, the sorted data set is imported into MATLAB through the readtable or readmatrix function to obtain the sample data set; the sorted data set is filtered and denoised to obtain a high-quality data set; The high-quality data set is processed by MATLAB to obtain the sample data set.

8. The lithium-ion battery health status assessment device based on image recognition according to claim 7, characterized in that: In the sampling voltage interval determination module, in the process of determining the voltage interval for sampling the constant current charging voltage data of the lithium battery through the capacity increment analysis strategy, the battery terminal voltage is differentiated relative to the capacity through the capacity increment analysis strategy to obtain the voltage interval for sampling the constant current charging voltage data of the lithium battery; the differential processing calculation formula is: Where Q represents the battery charge; V represents the battery voltage; k is the sampling time; and L is the step size of the difference.

9. The lithium-ion battery health status assessment device based on image recognition according to claim 8, characterized in that: In the Gram angle field map generation module, in the process of processing the sampled data by the Gram angle field strategy to generate the Gram angle field map, the sampled data is normalized to obtain normalized data; The normalized data is converted by an inverse cosine function to obtain angled data; the angled data is calculated by a trigonometric function and an angle formula to quantify the correlation between each pair of time points and generate the Gram angle field map; The mathematical expression of the normalization process is: In the formula, x~ i is the value obtained after normalization of the i-th sampling data; x i is the i-th sampling voltage data; x max is the maximum value in the sampled data sequence; x min is the minimum value in the sampled data sequence; The conversion expression of the arccosine function is: In the formula, r i is the normalized time index; φ i is the angle value of the i-th sampled data after arccosine mapping; N is the length of the sampled data sequence; The formula for calculating the angled data by trigonometric functions and angle formulas is: In the formula, φ i is the angle value of the i-th sampling data after arc cosine mapping.

10. The lithium-ion battery health status assessment device based on image recognition according to claim 9, characterized in that: In the battery health status assessment model construction and training module, in the process of training the battery health status assessment model through the setting loss function and the setting optimizer, the setting loss function is a mean square error loss function; the setting optimizer is a Rectified Adam optimizer.

Images